Association Between Use of Desloratadine and Risk of Seizures, Supraventricular Tachycardia, and Atrial Fibrillation Or Flutter: a Nordic Register-Based Study

Home , Mizolastine

DESLORATADINE PAGE 1 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Association between use of desloratadine and risk of seizures, supraventricular tachycardia, and atrial fibrillation or flutter: A Nordic register-based study

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE P A G E 2 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 PASS INF OR MATI ON Titl e Ass o ci ati o n b et w een use of desloratadine and risk of seizures, s u pr a v e ntri c ul ar t a c h y c ar di a, a n d atri al fi brill ati o n or fl utt er: A N or di c r e gist er -based study V e rsi o n i d e ntifi e r of t h e 1 fi nal st u dy re port Date of last versio n of t he 1 0 May 2019 fi nal st u dy re port E U P AS register nu mber E UP AS15038 Active s u bsta nce Desloratadine, A T C code R06 A X27; Phar macotherapeutic group: A nti hist a mi n es – H 1 a nt a g o nist Medicinal product A E RI US, AZ O M YR, and NE OCL ARIT Y N Product reference E U/ 1/ 0 0/ 1 6 0, A E RI U S E U/ 1/ 0 0/ 1 5 7, A Z O M Y R E U/1/00/161, NE OCL ARIT Y N Procedure nu mber E ME A/ H/ C/000313/ ME A/065, AE RI US E ME A/ H/C/000310/ ME A/065, AZ O M YR E ME A/ H/C/000314/ ME A/065, NE OCL ARIT Y N Marketing authorisation Merck Sharp & Doh me B. V. hol der(s) Waarder w e g 3 9 2031 B N Haarle m The Netherlands Joint P ASS N o Research question and T o d es cri b e t h e us e of d esl or at a di n e i n t h e g e n er al p o p ul ati o n; t o o bjectives describe the incidence rates of first seizure, supraventricular t a c h y c ar di a, atri al fi brill ati o n or fl utt er , a n d first r e c urr e nt s ei z ur e; and to exa mine the associations bet ween desloratadine exposure a n d ris k of first s ei z ur e, s u pr a v e ntri c ul ar t a c h y c ar di a, atri al fi brill ati o n or fl utt er, a n d first r e c urr e nt s ei z ur e. Co u ntry( -ies) of st u dy D e n m ar k, Fi nl a n d, and S weden A ut hor A n n ett e Kj ær Ers b øll P P D of S o ut h er n D en mark St u di estr æ d e 6 D K -1 4 5 5 Copenhagen K Den mark P h o n e: P P D E -ma il: P P D Merck Final Repository 1 0 -J U N -2019 ( R C A M) Date

(E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE P A G E 3 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 MAR KETIN G AUTHORISATION HOLDER ( S) M ar k eti n g a ut h oris ati o n Merck Sharp & Doh me B. V. h ol der(s) Waarder w e g 3 9 2031 B N Haarle m The Netherlands M A H contact person P P D Gl o b al R e g ul at or y Aff airs – R AI E M E A R e gi o n al Center Merck Sharp & Doh me ( Europe), I n c. Lynx Binnenhof 5 1200 Brussels, Belgiu m P h o n e: P P D F a x: P P D E -ma il: P P D

(E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE PAGE 4 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: 4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

text

TABLE OF CONTENTS

TABLE OF CONTENTS ...... 4 LIST OF TABLES ...... 7 LIST OF FIGURES ...... 8 LIST OF ANNEXES...... 9 1 ABSTRACT...... 10 2 LIST OF ABBREVIATIONS ...... 14 3 INVESTIGATORS ...... 15 4 OTHER RESPONSIBLE PARTIES...... 15 5 MILESTONES ...... 16 6 RATIONALE AND BACKGROUND ...... 17 7 RESEARCH QUESTION AND OBJECTIVES ...... 17 8 AMENDMENTS AND UPDATES...... 18 9 RESEARCH METHODS...... 18 9.1 Study design...... 18 9.2 Setting...... 21 9.3 Subjects ...... 21 9.4 Variables ...... 24 9.4.1 Exposure ...... 24 9.4.2 Outcome...... 26 9.4.3 Covariates ...... 27 9.4.4 Other variables...... 31 9.5 Data sources and measurement ...... 31 9.5.1 Study Procedures ...... 32 9.6 Bias ...... 33 9.7 Study size ...... 36 9.7.1 Fixed sample size...... 36 9.7.2 Proportion of current and non-current exposure time among DL users for the purpose of calculation of minimum detectable IRR ...... 36 9.7.3 Background event rate ...... 37 9.7.4 Number of years in the study...... 37

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 5 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: 4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

text 9.7.5 Statistical parameters and assumptions...... 38 9.7.6 Example of calculating the minimum detectable IRR ...... 38 9.8 Data transformation ...... 39 9.8.1 Data management...... 40 9.9 Statistical methods ...... 42 9.9.1 Main summary measures ...... 42 9.9.2 Main statistical methods ...... 42 9.9.2.1 Substudy 1: Descriptive analysis of exposure of DL...... 42 9.9.2.2 Substudy 2: First seizure...... 43 9.9.2.3 Substudy 3: SVT...... 43 9.9.2.4 Substudy 4: Atrial fibrillation/flutter ...... 44 9.9.2.5 Substudy 5: First recurrent seizure ...... 45 9.9.2.6 Additional information for Substudies 2A, 3A, 4A, and 5A ...... 45 9.9.2.7 Additional information for Substudies 2B, 3B, 4B, and 5B ...... 46 9.9.3 Missing values ...... 46 9.9.4 Sensitivity analyses...... 47 9.9.5 Post hoc analyses ...... 56 9.9.6 Amendments to the statistical analysis plan ...... 57 9.10 Quality control ...... 58 10 RESULTS ...... 59 10.1 Participants...... 59 10.1.1 Protection of Human Subjects ...... 59 10.2 Descriptive data...... 60 10.3 Outcome data in the general population...... 60 10.4 Main results ...... 61 10.5 Other analyses ...... 62 10.5.1 Supplementary analyses...... 62 10.5.2 Post hoc analyses ...... 65 11 DISCUSSION...... 66 11.1 Key results ...... 66 11.2 Limitations...... 68 11.3 Interpretation ...... 70

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 6 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: 4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

text 11.4 Generalisability ...... 75 12 OTHER INFORMATION ...... 75 13 CONCLUSION ...... 75 REFERENCES...... 77

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 7 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: 4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

text

LIST OF TABLES

Table 1 Overview of Substudies...... 19 Table 2 Overview of the Outcome Variables in the Study...... 27 Table 3 Information on whether potential confounding factors are included in the DAGs and the minimum sufficient adjustment sets, which will be used for confounder adjustment in the association studies...... 29 Table 4 Definition of confounders included in the minimum sufficient adjustment sets that will be used for confounder adjustment in the association studies...... 30 Table 5 Overview of national health registers in the Nordic countries of relevance for the present study...... 32 Table 6 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rate (IR) of seizures is based on Hauser & Beghi (2008). A 1-sided test, significance level of 5% and power of 80% have been used...... 38 Table 7 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rate (IR) of SVT is based on Orejarena et al. (1998). A 1-sided test, significance level of 5% and power of 80% have been used...... 39 Table 8 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rates (IR) for atrial flutter were not available; therefore, we base the calculations only on IR of A-fib obtained from Wilke et al. (2013). A 1-sided test, significance level of 5% and power of 80% have been used...... 39 Table 9 Use of loratadine, rupatadine, and DL in the Nordic countries ...... 48 Table 10 Non-sedative antihistamines marketed in the Nordic countries ...... 49 Table 11 Definition of confounders included in the primary and alternative minimum sufficient adjustment set that will be used for confounder adjustment in the association studies...... 51

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 8 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: 4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

text

LIST OF FIGURES

Figure 1 Example of current exposed periods versus unexposed periods for study subjects included in the association analyses (Substudies 2B, 3B, 4B, and 5B) ...... 24

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 9 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: 4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

text

LIST OF ANNEXES

ANNEX 1 Flow Diagram of Study Population ...... 80 ANNEX 2 Tables with the Main and Supplementary Analysis Results for the Desloratadine Study...... 82 ANNEX 3 Tables with the Post hoc Analysis Results for the Desloratadine Study ...... 108 ANNEX 4 Study Protocol...... 127 ANNEX 5 Statistical Analysis Plan...... 216 ANNEX 6 Variable Definitions...... 269

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 10 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

1 ABSTRACT Title

Association between use of DL and risk of seizures, supraventricular tachycardia, and atrial fibrillation or flutter: A Nordic register-based study

Keywords

desloratadine, atrial fibrillation or flutter, supraventricular tachycardia, seizure, Nordic register-based study

Rationale and background

A post-authorization safety study was conducted to assess the potential risk of desloratadine (DL) exposure on seizures, supraventricular tachycardia (SVT), and atrial fibrillation or flutter (A-fib/flu).

Research question and objectives

The objectives of this study were to explore the use of DL in the general population (Substudy 1); to describe the incidence rate of first seizure (Substudy 2A); to examine the associations between DL exposure and risk of first seizure (Substudy 2B); to describe the incidence rate of SVT (Substudy 3A); to examine the association between DL exposure and SVT (Substudy 3B); to describe the incidence rate of A-fib/flu (Substudy 4A); to examine the associations between DL exposure and A-fib/flu (Substudy 4B); to describe the incidence rate of first recurrent seizure (Substudy 5A); and to examine the association between DL exposure and first recurrent seizure (Substudy 5B).

Study design

This study is an observational, nationwide, register-based study using person-specific linkage of data from the national population registers from Denmark, Finland and Sweden (“Nordic countries”), including all individuals who redeemed a prescription of DL and all individuals with a registered diagnosis of seizure, SVT, or A-fib/flu. The original study design also included data from Norway, but due to unanticipated delays in data availability, the design was modified to include data from the remaining Nordic countries only.

Setting

The study population consists of a cohort of DL users, a cohort of individuals with seizures, a cohort of individuals with SVT, and a cohort of individuals with A-fib/flu. The general population of the three Nordic countries was used to derive estimates of the risk time by age, year, and country.

Subjects and study size, including dropouts

The sample size is fixed, as it consisted of all individuals in three Nordic countries (Denmark, Finland and Sweden) who redeemed at least one prescription for DL or who received a diagnosis of seizure, SVT, or A-fib/flu.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 11 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

The total number of incident DL users is 1,840,239 individuals. The number of individuals with a diagnosis of: first seizure is 96,220, first SVT is 87,499, first A-fib/flu is 427,370, and first recurrent seizure is 17,819, respectively.

Variables and data sources

The exposure variable of interest is DL for which current use (i.e., exposed period) is defined as the period after each redeemed prescription equal to the number of days’ supply plus a 4-week grace period to account for intermittent use. To minimize exposure misclassification, we used person time at least 26 weeks beyond the dispensing date of the last prescription as the “unexposed” reference period.

The outcome variables are first seizure, first SVT diagnosis, first A-fib/flu diagnosis, and first recurrent seizure. Directed acyclic graphs were developed to identify the minimum sufficient adjustment set of confounders to include in the association analysis of each outcome.

Data were obtained from nationwide population registers, including the national patient registers, the civil registration systems, and the prescription registers.

Data analysis

A descriptive analysis of DL use in the general population was performed. Incidence rates of seizure, SVT, A-fib/flu, and first recurrent seizure were calculated. Among persons ever dispensed DL, the associations between DL exposure and first seizure, SVT, A- fib/flu, and first recurrent seizure was evaluated using Poisson regression of incidence rates accounting for confounding factors. Additional supplementary analyses were performed.

Results

The incidence rate (IR) of DL users was 732.2 per 100,000 person-years. The IR per 100,000 person-years in the general population of: first seizure was 38.3, first SVT was 34.8, and first A-fib/flu was 170.0.

The results of the main analysis found a higher rate of first seizure in DL exposed person- time compared with DL unexposed person-time. Across all age groups, the magnitude of the adjusted incidence rate ratio (aIRR) for seizure is small (aIRR 1.15, 95% CI 1.03; 1.29) adjusted for age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. The overall finding of increased risk of first seizure is driven by increased IR of first seizure in the 0-to 4 (aIRR 1.47, 95% CI 1.16; 1.87) and in the 5-19-year-old (aIRR 1.32, 95% CI 1.09; 1.59) age categories when comparing DL exposed and DL unexposed person-time. DL exposure status was not associated with first seizure in patients 20 years old or older.

This PASS found no association between DL exposure status and incident SVT in the main analysis or in supplementary analyses (aIRR 1.03 95% CI 0.92; 1.15).

This study also found a slightly higher rate of first A-fib/flu in DL exposed compared with DL unexposed person-time (aIRR 1.08, 95% CI 1.02; 1.13) adjusted for age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status and (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 12 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

severe rhinitis. The overall finding of increased rate of first A-fib/flu is driven by increased IR of first A-fib/flu in the 20-64 years age category (aIRR 1.16, 95% CI 1.07; 1.27) when comparing DL exposed and DL unexposed person-time. The aIRR of first A- fib/flu was consistent across sensitivity analyses that varied the exposure definition but was substantially attenuated in an analysis that adjusted for an alternative set of confounders (aIRR 1.01, 95% CI 0.96; 1.06).

Discussion

There was an association between the IR of first seizure and first A-fib/flu during the DL exposed person-time as compared to the DL unexposed person-time in the age-, sex-, country- and calendar-year adjusted analyses as well as in the fully confounder-adjusted analyses. Adjustment for confounding variables (asthma, rhinitis, chronic urticaria) had only a small effect on the estimates.

This PASS found that DL exposed person-time was associated with seizure and A-fib/flu and no association between exposed DL person-time and risk of first SVT.

For first seizure, this association was largely driven by the younger (0-to 4 and 5-to 19- year old) age groups. The aIRR is about 1.5 for 0-to 4-year-olds and 1.3 for 5-to 19-year- old age groups comparing DL exposed person-time to unexposed person-time. The associations were largely consistent across sensitivity analyses. Taken together with prior pharmacovigilance data, there is reasonable evidence to suggest that seizure should be considered an adverse reaction to DL. Even if causality is assumed, however, the absolute rate difference is 42 per 100,000 person year (PY) in the 0-4 and 17 per 100,000 PY in the 5-19-year age groups, indicating absolute increases in risk are small.

For first A-fib/flu, the association with DL persisted after adjustment for preselected confounders (aIRR 1.08, 95% CI 1.02; 1.13). In age-stratified analyses, the association was strongest for patients aged 20-64 years (aIRR 1.16, 95% CI 1.07; 1.27) and was not seen in the elderly, in whom baseline risk of this outcome is known to be highest. Sensitivity analyses adjusting for an alternative set of confounders attenuated the aIRR. Thus, residual confounding could plausibly explain the association between DL and A- fib/flu. Evidence is insufficient to conclude that the association between current DL use and A-fib/flu is causal.

Marketing Authorisation Holder(s) Merck Sharp & Dohme B.V. Waarderweg 39 2031 BN Haarlem The Netherlands

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE P A GE 13 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 Na mes a n d affiliatio ns of pri nci pal i nvestigators A n n ett e Kj ær Ers b øll P P D U ni v ersit y of Southern Den mark St u di estr æ d e 6 D K -1455 Copenhagen K Den mark P h o n e: P P D E- m ail: P P D

(E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE P A GE 14 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 2 LIST OF ABBREVIATIONS A E Adverse event A- fi b Atri al fi brill ati o n A- fl u Atri al fl utt er A-fi b/fl u Atri al fi brill ati o n or atri al fl utt er aI R R A dj ust e d i n ci d e n c e r at e r ati o P P D A T C A n at o mi c al T h er a p e uti c C h e mi c al Cl assifi c ati o n S y st e m C H F C hr o ni c h e art f ail ur e C H MP Co m mittee for Medicinal Products for Hu man Use C OP D Chronic obstructive pul monary disease D A G Dir e ct e d a c y cli c gr a p h D D D D efi n e d d ail y d os e D L Desloratadine DS U R Develop ment safet y u p d at e r e p ort E M A European Medicines Agency EU European Union I C D I nt er n ati o n al Cl assifi c ati o n of Dis e as es I R I n ci d e n c e r at e I R R I n ci d e n c e r at e r ati o M A H M ar k eti n g a ut h ori z ati o n h ol d er M S D Merck Sharp & Doh me, a subsidiary of M er c k & C o., I n c. P P D NS AE N o n -serious adverse event P A S S P ost -a ut h oris ati o n s af et y st u di es PI C P ers o n al i d e ntifi c ati o n c o d e PPV Po siti v e pr e di cti v e v al u e PR AC Phar macovigilance Risk Assess ment Co m mittee PS U R P eri o di c s af et y u p d at e r e p ort P V Phar macovigilance P Y Person -years O T C O ver -t h e -counter Q B A Q u a ntit ati v e Bi as A n al ys is S A E Serious adverse event S A P St atisti c al a n al y sis pl a n S C RI D S elf -c o ntr oll e d ris k i nt er v al d esi g n S D St a n d ar d d e vi ati o n S V T Supraventricular tach yc ar di a

(E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE P A GE 15 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 3 INVESTIGATORS Pri n ci p al i n v esti g at or Annette Kj ær Ersbøll, Professor P P D U ni v ersit y of Southern Den mark St u di estr æ d e 6 D K -1455 Copenhagen K Den mark P h o n e: P P D E -m ail: P P D C o or di n ati n g Den mark : A n n ett e Kj ær Ers b øll, Pr of ess or, P P D investigator for each U ni v ersit y of Southe rn Den mark, Den mark country i n w hi c h t h e Fi nl a n d : Eero Pukkala, Professor, School of Health st u d y is t o b e p erf or m e d S ci e n c es, U ni v ersit y of Ta mpere, Finland S weden : Kristi a n B oli n, Pr of ess or , D e p art m e nt of E c o n o mi cs, L u n d U ni v ersit y , S weden Sponsor contacts P P D , Phar maco e pi d e mi ol o g y Merck Sharp & Doh me, Corp., a subsidiary of M er c k & C o., I n c. U G 1 D - 6 0 351 N Su mney t o w n Pi k e N ort h W al es, P A 1 9 4 5 4 U nit e d St at es P h o n e: P P D E -m ail: P P D Ot h er c o nt a cts N A Vendor/ Collaborator P P D E waldsgade 3 D K -2200 Copenhagen Den mark P h o n e: P P D w w w.appliedecono mics.dk E x p ert c o ns ult a nts P P D

Investigators K a us hi k S e n g u pt a, R es e ar c h assist a nt, P P D , U ni v ersit y of Southern Den mark, Den mark T h or a M ajl u n d Kj ær ulff, P h D F ell o w, P P D , U ni v ersit y of Southern Den mark, Den mark

4 OT HER RESP ONSIBLE PA R TI ES

N o n e.

(E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE P A GE 16 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 5 MILEST ONES

Mil est o n e Pl a n n e d A ct u al d at e Co m ment s d at e St art of d at a c oll e cti o n 0 9- N O V - 0 9- N O V - 2016 2016

E n d of d at a c oll e cti o n 0 9- N O V - 1 7- S E P - Fi n al d at a fil es r e c ei v e d t o 2016 2018 st art c o nstr u ct d at a s et fr o m three Nordic c o u ntri es St art of d at a a n al y sis 2 8- F e b - 1 5- F e b - Dr aft a n al yt i c t a bl es 2 0 1 9 2019 R e gistr ati o n i n t h e E U 16- Sep- 16- Sep- P A S r e gist er 2016 2016 P R A C Assess ment Report 2 7- M a y - 2 7- M a y - C CI f or t h e P ost - A ut h oris ati o n 2016 2 0 1 6 Measure

P ost - A ut h oris ati o n 2 8- S E P - 0 1- F e b - P R A C C CI Measure procedure M E A 2 0 1 9 2019 0 6 5. 4 (s u b mitt e d)

Fi n al r e p ort of st u d y 0 9- N O V - 1 0- J U N - Approved ne w deadline r es ults 2017 2019 C CI

(E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE PAGE 17 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

6 RATIONALE AND BACKGROUND The European Medicines Agency (EMA) requested that Merck Sharp and Dohme (MSD) consider options for a post-authorization safety study (PASS; Category 3 PV activity) to investigate whether there is an association between desloratadine (DL) use and seizures, supraventricular tachycardia (SVT), and atrial fibrillation or flutter (A-fib/flu) in the general population. DL is a prescription oral antihistamine approved in the European Union (EU) for the relief of symptoms associated with allergic rhinitis and urticaria in both adults and children. Since market authorization, there have been a small number of adverse event (AE) reports of seizures and supraventricular arrhythmias in patients taking DL, but the case reports do not permit evaluation of the association. Denmark, Finland, and Sweden (hereafter also referred to as “Nordic countries”) offer unique opportunities for such a study by means of the existence of centralized registration of activities in the healthcare sector, covering complete populations over many years.

7 RESEARCH QUESTION AND OBJECTIVES

The research question under study is to examine the associations between DL use and seizures, SVT, and A- fib/flu in the general population.

Primary hypotheses (stated as null-hypotheses)

 There is no association between current DL use and firstseizure.  There is no association between current DL use and SVT.  There is no association between current DL use and A-fib/flu.

Primary objectives

 Describe the IRs of the following outcomes in the general population:

 Incident diagnosis of seizure;  Incident diagnosis of SVT; and  Incident diagnosis of A-fib/flu.

 Compare the risk of incident seizure among persons while currently exposed to DL to the risk among the same persons while unexposed to DL for the total DL population and stratified by age, while adjusting for relevant confounding factors.

 Compare the risk of incident SVT among persons while currently exposed to DL to the risk among the same persons while unexposed to DL for the total DL population and stratified by age, while adjusting for relevant confoundingfactors.

 Compare the risk of incident A-fib/flu among persons while currently exposed to DL to the risk among the same persons while unexposed to DL for the total DL population and stratified by age, while adjusting for relevant confounding factors.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE P A GE 18 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 Secondary h y p ot h esis (st at e d as a n ull- h y p ot h esis)

o There is no association bet ween current D L use and first recurrent seizure.

Secondary o bj e cti v es

o A mong persons who had a first seizure, describe the I R of first r e c urr e nt s ei z ur e. o A mong persons who had a first seizure, co mpare the risk of first recurrent seizure w hil e c urr e ntl y e xposed t o D L t o t h e ris k a m o n g t h e s a m e pers o ns w hil e unexposed to D L , w hil e a dj usti n g f or r el e v a nt c o nf o u n di n g f a ct ors.

8 A MEND MENTS AND UPDAT E S

Nu mber D at e S e cti o n of st u d y A mend ment or Reason pr ot o c ol u p d at e 1 24- J A N -2019 N/ A U p d at e t o C CI St atisti c al A n al y sis Pl a n

9 RESEARCH METHODS

T e xt t h at c a m e dir e ctl y fro m the approved protocol a n d S A P ar e sho wn in grey t e xt boxes .

9. 1 Study design

The proposed stud y will b e a n o bs er v ati o n al ( n o n -e x p eri m e nt al), n ati o n wi d e, r e gist er - based study usi n g d at a fr o m t h e N or di c n ati o n al p o p ul ati o n r e gist ers. T h e ass o ci ati o ns bet ween current D L use and seizures, current D L use and S V T, current D L use and A - fi b/fl u, a n d c urr e nt D L us e a n d first r e c urr e nt s ei z ur e will b e ass ess e d i n a n al y s es usi n g person -s p e cifi c li n k a g e of d at a [ T a bl e 1 ]. T h e c o h ort of all D L us ers will b e us e d f or describing D L use (Substudy 1). T h e c o h orts of i n di vi d u als w h o h a v e h a d first s e i z ur e, S V T, or A -fi b/fl u will b e us e d t o e x a mi n e t h e I Rs of first s ei z ur e, S V T, a n d A -fi b/fl u i n t h e g e n er al p o p ul ati o n ( S u bst u di es 2 A, 3 A, a n d 4 A, r es p e cti v el y ). A c o h ort st u d y d esi g n a m o n g all D L us ers (i. e., a ris k i nt er v al d esi g n i n cl u di n g t h e D L -o nl y c o h ort: [ R ef. 5. 4: 03 QP D Q] ) will be used for the association bet ween current D L exposure and first s ei z ur e, S V T, a n d A -fi b/fl u ( S u bst u di es 2 B, 3 B, a n d 4 B, r es p e cti v el y). T h e c o h ort of i n di vi d u als w h o h a v e h a d fi rst r e c urr e nt s ei z ur e will b e us e d f or esti m ati n g t h e I R of first recurrent seizure ( Substudy 5 A). A c o h ort st u d y design a mong all D L users who have experienced a first seizure will be used for the association bet ween current D L exposure a n d first rec urre n t seizure ( Substudy 5 B).

(E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE PAGE 19 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038 Table 1 Overview of Substudies

Primary objectives Substudy Aim 1 Descriptive analysis of DL use in the general population 2A Descriptive analysis of IR of first seizure in the general population 2B Association between DL exposure and risk of first seizure 3A Descriptive analysis of IR of first SVT in the general population 3B Association between DL exposure and risk of first SVT 4A Descriptive analysis of IR of first A-fib/flu in the general population 4B Association between DL exposure and risk of A-fib/flu Secondary objectives 5A Descriptive analysis of IR of first recurrent seizure in the population of persons who have experienced first seizure 5B Association between DL exposure and risk of first recurrent seizure

For the main analyses in the present study of the association between current DL use and the outcomes of interest (Substudies 2B, 3B, 4B, and 5B), person time exposed to DL will be determined from dispensing records. “Current use” (i.e., “exposed” period) will be defined for each prescription as the sum of days’ supply plus a 4-week grace period to account for intermittent use and a possible wash-out effect. If a new DL prescription redemption occurs during an exposed period (either during the period equal to the sum of the days’ supply or the 4-week grace period), the exposure period extends from that date with a period equal to the sum of days’ supply in the newly redeemed prescription plus a 4-week grace period. Note: Days’ supply will be calculated from the quantity of tablets or amount of solution dispensed and the standard daily dose based on the age of the patient (i.e., 6–11 months: 1 mg/day; 12 months-5 years: 1.25 mg/day; 6–11 years: 2.5 mg/day; ≥12 years: 5 mg/day). Because the drug is used “as needed”, it is quite possible that there may still be exposed days in the period after the latest exposed period (i.e., after the exposed period that includes the sum of the days’ supply plus a 4-week grace period). To minimize exposure misclassification, we propose to use person time at least 26 weeks beyond the dispensing date of the prior prescription as the “unexposed” reference period. The “unexposed” period is actually a period with remote exposure to DL because dispensing of DL is a condition of entering the study population. We will refer to the “remote exposure” period as “unexposed” (For more details, see [Section 9.4.1] and [Figure 1]). The time between the exposed period and the unexposed period is considered neither exposed nor unexposed.

The advantage of using this design restricted to DL users, in which the same persons may have both exposed and unexposed periods, is that we reduce confounding due to time- independent factors associated with DL use.

The study is divided into six parts described below (For more details, see [Section 9.9]).

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 20 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

1. Substudy 1 A descriptive analysis of DL use in the general population. To describe DL use in the general population, we will identify both prevalent and incident users of DL and describe the distribution of the number of redeemed DL prescriptions for the total population and stratified by country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status.

2. Substudies 2A and 2B

A descriptive analysis of the IR of first seizure overall in the general population and stratified by country, sex, and age will be conducted (2A).

Among persons ever dispensed DL, the association between exposure to DL use and first seizure will be evaluated using Poisson regression of the IR of first seizure for the total population and stratified by age when accounting for confounding factors (2B).

3. Substudies 3A and 3B

Descriptive analysis of the IR of SVT overall in the general population and stratified by country, sex, and age will be conducted (3A).

Among persons ever dispensed DL, the association between exposure to DL use and first SVT will be evaluated using Poisson regression of the IR of first SVT for the total population and stratified by age when accounting for confounding factors (3B).

4. Substudies 4A and 4B

Descriptive analysis of the IR of A-fib/flu overall in the general population and stratified by country, sex, and age will be conducted (4A).

Among persons ever dispensed DL, the association between exposure to DL use and first A-fib/flu will be evaluated using Poisson regression of the IR of first A-fib/flu for the total population and stratified by age when accounting for confounding factors (4B).

5. Substudies 5A and 5B

Among persons who have experienced a first seizure, a descriptive analysis of the IR of first recurrent seizure will be conducted (5A).

Among persons ever dispensed DL and who have experienced a first seizure, the association between exposure to DL use and first recurrent seizure will be evaluated using Poisson regression of the IR of first recurrent seizure when accounting for confounding factors (5B). This analysis will only be done if there are at least 10 persons with recurrent seizures.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 21 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

6. Supplementary analyses In total, 10 supplementary analyses will be performed to examine the robustness of the results. These are described in further details in [Section 9.9.2.8].

9.2 Setting

The cohort of individuals with redeemed DL prescriptions will be identified from the four Nordic national prescription registers. Similarly, the cohort of all individuals with seizures (first seizure and first recurrent seizure, the cohort of individuals with SVT, and the cohort of individuals with A-fib/flu will be identified from the four Nordic national patient registers. The population is comprised of all individuals with DL prescriptions and of all individuals with seizures, SVT, or A-fib/flu in the four Nordic countries in the period 2001–2015 in Denmark and Finland, 2008–2015 in Norway, and July 2005–2015 in Sweden (Table I of [Annex 5]). Individuals for Substudies 2B, 3B, and 4B become eligible for the study cohort upon first dispensing of DL, and individuals in Substudy 5B become eligible upon first seizure if they have also redeemed at least one DL prescription (See [Section 9.3]). Data will be available until and including 2015 for all countries.

9.3 Subjects

Inclusion criteria

Substudy 1

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

 General population living in Denmark, Finland, Norway, and Sweden on 01 January of each year.

Substudy 2A

1.0 Individuals who have experienced a seizure during the study period (seizure cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

2.0 General population living in Denmark, Finland, Norway, and Sweden on 01 January of each year.

Substudy 2B

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 22 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Substudy 3A

 Individuals who have experienced a SVT during the study period (SVT cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

 General population living in Denmark, Finland, Norway, and Sweden on 01 January of each year.

Substudy 3B

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

Substudy 4A

 Individuals who have had an A-fib/flu diagnosis during the study period (A-fib/flu cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

 General population living in Denmark, Finland, Norway, and Sweden on 01 January of each year.

Substudy 4B

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

Substudy 5A

 Individuals who have experienced a first recurrent seizure during the study period and have residential location in Denmark, Finland, Norway, or Sweden.

Substudy 5B

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort), have experienced a seizure (persons included in both the DL and seizure cohorts), and have residential location in Denmark, Finland, Norway, or Sweden.

Exclusion criteria

Substudy 2A

 Individuals with a diagnosis of seizure, epilepsy, or prescriptions of antiepileptic medicine before entering the study period, as they have prevalent disease.

 Individuals with a diagnosis of malignant brain tumor or head trauma before the first seizure, as they are at high risk of seizures due to causes other than DL use.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 23 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Substudy 2B

 Individuals with a diagnosis of seizures, epilepsy, prescriptions of antiepileptic medicine, malignant brain tumor, or head trauma before redemption of first DL prescription, as they have prevalent disease or are at high risk of seizures due to causes other than DL use.

 Individuals with a brain tumor (benign and malignant), initiation of treatment with antiepileptic medicine, or head trauma occurring after beginning of DL use will be censored at date of first occurrence, as they are at high risk of seizures due to causes other than DL use.

Substudies 3A and 4A

 Individuals with a diagnosis of SVT or A-fib/flu before entering the study period, as they have prevalent disease.

 Individuals with a diagnosis of congenital pre-excitation syndrome (e.g., Wolff Parkinson White) before entering the study period, as they are at high risk of cardiac SVT or A-fib/flu due to causes other than DL use.

Substudies 3B and 4B

 Individuals with a diagnosis of SVT or A-fib/flu before use of DL, as they have prevalent disease.

 Individuals with a diagnosis of congenital pre-excitation syndrome (e.g., Wolff Parkinson White) before use of DL, as they are at high risk of SVT or A-fib/flu due to causes other than DL.

Substudy 5A

 Individuals with a diagnosis of malignant brain tumor or head trauma between the initial seizure and first recurrent seizure, as they are at high risk of recurrent seizures due to causes other than DL use.

Substudy 5B

 Individuals with a diagnosis of epilepsy, prescriptions of antiepileptic medicine, malignant brain tumor, or head trauma before redemption of first DL prescription and first seizure, as they have prevalent disease or are at high risk of recurrent seizures due to causes other than DL use.

 Individuals with a brain tumor (benign and malignant), initiation of treatment with antiepileptic medicine, or head trauma occurring after redemption of first DL prescription and first seizure will be censored at date of first occurrence, as they are at high risk of recurrent seizures due to causes other than DL use.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 24 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

9.4 Variables 9.4.1 Exposure

The main exposure of interest in the present study is DL use identified in the national prescription registers by use of Anatomical Therapeutic Chemical Classification System (ATC) code R06AX27. The exploratory analysis will describe DL use based on data about persons who have redeemed at least one DL prescription (Substudy 1). In the association studies (Substudies 2B, 3B, 4B, and 5B), DL use is considered a time-varying variable, as the same person can be both exposed and unexposed during the study period depending on the time period from last redeemed DL prescription and the amount of days’ supply redeemed at the last DL prescription redemption.

For the association analyses, person time exposed to DL will be determined from dispensing records, and period of current use will be defined for each prescription as days’ supply starting from the date of redemption plus a 4-week grace period to account for intermittent use and a possible wash-out effect. To minimize exposure misclassification, we propose to use person time at least 26 weeks beyond the dispensing date of the prior prescription as the “unexposed” reference period (in supplementary analysis 7, an alternative period of 52 weeks beyond the prior prescription is used to define the unexposed reference period). If a new DL prescription redemption occurs during an exposed period (either during the period equal to the sum of the days’ supply or the 4-week grace period), the exposure period extends from that date with a period equal to the sum of days’ supply in the newly redeemed prescription plus a 4-week grace period. In addition, a prescription redemption before 26 weeks after previous prescription redemption would mark the start of a new exposed period (and would not result in an unexposed period). [Figure 1] below provides an example of how persons included in the association analyses enter and exit exposure periods during the study period.

Days’ supply will be calculated from the quantity of tablets or amount of solution dispensed and the standard daily dose based on the age of the patient (i.e., 6–11 months: 1 mg/day; 12 months–5 years: 1.25 mg/day; 6–11 years: 2.5 mg/day; ≥12 years: 5 mg/day).

Figure 1 Example of current exposed periods versus unexposed periods for study subjects included in the association analyses (Substudies 2B, 3B, 4B, and 5B)

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 25 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Figure 1 illustrates how subjects in the study population (i.e., restricted to persons with at least one redeemed prescription of DL) enter and exit exposed and unexposed periods. An exposed period is a period starting from the day of redemption of a DL prescription and includes the following period equal to the number of days’ supply in the drug packages purchased (illustrated with dark grey boxes) and a 4-week grace period (illustrated with the light grey boxes). Unexposed periods, which start 26 weeks after the dispensing date of the last DL prescription, are depicted with white boxes. The time between the end of the exposed period and the unexposed period is considered neither exposed nor unexposed; these periods are illustrated with shaded boxes. Moreover, the date at which an outcome (i.e., seizure, recurrent seizure, SVT, or A-fib/flu) occurs (if occurring) is marked with a black triangle. For example, person 1 experienced the outcome before redeeming DL for the first time (date of entering the study); and therefore, this person will be excluded. Person 2 experienced the first outcome while in an exposed period and the recurrent outcome occurred in an unexposed period. Person 3 was exposed to DL twice, but did not experience an outcome. Finally, person 4 was exposed to DL once and experienced the outcome in a period where the person was neither exposed nor unexposed; therefore, this outcome will not count toward the analysis of the association.

In a supplementary analysis, an individual’s exposure status is categorized according to time since last DL dispensing (periods 0–4, 5–8, 9–16, and 17–26 weeks each compared with >26 week since last DL dispensing), and individuals are considered unexposed in the period beyond 26 weeks after a DL prescription redemption until next DL prescription redemption, the end of the study period, death, emigration, or occurrence of the outcome, whichever comes first.

Loratadine, the parent compound of DL, is also available via prescription and over–the- counter (OTC) in the Nordic countries. We think it is somewhat unlikely that patients would switch between DL and loratadine; however, it is possible that some of the time counted as unexposed in the analyses could actually be time exposed to loratadine. Moreover, other non-sedating prescription antihistamines are available in the Nordic countries, and if the effect on the outcomes is driven by use of non-sedating antihistamines in general rather than the specific effect of DL exposure, these drugs should be examined to elucidate potential misclassification of exposure. To explore exposure misclassification (i.e., the influence of loratadine and other non-sedating prescription antihistamines), on the associations under study), we will conduct supplementary analyses, which will be specified in the statistical analysisplan.

Regarding missing values in variables from the prescription registers, normally no registration is interpreted as the person does not use the drug (e.g., no registration of DL use is interpreted as the person has not used DL). Prescription drugs are not available from sources other than the pharmacies, and all drugs purchased at pharmacies are included in the prescription register (the only exception are drugs supplied from hospitals; however, less than 1% of DL is supplied by the hospitals). We do not expect missing information for data in the prescription registry that will be used to determine DL exposure (e.g., number of packs per redemption). However, we will examine data to see whether missing information occurs. Missing data handling will be determined after review of the data, but before data from different registers are linked together (e.g., handling of missing values for exposure data is blinded to outcome status and vice versa)

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 26 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

and will be described in a data control report, which will be submitted with the study report. 9.4.2 Outcome

Four outcome variables will be used in the present study: first seizure (Substudy 2B), first SVT (Substudy 3B), first A-fib/flu (Substudy 4B), and first recurrent seizure (Substudy 5B). Except where noted, the case definition of seizure excludes febrile seizure, a condition in infancy and early childhood attributed to fever. However, supplementary analysis 2 will evaluate the outcomes of febrile and non-febrile seizures in children. (The diagnosis of febrile seizure is generally not used in adults.) Information on outcome variables will be obtained from the Nordic national patient registers, including diagnostic and treatment information for patients treated at the secondary and tertiary hospital level in all four Nordic countries using International Classification of Diseases (ICD)-10 codes. We will exclude seizure cases if they are registered with brain tumor (benign and malignant), stroke, or acute drug intoxication or overdose of drugs during the same hospitalization as the seizure. [Table 2] gives an overview of the outcome variables in the study. For all outcome variables, we will include the primary diagnoses from the emergency room or inpatient settings registered in the Nordic national patient registers.

The National Patient Registers are used for reimbursement of services to the hospitals. It is mandatory to enter specific information (such as name, date, main diagnosis) for the record to be established. Due to the requirements of the national registries, we do not expect missing data on the outcome variables. In Denmark and Finland, it is only possible to register hospitalizations into the Nordic national patient registers if complete information on the primary diagnosis, date, and hospital department is entered into the registration system. Therefore, no missing values should occur for these variables. This procedure is most likely the same for Sweden and Norway. Data in the Danish National Patient Register is automatically checked for missing codes, inconsistencies between diagnosis and gender, incorrect digits and errors in the personal identification code (PIC). If an error is detected, the record is returned to the source hospital for correction [Ref. 5.4: 04C0C2]. No studies have examined the validity of seizures in the total population, but Vestergaard et al (2006) examined the validity of the discharge diagnosis of febrile seizure in children using the National Patient Register (ICD10 code R56.0) [Ref. 5.4: 04C0C7]. The positive predictive value (PPV) was 92.8% (95% CI: 88.8– 95.7%). The sensitivity (defined as completeness by the authors) is 71.5% (95% CI: 66.3–76.4%) [Ref. 5.4: 04C0C7]. The diagnosis of A-fib and A-flu has been validated in the Danish National Patient Register [Ref. 5.4: 04C0BZ]. The PPV for the combined diagnosis of A-fib/flu (I48) was 92.6%. Other studies have found even higher PPVs [Ref. 5.4: 04C0BS], [Ref. 5.4: 04C0BQ]. No studies have validated the diagnosis for SVT (I47); however, we find it likely that this diagnosis has approximately the same validity in the National Patient Register as seen for A-fib/flu.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE P A GE 27 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 T a bl e 2 Overvie w of the Outco me Variables in the Study. Substudy Description of outco me variable Substudy 2 B First s ei z ur e: A n i n ci d e nt c as e of s ei z ur e is a p ers o n wit h t h e first diagnosis of seizure Substudy 3 B Supraventricular tach yc ar di a ( S V T): A n i n ci d e nt c as e of S V T is a p ers o n wit h first di a g n osis of S V T Substudy 4 B Atri al fi brill ati o n or atri al fl utt er ( A -fi b/fl u): A n i n ci d e nt c as e of A- fi b/fl u is a p ers o n wit h first di a g n osis of A -fi b or first di a g n osis of A -fl u. T h es e t w o di a g n os es will b e c o m bi n e d i nt o a c o m p osit e e n d p oi nt Substudy 5 B First r e c urr e nt s ei z ur e: A first r e c urr e nt c as e of s ei z ur e is a p ers o n wit h t h e di a g n osis of t h e second seizure I C D-10 c o d es ar e list e d i n A n n e x 6 ( Pr ot o c ol A n n e x 4) . N ot e t h at t h e c as e d efi niti o n of seizure in these substudies excludes febrile seizure

9. 4. 3 Covariates

I n t h e pr es e nt st u d y, we have used directed acy cli c gr a p hs ( D A Gs), als o c all e d c a us al di a gr a ms, f or confounder selection. D A Gs are a well -accepted methodology f or usi n g c a us al k n o wl e d g e a n d a s et of f or m al m at h e m ati c al pri n ci pl es f or s el e cti n g w hi c h v ari a b les to adjust for when perfor ming association anal ys es [ R ef. 5. 4: 045 R C Q] . T h e y pr o vi d e a s ys t e m ati c w a y t o e x pl or e t h e r el ati o ns hi ps b et w e e n t h e e x p os ur es, o ut c o m es, a n d c o v ari at es ( u ni dir e cti o n al, bi dir e cti o n a l, c a us al) a n d f a cilit at e d e ali n g wit h a l ar g e nu mber of potential confounders. D A Gs help make the assu mptions underly i n g a n a n al ys is e x pli cit. T h e s el e cti o n of v ari a bl es n e e d e d f or c o nf o u n d er a dj ust m e nt t o o bt ai n a n u n bi as e d esti m at e of t h e ass o ci ati o n u n d er st u d y is c all e d t h e mi ni m u m s uffi ci e nt adjust ment set of confounders. We used the open source and freel y a v ail a bl e s oft w ar e D A Gitt y f or the develop ment of the D A Gs [ R ef. 5. 4: 045 W X9] . D A Gitt y h el ps t h e resea rcher visualize the structure of relevant variables for the association under study i n cl u d e d i n t h e D A G, as w ell as t o i d e ntif y t h e mi ni m u m s uffi ci e nt a dj ust m e nt s ets available for confounder adjust ment. I n t h e follo wing paragraphs, we present a brief desc ri pti o n of t h e D A G pr o c ess; h o w e v er, t h e f ull D A G pr o c ess is d es cri b e d i n [Annex 4].

T h e f a ct ors list e d i n t h e l eft -hand colu mn of [ Table 3] were considered potential confounders (i.e., candidate variables) for the association bet ween D L use and seizures, S V T, a n d/ or A -fi b/fl u b as e d o n t h e lit er at ur e, as w ell as c o ns ult ati o n wit h a gr o u p of cli ni c al e x p erts. T h e p ot e nti al c o nf o u n d ers w er e r e vi e w e d d uri n g a D A G m e eti n g h el d o n January 6, 2 0 1 5. At t his m e eti n g, t h e r el ati o ns hi ps a m o n g t h e diff er e nt f a ct ors w er e dis c uss e d, a n d D A Gitt y w as us e d t o h el p i d e ntif y t h e mi ni m u m s uffi ci e nt a dj ust m e nt s et of confounders to include in the association analysis of each outco me. Three D A Gs including potential confounders of the association bet ween D L use and seizures, S V T , a n d A -fi b/fl u, r es p e cti v el y , were developed by t h e D a nis h n ati o n al i n v esti g at ors, M S D, P P D , a n d t w o cli ni c al e x p erts. Aft er t h e D A G m e eti n g, w e c o ns ult e d a d er m at ol o gist to confir m the correctness of the developed D A Gs, especiall y r e g ar di n g t h e r el ati o ns hi ps b et w e e n urti c ari a a n d t h e other potential confounders. The D A Gs helped clarify w h et h er the potential confounders listed were confounders, colliders, or inter mediate variables a n d w hi c h of t h e p ot e nti al c o nf o u n d ers c o nstit ut e d a mi ni m u m s uffi ci e nt a d j ust m e nt s et. W e s el e ct e d t h e fi n al mi ni m u m s uffi ci e nt a dj ust m e nt s et, w hi c h will b e us e d t o a dj ust f or (E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE PAGE 28 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

confounding factors in the association studies based on whether we found that a given combination of confounders would be obtainable from the registers. The final minimum sufficient adjustment set for the association studies (i.e., 2B, 3B, 4B, and 5B) are seen in the right-hand columns of [Table 3].

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 29 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038 Table 3 Information on whether potential confounding factors are included in the DAGs and the minimum sufficient adjustment sets, which will be used for confounder adjustment in the association studies. Potential confounding factors Potential confounders included in the Confounders included in the selected DAG minimum sufficient adjustment set for each association study Seizures SVT A-fib/flu Seizures SVT A-fib/flu

Age (age will be derived as a categorical and a continuous variable: Yes Yes Yes Yes Yes Yes Age groups [0–4 years, 5–19 years, 20– 64 years, ≥65 years] and years) Sex (male versus female) Yes Yes Yes Yes Yes Yes Country of residence (Denmark, Yes Yes Yes Yes Yes Yes Finland, Norway, Sweden) Calendar year (years ranging from 2001– 2015) Yes Yes Yes Yes Yes Yes

Drug overdose (other than Yes Yes Yes No No No desloratadine) Drug and alcohol abuse Yes Yes Yes No No No Diabetes (both type 1 and type 2) Yes Yes Yes No No No Use of hypoglycemic agents (oral anti- Yes No No No No No diabetics, insulin) Hypertension No Yes Yes No No No Thyroidism (both hypo- and hyperthyroidism, e.g., Grave’s disease, No Yes Yes No No No thyrotoxicosis) Structural heart disease: Left ventricular hypertrophy, left ventricular systolic No Yes Yes No No No dysfunction, chronic heart failure (CHF)

Seasonality (i.e., winter [December- February], spring [March – May], Yes Yes Yes Yes Yes Yes summer [June-August], and autumn [September-November]) Asthmatic status Yes Yes Yes Yes Yes Yes

Disease severity of rhinitis Yes Yes Yes Yes Yes Yes Chronic obstructive pulmonary disease Yes Yes Yes No No No (COPD) Smoking Yes Yes Yes No No No Inflammatory disease No Yes Yes No No No Metastatic disease Yes No No No No No Infections No Yes Yes No No No Stroke Yes Yes Yes No No No Chronic urticaria Yes Yes Yes Yes Yes Yes Unspecific autoimmune disease Yes Yes Yes No No No Type 1 allergy Yes Yes Yes No No No Antihypertensive treatment No Yes Yes No No No

[Table 4] below shows the variables included in the minimum sufficient adjustment sets in all four association analyses. In addition to being confounders included in the association analyses, the variables in the minimum sufficient adjustment sets will be used to characterize the population using DL in the descriptive study of DL use (Substudy 1). The

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 30 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

table describes how each of the variables will be operationalized. ICD-10 codes and ATC codes that will be used for the definition of the variables can be found in [Annex 6].

Table 4 Definition of confounders included in the minimum sufficient adjustment sets that will be used for confounder adjustment in the association studies. Confounders included in the Definition Data source minimum sufficient adjustment sets Age For the descriptive analyses of DL use, age will be defined as the age of the Civil purchaser at the date of prescription redemption and stratified into age groups registration (0–4 years, 5–19 years, 20–64 years, ≥65 years). system In the association studies, age is a time varying confounder; and therefore, risk time will be split up in years of age. We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

Sex Sex is a time-independent confounder and will be included as males versus Civil females in both the descriptive and association analyses. registration system Country Country of residence is a time-independent confounder, as persons will be Civil excluded if emigrating the country. Country of residence will be included with registration four categories, i.e., Denmark, Finland, Norway, or Sweden in both the system descriptive and association analyses. Calendar year For the descriptive analyses of DL use, calendar year is defined as the year of Prescriptio prescription redemption. n register In the association studies, calendar year is a time varying confounder; and therefore, risk time will be split up in calendar years.

Seasonality For the descriptive analyses of DL use, seasonality is defined as the season Prescriptio (i.e., winter [December-February], spring [March – May], summer [June- n register August], and autumn [September-November]) when the DL prescriptions are redeemed. In the association studies, seasonality is a time varying confounder; and therefore, risk time will be split up in seasons defined as winter (December– February), spring (March–May), summer (June–August), and autumn (September–November).

Asthma status Both for the descriptive and association analyses, asthmatic status is defined as a National binary variable indicating whether or not a person has redeemed treatment for patient asthma defined as at least two prescriptions of inhalant steroids within a six- register and month period and/or contacts hospitals with a diagnosis of asthma (including prescriptio both primary and secondary diagnoses) during a five-year period before first DL n register exposure. To distinguish persons treated for chronic obstructive pulmonary disease from those treated for asthma, first registered asthma treatment has to be redeemed when the purchaser was 45 years or younger.

Disease severity Both for the descriptive and association analyses, severity of rhinitis will be Prescriptio of rhinitis defined as binary variable indicating whether or not a person has received n register treatment for severe rhinitis. Persons with severe rhinitis will be identified from the prescription register as persons who have redeemed immunotherapy at least once during a five-year period before first DL exposure.

Chronic urticaria Both for the descriptive and association analyses, chronic urticaria status is National defined as a binary variable indicating whether or not a person has a registered patient diagnosis of chronic urticaria in the five-year period before first DL exposure. register

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 31 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

9.4.4 Other variables We will also obtain information on immigration and emigration, as well as date of birth and death status from the civil registration systems to be able to calculate the IRs in Substudies 2B, 3B, 4B, and 5B.

The interpretation of register-based variables with regard to missing information was discussed in [Section 9.4.1] and [Section 9.4.2]. For a few of the included register-based variables (e.g., sex and date of birth), all included persons should have information. We do not expect missing information on these variables; however, we will examine data to see whether missing information occurs. Missing data handling will be determined after review of the data, but before data from different registers are linked together (blinded to exposure to DL and outcome status) and will be described in a data control report, which will be submitted along with the study report.

9.5 Data sources and measurement

The DL study will include register information from four Nordic countries – Denmark, Finland, Norway, and Sweden. In addition to a long history of collecting high quality information on births, deaths, immigration and emigration, disease incidence, and activities in the healthcare sector [Ref. 5.4: 00W4D0], exceptional opportunities to perform register-based research are driven by the unique PIC introduced in the Nordic countries in the 1960’s and available to all persons with permanent residence in the Nordic countries [Ref. 5.4: 00W4G3]. The PIC makes it possible to link information at the individual level from several registers for scientific research purposes. The national prescription registers and national patient registers within each of the Nordic countries capture all the individual encounters of purchasing prescribed DL and allow sufficient longitudinal data to differentiate between first and recurrent seizures and to identify incident SVT and A-fib/flu cases. Person-specific use of DL will be elucidated from the national prescription registers by obtaining information on redemption of DL prescription for each person [Ref. 5.4: 00W4CX]. Person-specific information on seizures, SVT, and A-fib/flu will be derived from the Nordic national patient registers.

[Table 5] presents key information on the population-based health registers in the Nordic countries of relevance for the present study. Data will be available until and including 2015 for all countries. Data will be applied for including 2016 and used if available. The data extraction period refers to the longest period for which data on exclusion variables (e.g., seizures, SVT, or A-fib/flu before baseline) can be obtained.

The actual study period for each country is influenced by the periods of available data in the relevant registers as well as disease and drug-free periods applied when analyzing the data.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 32 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038 Table 5 Overview of national health registers in the Nordic countries of relevance for the present study. Country Register Denmark Finland Norway Sweden National prescription 1995–2015 1994–2016 2004–2016 2005–2016 register National patient register 1977–2015 (1) 1967–2015 (2) 2008–2015 1987–2015 (3)

Civil registration system 1968–2015 1967–2016 1964–2015 1965–2015 2001–2015 2001–2015 2010–2015 2006–2015 Estimated study period (15 years) (15 years) (6 years) (11 years)

Data extraction period 1977–2015 (or 1967–2015 (or 2008–2015 (or 1987–2015 (or most recent data most recent data most recent data most recent available) available) available) data available)

(1) Contacts with outpatient clinics (incl. emergency departments) since 1995. (2) Contacts with outpatient clinics (incl. emergency departments) since 1998. (3) Contacts with outpatient clinics (incl. emergency departments) since 2001.

The national prescription registers include information on the date of prescription redemption, information on the purchaser, and information on the drug redeemed (e.g., ATC code, number of pills, daily dose, pack size, and number of packs purchased) [Ref. 5.4: 045TYK]. The Nordic national patient registers include diagnostic and treatment information for patients treated at the secondary and tertiary hospital level [Ref. 5.4: 045TZL], [Ref. 5.4: 045W09]. Clinical experts have been consulted on how to include information from the prescription and patient registers. Information on date of birth, immigration, emigration, and death will be obtained from the civil registration systems [Ref. 5.4: 00W4G3], [Ref. 5.4: 045W09].

9.5.1 Study Procedures

This is an observational, register-based study and pre-existing health-related national register data will be the sole data source. According to Danish, Finnish, Norwegian, and Swedish law, register- based studies can be carried out without consent from the individual subjects where the processing takes place for the sole purpose of carrying out statistical or scientific studies of significant public importance and where such processing is necessary in order to carry out these studies. It is an absolute requirement that the publication of statistical or scientific results may never reveal the identity of individuals or otherwise compromise data subjects. We will obtain approval by the data agencies in the four countries before data management and data analyses will be performed.

To identify incident disease events, we apply a disease-free run-in period as long as possible for the registers in Denmark, Sweden and Finland. However, as the national patient register was established in 2008 in Norway, we apply a two-year disease-free run- in period in Norway. Furthermore, since the prescription registers were established in 2004 and July 2005 in Norway and Sweden, respectively, we will apply a drug-free run-in period to account for the potential bias from left-truncation. This means that a person with prescription redemption of DL in Norway and Sweden within the first 6 months after

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 33 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

establishment of the prescriptions registers will not be included in the analysis. In summary and with the aim to clarify the study periods, the study period for all substudies (unless explicitly specified in the SAP [Annex 5]) is 2001–2015 for Denmark and Finland, 2006–2015 for Sweden (given the drug-free period of 6 months) and 2010–2015 for Norway (given the two-year disease-free run-in period).

9.6 Bias

The present study utilizes observational data from nationwide population registers covering the entire population meaning the total population is included and that loss to follow-up is minimal, limiting the impact of selection bias on results. Furthermore, observational data are extracted from health registers which are established and operated for purposes not immediately related to the present study. This minimizes information bias related to the differential misclassification of outcomes, but introduces other types of limitations with respect to quantifying exposure, outcome, and confounders as outlined below:

Exposure: Use of DL  The prescription registers capture information on purchasing of drugs from pharmacies. The actual adherence of prescribed drugs and consumption cannot be established in these databases. This is a major limitation because the drugs concerned are taken on an as needed basis to treat symptomatic disease. Thus, it is difficult to establish when persons are truly exposed to the drug. Therefore, supplementary analyses will be performed to characterize the relationship of prescribing and timing of events of interest by categorizing time since DL prescription to the outcomes of interest and by using 52 instead of 26 weeks beyond previous prescription redemption as start of an unexposed period.

Left-truncation is a potential bias in the study meaning that we have no information on DL use before the start of registration of prescriptions in each country. Left-truncation bias in relation to DL use is only a relevant issue for Sweden and Norway since DL was available from 2001, but the prescription registers in Norway and Sweden were established in 2004 and 2005, respectively [Table 4]. Therefore, we do not know whether a DL user in the first year of registration in Norway and Sweden is a long-term user (prevalent user) or a first-time user (incident user). To account for this potential bias, a person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included. This is not a limitation in Denmark and Finland since DL was first approved in 2001 and the prescription registers in Denmark and Finland were established in 1995 and 1994, respectively. However, left-truncation bias might also occur when persons are immigrating into the study population. Therefore, we also exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden.

 There is a possibility that patients could purchase DL OTC; however, because of reimbursement systems in the Nordic countries, it is actually less expensive to purchase prescription DL. In Norway and Sweden, DL has only been allowed for prescription sale; However, in Denmark, DL has been available for OTC sale since (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 34 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

2013, but OTC sales only counted for less than 4% of the total amount of DL purchased in 2013. A supplementary analysis will examine the associations under consideration restricting the time period to include only years where DL was not allowed for OTC sale.  Loratadine, the parent compound of DL, is also available via prescription and OTC in the Nordic countries. We think it is somewhat unlikely that patients would switch between DL and loratadine. However, it is possible that some of the time counted as unexposed in the analyses could actually be time exposed to loratadine or OTC DL. Moreover, other non-sedating prescription antihistamines are available in the Nordic countries, and if the effect on the outcomes is driven by non-sedating prescription antihistamines in general, rather than the specific effect of DL exposure only, these drugs should be examined to elucidate potential misclassification of exposure. To explore exposure misclassification (i.e., the influence of loratadine and other non- sedating prescription antihistamines on the associations under study), we will conduct supplementary analyses, which will be specified in the statistical analysis plan.

Outcome: Episodes of seizures, SVT, and A-fib/flu

 Seizures, SVT, or A-fib/flu not registered with a relevant diagnosis code in the national patient registers will not be included in the study. However, the completeness of registration is assumed high, since both a procedure for data control of information in the national patient registers is established, and hospitals have an incentive to register patients, as the financing of hospitals is based on the registration of patients treated and procedures performed. If under-recording exists, it would reduce the number of outcomes evaluated, but this under-recording is not likely to be differential (i.e., associated with periods of use or nonuse of DL) and introduce bias.  The coverage of calendar years in the national patient registers differ between the Nordic countries and may be insufficient to capture the complete history of seizures, SVT, or A- fib/flu at the individual level. Regarding seizures, recurrent seizures would be studied as primary seizures, if first seizure happened before start of registration.  Truncation is a potential bias in the study meaning that we have no information on seizures, SVT, and A-fib/flu before the start of registration of each of these variables. For Denmark, Finland, and Sweden, information on seizures, SVT, and A-fib/flu is available from 1977, 1967, and 1987, respectively. Thus, the impact of truncation bias on results from these countries is limited. However, information on seizures, SVT, and A-fib/flu in Norway is only available from 2008, which limits the study period of the Norwegian data further and increases the risk of truncation bias.  The present study aims at studying incident outcomes occurring in the community setting. To differentiate between incident and prevalent disease, we exclude persons with registered disease before baseline. In addition, we used surrogates of prevalence (i.e., redeemed prescriptions of antiepileptic medicine) to exclude persons with prevalent disease. To capture disease cases emerging in the community setting, we limit ascertainment to those sites of care where incident presentation would be present and use only primary diagnoses of hospitalization, as secondary diagnoses often represent conditions emerging duringhospitalization. (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 35 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Confounders

In the present study, we expect complete information on the central confounders of age, sex, calendar year, country, and seasonality, as these are measured as key variables in the registers and information on these variables has to be entered into the system to establish a record file. Information on the three confounders of severity of rhinitis, asthmatic status, and chronic urticaria status is challenging to obtain from the nationwide registers for many reasons. First of all, these conditions in general emerge in the primary health care sector where patients normally seek care from their general practitioner for treatment of symptoms. Second, truncation bias may, as with regard to the measures of exposure and outcome, also occur in relation to these three confounders in Norway as the patient register was established in 2008, making the follow-up period rather short. Third, different issues regarding codes and use of drugs for the three conditions have to be realized and discussed to understand the potential impact of residual confounders in the present study. We have discussed these issues with clinical experts and the essence of the discussion is given below.  Severity of rhinitis is measured as use of immunotherapy and the binary measure of severity of rhinitis will not capture different levels of disease severity. Therefore, we expect the sensitivity of the measure to be rather low and severity of rhinitis to be underreported in the present study. However, we do not expect misclassification of those persons who received immunotherapy, as it is assumed that persons receiving immunotherapy also have severe rhinitis. A QBA examining the impact of misclassification of severity of rhinitis on the results will beperformed.  Identifying persons with asthma by use of register data is challenging due to the fact that the drugs used for treatment of asthma overlap with the treatment of chronic obstructive pulmonary disease (COPD). In order not to misclassify persons with COPD into the group of persons with asthma, we identify persons with asthma as those initiating asthma treatment before the age of 45 years. However, this means that for the part of the population aged 45 years or older, we will only have information on asthmatic status from the patient register in which hospitalized cases are registered. These cases are more likely to have severe asthma than those identified through the prescription register meaning that residual confounders may be more likely to occur among the population older than 45 years than those younger. A QBA examining the impact of misclassification of asthma on the results will be performed.  Chronic urticaria is a very rare diagnosis given at highly specialized hospital units. This means that we expect a delay from onset of symptoms of chronic urticaria and the diagnosis of chronic urticaria of approximately 2–5 years. In the period between onset of disease and diagnosis, persons may receive antihistamines, including DL prescribed by the general practitioner, for their symptoms. Therefore, chronic urticaria may be underreported in the patient register and the sensitivity of the measure is expected to be somewhat low. To get an impression of how important this misclassification will be in the present study, we will look at how many of the persons with a diagnosis of chronic urticaria have redeemed prescriptions of DL during the five years prior to the date of diagnosis. A QBA examining the impact of misclassification of chronic urticaria on the results will be performed.

 As in any association study, there is always the possibility of unmeasured (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 36 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

confounders or residual confounding resulting from roughly categorized confounders that could affect the results of the study.

9.7 Study size

The proposed study will be performed in the framework of an observational design with the use of register-based data. The primary interest is to assess the association between current DL exposure in the general population and the outcomes (i.e., first seizure [Substudy 2B], first SVT [Substudy 3B], and first A-fib/flu [Substudy 4B]). For Substudies 2B, 3B, and 4B, a cohort study design among DL users will be used. The sample size in the present study will be fixed, as the study population for the association studies 2B, 3B, and 4B consists of all individuals in the four Nordic countries who have redeemed a DL prescription at least once. Hence, the aim of this section is to calculate the minimum detectable incidence rate ratio (IRR). Calculations concern Substudy 2B evaluating the association between DL use and first seizure, Substudy 3B evaluating the association between DL use and first SVT event, and Substudy 4B evaluating the association between DL use and first A-fib/flu. The actual annual IR of seizures, SVT, and A-fib/flu among current DL users is unknown and will be determined in the current study.

Minimal detectable IRRs were calculated based on a rearranged formula for sample size calculations developed by Bryant & Morganstein [Ref. 5.4: 045WWJ] (Formula provided in [Annex 5]). The following are the parameters needed to calculate the minimum detectable IRR: the number of DL users in each age group (i.e., the fixed sample size); the proportion of exposed time (current exposure) and unexposed time (non-current/remote exposure) for DL users; the background event rates of seizures, SVT, and A-fib/flu; and the number of years included in the study along with statistical parameters and assumptions included in the equation (i.e., the significance level, power, and 1-sided versus 2-sided tests).

9.7.1 Fixed sample size

The number of DL users from each country has been estimated based on national drug sale statistics [Annex 6]. The annual unique number of DL users in Denmark, Finland, and Sweden for age groups 0–4, 5–19, 20–64, and ≥65 years is estimated to be 20,000, 86,000, 215,000, and 44,500, respectively. Data for Norway on DL prescriptions and seizures/SVT/A-fib/flu are available for a shorter time period; therefore, Norway is not included in the calculations of the minimum detectable IRR. The number of annual unique DL users in the four countries is thereby slightly underestimated. 9.7.2 Proportion of current and non-current exposure time among DL users for the purpose of calculation of minimum detectable IRR

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 37 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

obtained from the feasibility study performed prior to developing the synopsis and the present protocol. For the minimum detectable IRR calculations, the mean number of prescriptions in each age group per year from the Danish data is used as an estimate of the average annual number of redemptions. The exposure period is then calculated as follows: (mean annual number of prescriptions in each age group*4 weeks (standard days’ supply)) + 4 weeks grace period to account for intermittent use and a possible wash-out effect. Beyond this period, individuals are assumed to be unexposed (i.e., non- current users). The average number of prescriptions per year was estimated to be 1.50, 1.62, 1.78, and 2.30 for age groups 0–4, 5–19, 20–64, and ≥65 years, respectively1.

Exposure periods for age groups 0–4, 5–19, 20–64, and ≥65 years were then calculated to be 10.00, 10.48, 11.12, 13.20 weeks, respectively. The proportion of exposed time in each age group is calculated as number of weeks exposed divided by number of weeks in a year (i.e., 52 weeks). For example, for age group 0–4 years, the proportion of exposed time is 10 weeks/52 weeks = 19%, and thus, the proportion of non-exposed time is 81%. Note: A simplified method was used to calculate exposed and unexposed person time for each age group for the purpose of estimating the minimum detectable IRR. The method for calculating exposed, unexposed, and neither exposed nor unexposed time for individuals for the study analysis is outlined in [Section 9.4.1].

1 By using Danish figures we might get conservative estimates of the minimal detectable IRR as mean number of prescriptions calculated for the feasibility study showed slightly lower means for Denmark compared to Sweden (figures from Finland and Norway were not available for these calculations). The average number of prescriptions per age group for the 20–64 and ≥65 year’s age groups was calculated as the mean number of annual prescriptions (2001-2012) per age group using data from Denmark for ages ≥16 years. For the 5–19 years age group, the calculation was based on data from Denmark for ages 16–19. For the 0–4 years age group, the calculation was based on Swedish data from the feasibility study, which showed a tendency for a lower number of prescriptions in this age group compared to the older age groups.

9.7.3 Background event rate

The background IR of seizures, SVT, and A-fib/flu was obtained from published studies [Ref. 5.4: 00W4CY], [Ref. 5.4: 03XNM4], [Ref. 5.4: 03XNL7]. IRs of seizure among unexposed individuals for age groups 0–4, 5–19, 20–64, and ≥65 years were estimated to be 65, 50, 40, and 40 per 100,000 person-years, respectively [Ref. 5.4: 00W4CY]. The IR of SVT is reported to be 13/100,000 person-years for persons aged 19 years or younger, 27/100,000 person-years for persons aged 20–64 years, and 122/100,000 person-years for persons aged 65 years or older [Ref. 5.4: 03XNM4]. Wilke et al. (2013) [Ref. 5.4: 03XNL7] reported the incidence of A-fib to increase markedly with age with 0.0016 cases/1000 person-years in children (<15 years) and approximately 30 cases/1000 person-years for persons aged 80 years or older. No data on the epidemiology of A-flu were available. 9.7.4 Number of years in the study

The maximum number of years (t) that an individual can be included in the study is 5 years for the age group 0–4 years and 12 years for the remaining age groups (≥5 years). The period of 12 years corresponds to the period where DL has been on the market. For the calculations of the minimum detectable IRR, we assume that individuals, on average, will be included in the study for a shorter period; and therefore, the maximum number of years included has been set to 2 years for ages 0–4 years and 6 years for ages ≥5 years.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 38 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

9.7.5 Statistical parameters and assumptions

The minimum incidence rate ratio, IRR>1, that can be detected is calculated using the following parameter values and assumptions:

Significance level, α: 0.05 Power, 1-β: 0.80 1-sided test

9.7.6 Example of calculating the minimum detectable IRR

The mean number of DL prescriptions in the age group 0–4 years is 1.50 prescriptions. This gives an exposure period of (1.50*4 weeks) + 4 weeks = 10 weeks. In this age group, 20,000 children have redeemed DL (i.e., the fixed sample size for this age group). Due to the exposure period of 10 weeks in this group, the proportion of the time exposed to DL is 10 weeks/52 weeks per year = 19%. The background annual rate of seizures is 65/100,000 in this age group. It is assumed that individuals in this age group will be included in the study for an average of 2 years. When using a 1-sided test, significance level of 5% and power of 80%, the minimum detectable IRR will be 3.0.

Table 6 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rate (IR) of seizures is based on Hauser & Beghi (2008). A 1-sided test, significance level of 5% and power of 80% have been used. Age Annual IR of seizures Mean number of Estimated number of Minimum detectable group among unexposed individuals with at weeks exposed based on IRR t=2 and 6 years (years) individuals per least one annual DL mean number of DL (for age 0-4 years and 100,000 person-years prescription in the prescriptions perperson age ≥5 years) Nordic countries 0–4 65 20,000 10.00 3.0 5–19 50 86,000 10.48 1.5 20–64 40 215,000 11.12 1.3 ≥65 40 44,500 13.20 1.7

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 39 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038 Table 7 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rate (IR) of SVT is based on Orejarena et al. (1998). A 1-sided test, significance level of 5% and power of 80% have been used. Age group Annual IR of SVT among Mean number of Estimated number of Minimum (years) unexposed individuals per individuals with weeks exposed based detectable IRR t=2 100,000 person- years at least one on mean number of and 6 years (for age annual DL DL prescriptions per 0- 4 years and age prescription in the person ≥5 years) Nordic countries 0–4 13 20,000 10.00 8.2 5–19 13 86,000 10.48 2.1 20–64 27 215,000 11.12 1.4 ≥65 122 44,500 13.20 1.4 Table 8 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rates (IR) for atrial flutter were not available; therefore, we base the calculations only on IR of A-fib obtained from Wilke et al. (2013). A 1-sided test, significance level of 5% and power of 80% have been used. Age Annual IR of A-fib Mean number of Estimated number of Minimum detectable group among unexposed individuals with at weeks exposed based on IRR t=2 and 6 years (years) individuals per 100,000 least one annual DL mean number of DL (for age 0- 4 years and person-years prescription in the prescriptions perperson age ≥5 years) Nordic countries 0–4 1.6 20,000 10.00 54.0 5–19 3.8 86,000 10.48 3.4 20–64 168.9 215,000 11.12 1.6 ≥65 2154.3 44,500 13.20 1.1

9.8 Data transformation

The outcome variable in Substudy 2 was first seizure, which was defined as a first seizure diagnosed in the study period. Seizure was identified in the patient registers among all inpatient hospitalizations and emergency room encounters or contacts with seizure as the primary diagnosis.

The outcome variable in Substudy 3 was first SVT, which was defined as a first SVT diagnosed in the study period. SVT was identified in the patient registers among all inpatient hospitalization and emergency room encounters or contacts with SVT as the primary diagnosis.

The outcome variable in Substudy 4 was first A-fib/flu, which was defined as a first A- fib/flu diagnosed in the study period. A-fib/flu was identified in the patient registers among all inpatient hospitalization and emergency room encounters or contacts with A- fib/flu as the primary diagnosis.

The outcome variable in Substudy 5 was first recurrent seizure, which was defined as a first recurrent seizure diagnosed in the study period among individuals who have had a first seizure. Seizure was identified in the patient registers among all inpatient hospitalization and emergency room encounters or contacts with seizure as the primary diagnosis.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 40 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

The construction of the potential confounding factors is described below:  Age: Age was categorized into 4 age categories (0-4 years, 5-19 years, 20-64, ≥ 65 years)

 Calendar time: 1-year categories (2001, 2002, …, 2015)

 Seasonality: Winter (December, January, February), spring (March, April, May), summer (June, July, August) and autumn (September, October, November)

 Country: Denmark, Finland, Sweden

All diseases included as potential confounders were identified in the patient registers among all hospital encounters (contacts) with specific primary or secondary diagnosis within the last 5 years before the first DL prescription redemption. See [Annex 6] for the specific ICD-10 codes used. The diseases included were asthma, chronic urticaria, rhinitis, diabetes, hypo/hyper-thyroidism, inflammatory disease, infections, and type 1 allergy.

All drugs included as potential confounders were identified in the prescription registers among all prescriptions with specific ATC-codes the last 5 years before the first DL prescription redemption. See [Annex 6] for specific ATC-codes. The drugs included were for treatment of asthma, severe rhinitis, diabetes, hypo/hyper-thyroidism, inflammatory disease and type 1 allergy.

9.8.1 Data management

The handling of data in the DL study involves six steps and requires applications and approvals for access to data in each of the four Nordic countries. In addition to the acquisition and management of data, a primary scientific coordinator will be responsible for the overall study and establishment of a joint Nordic study dataset. Four national scientific coordinators will be responsible for steps 1–4 in each country, whereas the Danish scientific coordinator also will be responsible for steps 5–6.

The handling of data is categorized into the following six steps:  All national scientific coordinators will apply to the relevant agencies for permission to perform the study and to get access to data.

 All national scientific coordinators will facilitate the construction of the study populations:

 Study population consisting of all DL users during the study period.

 Study population consisting of all individuals with seizure during the study period and the years prior to the study period.

 Study population consisting of all individuals with SVT during the study period and years prior to the study period.

 Study population consisting of all individuals with A-fib/flu during the study

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 41 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

period and years prior to the study period. All national scientific coordinators are responsible for acquiring and validating the datasets and will explore how the data can be combined from the registers. Data quality control includes but is not restricted to check for legal values for each categorical variable, check of consistency between dates (at least date of birth before all other dates and date of death after all dates), and check and advise on the handling of missing data. All national scientific coordinators will produce a data control report describing the checks performed. All national scientific coordinators will derive the final dataset from the data obtained from the registers by combining the registers according to the study designs. All national scientific coordinators will produce a data control report describing the checks performed of the final dataset, including reasons for modifications and exclusions. In this process, all national coordinators have to agree on the reasons for exclusion (e.g., missing value on crucial variables, chronological errors in the relation between dates, non-legal values of categorical variables, and extreme values of continuous variables). The national scientific coordinators should send the final data set and the data control reports that describe the data and provide suggestions on how to handle missing values and invalid codes to the Danish scientific coordinator. A template for the data control report will be provided by the Danish scientific coordinator and will include the following requirements:

1. Information on known misclassification of each variable (e.g., underreporting, low sensitivity or specificity, categorization with obvious invalid values, etc.).

2. Check for legal values for each categorical variable. Check for reasonable distribution of variables. Include advice on how to handle unexpected observations.

3. Check for reasonable minimum, maximum, and central tendency (median, mean) for each continuous variable. Check for outliers must be performed (e.g., exploratory plots, such as box-whiskers plot). Include advice on how to handle unexpected observations.

4. Check of chronological relation between date variables: At least date of birth before all other dates and date of death after all dates, but also reasonable relation of dates of diagnoses and prescriptions. Unexpected patterns should be described and solutions for handling such observations should beincluded.

5. Check of missing information on variables and include advice on how to handle such observations (e.g., delete observations with missing information, put missing observations into a specific category, etc.).

 The final datasets from Finland, Norway, and Sweden are transferred to Statistics Denmark where all subsequent data handling is done by the Danish scientific coordinator.

 The Danish scientific coordinator combines data as described by the document developed by all national scientific coordinators and the datasets from all countries will be joined into a combined analysis dataset. Relevant variables will be derived.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 42 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

 The Danish scientific coordinator will assess the data validity of all countries by logical checks, examination of extreme values, and missing data. It is important that identification numbers are maintained to facilitate linkage back to the original datasets to be able to check the data and for the sake of transparency.

9.9 Statistical methods

9.9.1 Main summary measures

Frequency distributions (number, percentage), mean and standard deviation, minimum, median and maximum values, and incidence rates were used as summary measures for descriptive analyses.

9.9.2 Main statistical methods

9.9.2.1 Substudy 1: Descriptive analysis of exposure of DL

A cohort study describing DL use in the general population will be performed. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescription registers will not be included in order to avoid left- truncation bias. In addition, to avoid analogous misclassification of prevalent dispensings with incident DL dispensings among immigrants, we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. Prevalent users will be defined as persons who have at least one prescription of DL in the period of interest (e.g., the entire study period or each year). Incident users are first time users of DL in the period of interest. The following descriptive analyses will be performed:

 The distribution of prevalent and incident users in the total population and stratified by country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status.

 Descriptive information on the mean, standard deviation, median, and maximum and minimum number of redeemed DL prescriptions in the total population and stratified by country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status.

In accordance with the remaining analyses, the number of individuals in the general population each year will be obtained from the NORDCAN database and used as an estimate of the number of person-years at risk assuming each person in the population each year contributes one person-year at risk [Ref. 5.4: 045VPW]. When calculating the IRs, the population at risk is defined as the population on July 1 of the given year for Denmark and Norway, the average of the population on January 1 and December 31 of the given year for Sweden and the average of the population on December 31 that year and December 31 the previous year for Finland (i.e., the population at risk in 2001 is the average of the population at risk on December 31, 2001 and December 31, 2000). When calculating the prevalence proportion, the denominator will be the population on January 1 of the given year. The 95% confidence intervals (CI) will be calculated for both IRs and prevalence proportions. (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 43 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

9.9.2.2 Substudy 2: First seizure Substudy 2A

A cohort study estimating the incidence of first seizures in the general population will be conducted. The cohort of individuals with seizures and the distribution of the population by age, year, and country will be used to derive estimates for the general population. The number of individuals in the general population on 01 January of each year will be obtained from the NORDCAN database and used as an estimate of the number of person years at risk assuming each person in the population on 01 January of each year contributes one person-year at risk [Ref. 5.4: 045VPW]. The IRs of seizure will be shown for the total population and stratified by country, sex, and age. The 95% CI for IRs will be calculated as:

where IR is the incidence rate, exp is the exponential function, and N new users is the number of new users [Ref. 5.4: 045VXZ].

Substudy 2B

A cohort study among ever-DL users analyzing the association between exposed and unexposed time and first seizures will be performed for the total population and stratified by age. Current use (i.e., “exposed” period) will be defined for each prescription as days’ supply plus a 4-week grace period to account for intermittent use and a possible wash-out effect. Unexposed time will be defined as the period starting 26 weeks from the dispensing date of the prior DL prescription until the next DL prescription redemption, if any. A person who has redeemed at least one prescription of DL is included in the cohort and enters the cohort at the date of first prescription redemption. Study individuals are followed until occurrence of a seizure; occurrence of one of the conditions described in [Section 9.3.2]; 31 December 2015; emigration; or death, whichever comes first. The association between exposure to DL and first seizure will be evaluated using Poisson regression of the IR of first seizure, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescription registers will not be included in order to account for truncation bias. In addition, we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) [Table 4]. We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

9.9.2.3 Substudy 3: SVT

Substudy 3A

A cohort study estimating the incidence of SVT in the general population will be performed. The cohort of DL users, the cohort of individuals with SVT, and the (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 44 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

distribution of the population by age, year, and country will be used to derive estimates in the general population. The number of individuals in the general population on 01 January of each year will be obtained from the NORDCAN database and used as an estimate of the number of person years at risk assuming each person in the population on 01 January of each year contributes one person-year at risk [Ref. 5.4: 045VPW]. The IRs of SVT will be shown for the total population and stratified by country, sex, and age. The confidence interval of IRs is calculated as outlined above under Substudy 2A.

Substudy 3B

A cohort study among ever-DL users analyzing the association between exposed and unexposed time (as defined in Substudy 2B) and first SVT will be performed for the total population and stratified by age. A person who has redeemed at least one prescription of DL is included in the cohort and enters the cohort at the date of first prescription redemption. Study individuals are followed until occurrence of an SVT; 31 December 2015; emigration; or death, whichever comes first. The association between exposure to DL use and first SVT will be evaluated using Poisson regression of the IR of first SVT, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescription registers will not be included in order to account for truncation bias. In addition, we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) [Table 4]. We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

9.9.2.4 Substudy 4: Atrial fibrillation/flutter

Substudy 4A

A cohort study estimating the incidence of A-fib/flu in the general population will be performed. The cohort of individuals with A-fib/flu and the distribution of the population by age, year, and country will be used to derive estimates in the general population. The number of individuals in the general population on 01 January of each year will be obtained from the NORDCAN database and used as an estimate of the number of person years at risk assuming each person in the population on 01 January of each year contributes one person-year at risk [Ref. 5.4: 045VPW]. The IRs of A-fib/flu will be shown for the total population and for each country, sex, and age. The confidence interval of IRs is calculated as outlined above under Substudy 2A.

Substudy 4B

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 45 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

between exposure to DL use and first A-fib/flu will be evaluated using Poisson regression of the IR of first A-fib/flu, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescription registers will not be included in order to account for truncation bias. In addition, we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar years, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) [Table 4]. We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

9.9.2.5 Substudy 5: First recurrent seizure

Substudy 5A

A cohort study estimating the incidence of first recurrent seizures (excluding febrile seizure) in the population of individuals who have experienced a first seizure will be conducted. The IRs of first recurrent seizures will be shown for the total population. The 95% confidence interval for IRs will be calculated as outlined above under Substudy 2A.

Substudy 5B

A cohort study among ever-DL users with a previous seizure analyzing the association between exposed and unexposed time (as defined in Substudy 2B) and first recurrence of seizures (excluding febrile seizure) will be performed. A person who has redeemed at least one prescription of DL and has experienced a seizure is included in the cohort and enters the cohort at the date of first seizure. Study individuals are followed until first recurrent seizure; occurrence of any condition mentioned in [Section 9.3.2]; 31 December 2015; emigration; or death, whichever comes first. The association between exposure to DL use and first recurrent seizure will be evaluated using Poisson regression of the IR of first recurrent seizure, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescription registers will not be included in order to account for truncation bias. In addition, we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar years, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) [Table 4]. We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

9.9.2.6 Additional information for Substudies 2A, 3A, 4A, and 5A

A specification has to be made regarding the number of persons at risk, which is not correctly stated in the protocol, where it is specified to be general population on January 1 of each year. More correctly, when calculating the IRs, the population at risk is defined as the population on July 1 of the given year for Denmark and Norway, the average of the

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 46 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

population on January 1 and December 31 of the given year for Sweden and the average of the population December 31 that year and December 31 the previous year for Finland (i.e., the population at risk in 2001 is the average of the population at risk December 31, 2001 and December 31, 2000).

9.9.2.7 Additional information for Substudies 2B, 3B, 4B, and 5B

Rationale for the choice of study design: Substudy 2B, 3B, 4B and 5B are historical cohort studies using prospectively collected data from nationwide population registers with a cohort defined by first use of DL. During the follow-up period, DL will be treated as a time-varying exposure. Using a variety of exposure definitions, all person-time will be categorized as exposed, unexposed, or unclear (period immediately following a course of current treatment). IRs will be calculated for each outcome of interest for each exposure category, and IRRs will be estimated for exposed person time compared to unexposed person-time aggregated across each cohort. In this sense, it is a risk interval design, but it differs from the type of self-controlled risk interval design (SCRID) which has frequently been used in vaccine safety research, in that it does not condition the comparison within individuals [Ref. 5.4: 03QPDQ]. A design similar to the one we propose was used to study the effect of cisapride on ventricular arrhythmias [Ref. 5.4: 04CD8S].

To account for potential time-varying confounders, the following time-varying covariates will be included in the Poisson analysis: age, seasonality, calendar year, asthma status, disease severity of rhinitis, and chronic urticaria. Because the same patients contribute to exposed and unexposed person-time, we expect that time constant covariates will be reasonably balanced between exposure groups—though if some prognostic factors were associated with longer or shorter time on drug, some confounding could occur. A SCRID analyzed by conditional Poisson regression might better control for time-constant confounding within individuals than the design we propose. However, our design avoids an important potential bias that may occur in a SCRID. In that approach, in order to be included in the conditional Poisson analysis, an individual needs to contribute both an exposed and a non-exposed period. This would exclude persons with only an exposed period, an analytic choice which could bias toward the null. If in fact there is a drug- associated risk, we would expect the outcome to occur during the first-ever exposure period for some persons included in the study. These persons would then be censored from the analysis at the time they experience the outcome and hence would only have had an exposed period. An analysis based only on those patients who have had at least one exposed and one unexposed period could therefore potentially underestimate the true risk. Underestimation of the true risk is likely to be a more important bias for this study than confounding by time constant factors; therefore, we have maintained our proposal to adjust for both time-constant and time-varying confounders in the analysis, rather than performing a conditional Poisson regression.

9.9.3 Missing values

In a register-based study, a missing value in a data set means that a value for an individual is missing because complete information was not entered into the register. Because the

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 47 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

current study ascertains only those patient outcomes that present for hospital-based care, it is possible that events that occurred in the community but did not require hospitalization or emergency care would not be captured. Furthermore, although the completeness of the patient registers is high, there is a risk that a hospital contact (encounter) is completely missing if it is not entered into the register. Therefore, there is a theoretical risk of under-ascertainment of study outcomes. However, there is no reason to believe that this under-ascertainment would vary by exposure status.

There are very few observations with missing values identified in the registers, e.g. missing information about strength of the drug.

Similarly, few observations with illegal values were identified, e.g. migration date with the day or month having values “00”.

Observations with missing values or with illegal values were excluded (Annex 1).

9.9.4 Sensitivity analyses

The following supplementary analyses will be conducted. Unless otherwise specified, supplementary analyses will be performed on the total population (i.e., not stratified by age) and using the primary exposure definition.

(1) alternative definitions of exposure based on time since last DL dispensing (periods 0- 4, 5–8, 9–16, and 17–26 weeks each compared with >26 week since last DL dispensing) for Substudies 2B, 3B, 4B, and 5B. Because as-needed medications such as antihistamines may not be taken daily, this approach reflects the clinical expectation that the probability of actual exposure on a given day is highest shortly after filling a prescription and diminishes with increasing time. Each time a DL prescription is refilled, the time since last dispensing will reset to 0.

(2) for Substudies 2B and 5B, differentiating between febrile and non-febrile seizures for children aged 0–4 years.

(3) for Substudies 2B, 3B, and 4B, analyses will exclude persons who have been diagnosed with chronic urticaria and/or have redeemed very high doses of DL (dose for chronic urticaria is typically 4 times the standard dose for allergic rhinitis). To operationalize the exclusion of persons with chronic urticaria and/or who redeems high doses of DL, we will exclude person time at risk for a person from the date of a diagnosis of chronic urticaria or date of DL prescription redemption of an amount of pills equal to at least twice the days’ supply of DL for the individual’s age in the period before next DL prescription redemption.

(4) the association analyses in Substudies 2B, 3B, and 4B will be stratified by countries to examine potential differences across countries.

(5) additional supplementary analyses, which will be specified in the SAP, will evaluate the potential effect of exposure misclassification in Substudies 2B, 3B, and 4B (i.e., evaluate use of loratadine and other non-sedating prescription antihistamines).

Loratadine and rupatadine are two drugs metabolized to DL in the liver. Loratadine was marketed in Denmark in 1988, Norway and Sweden in 1989, and Finland in 1995. As of (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 48 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

2013, when the most current data are available, rupatadine had not been marketed in the Nordic countries. However, if in future years of the study, which may extend to 2015 depending on data availability, rupatadine dispensing does appear in the registries, it will be included in supplementary analysis 5 [Table 9]. Loratadine is available both by prescription and OTC without a prescription. An average of 63% of loratadine purchased in the Danish population, 93% in Norway, 41% in Sweden and 72% in Finland from 2001 to 2013 was purchased by prescription and therefore also registered at the individual-level in the prescription register. The reason why some persons choose to purchase OTC drugs by a prescription is that prescription drugs are often subsidized and will therefore be less expensive for the purchaser. Non-sedative antihistamines include a range of different drugs (including DL and loratadine) used in different time periods across the Nordic countries, but in this supplementary analysis, they will be considered as one group [Table 10].

Table 9 Use of loratadine, rupatadine, and DL in the Nordic countries Country Drug Loratadine Rupatadine (R06AX28) Desloratadine (R06AX13) (R06AX27) Denmark Dispensing group OTC medicine also Only prescription. Only prescription sale in non-pharmacy 2001-2012. In 2012 outlets. and 2013 parallel- imported products are launched Date for first April 28th 1988 June 8th 2015 January 15th 2001 recorded use or launch date Sales by 60,965 (63 %) 0 39,157 (99.5%) prescription 2001-13 (DDD x 1000) Total sales 2001-13 97,265 (100 %) 0 39,347 (100 %) (DDD x 1000) Norway Dispensing group “Unntatt fra Only prescription. Only prescription reseptplikten” OTC Dispensing group 2001-2013. Dispensing medicine sale. “Prescription C” group “Prescription C”* Date for first July 10th 1989 Date for first launch January 15th 2001 recorded use or permission October 21st launch date 2013. No sale in 2013 and 2014 (not on the market). Sales by 116,440 (93 %) 0 152,514 (100 %) prescription 2004-13 (DDD x 1000) Total sales 2004-13 125,743 (100 %) 0 152,514 (100 %) (DDD x 1000) Sweden Dispensing group OTC medicine also Only prescription. Only prescription sale in non-pharmacy 2001-2013 except two outlets. brands, Desloratadine Sandoz and Flynise from Actavis (launched August 16th 2012 and April 16th 2015). Date for first February 3rd 1989 Date for first launch January 15th 2001. recorded use or permission September launch date 5th 2013. No sale in 2013 and 2014 (not on the market). Sales by 106,074 (41 %) 0 162,661 (100 %) prescription 2006-13 (DDD x 1000) Total sales 2006-13 258,640 (100 %) 0 162,661 (100 %) (DDD x 1000) (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 49 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038 Country Drug Loratadine Rupatadine (R06AX28) Desloratadine (R06AX13) (R06AX27) Finland Only prescription from 2001-2012. From 2013 Dispensing group Prescription and OTC Only prescription onwards, also OTC sale Date for first recorded use or May 8th 1995 --- January 15th 2001 launch date Sales by prescription 2001-13 67,937 (72 %) 0 93,076 (98%) (DDD x 1000) Total sales 2001-13 94,637 (100 %) 0 94,662 (100 %) (DDD x 1000) Data source Sweden: https://www.socialstyrelsen.se/statistik/statistikdatabas/lakemedel; https://lakemedelsverket.se/LMF/?type=product Table 10 Non-sedative antihistamines marketed in the Nordic countries Denmark (2001-2015) Finland (2001-2015) Norway (2001-2014) Sweden (2001-2015) Cetirizine R06AE07 Cetirizine R06AE07 Cetirizine R06AE07 Cetirizine R06AE07 Levocetirizine R06AE09 Levocetirizine R06AE09 Levocetirizine R06AE09 Terfenadine R06AX12 Terfenadine R06AX12 Terfenadine R06AX12 Loratadine R06AX13 Loratadine R06AX13 Loratadine R06AX13 Loratadine R06AX13 Acrivastine R06AX18 Acrivastine R06AX18 Acrivastine R06AX18 Ebastine R06AX22 Ebastine R06AX22 Ebastine R06AX22 Ebastine R06AX22 Mizolastine R06AX25 Mizolastine R06AX25 Mizolastine R06AX25 Fexofenadine R06AX26 Fexofenadine R06AX26 Fexofenadine R06AX26 Fexofenadine R06AX26 Desloratadine R06AX27 Desloratadine R06AX27 Desloratadine R06AX27 Desloratadine R06AX27 Rupatadine R06AX28 Rupatadine R06AX28 Bilastine R06AX29 (2013 - Bilastine R06AX29 2014) (2014) Data source Sweden: https://www.ehalsomyndigheten.se/ Data source Denmark: Medstat.dk Data source Norway: Norwegian Drug Wholesale Statistics, Norwegian Institute of Public Health Data source Finland: Kela.fi

Two sensitivity analyses will be performed for which a wider exposure definition will be used to evaluate the potential effect of exposure misclassification in Substudies 2B, 3B, and 4B. First, the exposure measure for this analysis will be defined as redeemed prescriptions of DL and loratadine grouped together. Second, the exposure measure will be further expanded to include all non-sedative antihistamines purchased by prescription in the Nordic countries during the study period. The prescription registers in Denmark, Norway and Sweden include all prescriptions purchased regardless of whether these are subsidized by the state, whereas the Finnish register only includes purchased prescriptions of subsidized drugs. Neither of these analyses will be able to capture OTC use of DL, loratadine, or other non-sedative antihistamines.

(6) a supplementary analysis for Substudies 2B, 3B, and 4B will evaluate whether the potential risk of the outcomes is higher following the first ever-DL prescription redemption compared to the second, third prescription redemption etc.

05822K DESLORATADINE PAGE 50 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

versus non-exposed) will be analyzed. The exposure variable (Exposure status) will be categorized as: currently DL unexposed; currently DL exposed following first prescription redemption; currently DL exposed following second prescription redemption; and currently DL exposed following third or more prescription redemption. The exposure variable will be analyzed as a fixed effect (as done in Substudies 2B, 3B and 4B) with 4 levels in the supplementary analyses (in Substudies 2B, 3B and 4B, the exposure variable has 2 categories corresponding to currently DL exposed and currently DL unexposed). If the exposure status variable is significant, pairwise comparisons between the 4 levels will be performed. If a pairwise comparison is performed, IRRs will be calculated. Pairwise comparisons with 1) the non-exposed category as reference and 2) with the currently DL exposed following first prescription redemption as reference are both of interest. Correspondingly, two sets of IRRs will be calculated, 1) one set of IRRs with the non- exposed category as reference and 2) one set of IRRs with the category currently DL exposed following first prescription redemption as the reference.

(7) a supplementary analysis for Substudies 2B, 3B, 4B, and 5B for which non-exposed periods start 52 weeks following the last prescription redemption.

A supplementary analysis that uses alternative definitions of exposed and unexposed time will be performed in which we will use a 52-week post supply definition, where an unexposed period starts 52 weeks beyond the latest prescription redemption (instead of 26 weeks after last prescription redemption as in the primary exposure definition).

(8) a supplementary analysis using an alternative adjustment set to examine the robustness of the study results. This adjustment set consists of age, sex, country, calendar year, seasonality, severity of rhinitis, asthmatic status, diabetes, hypo-/hyperthyroidism, inflammatory disease, infections, type 1 allergy. For the outcomes SVT and A-fib/flu, antihypertensive treatment will also be added.

The operationalization of these variables is presented in [Table 11] below. The alternative confounder adjustment set was chosen because it was assumed that the variables will be available in the registers and because this set could be used in all four analyses. There is an overlap between the operationalization of two of the confounders: severity of rhinitis and type 1 allergy, and if we find a strong correlation when examining the frequency distributions of these two variables, we will merge the variables together as one single variable and include it in the analysis.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 51 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038 Table 11 Definition of confounders included in the primary and alternative minimum sufficient adjustment set that will be used for confounder adjustment in the association studies. Confounders Confounders included in the included in the primary alternative Definition Data source minimum minimum sufficient sufficient adjustment set adjustment set

Age Age Age is a time varying confounder; and therefore, Civil risk time will be split up in years of age. We will registration evaluate whether age can be included as a system continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

Sex Sex Sex is a time-independent confounder and Civil will be included as males versus females in registration both the descriptive and association system analyses.

Country Country Country of residence is a time-independent Civil confounder, as persons will be censored if registration emigrating from the country. Country of system residence will be included with four categories, i.e., Denmark, Finland, Norway, or Sweden in both the descriptive and association analyses.

Calendar year Calendar year Calendar year is a time varying confounder; Prescription and therefore, risk time will be split up in register calendar years.

Seasonality Seasonality In the association studies, seasonality is a Prescription time varying confounder; and therefore, risk register time will be split up in seasons defined as winter (December–February), spring (March–May), summer (June–August), and autumn (September–November).

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 52 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Confounders Confounders included in the included in the primary alternative Definition Data source minimum minimum sufficient sufficient adjustment set adjustment set

Asthma status Asthma status Asthmatic status is defined as a binary National variable indicating whether or not a person patient has redeemed treatment for asthma defined register and as at least two prescriptions of inhalant prescription steroids within a six-month period and/or register contacts to hospitals with a diagnosis of asthma (including both primary and secondary diagnoses) during a five-year period before first DL exposure. To distinguish persons treated for chronic obstructive pulmonary disease from those treated for asthma, first registered asthma prescription has to be redeemed when the purchaser was 45 years or younger.

Disease severity Disease severity of Severity of rhinitis will be defined as a Prescription of rhinitis rhinitis binary variable indicating whether or not a register person has received treatment for severe rhinitis. Persons with severe rhinitis will be identified from the prescription register as persons who have redeemed immunotherapy at least once during a five- year period before first DL exposure.

Chronic urticaria --- Both for the descriptive and association National analyses, chronic urticaria status is defined patient as a binary variable indicating whether or register not a person has a registered diagnosis of chronic urticaria in the five-year period before first DL exposure.

--- Diabetes Diabetes is defined as a binary variable National indicating whether or not a person has a patient registered diagnosis of diabetes or has at register and least two prescription redemptions of prescription glucose-lowering drugs registered in the register prescription register in the five-year period before first DL exposure.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 53 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Confounders Confounders included in the included in the primary alternative Definition Data source minimum minimum sufficient sufficient adjustment set adjustment set

--- Hypo- Hypo-/hyperthyroidism is defined as a National /hyperthyroidism binary variable indicating whether or not a patient person has a registered diagnosis of hypo- register and /hyperthyroidism or has at least two prescription prescription redemptions of drugs for register treatment of hypo-/hyperthyroidism (i.e., Thycapsol and Eltroxin) registered in the prescription register in the five-year period before first DL exposure.

--- Inflammatory Inflammatory disease is defined as a binary National disease variable indicating whether or not a person patient has a registered diagnosis of inflammatory register and bowel disease, psoriasis, rheumatic prescription diseases, vasculitis, or sarcoidosis or has register redeemed at least two prescriptions of Daivonex (drug against psoriasis) in the five-year period before first DL exposure.

--- Infections Infection is defined as a binary variable National indicating whether or not a person has a patient registered diagnosis of lung infection register (pneumonia) or sinusitis in the five-year period before first DL exposure.

--- Type 1 allergy Type 1 allergy is defined as a binary National variable indicating whether or not a person patient has a registered diagnosis of acute urticaria, register and anaphylaxis, quinckes oedema or has at prescription least one prescription redemptions of register immunotherapy drugs registered in the prescription register in the five-year period before first DL exposure.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 54 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Confounders Confounders included in the included in the primary alternative Definition Data source minimum minimum sufficient sufficient adjustment set adjustment set

--- Antihypertensive Use of antihypertensive treatment will be National defined as binary variable indicating treatment prescription whether or not a person has received treatment for hypertension (ACE inhibitor, register angiotensin II receptor antagonist, calcium channel blocker, beta-blockers, alpha- blockers, thiazide (diuretic treatment), methyldopa, or moxonidine). Persons treated with antihypertensive drugs will be identified from the prescription register as persons who have redeemed antihypertensive prescription at least twice during a five-year period before first DL exposure.

(9) Quantitative Bias Analysis (QBA): the type of QBA to be used will be based on the extent of confounding identified in the empiric data, as well as feasibility considerations. If the adjusted estimate is within 10% of the crude estimate, a simple QBA at the summary level will be used; otherwise a QBA at the individual record level will be used to examine the potential impact of misclassification of the covariates rhinitis severity, asthmatic status, and chronic urticaria on the main study results. If due to computational limitations, it is not possible to run this analysis, an analysis restricted to data from Denmark only will be used. If none of these individual record level analyses are feasible, an alternative QBA based on summarized date on the crude association between exposure and outcome will be conducted.

QBA is used to quantify the amount of systematic error that may occur in a study [Ref. 5.4: 04C0BP, 04C0BT, 04CH6J]. In our study, adequate information on some of the potential confounders (i.e., severity of rhinitis, asthma status, and chronic urticaria), will be challenging to obtain from the sources available (i.e., national patient registers). The national patient registers contain data for contacts to the secondary and tertiary hospitals in the Nordic countries, but we do not have access to data on treatment of these conditions in the primary health care setting. We therefore propose to perform a QBA to try to estimate the effect of misclassification of these confounders on the results. QBA can be performed at either the summarized level or individual record level [Ref. 5.4: 04C0BT]. The advantage of performing QBA at the summarized level is that it is fast and simple to run (https://sites.google.com/site/biasanalysis/); however, QBA of summarized data is performed on crude estimates of the association under consideration, and therefore does not take into account the complexity of the statistical regression model. In contrast, using individual record level QBA methods make it possible to include information on other confounders while examining the impact of misclassification in the regression models [Ref. 5.4: 04CH6J]. The disadvantage of QBA

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 55 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

performed at the individual record level is that it requires substantial computational power to run these analyses on huge datasets.

Strategy of QBA analyses

We propose to select the type of QBA to be used based on the extent of confounding identified in the empiric data, as well as feasibility considerations. The QBAs will be applied for Substudies 2B, 3B and 4B. If the adjusted estimate is within 10% of the crude estimate, we will conduct a simple QBA at the summary level; otherwise we will conduct the QBA at the individual record level to examine the potential impact of misclassification of some of the covariates (i.e., severity of rhinitis, asthma status, and chronic urticaria) on the main study results. If it is not possible to run this analysis due to computational limitations, we propose an analysis restricted to use of data from Denmark only. If none of these individual record level analyses are feasible, we propose to conduct an alternative QBA based on summarized data on the crude association between exposure and outcome. We acknowledge that the interpretation of results from the latter analysis may be challenging due to simplification of the model (e.g., the lack of adjustment for the remaining confounders in the model). However, we judge this to be our best option with regard to quantifying the potential impact of misclassified covariates on the associations under study.

Sensitivity and specificity

Regarding the validity of the observed information on the three confounders, we probably have low sensitivity (i.e., the true positive rate or the proportion of persons with the diseases that are correctly identified as such). However, when persons have received the diagnosis for these conditions in the hospital setting, it is likely that they also have the condition, which means that the specificity is high (i.e., the true negative rate or the proportion of healthy people who are correctly identified as not having the condition). We find it unlikely that the potential misclassification of the three confounders is differential and dependent on either use of DL or future development of seizure, SVT or A-fib/flu.

Summary level approach

When performing the summary-level analysis we will obtain the observed crude association between DL and the three outcomes and specify 12 different scenarios of bias (i.e., define combinations of assumed sensitivity and specificity) to see how these scenarios would have affected the study results. The specificity will be set to 0.985, 0.990, 0.995, 0.999, and 1.0 and the sensitivity to 0.2, 0.4, 0.6, 0.8, and 1.0.

Individual record level approach

For the individual record level QBA, we will perform a probabilistic bias analysis in which the biased measure of association will be corrected for misclassification. Overall the principle is to estimate the positive and negative predictive values based on the observed data and the assumed distribution of the sensitivity and specificity. The predictive values are used to estimate the probabilities of the study population being correctly classified. Based on these probabilities a distribution of the variable is

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 56 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

generated. By sampling from this distribution, we construct a new variable assigning a new value of the covariate under consideration for each individual in the study population. We rerun the analysis using the new version of the variable to get an estimate corrected for the misclassification bias. The steps of this procedure are as follows:

The first step is to determine reasonable intervals for the biased parameters and distributions that reflect the uncertainty of the bias. We will apply a trapezoidal distribution to the parameters in which the minimum (min), lower mode (mode1), upper mode (mode2) and maximum (max) are defined. In the present study we will use the following parameters: specificity: min=0.985, mode1=0.990, mode2=0.995 max=1.00 and sensitivity: min=0.2, mode1=0.4, mode2=0.6 and max=0.8. Monte Carlo simulations are used to realize the distributions. A random number generator is used to sample values for sensitivity and specificity from the distributions. After iterating the analysis 999 times, we get a distribution of the results. From this distribution, the point estimate corrected for systematic error is the median and the 2.5th and 97.5th percentile can be interpreted as the 95% CI interval. Furthermore, we will calculate the total error of the model presented as the sum of the systematic error and the random error in the model.

(10) a supplementary analysis restricted to calendar time where misclassification due to OTC use does not exist (i.e., prior to OTC availability of DL in Denmark and Finland (only OTC sale in 2013). No OTC sale of DL has occurred in Norway and Sweden in the study period.

We will perform a supplementary analysis restricted to calendar time where misclassification due to OTC use could not exist. Table I shows data on OTC availability of DL in the Nordic countries. In Denmark and Finland, DL OTC sales did not occur until 2013 (0.5 % of the total sales from 2001-2013 in Denmark and 2% in Finland were OTC). No OTC sales of DL have occurred in Norway and Sweden in the study period). Hence the study period for this analysis will be 2001–2012 for Denmark and Finland, 2006–2013 for Sweden and 2010–2013 for Norway.

9.9.5 Post hoc analyses

A number of post hoc analyses were performed. Unless otherwise specified, the post hoc analyses were performed on the total population using the primary exposure definition.

(1) For Substudy 1, baseline characteristics of the incident DL users at the date of first prescription redemption was described, overall and stratified by country.

(2) For Substudy 2, the analysis was stratified by alternative age categories.

(3) For Substudy 2, the analysis was stratified by country and age group.

(4) For Substudy 4, the analysis was stratified by alternative age categories.

(5) For Substudy 4, the analysis was stratified by country and age group.

(6) For supplementary analysis 1, the analyses were stratified by country.

(7) For supplementary analysis 6.1, the analysis was stratified by country.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 57 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

(8) For supplementary analysis 6.3, the analysis was stratified by country. (9) A sensitivity analysis was run to replace the originally planned quantitative bias analysis (QBA), which was not possible to perform. Two QBAs were described in the SAP. The first proposed sensitivity analysis described was too computer intensive and could not be conducted on this large dataset. The second proposed sensitivity analysis described in the SAP assumes a standard cohort study with an exposed cohort and an unexposed cohort. However, in the present study design, risk-time for all individuals were partitioned into exposed and unexposed (not currently exposed) person-time. Thereby, all individuals were initially at inclusion exposed.

An alternative quantitative bias analysis was performed that evaluated the potential confounding effect of variables that were ascertained using a different time window than in the original analysis. The original confounder status (asthma, rhinitis, chronic urticaria) was defined by status in the 5-year period prior to first DL prescription redemption. In this sensitivity analysis, an alternative confounder definition was examined that was ascertained over a longer time interval. It was defined as status in the 5-year period prior to first DL prescription redemption and extending to 26 weeks after first DL prescription redemption. Yet another alternative assessment period extended ascertainment to 52 weeks after the first DL dispensing. Using this alternative confounder definition, this sensitivity analysis examines the potential effect of confounder misclassification, particularly the effect of false negatives (not having the disease diagnosis recorded before first DL prescription redemption). Previous research has shown that some prevalent conditions may be recognized and recorded in a delayed fashion, only after patients have had regular contact with the health care system [Ref. 5.4: 05750Y]. The frequency distribution of the number of incident DL users according to confounder status using the original definition and the alternative definitions. Analyses for Substudies 2B, 3B, 4B and 5B using the alternative confounder definitions were performed.

9.9.6 Amendments to the statistical analysis plan

Four deviations from the Statistical Analysis Plan are listed below:

1. Data from Norway were not included as data from Norway were not received in due time to be included in the final report.

2. Seizure in the main analyses was defined as non-febrile seizures. Only the ICD-10 system uses specific codes for febrile and non-febrile seizures. Therefore, incident seizures are defined based on data using the ICD-10 system, starting in 1994. As the study period starts in 2001, incident seizure can be defined based on no seizures in the previous six years, which is considered sufficient, [Table 5].

3. It was not possible to estimate the incidence rate of first recurrent seizure in the general population as the total numbers of individuals with a first seizure in the general population is unknown (Table 5A of [Annex 2]).

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 58 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

4. The planned quantitative bias analysis described in the Statistical Analysis Plan could not be performed. The first analysis described was too computer intensive. The second analysis assumes a standard cohort study (an exposed cohort and an unexposed cohort), and in the current study design patients were all exposed with person-time being divided into currently exposed and unexposed (not currently exposed). As an alternative, post hoc analysis 9 was performed with an alternative definition of the confounders, see description in [Section 9.9.5].

9.10 Quality control

By signing the protocol, the Sponsor agrees to be responsible for implementing and maintaining a quality management system with written development procedures and functional area standard operating procedures to ensure that studies are conducted and data are generated, documented, and reported in compliance with the protocol, accepted standards of Good Clinical and Pharmacoepidemiology Practice, and all applicable federal, state, and local laws, rules and regulations relating to the conduct of the study.

The study is register-based. As a result, data quality is difficult to ascertain directly. However, previous studies have examined the validity and quality of information in the Nordic registers. The Danish National Prescription Registry contains data of high quality, including detailed information on dispensed drugs, and as the register covers all prescription dispensed in Danish pharmacies, loss to follow-up is unlikely for individuals with permanent residence in Denmark [Ref. 5.4: 045TYK]. The completeness of the Norwegian and Swedish prescription registers is characterized as good [Ref. 5.4: 00W4CX], [Ref. 5.4: 046B3C]. Completeness of registration in the Norwegian prescription register is ensured by law and quality checks are carried out monthly to identify possible errors or inconsistencies [Ref. 5.4: 00W4CX]. The patient identity data are only missing for approximately 0.3% of all items in the Swedish prescription register [Ref. 5.4: 046B3C]. The Finnish prescription register has been described in detail and considered as excellent [Ref. 5.4: 0469YG]. Data quality in the Danish National Patient Register is overall assumed to be of high quality; however, validity of data depends on the diagnosis under consideration; the positive predictive values for diagnoses similar to those in this study are >90%. [Ref. 5.4: 04C0C2]. The Norwegian Patient Register has relatively good agreement with the Norwegian Cancer Register [Ref. 5.4: 0469WT] and the completeness and accuracy of the Finnish Patient Register has been evaluated as varying from satisfactory to very good for common diagnoses with positive predictive values ranging from 75%– 99% [Ref. 5.4: 046B36]. The validity of the Swedish Patient Register is high for many, but not all diagnoses. The positive predictive values of most diagnoses in the Swedish Patient Register compared to medical records ranges from 85– 95% [Ref. 5.4: 03RSH6].

The statistical analyses will be performed on servers at Statistics Denmark. The programming will be performed by two researchers independently, limiting programming errors. The statistical programs will be stored on the servers at Statistics Denmark.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 59 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

10 RESULTS In the results section, the terms “currently DL exposed periods” and “currently DL unexposed periods” have been described as “exposed follow-up time” and “unexposed follow-up time.”

In the results and discussion, the term “association” is used to describe relationships between DL and the study outcome that are statistically significant at the 0.05 level. An effect of similar magnitude, but that is not statistically significant will not be described as “associated”.

Furthermore, person-years have been abbreviated to PY. Incidence rates are expressed per 100,000 PY unless otherwise specified.

Prevalent users refer to individuals who have ever had a DL prescription redemption during the study period. The prevalent proportion is defined as the number of prevalent users divided by the population of the country at the midpoint of the year.

10.1 Participants

The flow diagram (Figure 1 of [Annex 1]) shows the selection of subjects in the study. In Substudy 1, individuals redeeming at least one prescription of DL during the study period in Denmark (2001–2015), Finland (2002–2015), and Sweden (2006–2015) were included, resulting in a total population of 1,912,193 prevalent DL users (Denmark, n=314,534; Finland, n=627,206; Sweden, n=970,453). From these individuals, patients having erroneous details in the migration registers (e.g., two consecutive immigrations or emigrations, last migration event before redemption of desloratadine being emigration), patients redeeming DL within 6 months of the first immigration event, and (in the case of Sweden) patients redeeming DL within 6 months of the establishment of the prescription registers were excluded. This resulted in a population of 1,840,239 new users of DL (Denmark, n=312,481; Finland, n=626,792; Sweden, n=900,966).

10.1.1 Protection of Human Subjects

This is an observational study with no administration of any therapeutic or prophylactic agent. Subjects observed in this study continues with the normal standard of care as provided by their personal physician. Pre-existing national register data were the sole data source.

According to Danish, Finnish, Norwegian, and Swedish law, register-based studies can be carried out without consent from the data subjects where the processing takes place for the sole purpose of carrying out statistical or scientific studies of significant public importance and where such processing is necessary in order to carry out these studies. It is an absolute requirement that the publication of statistical or scientific results may never reveal the identity of individuals or otherwise compromise data subjects. We obtained approval by the data agencies in the four countries before data management and data analyses will be performed.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 60 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

10.2 Descriptive data Descriptive analysis of the DL users (incident and prevalent) are included in Tables 1.1 and 1.2 in [Annex 2].

The IR per 100,000 PY, 95% CI of DL users in the general population of the three Nordic countries across 2001 and 2015 was 732.2 (95% CI 731.2; 733.3, Table 1.1 of [Annex 2]). The IRs of DL use in the Finnish (839.7) and Swedish (956.6) populations were more than double that of the Danish (378.8) population. In all three Nordic countries combined, IRs decreased with increasing age. Additionally, IRs were slightly higher among females (735.1) than males (729.3) while being considerably higher in spring (1194.2) and summer (865.8) than in autumn (450.5) and winter (418.4). The IR of DL use, in general, was higher in the later study years (after 2006 when the Swedish data are included). The IR of DL users in patients having asthma, severe rhinitis, and chronic urticaria was 105.4 (95% CI 105.0; 105.8), 3.2 (95% CI 3.2; 3.3), and 17.2 (95% CI 17.0; 17.3), respectively, Table 1.1 of [Annex 2].

The mean defined daily dose (DDD) of DL in the incident DL population was 207.80 (SD=418.02), Table 1.2 of [Annex 2]. The distribution of DDD was very skewed with a median DDD at 70 and range (minimum-maximum) from 1 to 19,970 DDD. Mean DDD was similar in males (207.08) and females (208.5), higher in Sweden (229.73) compared to Denmark (182.74) and Finland (188.77) and increased with increasing age as well as time (calendar year). A higher mean DDD of DL was observed in spring (147.08) and summer (128.07) compared to autumn (117.82) and winter (116.53). As expected from a clinical perspective, the mean DDD of DL used in patients with asthma, severe rhinitis, and chronic urticaria was higher than that in patients without these conditions.

10.3 Outcome data in the general population

The outcome data are tabulated in [Annex 2].

In Substudy 2A, first seizure occurred in 96,220 individuals from the general population in Denmark, Finland and Sweden, with an IR per 100,000 PY of 38.3, 95% CI 38.0; 38.5, Table 2A. The IRs of first seizure was stratified by country, sex, and age. The IR was highest in Sweden (n=50,210; IR 53.3, 95% CI 52.8; 53.8) followed by Finland (n=27,300; IR 36.6, 95% CI 36.1; 37.0) and Denmark (n=18,710; IR 22.7, 95% CI 22.4; 23.0). The IR was higher in males (40.1, 95% CI 39.8; 40.5) than females (36.5, 95% CI 36.1; 36.8). The highest IR of first seizures in the general population was observed in the 0–4 years age group (116.7, 95% CI 114.9; 118.4).

In Substudy 3A, the incidence of first SVT was observed in 87,499 individuals from the general population, with an IR of 34.8, 95% CI 34.6, 35.0 per 100,000 PY, Table 3A of [Annex 2]. Table 3A also shows the IRs of first SVT when stratified by country, sex, and age. The IR was the highest in Denmark (n=41,311; IR 50.1, 95% CI 49.6, 50.6) followed by Sweden (n=29,621; IR 31.5, 95% CI 31.1; 31.8) and Finland (n=16,567; IR 22.2, 95% CI 21.9; 22.5). The IR was similar in males (35.6, 95% CI 35.3; 35.9) and females (34.0, 95% CI 33.7; 34.4). The IR of first SVT increased with increasing age and was the highest in the 65 year and older age group (81.6, 95% CI 80.7; 82.4).

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 61 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

In Substudy 4A, the incidence of first A-fib/flu occurred in 427,370 individuals from the general population, with an IR per 100,000 PY of 170.0, 95% CI 169.5; 170.6, Table 4A of [Annex 2]. The IR was the highest in Sweden (187.3, 95% CI 186.4; 188.2), as compared with Denmark (157.2, 95% CI 156.3; 158.0) and Finland (162.5, 95% CI 161.6; 163.4). Among the incident A-fib/flu cases (n=427,370), 52.7% occurred in males and 47.4% in females. As clinically expected, the IR was higher in males (180.9, 95% CI 180.1; 181.6) than females (159.4, 95% CI 158.7; 160.1), and increased with increasing age. The IR of first A-fib/flu was markedly higher in the oldest (≥ 65 years) age group (703.8, 95% CI 701.3; 706.3) compared to the younger ones.The IR of first A-fib/flu was lower in the 20–64 years age group (80.1, 95% CI 79.6, 80.5) and considerably lower in the 0-to 4-year (0.6, 95% CI 0.5; 0.7) and 5-19 years (1.1, 95% CI 1.1; 1.3) age groups. In Substudy 5A, the total number of individuals experiencing first recurrent seizures in the population of individuals who previously experienced a first seizure was 17,819 (Table 5A of [Annex 2]) with 2,712, 4,500, and 10,607 individuals in Denmark, Finland and Sweden, respectively. It was not possible to estimate the IR of first recurrent seizure in the general population as the total numbers of individuals with a first seizure in the general population is unknown (see [Section 9.9.6]).

10.4 Main results

The main results are tabulated in [Annex 2].

The main results of Substudy 2B were that the IR of first seizure per 100,000 PY during exposed follow-up time (i.e., among those currently exposed to DL, 39.9, 95% CI 36.2; 44.0) was higher than during unexposed follow-up time (i.e., among those currently not exposed to DL, 30.1, 95% CI 28.8; 31.5, Table 2B1 of [Annex 2]). When analyzed within specific age categories, the IR was higher during exposed follow-up time as compared to during unexposed follow-up time in all the age categories except the 20–64 years age group, where the IRs were nearly identical. The IR of first seizure was the highest among those in the 0–4 years age group comparing DL exposed (114.8, 95% CI 94.4; 139.7) with DL unexposed (73.0, 95% CI 64.5; 82.5).

Crude analyses of the association between DL exposure and first seizure indicated that periods of exposed follow-up time were associated with higher incidence of first seizures when compared with periods of unexposed follow-up time (IR ratio (IRR) 1.32, 95% CI 1.19; 1.48, Table 2B2 of [Annex 2]). A pre-study exercise using directed acyclic graphs (DAGs) indicated that age, sex, country and calendar year be adjusted for in the semi- adjusted models, and age, sex country, calendar year, seasonality, asthmatic status, severe rhinitis, and chronic urticaria status in the fully adjusted models. The adjusted incidence rate ratio (aIRR) of first seizure when comparing exposed and unexposed follow-up time remained greater than 1 in the fully adjusted models, albeit with some attenuation of the estimates (aIRR 1.15, 95% CI 1.03; 1.29). Thus, in the fully adjusted model, a 15% higher IR of first seizure was observed during exposed follow-up time than during unexposed follow-up time.

The overall association between first seizure and DL exposure is driven by the association in the 0–to 4-year (aIRR 1.47, 95% CI 1.16; 1.87) and 5–to19-year (aIRR 1.32, 95% CI 1.09;1.59) age categories, as there was no association between exposure status and first seizure in the older age categories (20–64 and ≥ 65 years).

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 62 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

The main results of Substudy 3B indicate that there was no association between exposure status and the IRs of first SVT per 100,000 PY (exposed, 31.1,95% CI 28.2; 34.4, unexposed 31.6, 95% CI 30.3; 32.8, Table 3B1 of [Annex 2]). Although the IR of first SVT increased with age within age groups, the IR of first SVT between exposed and unexposed periods was similar in the age-stratified analyses. The IR of first SVT was higher in the oldest (≥ 65 years) age group (exposed follow-up time: 79.3, 95% CI 66.0; 95.3; unexposed follow-up time: 83.1, 95% CI 77.6; 89.0). All analyses (crude, semi- and fully adjusted) indicate no association between DL exposure status and IR of first SVT, Table 3B2 of [Annex 2]. There were too few observations to perform sub-group analyses for the 0–4 years age group. Adjustment for age, sex, country and calendar year, as well as further adjustment for the remaining confounders (asthma, rhinitis, chronic urticaria) had only a minor effect on the effect estimates.

The main results of Substudy 4B were that the IR of first A-fib/flu per 100,000 PY during exposed follow-up time (142.8, 95% CI 136.2; 149.6) was higher than during unexposed follow-up time (136.2, 95% CI 133.6, 138.8, Table 4B1). In fully adjusted analyses, an association between exposure status and first A-fib/flu was observed (aIRR 1.08, 95% CI 1.02, 1.13, Table 4B2 of [Annex 2]). In the age-stratified analysis, fewer than 10 cases of incident first A-fib/flu were observed during exposed and unexposed follow-up time in the 0–4 and 5–19 years age categories. Due to privacy restrictions no further analysis was performed in these two age categories. In the remaining two age categories (20–64 years and ≥65 years), the IR of first A-fib/flu was considerably higher in the oldest (≥ 65 years) age group (exposed 830.5, 95% CI 783.6; 880.2, unexposed 776.7, 95% CI 759.0; 794.8) compared to the 20–64 years age group (exposed 84.1, 95% CI 77.7; 91.1 unexposed 73.2, 95% CI 70.8; 75.8). This finding of higher rates of A-fib/flu in ≥65 years old age group is consistent with the known epidemiology of A-fib/flu, which increases with age.

A different age-related trend is seen in the DL and A-fib/flu analyses. In these analyses, there is an association between exposure status and first A-fib/flu for the 20–64 years age categories, but no association for the 65 year and older age group. In the fully adjusted analysis in the 20-64 years old age group, a 16% higher IR of first A-fib/flu was observed during the exposed follow-up time as compared to the unexposed follow-up time (aIRR 1.16, 95% CI 1.07; 1.27). Adjustment for age, sex, country and calendar year, as well as further adjustment for the remaining confounders (asthma, rhinitis, chronic urticaria) had only a minor effect on the IRR.

The main results of Substudy 5B indicate that the IR of first recurrent seizure during exposed follow-up time (8,601, 95% CI 6,634; 11,151) was higher than during unexposed follow-up time (5,644.8, 95% CI 5,097.3; 6,251.2, Table 5B1 of [Annex 2]). In the fully adjusted analysis, no association between exposure status and first recurrent seizure was observed (aIRR 1.26, 95% CI 0.94; 1.67), Table 5B2 of [Annex 2]. Adjustment for age, sex, country and calendar year decreased the IRR as compared with the crude model. However, further adjustment for the remaining confounders (asthma, rhinitis, chronic urticaria) had only a minor effect on the IRR.

10.5 Other analyses

10.5.1 Supplementary analyses

The results of the supplementary analyses are tabulated in [Annex 2].

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 63 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

In general, supplementary analyses support the findings of the main analyses. In supplementary analyses 1 (Table S1.1–S1.4 of [Annex 2]), an alternative definition of DL exposure based on the time since the last DL redemption was examined with respect to each study outcome. Associations between exposure status and first seizure were observed when comparing follow up time 0–4 with ≥ 27 weeks (aIRR 1.18, 95% CI 1.00; 1.40) and when comparing 9–16 with ≥ 27 weeks (aIRR 1.20, 95% CI 1.04; 1.38) since last DL prescription redemption,Table S1.1 of [Annex 2].Associations between exposure status and first A-fib/flu were observed between 0–4 and ≥ 27 weeks (aIRR 1.12, 95% CI 1.03; 1.21) and between 5–8 and ≥ 27 weeks (aIRR 1.11, 95% CI 1.02; 1.21) since the last DL prescription redemption, Table S1.3 of [Annex 2]. As in the main analyses, no association between the alternative DL exposure definition and first SVT or first recurrent seizure was seen, Tables S1.2, S1.4 of [Annex 2].

Supplementary analyses 2 (Tables S2.1–S2.2 of [Annex 2]) examined first seizures among the 0–4 years children, overall (first seizure independent of type: febrile or non- febrile) and stratified by type of first seizure. The association between exposure status and first seizure remained with higher IIRs for first seizure during exposed follow-up time compared to unexposed follow-up time, overall (aIRR 1.88, 95% CI 1.66; 2.13) and for febrile (aIRR 2.00, 95% CI 1.73; 2.30) and non-febrile seizures (aIRR 1.57, 95% CI 1.22; 2.01). The same analysis was repeated for the first recurrent seizure, overall and stratified by type of first seizure (a recurrent febrile seizure following a first febrile seizure; a recurrent non-febrile seizure following a first non-febrile seizure). There was an association between exposure status and first recurrent seizure for seizures overall (aIRR 2.22, 95% CI 1.53; 3.21) and for both febrile (aIRR 1.94, 95% CI 1.14; 3.30) and non- febrile seizures (aIRR 2.58, 95% CI 1.34; 4.97).

In supplementary analyses 3 (S3.1–S3.3), risk time was censored at date of diagnosis of chronic urticaria or date of high use of DL. When compared with the results of the main analysis (aIRR 1.15, 95% CI 1.03; 1.29), in this supplementary analysis the aIRR was attenuated (aIRR 1.08, 95% CI 0.95; 1.22) and DL exposure status was no longer associated with first seizure. There was no association between DL exposure status and first SVT in the main analyses (aIRR 1.03, 95% CI 0.92;1.15) or in the supplementary analysis (aIRR 0.92; 95% CI 0.81; 1.05). The association between DL exposure status and first A-fib/flu seen in the main analysis (aIRR 1.08, 95% CI 1.02; 1.13) remained in the supplementary analysis (aIRR 1.11, 95% CI 1.04; 1.17).

In supplementary analyses 4, country-specific analyses revealed that associations between exposure status and first seizure were observed only in Finland (aIRR 1.31, 95% CI 1.06; 1.63). No associations were observed in Denmark (aIRR 0.86, 95% CI 0.57; 1.31) or Sweden (aIRR 1.14, 95% CI 0.99; 1.31, Table S4.1 of [Annex 2]). Similarly, an association in the incidence rate ratio of first A-fib/flu between exposed and unexposed follow-up time was observed only in Denmark (aIRR 1.13, 95% CI 1.01; 1.27, Table S4.3 of [Annex 2]).

Two alternative exposure definitions were examined in supplementary analyses 5, Tables S5.1–S5.3 of [Annex 2]. First, the exposure definition was expanded to include both prescription redemptions of DL and loratadine. The second definition was expanded further to include other non-sedating antihistamines. As in the main analyses, an association between the alternative exposure definitions and first seizure and first A-

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 64 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

fib/flu remained. No association between the alternative exposure definitions and first SVT was observed. For both first seizure and first A-fib/flu, including all non-sedating antihistamines as part of exposure definition showed nearly identical results as the main analyses and as the analyses including loratadine and DL exposure altogether.

Supplementary analyses 6 stratified the exposure follow-up time according to the number of prescription redemptions, when comparing exposed to unexposed follow-up time. An association between exposure status and first seizure was observed following the first- ever DL prescription redemption (aIRR 1.22, 95% CI 1.03; 1.45), but not following the second (aIRR 1.26, 95% CI 0.98; 1.62) or three or more redemptions (aIRR 1.08, 95% CI 0.93; 1.26, Table S6.1 of [Annex 2]). Similarly, comparing exposed to unexposed follow- up time, an association between DL and first A-fib/flu was observed following the first- ever DL prescription redemption (aIRR 1.29, 95% CI 1.18; 1.42), but not the second (aIRR 1.06, 95% CI 0.93; 1.22) or the third or subsequent redemptions (aIRR 1.00, 95% CI 0.94; 1.07, Table S6.3). No association between exposure status and first SVT was seen, Table S6.2 of [Annex 2].

In supplementary analyses 7, an alternative definition of unexposed period starting 52 weeks after the last prescription redemption was examined. This analysis resulted in a similar result as found in the main analysis. An association between exposure status first seizure (aIRR 1.15 ,95% CI 1.02; 1.29) and first A-fib/flu (aIRR 1.08 95% CI 1.02; 1.13) persisted, Tables S7.1, S7.3 of [Annex 2], respectively. No association between DL exposure status and first SVT was observed (aIRR 1.06, 95% CI 0.95; 1.18, Table S7.2 of [Annex 2]).

In supplementary analyses 8, a sensitivity analysis was conducted using an alternative confounder adjustment set than was controlled for in the primary analyses. For all exposure-outcome comparisons, the alternative confounder set included age, sex, country, calendar year, seasonality, asthma status, diabetes, hypo/hyperthyroidism, inflammatory disease, infections, and type 1 allergy and severe rhinitis combined. Antihypertensive treatment was also controlled for in the analyses of first SVT and A-fib/flu.

Using the alternative confounder adjustment set, the IRR of first seizure was 14% higher during exposed follow-up time compared to unexposed follow-up time (aIRR 1.14, 95% CI 1.02; 1.28, Table S8.1 of [Annex 2]) compared to a 15% higher aIRR in the main analysis, Table 2B2. However, no association between exposure status was apparent in the analysis of first SVT (aIRR 0.98, 95% CI: 0.88, 1.09, Tables S8.2 of [Annex 2]). The 8% higher IRR of first A-fib/flu seen in the main analysis (Table 4B2 of [Annex 2]) is not seen in the supplementary analysis of first A-fib/flu using the alternative set of confounders (aIRR 1.01, 95% CI 0.96, 1.06, Table S8.3 of [Annex 2]).

Supplementary analyses 9 could not be performed, as described in [Section 9.9.5] and [Section 9.9.6].

Supplementary analyses 10 restricted the calendar time to periods without OTC sales of DL. While the magnitude of the aIRR is similar to the main analysis, there was no association between exposure status and first seizure (aIRR 1.14, 95% CI: 0.99, 1.31, Table S10.1 of [Annex 2]). As in the main analyses, no association was observed between exposure status and first SVT (aIRR 1.01, 95% CI: 0.88, 1.16; Tables S10.2 of

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 65 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

[Annex 2]). These analyses showed an association between exposure status and first A- fib/flu (aIRR 1.07, 95% CI 1.00, 1.14, Tables S10.3 of [Annex 2]).

10.5.2 Post hoc analyses

The results of post hoc analyses are presented in [Annex 3].

Post hoc analysis 1 shows the baseline characteristics of the incident DL users at the date of first DL prescription redemption overall (n=1,840,239) and stratified by country. The mean age among the incident DL users was 31.5 years (standard deviation (SD): 23.3) years, ranging from 30.9 (22.9) in Finland to 33.1 (23.7) years in Denmark, Post hoc Table 1 of [Annex 3]. Overall, there were slightly more females (50.7%) than males (49.3%) in the study population). Among the incident DL users, overall 14.4% (n=264,931) were registered with a diagnosis of asthma, 2.3% (n=43,097) with chronic urticaria, and 0.4% (n=8,102) with severe allergic rhinitis requiring immunotherapy within 5 years before first DL prescription redemption. A total of 40.8% and 29.6% of the DL users had the first prescription redemption in spring and summer, respectively.

Post hoc analysis 2 of first seizure using alternative age categories shows the IRR for first seizure by age when comparing exposed follow-up time to unexposed follow-up time: children age 0–2 years (aIRR 1.11, 95% CI 0.82; 1.52) and age 3–4 years (aIRR 1.38, 95% CI 0.93; 2.03) and between children age 5–10 years (aIRR 1.17, 95% CI 0.86; 1.59) and 11–19 years (aIRR 1.42, 95% CI 1.12; 1.81, Post hoc Table 2 of [Annex 3]). The wider confidence intervals may indicate that the lack of association in some age groups could be due to smaller sample sizes in each age and exposure category.

Post hoc analysis 3 presents IRR of first seizure stratified by country and age. An association between exposure status and first seizure was only observed in Finland (Denmark: aIRR 0.86, 95% CI 0.57; 1.31; Finland: aIRR 1.31, 95% CI 1.05; 1.63; Sweden: aIRR 1.14, 95% CI 0.99; 1.31, Post hoc Table 3 of [Annex 3]).

Post hoc analysis 4 of first A-fib/flu using alternative age categories with age group 20– 64 years divided into two categories age 20–44 years and age 45–65 years (Post hoc Table 4 of [Annex 3]). For the 20–64 years age category in the main analysis, the aIRR for the association between exposure status and first A-fib/flu was 1.16 (95% CI 1.07; 1.27). Evaluating the association stratified into two age categories, the aIRR was 1.13 (95% CI 0.91; 1.40) for the 20–44 years age category, and 1.03 (95% CI 0.94; 1.14) for the 45–65 years age category. The effect seen for the overall age category seems to be driven by the effect of DL on first A-fib/flu in the 20-to 44-year old age category.

Post hoc analysis 5 shows the relationship between exposure status and first A-fib/flu stratified by country and age. An association between exposure status and first A-fib/flu was seen in Denmark (aIRR 1.14, 95% CI 1.02; 1.28), but not in Finland (aIRR 1.06, 95% CI: 0.96; 1.17) or Sweden (aIRR 1.07, 95% CI 1.00; 1.15, Post hoc Table 5 of [Annex 3]).

Post hoc analysis 6 presents supplementary analysis S1.1 stratified by country for comparison of alternative exposure status 0-4 weeks versus ≥ 27 weeks after DL prescription redemption (Post hoc Table 6 of [Annex 3]). No association between exposure status and first seizure were seen between weeks 0-4 and ≥ 27 weeks after DL

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 66 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

prescription redemption in any of the three countries (Denmark: aIRR 1.24, 95% CI 0.73; 2.11; Finland: aIRR 1.35, 95% CI 0.98; 1.87; Sweden: aIRR 1.12, 95% CI 0.90; 1.39).

Post hoc analyses 7 and 8 present supplementary analyses S6.1 and S6.3 stratified by country and examine the alternative exposure status definition (first prescription redemption versus not currently DL exposed). An association between exposure status and first seizure was seen for Sweden (aIRR 1.24, 95% CI 1.00; 1.54), but not Denmark (aIRR 1.09, 95% CI 0.58; 2.02) or Finland (aIRR 1.24, 95% CI 0.87; 1.76, Post hoc Table 7 of [Annex 3]). An association between exposure status and first A-fib/flu was seen for all three countries (Post hoc Table 8 of [Annex 3]) with aIRRs for Denmark (aIRR 1.48, 95% C 1.18; 1.86), Finland (aIRR 1.30, 95% C 1.09; 1.55), and Sweden (aIRR 1.26, 95% C 1.11; 1.43, Post hoc Table 8 of [Annex 3]).

Post hoc analysis 9 is an alternative quantitative bias analysis. The originally planned quantitative bias analysis could not be performed (See explanation in [Section 9.9.5] and [Section 9.9.6]). Post hoc Tables 9.1.1 and 9.1.2 show the number of individuals who had asthma, rhinitis or chronic urticaria diagnosed using the original definition (i.e., a diagnosis in the 5-year period prior to the first DL prescription redemption) and number of individuals who had asthma, rhinitis or chronic urticaria diagnosed using the alternative ascertainment periods (i.e., a diagnosis in the 5-year period prior to the first DL prescription redemption and until either 26 or 52 weeks after date of first DL prescription redemption). An extra 20–50% individuals had a diagnosis of asthma, rhinitis or chronic urticaria using the alternative definitions. In post hoc Tables 9.2.1 to 9.2.2 of [Annex 3], the analyses 2B, 3B, 4B and 5B were repeated using the alternative confounder definitions for asthma, rhinitis or chronic urticaria with more individuals being diagnosed with the confounding variables due to the extension of the time period that a patient could qualify as having these conditions. Although a larger number of DL users have these confounders when the alternative confounder definitions were used, only minor effects on the aIRR for analyses 2B, 3B, 4B and 5B were seen.

11 DISCUSSION

11.1 Key results

This study aimed to describe use of DL during the study period in the general population and stratified by age, sex calendar year, seasonality and comorbidity status. Overall there were 1.8 million incident users of DL during the study period and 251 million person- years at risk. Use of DL increased over the study period ranging from 7,277 incident users in 2001 to 204,307 incident users in 2015. The IR of DL use was greatest in the spring (IR 1194.2, 95% CI 1191.5; 1197.0) and lowest in the winter (IR 418.4, 95% CI, 416.8; 420.0), consistent with the seasonality of allergic rhinitis.

The IR of DL use was slightly higher for females (IR 735.1, 95% CI 733.6; 736.6) than males (IR 729.3). The IR of DL use decreased with increasing age, 0-4 year (IR 2079.1), 5-19 years (IR 884.4, 95% CI 881.7 887.2), 20-64 years (IR 648.5, 95% CI 647.2; 649.8), and ≥65 years (IR 416.7, 95% CI 414.8; 418.6). The IR of use among patients with registrations: for asthma was 105.4, for severe rhinitis registrations was 3.2, and for chronic urticaria was 17.2.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 67 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Other key findings relating to the study objectives were that that the IR of DL use was much higher in Finland (IR 839.7, 95% CI 837.7; 841.8) and Sweden (IR 956.6, 95% CI 954.6; 958.6) as compared with Denmark (IR 378.8, 95% CI 377.5; 380.1) and the mean number of defined daily doses was similar in Denmark (182.74 ± 492.96) and Finland (188.77 ± 363.63), but larger in Sweden (229.73 ± 423.90).

This study also aimed to describe the IR of the study outcomes in the general population. In the general population, the IR per 100,000 PY of: seizure was 38.3, SVT was 34.8, and A-fib/flu was 170.0. In this study, the observed patterns in the age-stratified IRs for seizure and A-fib/flu coincide with the known epidemiology of seizure and A-fib/flu, which are highest in the youngest and oldest age groups, respectively.

The prespecified study questions of this PASS set out to examine the association between current DL exposure and first seizure, first SVT and first A-fib/flu. This study found a higher rate of first seizure in DL exposed person-time compared with DL unexposed person-time. Across all age groups, the magnitude of the aIRR for seizure is small (aIRR 1.15, 95% CI 1.03; 1.29) adjusted for confounders selected in the DAG analysis. Substudy 2 showed a higher IR of first seizure during exposed follow-up time (IR=39.9) as compared with unexposed follow-up time (IR=30.1). The overall finding of increased risk of first seizure is driven by increased IR of first seizure in the 0- to 4 (aIRR 1.47, 95% CI 1.16; 1.87) and in the 5-to 19- year-old (aIRR 1.32, 95% CI 1.09; 1.59) age categories.When comparing DL exposed and unexposed periods, exposure status was not associated with first seizure in patients ≥ 20 years. The association between DL and first seizure is consistent across sensitivity analyses that varied exposure definitions and adjusted for alternative sets of confounders.

This study found no association between DL exposure status and incident SVT. Substudy 3 showed no association between DL exposure and first SVT (aIRR 1.03, 95% CI 0.92; 1.15) adjusted for confounders selected in the DAG exercise. This finding was consistent across multiple subgroups, substudies and sensitivity analyses.

This PASS found a slightly higher rate of first A-fib/flu in DL exposed person-time compared with DL unexposed person-time. The main finding of Substudy 4 found an aIRR of first A-fib/flu of 1.08 (95% CI 1.02; 1.13) adjusted for confounders that were identified in the DAG exercise. The overall finding of increased IR of first A-fib/flu was driven by increased IR of first A-fib/flu in the 20-64 years age category (aIRR 1.16, 95% CI 1.07; 1.27) when comparing DL exposed and DL unexposed person-time. There were too few events in patients < 20 years of age to conduct the analysis, and there was no association between exposure status and first A-fib/flu in patients ≥ 65 years. The higher aIRR of first A-fib/flu was consistent across sensitivity analyses that varied exposure definitions but was substantially attenuated in an analysis that adjusted for an alternative set of confounders (aIRR 1.01, 95% CI 0.96; 1.06) than those used in the main analysis. This attenuation suggests that the association between DL and A-fib/flu found in the main analysis may be due to residual confounding.

Finally, Substudy 5 showed found no association between exposure status and first recurrent seizure, IR=8,602 per 100,000 PY exposed follow-up time and IR 5,645 unexposed follow-up time, with an aIRR 1.26 (95% CI 0.94; 1.67) adjusted for confounders selected in the DAG exercise.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 68 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

11.2 Limitations Variables used for inclusion and exclusion criteria

In Substudies 2B and 5B, seizure outcomes for which a drug overdose was also registered at the same hospitalization are excluded. In general, drug overdoses are likely to be underreported in the registers, as the registers often do not have sufficient detail to differentiate between overdoses from different types of drugs. Therefore, it was not possible to distinguish between overdoses from DL and overdoses from other drugs. Because overdoses of DL are rare compared to other drug overdoses, persons with any drug overdose registered during the same hospitalization as the seizure were excluded from analyses 2B and 5B. As a result, this study cannot determine whether the seizure outcomes were specifically associated with DL overdoses.

Sample size

Based on the assumptions made, the sample size was not adequate in the youngest age group for the assessment of the potential association between DL use and A-fib/flu. Also, in this age group only relatively strong associations (IRR=8.2) can be properly assessed for the association between DL use and SVT.

Lack of data from Norway

Data from Norway were not available for the study in time to be included in the analyses. This is a limitation of the study population and sample size. Inclusion of the Norwegian data would have had a negligible impact on the minimum detectable IRRs and would not have an effect on the power of the study as described in the protocol.Because the study period is shorter for Norway (2008-2015) than for the other 3 countries (2001-2015), the sample size estimates and the minimum detectable IRR in the original protocol were calculated without Norway and would still be applicable when not including Norway in the analyses. Therefore, this limitation is presumed to have a small effect on the results.

Prescription redemption versus consumption

It is likely that in this study there was exposure misclassification. The prescription registers contain information about redeemed prescriptions, but no information about actual use of the drug. It is difficult to ascertain when patients were actually taking DL because it is a medication taken on an as needed basis. Therefore, lack of information about actual date of use of the drug is considered a major limitation. To address this limitation, multiple sensitivity analyses were conducted under the assumption that the closer to the date of the prescription redemption the more likely a patient is to be taking the drug. Further, sensitivity analyses also examined associations with the study outcomes stratified by repeated prescription redemption, as repeated prescription redemption likely indicates medication use. Overall, the sensitivity analyses with alternative definitions of exposure status supported the results obtained in the main analyses. The robustness of the findings to alternative definitions of exposure is reassuring.

Seizure diagnoses

The incidence rate of seizure in the general population in Sweden was nearly double the size as compared to Denmark and Finland. The same pattern was seen in the incidence (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 69 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

rate of seizure in the currently DL unexposed periods. Seizure in the present study was defined using ICD-10 system with codes R560 and R568 for febrile and non-febrile seizures, respectively. These codes are used for diagnosing symptoms. However, a seizure diagnosis can also be coded using G-codes, used for diagnosing neurological diseases. Seizure has not been defined in the present study based on G-codes. No information is available about how often the R- and G-codes are used. However, there is no reason to believe that within a country, there would have been differential misclassification of seizure by exposure status.

The number of individuals with a recurrent seizure was limited. Therefore, the analyses of time to a recurrent seizure were sensitive to small changes. The power of these analyses was limited.

Lack of migration history in Finland

Migration history for the Finnish population was not available for the study. Therefore, individuals with a DL prescription redemption within the first 6 months after immigration could not be identified as potential prevalent users, who should have been excluded in the analysis. In Denmark and Sweden, the number of individuals excluded due to their migration history was small and the consequence of lacking migration history for Finland is also most likely small.

Lack of information on potential confounders

Potential confounders were identified using DAG methodology prior to initiating the analysis of data. Clinical experts were involved in DAG exercise. Two sets of potential confounders were identified for each outcome. One set was used for all main analyses; the second set was used in a supplementary analysis.

After performing the PASS, additional potential confounders have been considered. Concomitant medication use was not considered as a potential confounder when the DAG was derived.Data on concomitant medication use were not requested at the time of application for register data. The prespecified analysis only stipulated the extraction of specific medication use (DL, loratadine, non-sedating antihistamines, or anti-hypertensive medication) from the register data and therefore, other medication use could not be controlled for in the analysis. Potential confounders, such as, use of sympathomimetic decongestant drugs, which may be associated with A-fib/flu and may also be associated with initial symptomatic treatment of allergic rhinitis. For example, a published case report describes a patient with A-fib induced by use topical nasal decongestants (tramazoline) [Ref. 5.4: 057520]. A 2012 review of studies of drug induced atrial fibrillation identified other potential proarrhythmic drugs including, non-steroidal anti- inflammatory drugs (NSAIDS), corticosteroids, respiratory medications (oral steroids, oral beta-agonists, theophyllines), alendronate; these drugs had relative risks ranging from 1.3 for beta-agonists and 6.1 for high-dose pulse glucocorticoid therapy [Ref. 5.4: 05750T] It is conceivable, then, that the observed association between DL and A-fib/flu could be confounded by unmeasured medication use, particularly decongestants.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 70 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Quantitative bias analysis The originally planned quantitative bias analysis could not be performed, as explained in [Section 9.9.5]. An alternative analysis was performed to evaluate the consequence of false negative diagnoses of the confounding variables. This alternative analysis showed that, although a larger number of individuals would have been diagnosed with asthma, rhinitis or chronic urticaria using a longer interval to ascertain confounders, no substantial change was seen on the association between DL exposure periods and the outcomes.

11.3 Interpretation

This study found an association between DL exposure and two study outcomes, first seizure (aIRR 1.15, 95% CI 1.03; 1.29) and first A-fib/flu (aIRR 1.08, 95% CI 1.02, 1.13), that could not readily be explained by chance. No association with DL exposure and SVT was found. As with all epidemiological studies, the central challenge in interpreting an observed association is to judge whether it represents a causal effect versus a spurious finding, which might be explained by bias or confounding. In the discussion that follows, we present the evidence for and against a causal interpretation, for seizure and then for A-fib/flu largely based on Hill’s criteria for causation [Ref. 5.4: 05752B].

There are several components of Hill’s criteria for causation: strength of association, consistency, temporality, biological gradient, plausibility/coherence, specificity, and experiment. Selected components are discussed in turn below.

Seizure

Strength of association: The results of this PASS indicated a small-to-modest strength of association between DL exposure and seizures. The results of the main analysis show 47% increase in first seizure in 0-to 4-year-olds and 32% increase in first seizure in 5- to19 year-olds when comparing exposed with unexposed person-time, adjusting for the confounders selected in the DAG exercise.

Consistency: While there is variation in the effect estimates for this association in these age categories across countries, the association persisted in multiple sensitivity analyses that varied the exposure definitions and confounders that were adjusted for in the analyses. The presence of the finding across multiple sensitivity analyses indicates a consistent effect.

Specificity: Specificity means that a specific cause results in a specific outcome and that a specific outcome results from a single cause. There is no evidence of specificity, as there are many causes of seizures.

Temporality: As patients with a prior history of seizure were excluded, the results of this PASS show a temporal relationship between DL exposure and seizures.

Biologic gradient: While this study did not specifically examine dose of DL, a higher aIRR was found in time periods (0-4 weeks since last prescription) compared with ≥ 27 weeks since last prescription. It is clinically reasonable to assume that the probability of actual exposure to an as-needed medication is higher in the first four weeks following its dispensing than times more remote from the dispensing. (EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE P A GE 71 EPI DE MI OL O G Y N O.: EP 07044.002 PROTOCOL NO/A MEND MEN T N O.: M K -4117 -2 0 3 E U P AS RE GISTE R N O./ EUDRACT NO .: 15038 F urt her , t h e re was an association bet ween D L e xposure and seizure f oll o wi n g t h e first D L pr es cri pti o n co mpared with n o n- exposed person -ti m e . T h er e w a s n o ass o ci ati o n b et w e e n D L a n d first s ei z ur e f oll o wi n g t h e s e c o n d or t hir d pr es cri pti o n s co mpared with non - exposed person -ti m e . T w o p ot e nti al i nt er pr et ati o ns ar e c o nsist e nt wit h t his o bs er v ati o n. O n e is t h at t h e ti m e aft er first dis p e nsi n g r e pr es e nts a p eri o d w h e n a ct u al us e of t h e dr u g w as m ost li k el y . T his fi n di n g is als o c o nsist e nt wit h t h e c o n c e pt of d e pl eti o n of s us c e pti bl es; it is p ossi bl e t h at p ati e nts w h o ar e pr o n e t o D L - ass o ci at e d s ei z ur e e x p eri e n c e t h e s y m pt o ms e arl y a n d t h e n , st o p t a ki n g t h e dr u g .

Pl a usi bilit y/ c o h er e n c e: W hil e a m e c h a nis m f or a cti o n h as n ot b e e n est a blis h e d t o d et er mi n e pl a usi bilit y /coherence, e xisti n g lit er at ur e a n d p ost -marketing reports i n di c at e t h at s o m e p ati e nts wit h a hist or y of pre -e xisti n g s ei z ur es h a d seiz ures t h at worsened when D L was introduced and did not recur when D L w as dis c o nti n u e d [ R ef. 5. 4: 045 V R5] . T his s c e n ari o i n di c at es a p ositi v e c h all e n g e – de- c h all e n g e t est a n d l e n ds cr e d e n c e t o pl a usi bilit y .

Experi ment: C CI

Phar macovigilance and pri or lit er at ur e

I n 2 0 0 6, based on phar macovigilance and i nf or m ati o n fr o m spontaneous reports, the M A H a d d e d s ei z ur e t o t h e list of a d v ers e r e a cti o ns i n t h e E U Su m mary of Pr o d u ct c h ar a ct eristi cs [ R ef. 5. 4: 0578S2] .

I n 2 0 1 3, a p u bli c ati o n [ R ef. 5. 4: 045 V R5] describing 4 case reports of seizures in children wit h a f a mil y hist or y of seizures or other relevant medical history i n di c at e d t h at s o m e p ati e nts wit h a hist or y of pr e -e xisti n g s ei z ur es had seizures that worsened when D L w as introduced and did not recur when D L was discontinued . T his s c e n ari o i n di c at es a p ositi v e c h all e n g e – de- c h all e n g e t est. Subsequent t o t his p u bli c ati o n, b ut pri or t o t h e c o m pl eti o n of t his P A S S, a n u p d at e t o t h e S p e ci al Warnings and Precautions Section of t h e D L label for convulsions was added i n 2017 [ R ef. 5. 4: 045 V R5] , [ R ef. 5. 3. 6: 04 NF N0] . The European SP C st at es: “ Desloratadine should be ad ministered wit h c a uti o n i n p ati e nts wit h m e di c al or f a mili al hist or y of s ei z ur es, a n d m ai nl y y o u n g children, being more susceptible to develop ne w seizures under desloratadine treat ment. Healthcare providers may conside r dis c o nti n ui n g d esl or at a di n e i n p ati e nts w h o e x p eri e n c e a s ei z ur e w hil e o n tr e at m e nt . ”

(E U G UI D A NCE: 23 J A N U AR Y 2013 E M A/738724/2012)

05822 K DESLORATADINE PAGE 72 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Summary This PASS indicates a small-to-modest association between DL exposure and seizures, particularly in children 0-4 years of age. This association is largely consistent across substudies and sensitivity analyses that varied the DL exposure definition and confounder set considered in the analysis. Pharmacovigilance data and results from this PASS it is reasonable to consider seizures as an adverse reaction to DL.

Public Health Implications

Based on the magnitude of association observed in this study, the safety risk posed by DL exposure on a population basis would be expected to be small. More specifically, among children 0-4 years of age, the absolute increase in incidence rate of first seizure in currently DL exposed person-time compared to not currently exposed person-time was 41.8 per 100,000 PY. Based on this rate difference, assuming a causal relationship and no residual confounding, one would expect 1 additional incident seizure for every 2,392 children aged 0-4 years using DL for one year. Among children and teenagers 5-19 years of age, the absolute increase in incidence rate of first seizure in currently DL exposed person-time compared to not currently exposed person-time was 17.3 per 100,000 PY. Based on this rate difference, assuming causal relationship and no residual confounding, one would expect 1 additional incident seizure for every 5,870 children or teenagers aged 5-19 years receiving DL for one year. The foregoing statements are true assuming continual use. While patients using DL may have more severe allergy symptoms than patients using over-the-counter medications, DL is used for symptomatic treatment and is seldom used as continual year-round therapy.

A-fib/flu

Strength of association: This PASS found a small association between DL exposure and A-fib/flu. The results of the main analysis show 8% increase in first A-fib/flu overall and 16% increase in first A-fib/flu in 20-to 64-year-olds when comparing exposed with unexposed person-time and adjusting for confounders selected in the DAG exercise.

Consistency: While the results of this study consistently show an association between DL exposed person-time and A-fib/flu in the 20-to 64-year age group in both the main analysis and sensitivity analyses that vary the exposure definition, this association is not seen in the 65 year and older age category. In the general population, overall rate of A- fib/flu is highest in the 65 years and older age category, the age category that is presumably most vulnerable to A-fib/flu. It is difficult to understand why a true causal effect would not be observed in the age category with the highest underlying risk. Further, the overall attenuation of effect and lack of association between DL exposed person-time and A-fib/flu when adjusting for an alternative confounder set suggests residual confounding may be driving the results observed in the main analysis. The lack of consistency, as evidenced by the lack of association in the 65 year and older age category and lack of association with adjustment for an alternative confounder set, does not support a causal interpretation.

Specificity: Specificity means that a specific cause results in a specific outcome and that a specific outcome results from a single cause. In this instance, specificity does not apply, as A-fib/flu may result from multiple risk factors and causes.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 73 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Temporality: As patients with a recorded prior history of A-fib/flu were excluded, the results of this PASS indicate a temporal relationship between DL exposure and A-fib/flu. DL use occurs prior to the outcome.

Biologic gradient: While dose was not specifically examined to assess the presence of a biological gradient, there was a higher magnitude of association after first DL prescription compared with the main analysis. This association is not seen following the second or third DL prescription, A separate sensitivity analysis shows that the association between DL and A-fib/flu wanes after time since last prescription redemption increases from 0-8 weeks to ≥ 9 weeks.

Plausibility/coherence: While no specific mechanistic studies were undertaken, review of preclinical data evaluating the electrophysiological properties of the drug did not find a pro-arrhythmic mechanism suggested by pharmacology [Ref. 5.3.6: 03XD75]. Thus, there is no known mechanism of action that supports the plausibility/coherence of this association.

Experiment: In this PASS, an association between DL exposure and A-fib/flu was only seen in the 20 to-64-year-old age group. No signal for A-fib/flu was observed in DL clinical trial data [Ref. 5.3.6: 00LZQ9], [Ref. 5.3.6: 03XD75].

Pharmacovigilance and prior literature

The Merck Adverse Event Reporting and Review System database was searched through 31-Jan-2019, for reports related to DL and DL + pseudoephedrine. There were 18 events listed as A-fib (12) or flutter (6) listed with DL and no events for DL + pseudoephedrine. Half of the reports for A-fib contained insufficient information and the other half were confounded by concurrent conditions/medical history and/or concomitant medications. Approximately 83% of the post marketing reports identified for DL and flutter had limited information. One remaining case confounded by a previous history of arrhythmias.

The spontaneous reports included no instances of positive rechallenges for A-fib/flu.

The review of the literature search results did not identify any published clinical articles reporting a clinically significant association between DL and A-fib/flu.

Review of post-marketing cases and the available evidence in the published literature does not support an association between DL and A-fib/flu.

Summary

This PASS found a small association between DL exposure and A-fib/flu. Evidence is insufficient to conclude that the association between current DL use and A-fib/flu is causal. Several factors argue against a causal interpretation for the association between DL and A-fib/flu, including the strength of association and lack of consistency in the findings in subgroup and sensitivity analyses. The lack of association when adjusting for an alternative confounder set indicates that residual confounding may influence results.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 74 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

Alternative explanations Given that interpretation of the A-fib/flu findings through the lens of Hill’s criteria are inconclusive, it is helpful to consider alternative explanations that may help explain A- fib/flu findings.

As noted in the limitations section, this study lacked information on some potential confounders. Several factors that could confound the relationship between DL and A- fib/flu were not controlled for in the analysis. First, concomitant medication use was not captured.In particular, use of sympathomimetic drugs or other stimulants, which may be associated with A-fib/flu and may also be associated with initial symptomatic treatment of allergic rhinitis. Confounding by sympathomimetic use may explain why no association was observed in the 65 year and older age group, in whom treatment with sympathomimetics is relatively contraindicated. This confounder was considered post hoc, when conducting the analysis of the study, and therefore, no data are available to include concomitant medication use as a confounder.

Second, clinically relevant comorbidities and risk factors for A-fib/flu were not adjusted for in the main analysis. For example, in supplementary analysis 8, an alternate confounder set was considered, which included seasonality, asthma status, diabetes, hypo/hyper-thyroidism, inflammatory disease, infections, type 1 allergy and severe rhinitis combined and anti-hypertensive treatment. After adjustment for this alternate confounder set, DL was no longer associated with A-fib/flu (aIRR 1.01, 95% CI 0.96, 1.06, Table S8.3). This attenuation suggests that residual confounding may affect the A- fib/flu results from the main analysis, and the alternative confounder set may be more clinically relevant with respect to A-fib/flu.

Public health implications

This PASS found a weak association between DL exposed person-time and A-fib/flu patients in 20-to 64-year age group. Given the small magnitude of association and no association in older age groups where the background risk is greatest, it is difficult to conclude that the association is causal. The possibility of residual confounding is also a plausible explanation for the results as an alternative confounder set attenuates these results. However, if one were to assume that the observed association is causal, the absolute risk difference is modest, which means that the risk of A-fib/flu attributable to DL exposure would be small. Among individuals 20-64 years of age currently exposed to DL, the absolute overall increase in first A-fib/flu compared to individuals currently not exposed was 17.3 per 100,000 PY.

Study Strengths

The main strength of this study is that it includes nationwide population register data covering the entire population from three Nordic countries, and includes more than 1.8 million DL users. In addition to the large sample size, these data have minimal patients lost-to-follow-up and limited selection bias, as these data cover the entire country. This has relatively long follow-up time especially for the Danish and Finnish populations. Further, by using patients who were all initially exposed to DL, the study design minimizes confounding by patient factors that do not vary over time. Data include individual-level information on potential confounders for all individuals included. The

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 75 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

quality of the registers is high due to the completeness and validity of the registered data [Ref. 5.4: 045TYK], [Ref. 5.4: 045TZL], [Ref. 5.4: 00W4G3].

The study outcomes have been previously validated. Previous studies validated the definition used for febrile seizure in Danish National Patient Register and reported a positive predictive value of 93%. Additionally, the A-fib/flu definition was validated in Danish National Patient Register and had a positive predictive value of 93%.

11.4 Generalisability

In this study, nation-wide Nordic registers were used with information on prescription drugs and hospital discharges. Since all registers have high completeness and high validity and all DL users and all individuals with seizure, SVT or A-fib/flu were included, the results have high external validity with generalizable results to Nordic and other European populations.

12 OTHER INFORMATION

None.

13 CONCLUSION

This study was undertaken to evaluate the potential associations between DL and four outcomes, seizure, SVT, A-fib/flu and first recurrent seizure using Nordic national register data from Denmark, Finland, and Sweden. Using a register-based cohort design, the study enrolled new users of DL and compared periods of presumed current use with periods of presumed non-use, based on time since last dispensing. The study identified a small association with incident seizure overall (aIRR 1.15, 95% CI 1.03; 1.29), adjusted for age-, sex, country- calendar year, asthma status, severe rhinitis status, and chronic urticaria. This finding was largely driven by associations in the younger (0-to 4 and 5-to 19-year old) age groups. For first seizure, aIRR is about 1.5 for 0-to 4-year-olds and 1.3 for 5-to 19-year-old age groups comparing exposed to unexposed PY. The associations were largely consistent across sensitivity analyses. Although the magnitude of association between seizure and DL found in this report is not strong, the findings are consistent with pharmacovigilance data and for this reason seizure should be considered an adverse drug reaction to DL. Even if causality is assumed, however, the absolute rate difference is 42 per 100,000 PY in the 0-4 and 17 per 100,000 PY in the 5-to 19-year age groups, indicating absolute increases in risk are small.

The study found no association between current use of DL and risk of first SVT.

The study also found an association between current use of DL and risk of first A-fib/flu that persisted after adjustment for preselected confounders (aIRR 1.08, 95% CI 1.02; 1.13). In analyses stratified by age, the association was strongest for patients aged 20-64 years (aIRR 1.16, 95% CI 1.07; 1.27) and was not seen in the elderly, in whom the baseline risk of this outcome is known to be highest. Several factors argue against a causal interpretation for the association between DL and A-fib/flu. This association was not observed in the 65 year and older age group where background risk is highest, was small in magnitude, and could be plausibly explained by residual confounding as evidenced by the attenuation with sensitivity analysis adjusting for alternative

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 76 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

confounders. Evidence is insufficient to conclude that the association between current DL use and A-fib/flu is causal.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 77 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

REFERENCES [Ref. 5.3.6: 00LZQ9] Global Clinical Development Worldwide Marketing Application: Integrated analysis of safety desloratadine tablets/syrup/reditabs, 20-Mar-2009. 773 p. [Ref. 5.3.6: 00V5KT] Global Clinical Development Agency Response: SCH 34117 FU2 048.2 - pharmacovigilance, 20-Mar-2012. 7 p. [Ref. 5.3.6: 03XD75] Global Clinical Safety and Pharmacovigilance Response Document: Comprehensive safety summary - desloratadine, 14- Jul-2014. 112 p. [Ref. 5.3.6: 04NFN0] Global Clinical Safety and Pharmacovigilance Label Update: 2.5 Clinical overview MK-4117 desloratadine, 12-Apr-2017. 7 p. [Ref. 5.4: 00W4CX] Furu K, Wettermark B, Andersen M, Martikainen JE, Almarsdottir AB, Sorensen HT. The nordic countries as a cohort for pharmacoepidemiological research. Basic Clin Pharmacol 2009;106:86-94. [Ref. 5.4: 00W4CY] Hauser WA, Beghi E. First seizure definitions and worldwide incidence and mortality. Epilepsia 2008;49(Suppl 1):8-12. [Ref. 5.4: 00W4D0] Thygesen L, Daasnes C, Thaulow I, Bronnum-Hansen H. Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving. Scan J Public Health 2011;39(Suppl 7):12-6. [Ref. 5.4: 00W4G3] Pedersen CB. The Danish Civil Registration System. Scan J of Public Health 2011;39(7 suppl):22-5. [Ref. 5.4: 03QPDQ] Glanz JM, McClure DL, Xu S, Hambidge SJ, Lee M, Kolczak MS, et al. Four different study designs to evaluate vaccine safety were equally validated with contrasting limitations. J Clin Epidemiol 2006;59:808-18. [Ref. 5.4: 03RSH6] Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim J-L, Reuterwall C, et al. External review and validation of the Swedish national inpatient register. BMC Public Health 2011;11:1-16. [Ref. 5.4: 03XNL7] Wilke T, Groth A, Mueller S, Pfannkuche M, Verheyen F, Linder R, et al. Incidence and prevalence of atrial fibrillation: an analysis based on 8.3 million patients. Europace 2013;15:486- 93. [Ref. 5.4: 03XNM4] Orejarena LA, Vidaillet H Jr, DeStefano F, Nordstrom DL, Vierkant RA, Smith PN, et al. Paroxysmal supraventricular tachycardia in the general population. J Am Coll Cardiol 1998;31(1):150-7. [Ref. 5.4: 045RCQ] Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999 Jan;10(1):37-48. [Ref. 5.4: 045TYK] Kildemoes HW, Sorensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health. 2011 Jul;39(7 Suppl):38-41. [Ref. 5.4: 045TZL] Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health. 2011 Jul;39(7 Suppl):30-3.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 78 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

[Ref. 5.4: 045VPW] The NORDCAN project [Internet]. Copenhagen: Association of Nordic Cancer Registries; c2009 [updated 2014 Dec 17]. Available from: http://www- dep.iarc.fr/NORDCAN/english/frame.asp. [Ref. 5.4: 045VR5] Cerminara C, El-Malhany N, Roberto D, Lo Castro A, Curatolo P. Seizures induced by desloratadine, a second-generation antihistamine: clinical observations. Neuropediatrics. 2013 Aug;44(4):222-4. [Ref. 5.4: 045VXZ] Kirkwood BR, Sterne JA. Essential Medical Statistics. 2nd ed. Malden (MA): Blackwell Publishing; 2003. [Ref. 5.4: 045W09] Pukkala E. Nordic biological specimen bank cohorts as basis for studies of cancer causes and control: quality control tools for study cohorts with more than two million sample donors and 130,000 prospective cancers. In: Dillner J, editor. Methods in biobanking. New York: Springer; 2011. p. 61-112. [Ref. 5.4: 045WWJ] Bryant E, Morganstein DR. Sample size determination for longitudinal surveys. Survey Research Methods Section, American Statistical Association Meeting; 1987 Aug 17-20; San Francisco, CA: 1987. [Ref. 5.4: 045WX9] Textor J, Hardt J, Knuppel S. DAGitty: a graphical tool for analyzing causal diagrams. Epidemiology. 2011 Sep;22(5):745. [Ref. 5.4: 0469WT] Bakken IJ, Gystad SO, Christensen OO, Huse UE, Laronningen S, Nygard J, et al. Comparison of data from the Norwegian Patient Register and the Cancer Registry of Norway. Tidsskr Nor Laegeforen. 2012 Jun 12;132(11):1336-40. [Ref. 5.4: 0469YG] Klaukka T. The Finnish database on drug utilization. Nor Epidemiol. 2001;11(1):19-22. [Ref. 5.4: 046B36] Sund R. Quality of the Finnish Hospital Discharge Register: a systematic review. Scand J Public Health. 2012 Aug;40(6):505- 15. [Ref. 5.4: 046B3C] Wettermark B, Hammar N, Fored CM, Leimanis A, Otterblad Olausson P, Bergman U, et al. The new Swedish Prescribed Drug Register--opportunities for pharmacoepidemiological research and experience from the first six months. Pharmacoepidemiol Drug Saf. 2007 Jul;16(7):726-35. [Ref. 5.4: 04C0BP] Fox MP, Lash TL, Greenland S. A method to automate probabilistic sensitivity analyses of misclassified binary variables. Int J Epidemiol. 2005 Dec;34(6):1370-6. [Ref. 5.4: 04C0BQ] Frost L, Vestergaard P. Alcohol and risk of atrial fibrillation or flutter: a cohort study. Arch Intern Med. 2004 Oct 11;164(18):1993-8. [Ref. 5.4: 04C0BS] Frost L, Andersen LV, Vestergaard P, Husted S, Mortensen LS. Trend in mortality after stroke with atrial fibrillation. Am J Med. 2007 Jan;120(1):47-53. [Ref. 5.4: 04C0BT] Lash TL, Fox MP, Fink AK. Statistics for biology and health. Gail M, Krickeberg K, Samet J, Tsiatis A, Wong W, editors. Dordrecht: Springer; Applying quantitative bias analysis to epidemiologic data; 2009.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 79 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

[Ref. 5.4: 04C0BZ] Rix TA, Riahi S, Overvad K, Lundbye-Christensen S, Schmidt EB, Joensen AM. Validity of the diagnoses atrial fibrillation and atrial flutter in a Danish patient registry. Scand Cardiovasc J. 2012 Jun;46(3):149-53. [Ref. 5.4: 04C0C2] Schmidt M, Schmidt SA, Sandegaard JL, Ehrenstein V, Pedersen L, Sorensen HT. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol. 2015 Nov 17;7:449-90. [Ref. 5.4: 04C0C7] Vestergaard M, Obel C, Henriksen TB, Christensen J, Madsen KM, Ostergaard JR, et al. The Danish National Hospital Register is a valuable study base for epidemiologic research in febrile seizures. J Clin Epidemiol. 2006 Jan;59(1):61-6. [Ref. 5.4: 04CD8S] Enger C, Cali C, Walker AM. Serious ventricular arrhythmias among users of cisapride and other QT-prolonging agents in the United States. Pharmacoepidemiol Drug Saf. 2002 Sep;11(6):477-86. [Ref. 5.4: 04CH6J] Lash TL, Fink AK. Semi-automated sensitivity analysis to assess systematic errors in observational data. Epidemiology. 2003 Jul;14(4):451-8. [Ref. 5.4: 05750T] Tamargo J, Caballero R, Delpon E. Drug-induced atrial fibrillation. Expert Opin Drug Saf. 2012;11(4):615-34. [Ref. 5.4: 05750Y] Lewis JD, Schinnar R, Bilker WB, Wang X, Strom BL. Validation studies of the health improvement network (THIN) database for pharmacoepidemiology research. Pharmacoepidemiol Drug Saf. 2007;16:393-401. [Ref. 5.4: 057520] Wieneke H. Induction of atrial fibrillation by topical use of nasal decongestants [letter]. Mayo Clin Proc. 2016 Jul;91(7):977. [Ref. 5.4: 05752B] Schunemann H, Hill S, Guyatt G, Akl EA, Ahmed F. The GRADE approach and Bradford Hill's criteria for causation. J Epidemiol Community Health. 2011;65:392-5. [Ref. 5.4: 0578S2] European Medicines Agency. Aerius: procedural steps taken and scientific information after the authorization. London (England): European Medicines Agency (EMA); 2018. 18 p. EMA/472210/2018.

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K DESLORATADINE PAGE 80 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

ANNEX 1 Flow Diagram of Study Population

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K PAGE 81

Annex 1 Flow D1 FlowAnnex S ofiagram tudy P opulation

Denmark Finland Sweden 2001-2016 2002-2016 2006-2016

Prevalent DL users PrevalentDL

N=314,534 N=627,206 N=970,453

Exclusion of Exclusion N=2, 053 N=414 N=69,487 prevalentusers DL individualswith erroneous data

Incident DL users DLIncident

N=312,481 N=626,792 N=900,966 Outcomes Non-febrileseizure 18,710 27,300 50,210 Supraventriculartachycardia 41,311 16,567 29,621 Atrialfibrillation flutter or 129,659 121,296 176,415 Recurrent seizureRecurrent 2,712 4,500 10,607

Final combined data of indicent DL users indicentDLof datacombinedFinal

N=1,840,239 N=1,840,239 Outcomes Non-febrileseizure 96,220 Supraventriculartachycardia 87,499 Atrialfibrillation flutter or 427,370 Recurrent seizureRecurrent 17,819

057Y7G

05822K DESLORATADINE PAGE 82 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

ANNEX 2 Tables with the Main and Supplementary Analysis Results for the Desloratadine Study

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K PAGE 83

Annex 2 Tables with the Main and Supplementary Analysis Results for the Desloratadine Study

Overview of the analyses in the desloratadine study

Table 0. Overview of the analyses in the desloratadine study Analyses Description Prevalent users and incident users in the total population and stratified country, byagegroups, users thepopulationincident andstratifiedandPrevalent totalusers in sex,year, calendar seasonality, asthmatic status, severe rhinitis and chronic urticaria status status,rhinitis urticariachronicandasthmaticseasonality,severe Substudy 1: DL use Substudy1:DL DDD of DL for the total population and stratified country, asthmaticseasonality,by DLagesex, year,groups, forstatus,thepopulationDDD stratifiedcalendarandof total severe rhinitisurtichronicandsevere cariastatus Substudy2 A ratesofIncidenceseizure B DLseizuresand A ratesofIncidenceSVT Substudy3 B DLSVTand A ratesofIncidenceA - fib/flu Substudy4 B DLAand - fib/flu A seizure ratesofIncidencerecurrentfirst Substudy5 B DL seizures recurrentfirstand 1 ofAlternativeDLdefinitionexposure 2 febrile Stratifiednon and by - aged0amongseizureschildren febrile – 4 3 or DLhigh usedof withurticariahavechronicExclusion ofpersonsdiagnoseddoses 4 countryStratified by 5 misclassification effect the ofAnalysis potentialexposureof Supplementary analysesSupplementary 6 afterRisk first - (RX)etc. (RX) comparedfourth third,second, to everDL redemptionprescription 7 nonDefinition of - changed52to periods weeks exposed lastfollowingredemption prescriptionthe 8 adjustment ofofalternativeUseconfounders an set 9 bias misclassification of Quantitativeanalysisconfounder 10 withovernoanalyses Restrictingperiodto - the - saleofDL counter A- atrialflutter;fib/flu= or fibrillation DDD= dose dailydefined ; SVT=supraventriculartachycardiaDL=desloratadine;

1 1

05822K PAGE 84

Substudy 1: Descriptive analysis of desloratadine use

Table 1.1 Incidence rates and prevalence proportions of desloratadine (DL) users in the general population between 2001 and 2015, for the total population and stratified by age groups, country, sex, calendar year, seasonality, asthmatic status, severe rhinitis, and chronic urticaria status$

Incidence ofDLusers Prevalence DLusers of

Number of ofNumber Person - IR ofNumber PP General General incident incident at years 100,000(per CI95% prevalent (n/100,000 CI95% population # users risk* person - years) users persons) Total Totalpopulation 1,840,239 251,322,682 732 . 2 731 . 2 733 .3 1,912,193 259,619,683 736.5 735. 5 737. 6 0 – 4years 300,787 14,467,424 2 , 079 .1 2 , 071 .7 2086 . 5 304,190 14,923,274 2,038. 4 2,031.1 2,045.6 5 – 19years 392,812 44,413,452 884 . 4 881 .7 887 . 2 467,135 46,091,298 1,013.5 1,010.6 1,016.4 Age groupsAge 20 – years64 964,459 148,723,014 648 .5 647 . 2 649 .8 1,048,863 153,820,554 681. 9 680. 6 683. 2 ≥65years 182,181 43,718,792 416 . 7 414 . 8 418 . 6 219,921 44,784,557 491. 1 489.0 493.1 Denmark 312,481 82,498,274 378 .8 377 . 5 380 . 1 314,534 82,313,321 382.1 380.8 383. 5 Country Finland 626,792 74,640,876 839 . 7 837 . 7 841 . 8 627,206 74,495,114 841.9 839.9 844.0 Sweden 900,966 94,183,532 956 . 6 954 . 6 958 .6 970,453 102,811,249 943.9 942.0 945.8 Males 907,152 124,391,577 729 .3 727 .8 730 .8 937,682 128,464,005 729.9 728.4 731.4 Sex Females 933,087 126,931,105 735 . 1 733 . 6 736 . 6 974,511 131,155,678 743.0 741.5 744.5 2001 7,277 5,355,081 135 . 9 132 . 8 139 .1 7,306 5,346,213 136. 7 133.6 139.8 2002 23,119 10,574,862 218 . 6 215 . 8 221 .5 25,461 10,558,979 241.1 238. 2 244.1 2003 56,120 10,600,195 529 . 4 525 .1 533 . 8 63,942 10,587,529 603.9 599. 3 608.6 2004 53,901 10,629,356 507 . 1 502 . 8 511 .4 76,113 10,614,776 717. 1 71 2 . 0 722. 2 2005 67,084 10,662,082 629 .2 624 . 4 63 4.0 163,079 19,657,280 829.6 825. 6 833. 7 2006 180,625 19,781,356 913 . 1 908 . 9 917 . 3 264,013 19,736,511 1,337. 7 1,332.6 1,342.8 2007 146,180 19,904,899 734 .4 730 . 6 738 .2 278,995 19,843,128 1,406.0 1,400.8 1,411.2 Calendar yearCalendar 2008 144,639 20,022,080 722 . 4 718 .7 726 . 1 300,322 19,963,490 1,504. 4 1, 49 9 . 0 1,509. 8 2009 149,661 20,156,844 742 .5 738 . 7 746 .3 323,970 20,089,462 1,612.6 1,607. 1 1,618.2 2010 148,942 20,285,317 734 . 2 730 . 5 73 8.0 339,899 20,221,081 1,680.9 1,675. 3 1,686. 6 2011 141,827 20,404,361 695 .1 691 .5 698 . 7 347,183 20,344,839 1,706. 5 1,700.8 1,712. 2 2012 132,126 20,520,448 643 .9 640 . 4 647 .4 327,271 20,462,405 1,599. 4 1,593.9 1,604. 9 2013 155,461 20,646,294 75 3.0 749 . 2 756 . 7 377,916 20,583,371 1,836.0 1,830. 2 1,841.9 2014 228,970 20,795,444 1 , 101 .1 1 , 096 . 6 1105 .6 502,007 20,720,869 2,422.7 2,416.0 2,429.4 2015 204,307 20,984,063 973 . 6 969 . 4 977 .9 498,362 20,889,754 2,385. 7 2,379. 1 2,392.3 Winter 262,876 62,830,671 418 .4 416 .8 4 20.0 528,365 64,904,921 814. 1 811. 9 816. 3 Spring 750,350 62,830,671 1 , 194 . 2 1 , 191 . 5 119 7.0 1,052,834 64,904,921 1,622.1 1,619.0 1,625.2 Seasonality Summer 543,995 62,830,671 865 . 8 863 . 5 868 . 1 944,525 64,904,921 1,455.2 1,452.3 1,458. 2 Autumn 283,018 62,830,671 450 . 5 448 .8 452 . 1 574,389 64,904,921 88 5 . 0 882. 7 887. 3

2 2

05822K PAGE 85

Asthma 264,931 251,322,682 105 . 4 105 . 0 105 . 8 - - - - Severe rhinitisSevere 8,102 251,322,682 3 . 2 3 .2 3 .3 - - - - Chronic Chronicurticaria 43,097 251,322,682 17 .2 1 7.0 17 . 3 - - - - IR=incidence proportionrate;PP=prevalenceIR=incidence intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete * *Person - at risk years sizeofonestimated are by the useof populationDL redemptionprescription country,eachJulyfor 1in 5yeareach in - agegroups year obtained from database from theobtained NORDCAN # # is populationsizeof estimated byGeneralthe population usesize on 1in countryeachJanuary in 5yeareach - agegroups year obtained NORDCANfrom the database

3 3

05822K PAGE 86

Table 1.2 Distribution of the number of defined daily dose (DDD) of DL in the total population and stratified by age, sex, country, calendar year, seasonality, asthmatic status, severe rhinitis, and chronic urticaria status among incident DL users$ Per incident DL user DLincidentPer

N Mean SD Median Min - max Total DL population DLTotal of DLDDD 1,840,239 207.80 418.02 70 1 – 19,970 0 – 4years 300,787 79.06 94.58 48 12– 6,096 5 – 19years 450,993 182.99 321.01 72 3 – 9,560 Age groupsAge 20 – 64years 999,913 228.61 442.4 8 100 1 – 19,070 ≥ 65years 206,293 230.27 469.30 90 3– 12,360 Denmark 312,481 182.7 4 492.9 6 48 10 – 19,970 Country Finland 626,792 188.77 363.63 60 1 – 11,030 Sweden 900,966 229.73 423.90 100 3 – 18,740 Males 907,152 207.08 404.9 2 80 1 – 15,810 Sex Females 933,087 208. 50 430.37 60 3 – 19,970 2001 7,277 44.13 55.45 30 10 – 900 2002 25,346 51.35 75.70 30 10 – 2,500 2003 63,806 65.02 80.55 30 10 – 3,500 2004 75,987 74.27 87.17 30 10 – 1,500 2005 100,270 77.34 88.34 40 10 – 1,700 2006 227,5 88 82.88 85.28 60 10 – 1,900 2007 248,860 90.67 93.54 60 10 - 1,760 Calendar yearsCalendar 2008 273,040 94.89 99.32 60 3 - 2,720 2009 299,192 103.13 105.22 72 1 – 3,560 2010 317,453 104.18 105.85 90 1 – 2,600 2011 326,499 107.68 108.12 90 1 – 3,420 2012 307,724 110.54 111.59 96 10 – 3,500 2013 359,083 119.15 119.20 100 10 – 2,990 2014 482,086 120.27 118.02 100 6 – 3,200 2015 478,880 129.53 126.62 100 4 – 4,500 Winter 491,769 116.53 158.10 60 1 – 5,140 Spring 1,0 10,183 147.08 185.02 96 1 – 5,830 Seasonality Summer 891,480 128.07 167.08 80 1 – 5,290 Autumn 529,082 117.82 164.17 60 1 – 6,320 No registration 1,575,308 190.70 392.80 60 1 – 19,970 Asthma Asthma 264,931 309.45 533.18 100 3 - 12,800 Severe rhinitisSevere registrationNo 1,8 32,137 206.96 416.52 64 1 – 19,970 4 4

05822K PAGE 87

rhinitisSevere 8,102 397.63 649.65 16 0 10 – 10,900 No registrationNo 1,797,142 205.93 413.31 60 1 – 19,970 Chronic urticariaChronic Chronic urticariaChronic 43,097 285.61 576.17 100 6 - 11,190 SD=Standard deviationSD=Standard $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete

5 5

05822K PAGE 88

Substudy 2A: First seizure Table 2A. Number of persons with a first-time diagnosis of seizures, risk time and incidence rate of seizures, overall and stratified by age groups, country and sex$ First ever everseizure First Variable Level N timeRisk (years)# IR (p er100,000 person - years ) CI95% Overall 96,220 251,322,682 38.3 38.0 38.5 0 – 4 16,878 14,467,424 116.7 114.9 118.4 5 – 19 16,885 44,413,452 38.0 37.4 38.6 Age Age groups 20 – 64 42,545 148,723,014 28.6 28.3 28.9 ≥65 19,912 43,718,792 45.5 44.9 46.2 Denmark 18,710 82,498,274 22.7 22.4 23.0 Country Finland 27,300 74,640,876 36.6 36.1 37.0 Sweden 50,210 94,183,532 53.3 52.8 53.8 Males 49,934 124,391,577 40.1 39.8 40.5 Sex Females 46,286 126,931,105 36.5 36.1 36.8 IR=incidence rateIR=incidence ; intervalCI=confidence $ For a complete listseeofpleasecompletecodes, a$For 6Annex # # estimated timeisRisk country,eachsize of5Julyonfor byyear1ineachthe use in population - NORDCANdatabasefromage groups theobtained year

6 6

05822K PAGE 89

Substudy 2B: Desloratadine and first seizure Table 2B1. Number of first seizure cases, risk time and incidence rate across persons currently and not currently exposed to desloratadine (DL)$ Risk time timeRisk Age groupsAge Exposurestatus Persons(N)included (N)seizures First 100,000(perIR person - years) CI95% (years) Currently DL exposed CurrentlyDL 397 995,212 39.9 36.2;44.0 All 1,561,410 Not currently DL exposed NotcurrentlyDL 1,900 6,307,9798 30.1 28.8;31.5 Currently DL exposed CurrentlyDL 100 87,091 114.8 94.4;139.7 0– 4years 291,345 Not currently DL exposed NotcurrentlyDL 255 349,496 73.0 64.5;82.5 Currently DL exposed CurrentlyDL 142 249,777 56.9 48.2;67.1 5– 19years 558,412 Not currently DL exposed NotcurrentlyDL 693 1,752,315 39.6 36.7;42.6 Currently DL exposed CurrentlyDL 115 566,997 20.3 16.9;24.4 20– years64 909,847 Not currently DL exposed NotcurrentlyDL 728 3,600,928 20.2 18.8;21.7 Currently DL exposed CurrentlyDL 40 91,346 43.8 32.1;59.7 ≥65 years 187,695 Not exposed currentlyDL 224 605,239 37.0 32.5;42.2 IR=incidence rateIR=incidence ; intervalCI=confidence $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete

Table 2B2. Analysis of the association between desloratadine (DL) exposure and first seizure$,a Crude Crudemodel Semi - model* adjusted adjustedFully model** Age groupsAge Exposurestatus IRR CI95% p - value IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 1. 32 1.19;1.48 1.15 1.03;1.29 1.15 1.03;1.29 All All <0.001 0.013 0.013 Not currently DL exposed NotcurrentlyDL 1 Ref 1 Ref 1 Ref Currently DL exposed CurrentlyDL 1.57 1.25;1.98 1.46 1.15;1.84 1.47 1.16;1.87 0– 4years <0.001 0.002 0.002 Not currently DL exposed NotcurrentlyDL 1 Ref 1 Ref 1 Ref Currently DL exposed CurrentlyDL 1.44 1.20;1.72 1.30 1.08;1.56 1.32 1.09; 1.59 5– 19 years <0.001 0.008 0.006 Not currently DL exposed NotcurrentlyDL 1 Ref 1 Ref 1 Ref Currently DL exposed CurrentlyDL 1.00 0.82;1.22 0.90 0.74;1.10 0.88 0.72;1.08 20– years64 0.97 0.29 0.22 Not currently DL exposed NotcurrentlyDL 1 Ref 1 Ref 1 Ref Currently DL exposed CurrentlyDL 1.18 0.85;1.66 1.10 0.78;1.54 1.12 0.80;1.58 ≥65 years 0.34 0.60 0.51 Not currently DL exposed NotcurrentlyDL 1 Ref 1 Ref 1 Ref IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years Years2001 a – 200 a4 combined re toofewto observations due to convergenceobtain * *A anddjusted sex,year forcountrycalendarage, A** djusted country, set:asthmaticseasonality,the adjustment sex,sufficient year, age,stminimumin calendar for confounders atus, rhinitis,severe urticariachronicstatus. Duetooto few performed rhinitis,withseverewasfew individualsadjustmentno age groupthestratified analysesin by .

7 7

05822K PAGE 90

Substudy 3A: First supraventricular tachycardia

Table 3A. Number of persons diagnosed with supraventricular tachycardia (SVT), risk time and incidence rate of SVT, overall and stratified by age groups, country and sex$ First ever everSVT First Variable Level N tRisk ime #(years) 100,000(perIR person - years) CI95% Overall 87,499 251,322,682 34.8 34.6 35.0 0 – 4 910 14,467 6.3 5.9 6.7 5 – 19 3,961 44,413,452 8.9 8.6 9.2 Age groupsAge 20 – 64 46,965 148,723,014 31.6 31.3 31.9 ≥65 35,663 43,718,792 81.6 80.7 82.4 Denmark 41,311 82,498,274 50.1 49.6 50.6 Country Finland 16,567 74,640,876 22.2 21.9 22.5 Sweden 29,621 94,183,532 31.5 31.1 31.8 Males 44,279 124,391,577 35.6 35.3 35.9 Sex Females 43,220 126,931,105 34.0 33.7 34.4 IR=incidence rateIR=incidence ; intervalCI=confidence $ For a complete listseeofpleasecompletecodes, a$For 6Annex # # eachsizeof Julyonestimated by1inthe use timeis Riskpopulation country, 5foryeareach in - NORDCANdatabasefromage groups theobtained year

8 8

05822K PAGE 91

Substudy 3B: Desloratadine and supraventricular tachycardia Table 3B1. Number of supraventricular tachycardia (SVT) cases, risk time and incidence rate across persons currently and not currently exposed to desloratadine (DL)$ Age groupsAge Exposurestatus Persons(N)included SVT(N) time(years)Risk 100,000(perIR person - years) CI95% Currently DL exposed CurrentlyDL 384 1,233,027 31.1 28.2;34.4 All 1,833,337 Not currently DL exposed NotcurrentlyDL 2,480 7,861,683 31.6 30.3;32.8 Currently DL exposed CurrentlyDL ≤10 - - - 0– # 4years 300,679 Not currently DL exposed NotcurrentlyDL ≤10 - - - Currently DL exposed CurrentlyDL 27 270,264 10.0 14.66.9; 5– 19years 594,480 Not currently DL exposed NotcurrentlyDL 194 1,890,300 10.3 11.88.9; Currently DL exposed CurrentlyDL 240 728,790 32.9 29.0;37.4 20– 64 years 1,103,449 Not currently DL exposed NotcurrentlyDL 1,469 4,626,640 31.8 30.2;33.4 Currently DL exposed CurrentlyDL 114 143,682 79.3 66.0;95.3 ≥65 years 278,186 Not currently DL exposed NotcurrentlyDL 815 980,353 83.1 77.6;89.0 IR=incidence rateIR=incidence ; intervalCI=confidence $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete # Too few analysisfew #the observationsTooperform to

Table 3B2. Analysis of the association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT)$,a Crudemodel Semi - model* adjusted adjustedFully model** Age groupsAge Exposurestatus IRR CI95% p - value IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 0.99 0.89;1.10 1.02 0.92;1.14 1.03 0.92;1.15 All All 0.81 0.69 0.61 Not exposed currentlyDL 1 ref 1 ref 1 ref Currently DL exposed CurrentlyDL ------0– # 4years - - - Not currently DL exposed NotcurrentlyDL 1 ref 1 ref 1 ref Currently DL exposed CurrentlyDL 0.97 0.65;1.46 1.00 0.66;1.50 1.02 0.67;1.55 5– 19 years 0.90 0.98 0.92 Not currently Notcurrently exposed DL 1 ref 1 ref 1 ref Currently DL exposed CurrentlyDL 1.04 0.90;1.19 1.06 0.92;1.21 1.07 0.93;1.23 20– years64 0.60 0.44 0.38 Not currently DL exposed NotcurrentlyDL 1 ref 1 ref 1 ref Currently DL exposed CurrentlyDL 0.95 0.78;1.16 0.95 0.78;1.16 0.95 0.78;1.15 ≥65 years 0.64 0.60 0.58 Not currently DL exposed NotcurrentlyDL 1 ref 1 ref 1 ref IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years2001 – 200 4 combined are to toofewtoconvergence observationsobtain due * *A anddjusted sex,year forcountrycalendarage, **A djusted country, set:asthmaticseasonality,the adjustment sex,sufficient year, age,stminimumin calendar for confounders severatus, e rhi nitis, urticariachronicstatus # Too few few# observationsTooperfor to m the analysis

9 9

05822K PAGE 92

Substudy 4A: Atrial fibrillation and flutter

Table 4A. Number of persons diagnosed with atrial fibrillation or flutter (A-fib/flu), risk time and incidence rate of A-fib/flu overall and stratified by age group, country and sex$ First ever everA First - fib/flu Variable Level N timeRisk #(years) 100,000(perIR person - years) CI95% Overall 427,370 251,322,682 170.0 169.5 170.6 0 – 4 83 14,467,424 0.6 0.5 0.7 5 – 19 510 44,413,452 1.1 1.1 1.3 Age groupsAge 20 – 64 119,101 148,723,014 80.1 79.6 80.5 ≥65 307,676 43,718,792 703.8 701.3 706.3 Denmark 129,659 82,498,274 157.2 156.3 158.0 Country Finland 121,296 74,640,876 162.5 161.6 163.4 Sweden 176,415 94,183,532 187.3 186.4 188.2 Males 224,994 124,391,577 180.9 180.1 181.6 Sex Females 202,376 126,931,105 159.4 158.7 160.1 IR=incidence rateIR=incidence ; intervalCI=confidence $ For a complete listseeofpleasecompletecodes, a$For 6Annex # # country,eachsize of 5Julyonestimatedfor byyear1ineachthe use timeinis Riskpopulation - fromagegroupsobtained year NORDCANdatabase the

05822K PAGE 93

Substudy 4B: Desloratadine and atrial fibrillation or flutter Table 4B1. Number of atrial fibrillation or flutter (A-fib/flu) cases, risk time and incidence rate across persons currently and not currently exposed to desloratadine (DL)$ Risk time timeRisk 100,000(perIR person - Age groupsAge Exposurestatus Persons(N)included A- (N)fib/flu CI95% (years) years) Currently DL exposed CurrentlyDL 1,749 1,225,076 142.8 136.2;149.6 All 1,822,783 Not currently DL exposed NotcurrentlyDL 10,629 7,804,139 136.2 133.6;138.8 Currently DL exposed CurrentlyDL ≤10 - - - 0– # 4years 300,739 Not currently DL exposed NotcurrentlyDL ≤10 - - - Currently DL exposed CurrentlyDL ≤10 - - - 5– #19 years 594,727 Not currently DL exposed NotcurrentlyDL ≤10 - - - Currently DL exposed CurrentlyDL 612 727,547 84.1 77.7;91.1 20– years64 1,102,093 Not currently DL exposed NotcurrentlyDL 3,383 4,618,803 73.2 70.8;75.8 Currently DL exposed CurrentlyDL 1,136 136,785 830.5 783.6;880.2 ≥65 years 267,085 Not currently DL exposed NotcurrentlyDL 7,219 929,496 776.7 759.0;794.8 IR=incidence rateIR=incidence ; intervalCI=confidence $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete # Too few analysisfew #the observationsTooperform to

Table 4B2. Analysis of the association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu)$,a Crudemodel Semi - model* adjusted adjustedFully model** Age groupsAge Exposurestatus IRR CI95% p - value IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 1.05 1.00;1.10 1.07 1.02;1.13 1.08 1.02;1.13 All All 0.070 0.007 0.005 Not currently DL exposed NotcurrentlyDL 1 ref 1 ref 1 ref Currently DL exposed CurrentlyDL ------0– # 4years - - Not currently DL exposed NotcurrentlyDL 1 ref 1 ref 1 ref Currently DL exposed CurrentlyDL ------5– #19 years - - Not currently DL exposed NotcurrentlyDL 1 ref 1 ref 1 ref Currently DL exposed CurrentlyDL 1.15 1.05;1.25 1.14 1.05;1.25 1.16 1.07;1.27 20– years64 0.002 0.003 <0.001 Not currently DL exposed NotcurrentlyDL 1 ref 1 ref 1 ref Currently DL exposed CurrentlyDL 1.07 1.00;1.14 1.05 0.98;1.11 1.04 0.98;1.11 ≥65 years 0.037 0.17 0.24 Not currently DL exposed NotcurrentlyDL 1 ref 1 ref 1 ref IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years2001 – 200 6 combined are to toofewtoconvergence observationsobtain due * *A djusted for and sex,year countrycalendarage, **A djusted for country, set:asthmaticseasonality,severthe adjustment sex,sufficient year, age,status,minimumin calendar confounders e rhi nitis, urticariachronicstatus # Too few analysisfew #the observationsTooperform to

05822K PAGE 94

Substudy 5A: First recurrent seizure

Table 5A. Number of persons with a first recurrent diagnosis of seizures $ First recurrent recurrent seizure First

N time(years)Risk # 100,000(perIR person - years) CI95% Total Totalpopulation 17,819 - - - Denmark 2,712 - - - Country Finland 4,500 - - - Sweden 10,607 - - - IR=incidence rateIR=incidence ; intervalCI=confidence $ For a complete listseeofpleasecompletecodes, a$For 6Annex # Estimates of risk time to the first recurrent seizure (approximated by the number of persons with seizure) is notavailablepersonswithseizure)by the of number (approximatedseizurerecurrent theoftimeto risk first#Estimates NORDCAN in

05822K PAGE 95

Substudy 5B: Desloratadine and first recurrent seizure Table 5B1. Number of first recurrent seizure cases, risk time and incidence rate across persons currently and not currently exposed to desloratadine (DL)$ Exposure Exposurestatus Persons(N)included recurrent First (N)seizure time(years)Risk 100,000(perIR person - years) CI95% Currently DL exposed CurrentlyDL 57 662 8 , 601.5 6 , 634.8;11 , 151.1 2,530 Not currently DL exposed NotcurrentlyDL 369 6 ,537 5 , 644.8 5 , 097. 3 6; , 251.2 IR=incidence rateIR=incidence ; intervalCI=confidence $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete

Table 5B2. Analysis of the association between desloratadine (DL) exposure and first recurrent seizure$,a Crudemodel Semi - model* adjusted adjustedFully model** Exposure Exposurestatus IRR CI95% p - value IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 1.52 1.15;2.01 1.27 0.96;1.68 1.26 0.94;1.67 0.005 0.11 0.13 Not currently DL exposed NotcurrentlyDL 1 r ef 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years Years2001 a – 200 3 combined are to toofewtoconvergence observationsobtain due * *A anddjusted sex,year forcountrycalendarage, **A djusted set:the adjustmentsufficient age,minimumin for confounders country, asthmaticseasonality,sever sex,year, status, calendar e rhi nitis, urticariachronicstatus

05822K PAGE 96

Supplementary analysis 1

Table S1.1 Analysis of the association between desloratadine (DL) exposure and first seizure when using an alternative exposure categorization$ Crudemodel adjustedFully model* Age groupsAge Exposurestatus (N)seizures First time(years)Risk IRR CI95% P - value IRR CI95% P - value 0 – 4weeks 145 348,682 1.38 1.17;1.64 1.18 1.00;1.40 5 – 8weeks 124 307,677 1.34 1.12;1.61 1.15 0.96;1.38 All 9 – 16weeks 219 519,387 1.40 1.22;1.61 <0.001 1.20 1.04;1.38 0.042 17 – weeks26 199 527,658 1.25 1.08;1.45 1.05 0.91;1.22 ≥27 weeks 1,900 6,307,978 1 r ef 1 r ef IRR= Incidence rate ratio; CI=confidence intervalIncidenceCI=confidence IRR= rateratio; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmaticsever e rhi nitis, status urticariachronic

Table S1.2 Analysis of the association between desloratadine (DL) exposure and supraventricular tachycardia (SVT) when using an alternative exposure categorization$ Crudemodel adjustedFully model* Age groupsAge Exposurestatus SVT(N) time(years)Risk IRR 95% CI P - value IRR CI95% P - value 0 – 4weeks 119 432,484 0.87 0.73;1.05 0.89 0.74;1.07 5 – 8weeks 115 378,782 0.96 0.80;1.16 1.02 0.84;1.23 All 9 – 16weeks 203 633,550 1.02 0.88;1.17 0.60 1.10 0.95;1.27 0.14 17 – weeks26 207 633,766 1.04 0.90; 1.19 1.15 1.00;1.33 ≥27 weeks 2,480 7,861,683 1 r ef 1 r ef IRR= Incidence rate ratio; CI=confidence intervalIncidenceCI=confidence IRR= rateratio; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient status,calendar country, set: asthmaticseasonality,confoundersthesever adjustment sex,sufficient adjusted year, age, minimum*Allin analyses for e rhi nitis, status urticariachronic

Table S1.3 Analysis of the association between desloratadine (DL) exposure and atrial fibrillation or flutter (A-fib/flu) when using an alternative exposure categorization$ Crudemodel adjustedFully model* Age groupsAge Exposurestatus A - (N)fib/flu time(years)Risk IRR CI95% P - value IRR CI95% P - value 0 – 4weeks 652 429,612 1.11 1.03;1.21 1.12 1.03;1.21 5 – 8weeks 549 376,360 1.07 0.98;1.17 1.11 1.02;1.21 All 9 – 16weeks 779 629,743 0.91 0.84;0.98 <0.001 0.99 0.92;1.06 0.012 17 – weeks26 730 630,457 0.85 0.79;0.92 1.01 0.94;1.09 ≥27 weeks 10,629 7,804,139 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmaticsever e status rhinitis,urticariachronic

05822K PAGE 97

Table S1.4 Analysis of the association between desloratadine (DL) exposure and first recurrent seizure when using an alternative exposure categorization$ Crudemodel adjustedFully model* First recurrent recurrent First Age groupsAge Exposurestatus time(years)Risk IRR CI95% P- value IRR CI95% P- value seizures (N)seizures 0 – 4weeks 17 703 0.81 0.50;1.32 0.67 0.80 0.49;1.31 0.61 5 – 8weeks 18 605 1.00 0.62;1.60 0.98 0.61;1.59 All 9 – 16weeks 33 984 1.13 0.79;1.61 1.10 0.77;1.59 17 – weeks26 23 963 0.80 0.53;1.22 0.77 0.50;1.19 ≥27 weeks 369 12,390 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient status,calendar country, set: asthmaticseasonality,confoundersthesever adjustment sex,sufficient adjusted year, age, minimum*Allin analyses for e rhinitis,uchronic statusrticaria

05822K PAGE 98

Supplementary analysis 2 Table S2.1 Analysis of the association between desloratadine (DL) exposure and first febrile seizure and first non-febrile seizure among children aged 0–4 years$, a Crudemodel adjustedFully model* First seizures seizures First timeRisk Febrile Febrilestatus Exposurestatus IRR CI95% p- value IRR CI95% p- value (N) (years) Currently DL exposed CurrentlyDL 385 86,442 1.76 1.56;1.99 1.88 1.66;2.13 All <0.001 <0.001 Not currently DL exposed NotcurrentlyDL 875 346,513 1 r ef 1 r ef Currently DL exposed CurrentlyDL 291 86,2 5 3 1.80 1.57;2.07 2.00 1.73;2.30 Febrile Febrileseizures <0.001 <0.001 Not currently DL exposed NotcurrentlyDL 647 345,9 60 1 r ef 1 r ef Currently DL exposed CurrentlyDL 94 86,12 8 1.66 1.30;2.11 1.57 1.22;2.01 Non - seizures febrile <0.001 <0.001 Not currently DL exposed NotcurrentlyDL 228 345,9 33 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years Years2001 a – 200 4 combined are toofewto due to observationsconvergence obtain *All analyses adjusted for confounders in the minimum sufficient asseasonality, sex, year,set: confoundersthe adjustment sufficientcalendar adjusted country,minimum*Allin analyses for thmaticstatus, chronicstatusurticaria too(duefewto individuals with severe rhinitis,withsevereindividualsadno wasperformed) justment

Table S2.2 Analysis of the association between desloratadine (DL) exposure and first recurrent febrile seizure and first recurrent non- febrile seizure among children aged 0–4 years$, a Crudemodel adjustedFully model* First recurrent recurrent First timeRisk Febrile Febrilestatus Exposurestatus IRR CI95% p- value IRR CI95% p- value seizures (N)seizures (years) Currently DL exposed CurrentlyDL 35 189 2.32 1.61;3.33 <0.001 2.22 1.53;3.21 <0.001 All Not currently DL exposed NotcurrentlyDL 178 2,229 1 r ef 1 r ef Currently DL exposed CurrentlyDL 16 142 1.90 1.12;3.21 0.027 1.94 1.14;3.30 0.025 Febrile Febrileseizures Not currently DL exposed NotcurrentlyDL 109 1,845 1 r ef 1 r ef Currently DL exposed CurrentlyDL 13 53 2.46 1.32;4.57 0.009 2.58 1.34;4.97 0.009 Non - seizures febrile Not currently DL exposed NotcurrentlyDL 43 433 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years Years2001 a – 200 6 combined are to toofewtoconvergence observationsobtain due *All analyses adjusted for confounders in the minimum sufficient asseasonality, sex, year,set: confoundersthe adjustment sufficientcalendar adjusted country,minimum*Allin analyses for thmaticstatus, chronicstatusurticaria too(duefewto individuals with severe rhinitis,withseverewasindividualsadjustmentperformed)no

05822K PAGE 99

Supplement ary analysis 3 Table S3.1 Analysis of the association between desloratadine (DL) exposure and first seizure when censuring risk time when persons are diagnosed with chronic urticaria or have high use of DL$, a Crudemodel adjustedFully model* Exposure Exposurestatus (N)seizures First time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 295 792,730 1.24 1.10;1.41 <0.001 1.08 0.95;1.22 0.27 Not currently DL exposed NotcurrentlyDL 1,734 5,795,206 1 r ef 1 r ef IRR= Inciden ce rateratio ; intervalCI=confidence $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years Years2001 a – combined 2002are to toofewtoconvergence observationsobtain due *All analyses adjusted for confounders in the minimum sufficient status,calendar country, set: asthmaticseasonality,confoundersthesever adjustment sex,sufficient adjusted year, age, minimum*Allin analyses for e rhinitis too(duefewto individuals with chronic urticaria, no adjustment was wasadjustmentperformed)no withurticaria,chronic

Table S3.2 Analysis of the association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) when censuring risk time when persons are diagnosed with chronic urticaria or have high use of DL$, a Crudemodel adjustedFully model* Exposure Exposurestatus SVT(N) time(years)Risk IRR CI95% p - value IRR* CI95% p - value Currently CurrentlyDL exposed 258 961,688 0.86 0.75;0.98 0.017 0.92 0.81;1.05 0.22 Not currently DL exposed NotcurrentlyDL 2,249 7,192,964 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a a Age groups0 - 5and 4 - toobservationsconvergence obtain dueto toofewcombinedare 19 *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmaticsever e rhinitis too(duefewto individuals with chronic urticaria, no adjustment was wasadjustmentperformed)no withurticaria,chronic

Table S3.3 Analysis of the association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) when censuring risk time when persons are diagnosed with chronic urticaria or have high use of DL$, a Crudemodel adjustedFully model* Exposure Exposurestatus A - (N)fib/flu time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 1,316 955,939 1.02 0.97;1.08 0.46 1.11 1.04;1.17 0.001 Not currently DL exposed NotcurrentlyDL 9,616 7,140,297 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a a Age groups0 5--4, 19 20and - 64 toobservationsconvergence obtain dueto toofewcombinedare *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmaticsever e rhinitis too(duefewto individuals with chronic urticaria, no adjustment was wasadjustmentperformed)no withurticaria,chronic

05822K PAGE 100

Supplementary analysis 4

Table S4.1 Analysis of the association between desloratadine (DL) exposure and first seizure for the total population and stratified by country$,a Crudemodel adjustedFully model* Country Exposurestatus (N)seizures First time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 397 995,211 1.32 1.19;1.48 1.15 1.03;1.29 All All <0.001 0.013 Not currently DL exposed NotcurrentlyDL 1,900 6,307,978 1 r ef 1 r ef Currently DL exposed CurrentlyDL 25 134,392 0.96 0.63;1.44 0.86 0.57;1.31 Denmark Denmark 0.83 0.47 Not currently DL exposed NotcurrentlyDL 243 1,247,493 1 ref 1 r ef Currently DL exposed CurrentlyDL 103 326,088 1.37 1.11;1.68 1.31 1.06;1.63 Finland 0.005 0.018 Not exposed currentlyDL 591 2,556,979 1 ref 1 r ef Currently DL exposed CurrentlyDL 269 534,731 1.18 1.03;1.35 1.14 0.99;1.31 Sweden Sweden 0.016 0.065 Not currently DL exposed NotcurrentlyDL 1,066 2,503,506 1 ref 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years Years2001 a - 200 3 combined are toofewtoconvergence observationsobtain to due *All analyses adjusted for confounders in the minimum sufficient the calendartotal country set: (onlyofasthmaticseasonality,confoundersthe analysis adjustment thesex, population),sufficient year,adjusted age,minimumfor*Allin analyses for status, severstatus, e rhi urticariachronicnitis, status

Table S4.2 Analysis of the association between desloratadine (DL) exposure and supraventricular tachycardia (SVT) for the total population and stratified by country$,a Crudemodel adjustedFully model* Country Exposurestatus SVT(N) time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 384 1,233,027 0.99 0.89;1.10 1.03 0.92;1.15 All 0.81 0.61 Not currently NotcurrentlyDL exposed 2,480 7,861,683 1 ref 1 r ef Currently DL exposed CurrentlyDL 100 182,830 0.99 0.81;1.22 0.99 0.81;1.22 Denmark 0.94 0.95 Not currently DL exposed NotcurrentlyDL 928 1,682,316 1 ref 1 r ef Currently DL exposed CurrentlyDL 93 406,709 1.14 0.92;1.42 1.10 0.88;1.38 Finland 0.24 0.41 Not currently DL exposed NotcurrentlyDL 634 3,164,811 1 ref 1 r ef Currently DL exposed CurrentlyDL 191 643,488 0.97 0.83;1.14 1.03 0.88;1.21 Sweden 0.75 0.72 Not currently DL exposed NotcurrentlyDL 918 3,014,557 1 ref 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Years Years2001 a - 200 3 combined are toofewtoconvergence observationsobtain to due *All analyses adjusted for confounders in the minimum sufficient the calendartotal country set: (onlyofasthmaticseasonality,confoundersthe analysis adjustment thesex, population),sufficient year,adjusted age,minimumfor*Allin analyses for status, severstatus, e rhi urticariachronicnitis, status forwasthe individualsadjustmentperformednostratified analyses by Duerhinitis,countrywithtooseverefewto .

05822K PAGE 101

Table S4.3 Analysis of the association between desloratadine (DL) exposure and atrial fibrillation of flutter (A-fib/flu) for the total population and stratified by country$,a Crudemodel adjustedFully model* Country Exposurestatus A - (N)fib/flu time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 1,749 1,225,076 1.05 1.00;1.10 1.07 1.02;1.13 All 0.070 0.008 Not currently DL exposed NotcurrentlyDL 10,629 7,804,139 1 ref 1 r ef Currently DL exposed CurrentlyDL 340 182,168 1.10 0.98;1.23 1.13 1.01; 1.27 Denmark Denmark 0.11 0.037 Not currently DL exposed NotcurrentlyDL 2,850 1,673,548 1 ref 1 r ef Currently DL exposed CurrentlyDL 468 404,243 1.08 0.98;1.19 1.06 0.96;1.17 Finland 0.11 0.27 Not currently DL exposed NotcurrentlyDL 3,357 3,142,213 1 ref 1 r ef Currently DL exposed CurrentlyDL 941 638,665 1.00 0.93;1.07 1.07 1.00;1.15 Sweden 0.90 0.063 Not currently DL exposed NotcurrentlyDL 4,422 2,988,378 1 ref 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Age Age a groups0 – 54and – years19 combinedare yearsand 2001- combined are2003 toconvergenceobservationsobtain to due toofew *All analyses adjusted for confounders in the minimum sufficient thet country set: (onlyofconfoundersthe analysis adjustment the sufficient adjusted age,minimumfor*Allin analyses for calendarotal asthmaticseasonality, sex,population), year, s urticariachronic statustatus, stratified analyses by rhinitis,Duecountrywith fortooseverewasfewtheto individualsperformedadjustmentno .

05822K PAGE 102

Supplementary analysis 5

Table S5.1. Analysis of the association between alternative exposure definitions (desloratadine (DL) and loratadine or non-sedating antihistamines) and first seizure$ Crudemodel adjustedFully model* Exposure Exposuredefinition Exposurestatus (N)seizures First time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently Currentlyexposed 551 1,280,565 1.38 1.26;1.51 <0.001 1.17 1.06;1.28 0.002 DL, loratadine DL, Not currently Notcurrentlyexposed 2,760 8,831,406 1 ref 1 r ef Currently Currentlyexposed 777 1,782,633 1.27 1.18;1.37 <0.001 1.17 1.08;1.27 <0.001 Non - antihistamines sedating Not currently Notcurrentlyexposed 3,805 11,107,294 1 ref 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient status,calendar country, set: asthmaticseasonality,confoundersthesever adjustment sex,sufficient adjusted year, age, minimum*Allin analyses for e rhi nitis, status urticariachronic

Table S5.2. Analysis of the association between alternative exposure definitions (desloratadine (DL) and loratadine or non-sedating antihistamines) and first supraventricular tachycardia (SVT)$ Crudemodel adjustedFully model* Exposure Exposuredefinition Exposurestatus SVT(N) time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently Currentlyexposed 529 1,589,693 0.95 0.87;1.04 0.28 1.01 0.92;1.11 0.79 DL, loratadine DL, Not currently Notcurrentlyexposed 3,849 11,005,788 1 ref 1 r ef Currently Currentlyexposed 922 2,235,920 0.92 0.86;0.99 0.017 1.02 0.95;1.09 0.59 Non - antihistamines sedating Not currently Notcurrentlyexposed 6,226 13,893,526 1 ref 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmaticsever e rhinitis,urchronic ticaria status

Table S5.3. Analysis of the association between alternative exposure definitions (desloratadine (DL) and loratadine or non-sedating antihistamines) and first atrial fibrillation or flutter (A-fib/flu)$, a Crudemodel adjustedFully model* Exposure Exposuredefinition Exposurestatus A - (N)fib/flu time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently Currentlyexposed 2,521 1,577,288 1.09 1.05;1.14 <0.001 1.09 1.04;1.14 <0.001 DL, loratadine DL, Not currently Notcurrentlyexposed 15,963 10,919,605 1 ref 1 r ef Currently Currentlyexposed 4,207 2,215,175 1.03 1.00;1.07 0.043 1.08 1.05;1.12 <0.001 Non - antihistamines sedating Not currently Notcurrentlyexposed 25,264 13,760,401 1 ref 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a a Age groups0 - 5and 4 - toobservationsconvergence obtain dueto toofewcombinedare 19 *All analyses adjusted for confounders in the minimum sufficient status,calendar country, set: asthmaticseasonality,severconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for e rhi nitis, status urticariachronic

05822K PAGE 103

Supplementary analysis 6 Table S6.1 Analysis of the association between desloratadine (DL) exposure and first seizure when using an alternative exposure categorization$, a Crude model adjustedFully model* Exposure Exposurestatus** (N)seizures First time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed following exposed followingredemption CurrentlyprescriptionfirstDL 148 305,352 1.61 1.901.36; 1.22 1.03;1.45 Currently CurrentlyDL exposedredemptionprescriptionfollowing second 66 147,360 1.49 1.16;1.90 1.26 0.98;1.62 Currently DL exposed following exposedprescriptionor following Currently thirdmoreDL 1.12 0.96;1.30 <0.001 0.049 183 542,500 1.08 0.93;1.26 redemption Not currently DL exposed NotcurrentlyDL 1,900 6,307,978 1 ref 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmaticsever e rhi nitis, status urticariachronic a a Year s 2001 - 2002 are combined toofewtoconvergence observationsobtain to due

Table S6.2 Analysis of the association between desloratadine (DL) exposure and SVT when using an alternative exposure categorization$ Crudemodel adjustedFully model* Exposure Exposurestatus SVT(N) time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed following exposed followingredemption CurrentlyprescriptionfirstDL 89 357,295 0.79 0.64;0.98 0.97 0.77;1.20 Currently DL exposed following exposedredemptionprescription following Currently secondDL 38 177,008 0.68 0.49;0.94 0.77 0.56;1.06 Currently DL exposed following exposedprescriptionor following Currently thirdmoreDL 1.17 1.03;1.56 <0.001 0.14 257 698,725 1.11 0.97;1.26 redemption Not currently DL exposed NotcurrentlyDL 2,480 7,861,683 1 ref 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for y, status, asthmaticsever e status rhinitis,urticariachronic

Table S6.3 Analysis of the association between desloratadine (DL) exposure and A-fib/flu when using an alternative exposure categorization$, a Crudemodel adjustedFully model* Exposure Exposure status A - (N)fib/flu time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed following exposed followingredemption CurrentlyprescriptionfirstDL 494 355,370 1.02 0.93;1.12 1.29 1.18;1.42 Currently DL exposed following exposedprescription following Currently secondDL redemption 216 176,047 0.90 0.79;1.03 1.06 0.93;1.22 Currently DL exposed following exposedprescriptionor following Currently thirdmoreDL 1.10 1.03;1.17 0.011 <0.001 1,039 693,659 1.00 0.94;1.07 redemption Not currently DL exposed NotcurrentlyDL 10,629 7,804,139 1 ref 1 r ef IRR= Inciden ce rate ratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Age Age a groups0 – 54and – years combinedare19 convergenceobservationsobtainto due toofewto *All analyses adjusted for confounders in the minimum sufficient set: confoundersthe adjustment sufficient adjusted minimum*Allin analyses for status,calendarcountry, asthmaticseasonality,sever sex, year, age, e rhi nitis, status urticariachronic

05822K PAGE 104

Supplementary analysis 7 Table S7.1 Analysis of the association between desloratadine (DL) exposure and first seizure when non-exposed periods start 52 weeks from the last prescription redemption$ Crudemodel adjustedFully model* Exposure Exposurestatus** (N)seizures First time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 397 1,002,347 1.36 1.22;1.52 1.15 1.02;1.29 <0.001 0.018 Not currently DL exposed NotcurrentlyDL 1,523 5,241,067 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient status,calendar country, set: asthmaticseasonality,confoundersthesever adjustment sex,sufficient adjusted year, age, minimum*Allin analyses for e rhi nitis, status urticariachronic ** Year 2001Year withis2002** combined toofewtoconvergence observationsobtain to due

Table S7.2 Analysis of the association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) when non-exposed periods start 52 weeks from the last prescription redemption$ Crudemodel adjustedFully model* Exposure Exposurestatus SVT(N) time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 386 1,241,262 0.99 0.88;1.10 1.06 0.95;1.18 0.80 0.33 Not currently DL exposed NotcurrentlyDL 2,077 6,585,193 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All *Allanalyses scalendar country, set: asthmaticseasonality,confoundersthe adjustment sex,sufficient year,adjusted age,minimumin for tatus,sever e rhi nitis, status urticariachronic

Table S7.3 Analysis of the association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) when non-exposed periods start 52 weeks from the last prescription redemption $,a Crudemodel adjustedFully model* Exposure Exposurestatus A - (N)fib/flu time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 1,758 1233299 1.01 0.96;1.07 1.08 1.02;1.13 0.64 0.007 Not currently DL exposed NotcurrentlyDL 9,203 6534292 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Age Age a groups0 – 4and 5– years combinedare19 convergenceobservationsobtainto due toofewto *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmaticsever e rhi nitis, status urticariachronic

05822K PAGE 105

Supplementary analysis 8

Table S8.1 Analysis of the association between desloratadine (DL) exposure and first seizure when adjusting for an alternative confounder adjustment set$, a Crudemodel adjustedFully model* Exposure Exposurestatus** First (N)seizures time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 397 995,212 1.32 1.19;1.48 1.14 1.02;1.28 <0.001 0.020 Not currently DL exposed NotcurrentlyDL 1,900 6,307,978 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar sex, set: asthmaseasonality,confoundershypo/hyperthestatus, adjustment sufficientcountry, year,adjusted age, minimum*Allin analysesdiabetes, for - thyroidism, ons, 1allergytypeandinfecti inflammatory disease, rhinitiscombinedsevereand a Year s 2001 - 2002 are combined toofewtoconvergence observationsobtain to due

Table S8.2 Analysis of the association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) when adjusting for an alternative confounder adjustment set$ Crudemodel adjustedFully model* Exposure Exposurestatus SVT(N) time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 384 1,233,027 0.99 0.89;1.10 0.98 0.88;1.09 0.81 0.68 Not currently DL exposed NotcurrentlyDL 2,480 7,861,683 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar sex, set: asthmaseasonality,confoundershypo/hyperthestatus, adjustment sufficientcountry, year,adjusted age, minimum*Allin analysesdiabetes, for - thyroidism, inflammatory disease, infections, infections, inflammatory disease, 1allergytypea ,rhinitiscombinedsevere and treatment ndantihypertensive

Table S8.3 Analysis of the association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) when adjusting for an alternative confounder adjustment set$, a Crudemodel adjustedFully model* Exposure Exposurestatus A - (N)fib/flu time(years)Risk IRR CI95% p - value IRR CI95% p - value Currently DL exposed CurrentlyDL 1,749 1,225,076 1.05 1.00;1.10 1.01 0.96;1.06 0.007 0.74 Not currently DL exposed NotcurrentlyDL 10,629 7,804,139 1 r ef 1 r ef IRR= Inciden ce rate ratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete a Age Age a groups0 – 54and – years combinedare19 convergenceobservationsobtainto due toofewto *All analyses adjusted for confounders in the minimum sufficient calendar sex, set: asthmaseasonality,confoundershypo/hyperthestatus, adjustment sufficientcountry, year,adjusted age, minimum*Allin analysesdiabetes, for - thyroidism, inflammatory disease, infections, infections, inflammatory disease, 1allergytypea ,rhinitiscombinedsevere and treatment ndantihypertensive

05822K PAGE 106

Supplementary analysis 9

Analysis Analysis notperformed

05822K PAGE 107

Supplementary analysis 10 Table S10.1 Analysis of the association between desloratadine (DL) exposure and first seizure in periods without over-the-counter sales$ Crudemodel adjustedFully model* Exposure status ** Exposure status (N)seizures First time(years)Risk IRR CI95% p - value IRR* CI95% p - value Currently DL exposed CurrentlyDL 262 666,315 1.34 1.17;1.53 1.14 0.99;1.31 <0.001 0.06 8 Not currently DL exposed NotcurrentlyDL 1,115 3,785,770 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmatic sever e status. rhinitis,urticariachronic ** Year 2001Year** toofewexcludedto observationsconvergence obtain dueto

Table S10.2 Analysis of the association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) in periods without over-the-counter sales$ Crudemodel adjustedFully model* Exposure Exposurestatus SVT(N) time(years)Risk IRR CI95% p - value IRR* CI95% p - value Currently DL exposed CurrentlyDL 251 812,279 0.96 0.84;1.10 1.01 0.88;1.16 0.58 0.87 Not currently DL exposed NotcurrentlyDL 1,498 4,667,806 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for y, status, asthmaticsever e status. rhinitis,urticariachronic

Table S10.3 Analysis of the association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) in periods without over-the-counter sales$ Crudemodel Fully adjustedmodel* Exposure Exposurestatus A - (N)fib/flu time(years)Risk IRR CI95% p - value IRR* CI95% p - value Currently DL exposed CurrentlyDL 1,122 807,531 1.05 0.98;1.12 1.07 1.00;1.14 0.16 0.036 Not currently DL exposed NotcurrentlyDL 6,151 4,635,386 1 r ef 1 r ef IRR= Inciden ce rateratio intervalCI=confidence ; $ For a complete codes, a list6$ForseeAnnex ofpleasecomplete *All analyses adjusted for confounders in the minimum sufficient calendar country, set: seasonalitconfoundersthe adjustment sex,sufficient year,adjusted age,minimum*Allin analyses for status,y, asthmaticsever e rhinitis,chronic status.urticaria

05822K DESLORATADINE PAGE 108 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

ANNEX 3 Tables with the Post hoc Analysis Results for the Desloratadine Study

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K PAGE 109

Annex 3 Tables with Post hoc Analysis Results for the Desloratadine Study

Overview of the post hoc analyses in the desloratadine study

Post hoc Table 0. Overview of the post hoc analyses in the desloratadine study Post hoc Analyses Description Substudy 1 Baseline characteristics of the incident desloratadine users DL and seizures stratified by alternative age categories Substudy 2 B DL and seizures stratified by country and age Substudy 3 B DL and SVT stratified by country and age DL and A-fib/flu stratified by alternative age categories Substudy 4 B DL and A-fib/flu stratified by country and age A Incidence rates of first recurrent seizure Substudy 5 B DL and first recurrent seizures 1 Alternative definition of DL exposure, comparison of 0-4 weeks versus ≥27 weeks, stratified by country Supplementary analyses 6 Risk after first-ever DL prescription redemption (RX) compared to second, third, fourth (RX) etc. 9 Quantitative bias analysis A-fib/flu= atrial fibrillation or flutter; DDD= defined daily dose; DL=desloratadine; SVT=supraventricular tachycardia

05822K PAGE 110

Post hoc Table 1 Baseline characteristics of the incident desloratadine users at the date of first desloratadine prescription redemption$ Overall Denmark Finland Sweden Overall N (%) 1,840,239 (100) 312,481 (17.0) 626,792 (34.1) 900,966 (49.0) Male 907,152 (49.3) 122,238 (42.6) 363,766 (58.0) 410,148 (45.5) Gender Female 933,087 (50.7) 179,243 (57.4) 263,026 (42.0) 490,818 (54.5) Mean (SD) 31.53 (23.3) 33.07 (23.7) 30.89 (22.9) 31.44 (23.4) Age Median (IQR) 30 (10, 49) 32 (12, 51) 30 (8, 49) 29 (10, 49) 0-4 years 300,787 (16.3) 46,546 (14.9) 120,967 (19.3) 133,274 (14.8) 5-19 years 392,812 (21.4) 63,820 (20.4) 113,448 (18.1) 215,544 (23.9) Age categories, mean (SD) 20-64 years 964,459 (52.4) 165,129 (52.8) 341,228 (54.4) 458,102 (50.9) ≥65 years 182,181 (9.9) 36,986 (11.8) 51,149 (8.2) 94,046 (10.4) 2001 7,277 (0.4) 7,277 (2.3) - - 2002 23,119 (1.3) 21,561 (6.9) 1,558 (0.3) - 2003 56,120 (3.1) 20,956 (6.7) 35,164 (5.6) - 2004 53,901 (2.9) 18,341 (5.9) 35,560 (5.7) - 2005 67,084 (3.7) 20,416 (6.5) 46,668 (7.5) - 2006 180,625 (9.8) 19,233 (6.2) 54,354 (8.7) 107,038 (11.9) 2007 146,180 (7.9) 19,681 (6.3) 52,143 (8.3) 74,356 (8.3) Calendar year 2008 144,639 (7.9) 20,001 (6.4) 52,447 (8.4) 72,191 (8.0) 2009 149,661 (8.1) 22,083 (7.1) 51,762 (8.3) 75,816 (8.4) 2010 148,942 (8.1) 22,940 (7.3) 54,392 (8.7) 71,610 (8.0) 2011 141,827 (7.7) 25,878 (8.3) 48,858 (7.8) 67,091 (7.5) 2012 132,126 (7.2) 23,592 (7.6) 36,691 (5.9) 71,843 (8.0) 2013 155,461 (8.5) 23,246 (7.4) 46,736 (7.5) 85,479 (9.5) 2014 228,970 (12.4) 25,966 (8.3) 62,269 (9.9) 140,735 (15.6) 2015 204,307 (11.1) 21,310 (6.8) 48,190 (7.7) 134,807 (15.0) Winter 262,876 (14.3) 45,921 (14.7) 84,938 (13.6) 132,017 (14.7) Spring 750,350 (40.8) 101,610 (32.5) 276,863 (44.2) 371,877 (41.3) Season Summer 543,995 (29.6) 112,843 (36.1) 177,404 (28.3) 253,748 (28.2) Autumn 283,018 (15.4) 52,107 (16.7) 87,587 (14.0) 143,324 (15.9) Yes 264,931 (14.4) 43,230 (13.8) 82,895 (13.2) 138,806 (15.4) Asthma No 1,575,308 (85.6) 269,251 (86.2) 543,897 (86.8) 762,160 (84.6) Yes 8,102 (0.4) 5,391 (1.7) 276 (0.04) 2,435 (0.3) Severe rhinitis No 1,832,137 (99.6) 307,090 (98.3) 626,516 (99.1) 898,531 (99.7) Yes 43,097 (2.3) 3,729 (1.2) 7,600 (1.2) 31,768 (3.5) Chronic urticaria No 1,797,142 (97.7) 308,752 (98.8) 619,192 (98.8) 869,198 (96.47) IQR: interquartile range; SD=standard deviation $ For a complete list of codes, please see Annex 6

05822K PAGE 111

Post hoc Table 2 (Table 2B2). Analysis of the association between desloratadine (DL) exposure and first seizure using alternative age categories$,a First seizures (N) Risk time (years) Crude model Fully adjusted model* Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 65 42,800 1.20 0.88; 1.61 1.11 0.82; 1.52 0-2 years 0.25 0.50 Not currently DL exposed 120 94,433 1 ref 1 ref Currently DL exposed 35 44,291 1.49 1.03; 2.17 1.38 0.93; 2.03 3-4 years 0.042 0.12 Not currently DL exposed 135 255,063 1 ref 1 ref Currently DL exposed 52 104,106 1.32 0.99; 1.77 1.17 0.86; 1.59 5-10 years 0.071 0.33 Not currently DL exposed 322 851,817 1 ref 1 ref Currently DL exposed 90 145,671 1.50 1.19; 1.89 1.42 1.12; 1.81 11-19 years <0.001 0.005 Not currently DL exposed 371 900,499 1 ref 1 ref Currently DL exposed 115 566,997 1.00 0.82; 1.22 0.88 0.72; 1.08 20-64 years # 0.97 0.22 Not currently DL exposed 728 3,600,928 1 ref 1 ref Currently DL exposed 40 91,346 1.18 0.85; 1.66 1.12 0.80; 1.58 ≥65 years # 0.34 0.51 Not currently DL exposed 224 605,239 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 # Original age categories a Years 2001-2006 are combined due to too few observations to obtain convergence * Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

05822K PAGE 112

Post hoc Table 3 (Table 2B2). Analysis of the association between desloratadine (DL) exposure and first seizure$,a stratified by country and age group Denmark Crude model Fully adjusted model* Age groups Exposure status First seizure (N) Risk time (years) IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 25 134,393 0.96 0.63; 1.44 0.83 0.86 0.57; 1.31 All 0.47 Not currently DL exposed 243 1,247,493 1 ref 1 ref Currently DL exposed ≤10 ------0–4 years # 0.19 Not currently DL exposed 28 62,981 1 ref 1 ref Currently DL exposed ≤10 ------5–19 years # 0.85 Not currently DL exposed 72 276,534 1 ref 1 ref Currently DL exposed ≤10 ------20–64 years # - Not currently DL exposed 107 774,532 1 ref 1 ref Currently DL exposed ≤10 ------≥65 years # - Not currently DL exposed 36 133,447 1 ref 1 ref

Finland Crude model Fully adjusted model* Age groups Exposure status First seizure (N) Risk time (years) IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 103 326,088 1.37 1.11; 1.68 1.31 1.05; 1.63 All 0.005 0.018 Not currently DL exposed 591 2,556,979 1 ref 1 ref Currently DL exposed 26 36,974 1.21 0.78; 1.87 1.16 0.74; 1.84 0–4 years 0.40 0.52 Not currently DL exposed 93 159,793 1 ref 1 ref Currently DL exposed 35 71,570 1.65 1.15; 2.35 1.69 1.16; 2.45 5–19 years 0.010 0.009 Not currently DL exposed 227 764,397 1 ref 1 ref Currently DL exposed 31 191,829 1.10 0.76; 1.61 1.11 0.75; 1.63 20–64 years 0.61 0.61 Not currently DL exposed 208 1,420,102 1 ref 1 ref Currently DL exposed 11 25,715 1.44 0.76; 2.74 1.46 0.76; 2.80 ≥65 years 0.28 0.28 Not currently DL exposed 63 212,688 1 ref 1 ref

05822K PAGE 113

Sweden Crude model Fully adjusted model* Age groups Exposure status First seizure (N) Risk time (years) IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 269 534,731 1.18 1.03; 1.35 1.14 0.99; 1.31 All 0.016 0.065 Not currently DL exposed 1,066 2,503,506 1 ref 1 ref Currently DL exposed 65 38,560 1.59 1.19; 2.14 1.56 1.15; 2.12 0–4 years 0.003 0.005 Not currently DL exposed 134 126,723 1 ref 1 ref Currently DL exposed 98 1,452,238 1.22 0.98; 1.52 1.23 0.98; 1.55 5–19 years 0.087 0.087 Not currently DL exposed 394 711,385 1 ref 1 ref Currently DL exposed 79 299,414 0.90 0.71; 1.14 0.90 0.70; 1.15 20–64 years 0.38 0.38 Not currently DL exposed 413 1,406,294 1 ref 1 ref Currently DL exposed 27 51,519 1.09 0.72; 1.65 1.12 0.73; 1.72 ≥65 years 0.70 0.60 Not currently DL exposed 125 259,105 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence * Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed # ≤10 individuals with seizures among currently DL exposed

05822K PAGE 114

Post hoc Table 4 (Table 4B2). Analysis of the association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) using alternative age categories$,a Crude model Fully adjusted model** Age groups Exposure status First seizures (N) Risk time (years) IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed ≤10 - - - - - 0–4 years # - - Not currently DL exposed ≤10 - 1 ref 1 ref Currently DL exposed ≤10 - - - - - 5–19 years # - - Not currently DL exposed ≤10 - 1 ref 1 ref Currently DL exposed 103 395,618 1.18 0.96; 1.46 1.13 0.91; 1.40 20–44 years 0.12 0.28 Not currently DL exposed 597 2,710,611 1 ref 1 ref Currently DL exposed 509 331,929 1.05 0.96; 1.15 1.03 0.94; 1.14 45–64 year 0.31 0.50 Not currently DL exposed 2,786 1,908,193 1 ref 1 ref Currently DL exposed 1,136 136,785 1.07 1.00; 1.14 1.04 0.98; 1.11 ≥65 years 0.037 0.24 Not currently DL exposed 7,219 929,496 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2006 are combined due to too few observations to obtain convergence * Adjusted for age, sex, country and calendar year **Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed # Too few observations to perform the analysis

05822K PAGE 115

Post hoc 5 (Table 4B2). Analysis of the association between desloratadine (DL) exposure and first A-fib/flu$,a stratified by country and age group Denmark Crude model Fully adjusted model* Age groups Exposure status A-fib/flu (N) Risk time (years) IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 340 182,168 1.10 0.98; 1.23 1.14 1.02; 1.28 All 0.11 0.026 Not currently DL exposed 2,850 1,673,548 1 ref 1 ref Currently DL exposed ≤10 - - - - - 0–4 years # - - Not currently DL exposed ≤10 - 1 ref 1 ref Currently DL exposed ≤10 - - - - - 5–19 years # - - Not currently DL exposed ≤10 - 1 ref 1 ref Currently DL exposed 117 106,328 1,33 1.09; 1.61 1.38 1.13; 1.68 20–64 years 0.006 0.002 Not currently DL exposed 875 1,054,346 1 ref 1 ref Currently DL exposed 222 27,014 1.03 0.89; 1.18 1.05 0.91; 1.20 ≥65 years 0.73 0.52 Not currently DL exposed 1,971 245,895 1 ref 1 ref

Finland Crude model Fully adjusted model* Age groups Exposure status A-fib/flu (N) Risk time (years) IRR 95% CI p-value IRR 95% CI p-value All Currently DL exposed 468 404,243 1.09 0.98; 1.19 0.11 1.06 0.96; 1.17 0.24 Not currently DL exposed 3,357 3,142,213 1 ref 1 ref 0–4 years # Currently DL exposed ≤10 ------Not currently DL exposed ≤10 - 1 ref 1 ref 5–19 years # Currently DL exposed ≤10 ------Not currently DL exposed ≤10 - 1 ref 1 ref 20–64 years Currently DL exposed 168 250,112 1.08 0.92; 1.27 0.37 1.08 0.91; 1.27 0.37 Not currently DL exposed 1,143 1,832,931 1 ref 1 ref ≥65 years Currently DL exposed 300 39,791 1.12 1.00; 1.27 0.061 1.05 0.93; 1.19 0.42 Not currently DL exposed 2,208 329,215 1 ref 1 ref

05822K PAGE 116

Sweden Crude model Fully adjusted model* Age groups Exposure status A-fib/flu (N) Risk time (years) IRR 95% CI p-value IRR 95% CI p-value All Currently DL exposed 941 638,665 1.00 0.93; 1.07 0.90 1.07 1.00; 1.15 0.060 Not currently DL exposed 4,422 2,988,378 1 ref 1 ref 0–4 years # Currently DL exposed ≤10 ------Not currently DL exposed ≤10 - 1 ref 1 ref 5–19 years # Currently DL exposed ≤10 ------Not currently DL exposed ≤10 - 1 ref 1 ref 20–64 years Currently DL exposed 327 371,107 1.12 0.99; 1.26 0.074 1.21 1.07; 1.37 0.003 Not currently DL exposed 1,365 1,731,526 1 ref 1 ref ≥65 years Currently DL exposed 614 69,980 1.02 0.94; 1.12 0.61 1.02 0.93; 1.11 0.66 Not currently DL exposed 3,040 354,386 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2006 are combined due to too few observations to obtain convergence # ≤10 individuals with seizures among currently DL exposed * Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed

05822K PAGE 117

Post hoc Table 6 (Table S1.1) Analysis of the association between desloratadine (DL) exposure and first seizure when using an alternative exposure categorization$, a, comparison of 0-4 weeks versus ≥27 weeks, stratified by country Comparison of exposure status 0-4 weeks versus ≥27 weeks Crude model Fully adjusted model* Age groups Country Exposure status First seizure (N) Risk time (years) IRR 95% CI IRR 95% CI 0-4 weeks 15 55,495 1.40 0.83; 2.36 1.24 0.73; 2.11 Denmark ≥27 weeks 243 1,247,493 1 ref 1 ref 0-4 weeks 40 117,352 1.47 1.07; 2.03 1.35 0.98; 1.87 All Finland ≥27 weeks 591 2,556,979 1 ref 1 ref 0-4 weeks 90 175,835 1.20 0.97; 1.49 1.12 0.90; 1.39 Sweden ≥27 weeks 1,066 2,503,506 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2003 are combined due to too few observations to obtain convergence * Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severe rhinitis, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

05822K PAGE 118

Post hoc Table 7 (Table S6.1) Analysis of the association between desloratadine (DL) exposure and first seizure when using an alternative exposure categorization$, a comparison of first prescription redemption versus not currently DL exposed, stratified by country Comparison of exposure status first prescription redemption versus not currently DL exposed Crude model Fully adjusted model* Country First seizure (N) Risk time (years) IRR 95% CI IRR 95% CI 11 38,418 1.47 0.80; 2.69 1.09 0.58; 2.02 Denmark 243 1,247,493 1 ref 1 ref 36 104,621 1.49 1.06; 2.08 1.24 0.87; 1.76 Finland 591 2,556,979 1 ref 1 ref 101 162,313 1.46 1.19; 1.79 1.24 1.00; 1.54 Sweden 1,066 2,503,506 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2002 are combined due to too few observations to obtain convergence * Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severe rhinitis, chronic urticaria status.

Post hoc Table 8 (Table S6.3) Analysis of the association between desloratadine (DL) exposure and A-fib/flu when using an alternative exposure categorization$, comparison of first prescription redemption versus not currently DL exposed, stratified by country Comparison of exposure status first prescription redemption versus not currently DL exposed Crude model Fully adjusted model* Country* A-fib/flu (N) Risk time (years) IRR 95% CI IRR 95% CI 80 47,094 1.00 0.80; 1.25 1.48 1.18; 1.86 Denmark 2,850 1,673,548 1 ref 1 ref 139 122,267 1.06 0.90; 1.26 1.30 1.09; 1.55 Finland 3,357 3,142,213 1 ref 1 ref 275 186,009 1.00 0.88; 1.13 1.26 1.11; 1.43 Sweden 4,422 2,988,378 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2003 are combined and age groups 0-4 and 5-19 are combined due to too few observations to obtain convergence * Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severe rhinitis, chronic urticaria status.

05822K PAGE 119

Post hoc Table 9.1.1 Confounder status in relation to date of first DL prescription redemption for the confounders: asthma, chronic urticaria and severe rhinitis$ Confounder status in relation to date of first DL Confounder status in relation to date of first DL Number of individuals Number of individuals 5 years before first DL 26 weeks after first DL 5 years before first DL 26 weeks after first DL Yes No Yes No DL population Total population, N=1,840,239 Population age≥5 years, N=1,539,452 Asthma Yes 264,931 118,635 146,296 215,308 92,641 122,667 No 1,575,308 53,986 1,521,352 1,324,144 41,217 1,282,927 Chronic urticaria Yes 43,097 7,223 35,874 33,031 6,046 26,985 No 1797,142 28,910 1,768,232 1,506,421 22,313 1,484,108 Severe rhinitis Yes 8,102 3,421 4,681 8,098 3,418 4,680 No 1,832,137 4,283 1,827,854 1,531,354 4,272 1,527,082 DL population with first seizure as Total population, N=1,561,410 Population age≥5 years, N=1,270,065 outcome (Table 2B) Asthma Yes 224,336 102,594 121,742 176,963 77,812 99,151 No 1,337,074 47,922 1,289,152 1,093,102 35,572 1,057,530 Chronic urticaria Yes 36,529 5,920 30,609 26,855 4,779 22,076 No 1,524,881 24,631 1,500,250 1,243,210 18,266 1,224,944 Severe rhinitis Yes 6,972 2,997 3,975 6,969 2,995 3,974 No 1,554,438 3,789 1,550,649 1,263,096 3,779 1,259,137 DL population with SVT as Total population, N=1,833,337 Population age≥5 years, N=1,532,658 outcome (Table 3B) Asthma Yes 264,011 118,314 145,697 214,419 92,335 122,084 No 1,569,326 53,841 1,515,485 1,318,239 41,107 1,277,132 Chronic urticaria Yes 42,950 7,200 35,750 32,887 6,023 26,864 No 1,790,387 28,790 1,761,597 1,499,771 22,197 1,477,574 Severe rhinitis Yes 8,078 3,408 4,670 8,074 3,405 4,669 No 1,825,259 4,272 1,820,987 1,524,584 4,261 1.520,323 DL population with A-fib/flu as Total population, N=1,822,783 Population age≥5 years, N=1,522,044 outcome (Table 4B) Asthma

05822K PAGE 120

Yes 263,296 118,184 145,112 213,686 92,196 121,490 No 1,559,487 53,791 1,505,696 1,308,358 41,053 1,267,305 Chronic urticaria Yes 42,9890 7,162 35,618 32,715 5,985 26,730 No 1,780,003 28,695 1,751,308 1,489,329 22,099 1,467,230 Severe rhinitis Yes 8,072 3,407 4,665 8,068 3,404 4,664 No 1,814,711 4,273 1,810,438 1,513,976 4,262 1,509,714 DL population with first recurrent Total population, N=2,530 Population age≥5 years, N=1,806 seizure as outcome (Table 5B) Asthma Yes 455 252 203 275 141 134 No 2,075 105 1,970 1,531 59 1,472 Chronic urticaria Yes 76 ≤10 ≤10 37 ≤10 ≤10 No 2,454 ≤10 ≤10 1,769 ≤10 ≤10 Severe rhinitis Yes ≤10 ≤10 ≤10 ≤10 ≤10 ≤10 No ≤10 ≤10 ≤10 ≤10 ≤10 ≤10 $ For a complete list of codes, please see Annex 6

05822K PAGE 121

Post hoc Table 9.1.2 Confounder status in relation to date of first DL prescription redemption for the confounders: asthma, chronic urticaria and severe rhinitis$ Confounder status in relation to date of first DL Confounder status in relation to date of first DL Number of individuals Number of individuals 5 years before first DL 52 weeks after first DL 5 years before first DL 52 weeks after first DL Yes No Yes No DL population Total population, N=1,840,239 Population age≥5 years, N=1,539,452 Asthma Yes 264,931 143,157 121,774 215,308 112,104 103,204 No 1,575,308 73,901 1,501,407 1,324,144 54,851 1,269,293 Chronic urticaria Yes 43,097 8,123 34,974 33,031 6,768 26,263 No 1797,142 33,207 1,763,935 1,506,421 25,700 1,480,721 Severe rhinitis Yes 8,102 4,306 3,796 8,098 4,302 3,796 No 1,832,137 6,599 1,825,538 1,531,354 6,570 1,524,784 DL population with first seizure as Total population, N=1,561,410 Population age≥5 years, N=1,270,065 outcome (Table 2B) Asthma Yes 224,336 123,725 100,611 176,963 94,090 82,873 No 1,337,074 65,402 1,271,672 1,093,102 46,931 1,046,171 Chronic urticaria Yes 36,529 6,692 29,837 26,855 5,381 21,474 No 1,524,881 28,274 1,496,607 1,243,210 21,033 1,222,177 Severe rhinitis Yes 6,972 3,778 3,194 6,969 3,775 3,194 No 1,554,438 5,859 1,548,579 1,263,096 5,831 1,257,265 DL population with SVT as Total population, N=1,833,337 Population age≥5 years, N=1,532,658 outcome (Table 3B) Asthma Yes 264,011 142,757 121,254 214,419 111,724 102,695 No 1,569,326 73,740 1,495,586 1,318,239 54,695 1,263,544 Chronic urticaria Yes 42,950 8,098 34,852 32,887 6,743 26,144 No 1,790,387 33,073 1,757,314 1,499,771 25,570 1,474,201 Severe rhinitis Yes 8,078 4,293 3,786 8,074 4,288 3,786 No 1,825,259 6,581 1,818,678 1,524,584 6,552 1,518,032 DL population with A-fib/flu as Total population, N=1,822,783 Population age≥5 years, N=1,522,044 outcome (Table 4B)

05822K PAGE 122

Asthma Yes 263,296 142,750 120,726 213,686 111,525 102,161 No 1,559,487 73,637 1,485,850 1,308,358 54,588 1,253,770 Chronic urticaria Yes 42,9890 8,059 34,721 32,715 6,704 26,011 No 1,780,003 32,972 1,747,031 1,489,329 25,466 1,463,863 Severe rhinitis Yes 8,072 4,290 3,782 8,068 4,286 3,782 No 1,814,711 6,583 1,808,128 1,513,976 6,554 1,507,422 DL population with first recurrent Total population, N=2,530 Population age≥5 years, N=1,806 seizure as outcome (Table 5B) Asthma Yes 455 299 156 275 166 109 No 2,075 144 1,931 1,531 81 1,450 Chronic urticaria Yes 76 12 64 37 ≤10 ≤10 No 2,454 55 2,399 1,769 ≤10 ≤10 Severe rhinitis Yes ≤10 ≤10 ≤10 ≤10 ≤10 ≤10 No ≤10 ≤10 ≤10 ≤10 ≤10 ≤10 $ For a complete list of codes, please see Annex 6

05822K PAGE 123

Post hoc Table 9.2.1 Alternative confounder status defined using 5 years before and 26 weeks after first DL prescription redemption. Analysis of the association between desloratadine (DL) exposure and first seizure$,a Fully adjusted model* Original confounder definition Alternative confounder definition Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 1.15 1.03; 1.29 1.16 1.04; 1.30 All 0.013 0.012 Not currently DL exposed 1 Ref 1 ref Currently DL exposed 1.47 1.16; 1.87 1.50 1.17; 1.91 0–4 years 0.002 0.002 Not currently DL exposed 1 Ref 1 ref Currently DL exposed 1.32 1.09; 1.59 1.32 1.09; 1.59 5–19 years 0.006 0.006 Not currently DL exposed 1 Ref 1 ref Currently DL exposed 0.88 0.72; 1.08 0.89 0.72; 1.08 20–64 years 0.22 0.23 Not currently DL exposed 1 Ref 1 ref Currently DL exposed 1.12 0.80; 1.58 1.13 0.80; 1.59 ≥65 years 0.51 0.49 Not currently DL exposed 1 Ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence *Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

Post hoc Table 9.2.2 Alternative confounder status defined using 5 years before and 52 weeks after first DL prescription redemption. Analysis of the association between desloratadine (DL) exposure and first seizure$,a Fully adjusted model* Original confounder definition Alternative confounder definition Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 1.15 1.03; 1.29 1.16 1.03; 1.30 All 0.013 0.012 Not currently DL exposed 1 Ref 1 ref Currently DL exposed 1.47 1.16; 1.87 1.49 1.17; 1.90 0–4 years 0.002 0.002 Not currently DL exposed 1 Ref 1 ref Currently DL exposed 1.32 1.09; 1.59 1.32 1.09; 1.59 5–19 years 0.006 0.006 Not currently DL exposed 1 Ref 1 ref Currently DL exposed 0.88 0.72; 1.08 0.88 0.72; 1.08 20–64 years 0.22 0.22 Not currently DL exposed 1 Ref 1 ref Currently DL exposed 1.12 0.80; 1.58 1.13 0.80; 1.60 ≥65 years 0.51 0.49 Not currently DL exposed 1 Ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence * Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

05822K PAGE 124

Post hoc Table 9.3.1 Alternative confounder status defined using 5 years before and 26 weeks after first DL prescription redemption. Analysis of the association between desloratadine (DL) exposure and SVT$,a Fully adjusted model* Original confounder definition Alternative confounder definition Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 1.03 0.92; 1.15 1.04 0.93; 1.16 All 0.61 0.52 Not currently DL exposed 1 ref 1 ref Currently DL exposed - - - - 0–4 years - - Not currently DL exposed 1 ref 1 ref Currently DL exposed 1.02 0.67; 1.55 1.04 0.68; 1.58 5–19 years 0.92 0.86 Not currently DL exposed 1 ref 1 ref Currently DL exposed 1.07 0.93; 1.23 1.08 0.94; 1.24 20–64 years 0.38 0.30 Not currently DL exposed 1 ref 1 ref Currently DL exposed 0.95 0.78; 1.15 0.95 0.78; 1.15 ≥65 years 0.58 0.58 Not currently DL exposed 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence *Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

Post hoc Table 9.3.2 Alternative confounder status defined using 5 years before and 52 weeks after first DL prescription redemption. Analysis of the association between desloratadine (DL) exposure and SVT $,a Fully adjusted model* Original confounder definition Alternative confounder definition Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 1.03 0.92; 1.15 1.04 0.93; 1.16 All 0.61 0.53 Not currently DL exposed 1 ref 1 ref Currently DL exposed - - - - 0–4 years - - Not currently DL exposed 1 ref 1 ref Currently DL exposed 1.02 0.67; 1.55 1.04 0.68; 1.57 5–19 years 0.92 0.87 Not currently DL exposed 1 ref 1 ref Currently DL exposed 1.07 0.93; 1.23 1.08 0.94; 1.24 20–64 years 0.38 0.31 Not currently DL exposed 1 ref 1 ref Currently DL exposed 0.95 0.78; 1.15 0.94 0.77; 1.15 ≥65 years 0.58 0.57 Not currently DL exposed 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence *Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

05822K PAGE 125

Post hoc Table 9.4.1. Alternative confounder status defined using 5 years before and 26 weeks after first DL prescription redemption. Analysis of the association between desloratadine (DL) exposure and A-fib/flu$,a Fully adjusted model* Original confounder definition Alternative confounder definition Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 1.08 1.02; 1.13 1.08 1.03; 1.14 All 0.005 0.004 Not currently DL exposed 1 ref 1 ref Currently DL exposed - - - - 0–4 years - - Not currently DL exposed 1 ref 1 ref Currently DL exposed - - - - 5–19 years - - Not currently DL exposed 1 ref 1 ref Currently DL exposed 1.16 1.07; 1.27 1.17 1.07; 1.28 20–64 years <0.001 <0.001 Not currently DL exposed 1 ref 1 ref Currently DL exposed 1.04 0.98; 1.11 1.04 0.97; 1.11 ≥65 years 0.24 0.25 Not currently DL exposed 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence *Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

Post hoc Table 9.4.2. Alternative confounder status defined using 5 years before and 52 weeks after first DL prescription redemption. Analysis of the association between desloratadine (DL) exposure and A-fib/flu $,a Fully adjusted model* Original confounder definition Alternative confounder definition Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 1.08 1.02; 1.13 1.08 1.02; 1.14 All 0.005 0.004 Not currently DL exposed 1 ref 1 ref Currently DL exposed - - - - 0–4 years - - Not currently DL exposed 1 ref 1 ref Currently DL exposed - - - - 5–19 years - - Not currently DL exposed 1 ref 1 ref Currently DL exposed 1.16 1.07; 1.27 1.18 1.08; 1.28 20–64 years <0.001 <0.001 Not currently DL exposed 1 ref 1 ref Currently DL exposed 1.04 0.98; 1.11 1.04 0.97; 1.10 ≥65 years 0.24 0.27 Not currently DL exposed 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence * Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

05822K PAGE 126

Post hoc Table 9.5.1 Alternative confounder status defined using 5 years before and 26 weeks after first DL prescription redemption. Analysis of the association between desloratadine (DL) exposure and first recurrent seizure$,a Fully adjusted model* Original confounder definition Alternative confounder definition Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 1.26 0.94; 1.67 1.24 0.93; 1.65 All 0.13 0.15 Not currently DL exposed 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence *Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

Post hoc Table 9.5.2 Alternative confounder status defined using 5 years before and 52 weeks after first DL prescription redemption. Analysis of the association between desloratadine (DL) exposure and first recurrent seizure$,a Fully adjusted model* Original confounder definition Alternative confounder definition Age groups Exposure status IRR 95% CI p-value IRR 95% CI p-value Currently DL exposed 1.26 0.94; 1.67 1.25 0.94; 1.66 All 0.13 0.14 Not currently DL exposed 1 ref 1 ref IRR= Incidence rate ratio; CI=confidence interval $ For a complete list of codes, please see Annex 6 a Years 2001–2004 are combined due to too few observations to obtain convergence *Adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, chronic urticaria status. Due to too few individuals with severe rhinitis, no adjustment was performed.

05822K DESLORATADINE PAGE 127 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

ANNEX 4 Study Protocol

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K P A GE 128

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 1 E pi d e mi ol o g y N o.: E P 0 7044.002

PASS I NF OR MATI O N

Titl e Association bet ween use of desloratadine and risk of seizures , s u prave ntric ular t ac h yc ar di a, a n d atri al fi brill ati o n or fl utt er : A N or dic register -based study Pr ot oc ol v ersi o n i d e ntifi er 1. 1 D at e of last v ersi o n of p r ot oc ol 1 6 February 2016 E U P A S register nu mber Study n ot registere d Acti v e substance D esl or at a di n e, A T C code R06AX27 ; Phar macotherapeutic gr o u p: A nti hist a mi n es – H1 antagonist M e dici n al pr o d uct AERI US, AZ O MYR, and NE OCLARITY N Pr o d uct reference E U/1/00/160, AERI US E U/1/00/157, AZ O MYR E U/1/00/161, NE OCLARITY N Procedure nu mber E MEA/H/C/000313 / MEA/065 , A E RI U S E MEA/H/C/000310 / MEA/065 , AZ O MYR E MEA/H/C/000314 / MEA/065 , NE OCLARITY N Marketi ng auth orizati on Merck Sharp & Doh me Ltd. h ol d er ( MAH) Hertford Road, Hoddesdon Hertfordshire E N11 9B U United Kingdo m Joint P ASS N o Researc h q u esti o n a n d To describe the use of desloratadine in the general population ; o bj ecti v es t o describe the incidence rates of first s eiz ur e , s u prave ntric ular tachycardia , atri al fi brill ati o n or fl utt er , a n d first rec urre nt seiz ure ; and to exa mine the associations bet ween desloratadine exposure and risk of first s eiz ur e , s u prave ntric ular tachycardia , atri al fi brill ati o n or fl utt er , a n d first rec urre nt seiz ure . C o u ntries of st u dy Den mark, Finland, Nor way , a n d S weden A ut h or Annette Kj ær Ersbøll P P D University of Southern Den mark Øster Fari magsgade 5 A D K -1353 Copenhagen K , Den mark P h o n e: P P D E -m ail: P P D

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K P A GE 129

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 2 E pi d e mi ol o g y N o.: E P 0 7044.002

MARKETING AUTH ORIZATI ON H OLDER

Marketi ng auth orizati on Merck Sharp & Doh me Ltd. h ol d er Hertford Road, Hoddesdon Hertfordshire E N11 9B U United Kingdo m M A H contact person P P D Gl o bal Reg ulat ory Affairs – R AI E ME A Regional Center Merck Sharp & Doh me (Europe), Inc. Lynx Binnenhof 5 1200 Brussel s, B el gi u m P h o n e: P P D F a x: P P D E- m ail: P P D

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K PAGE 130

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 3 Epidemiology No.: EP07044.002

1 TABLE OF CONTENTS

1 TABLE OF CONTENTS ...... 3 List of Tables ...... 6 List of Figures...... 7 2 LIST OF ABBREVIATIONS ...... 8 3 RESPONSIBLE PARTIES ...... 9 4 ABSTRACT...... 10 5 AMENDMENTS AND UPDATES...... 13 6 MILESTONES ...... 14 7 RATIONALE AND BACKGROUND ...... 15 8 RESEARCH QUESTION AND OBJECTIVES ...... 16 9 RESEARCH METHODS...... 17 9.1 Study design...... 17 9.2 Setting ...... 19 9.2.1 Inclusion criteria ...... 19 9.2.2 Exclusion criteria ...... 20 9.3 Variables...... 22 9.3.1 Exposure ...... 22 9.3.2 Outcomes ...... 24 9.3.3 Covariates ...... 26 9.3.4 Other variables ...... 29 9.4 Data sources...... 29 9.4.1 Study procedures...... 30 9.5 Study size ...... 30 9.5.1 Fixed sample size...... 31 9.5.2 Proportion of current and non-current exposure time among DL users for the purpose of calculation of minimum detectable IRR ...... 31 9.5.3 Background event rate ...... 32 9.5.4 Number of years in the study...... 32 9.5.5 Statistical parameters and assumptions...... 32 9.5.6 Example of calculating the minimum detectable IRR ...... 33

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 131

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 4 Epidemiology No.: EP07044.002

9.6 Data management ...... 34 9.7 Data analysis...... 35 9.7.1 Substudy 1: Descriptive analysis of exposure of DL...... 36 9.7.2 Substudy 2: First seizure...... 36 9.7.3 Substudy 3: SVT ...... 37 9.7.4 Substudy 4: Atrial fibrillation/flutter ...... 37 9.7.5 Substudy 5: First recurrent seizure ...... 38 9.7.6 Supplementary analyses...... 39 9.8 Quality control ...... 40 9.9 Limitations of the research methods...... 41 9.10 Other aspects ...... 44 10 PROTECTION OF HUMAN SUBJECTS ...... 45 11 MANAGEMENT AND REPORTING OF ADVERSE EVENTS/ADVERSE REACTIONS...... 46 11.1 Definition of adverse event...... 46 11.2 Definition of serious adverse event...... 46 11.3 Other relevant safety information...... 46 11.4 Causality assessment...... 47 11.5 Adverse event reporting ...... 47 12 PLANS FOR DISSEMINATING AND COMMUNICATING STUDY RESULTS ...... 48 13 REFERENCES...... 49 ANNEX 1. LIST OF STAND-ALONE DOCUMENTS...... 51 ANNEX 2. ENCEPP CHECKLIST FOR STUDY PROTOCOLS ...... 52 ANNEX 3. ABOUT DESLORATADINE ...... 59 ANNEX 4. ICD-10 CODES AND ATC CODES...... 60 ANNEX 5. MINIMUM DETECTABLE INCIDENCE RATE RATIO CALCULATION...... 61 ANNEX 6. DIRECTED ACYCLIC GRAPHS (DAGs) FOR THE STUDY ON DESLORATADINE AND RISK OF FIRST SEIZURE, FIRST RECURRENT SEIZURE, ATRIAL FIBRILLATION OR FLUTTER, AND SUPRAVENTRICULAR TACHYCARDIA...... 62 ANNEX 7. ADMINISTRATIVE AND REGULATORY DETAILS ...... 82 ANNEX 8. PRAC ENDORSEMENT...... 86 EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 132

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 5 Epidemiology No.: EP07044.002

ANNEX 9. QUALIFIED PERSON FOR PHARMACOVIGILANCE (QPPV) ...... 87 ANNEX 10. SIGNATURE PAGE ...... 88

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 133

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 6 Epidemiology No.: EP07044.002

LIST OF TABLES

Table 1 Overview of Substudies...... 17 Table 2 Overview of the Outcome Variables in the Study...... 26 Table 3 Information on whether potential confounding factors are included in the DAGs and the minimum sufficient adjustment sets, which will be used for confounder adjustment in the association studies...... 27 Table 4 Definition of confounders included in the minimum sufficient adjustment sets that will be used for confounder adjustment in the association studies...... 28 Table 5 Overview of national health registers in the Nordic countries of relevance for the present study...... 30 Table 6 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rate (IR) of seizures is based on Hauser & Beghi (2008) [2]. A 1-sided test, significance level of 5% and power of 80% have been used...... 33 Table 7 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rate (IR) of SVT is based on Orejarena et al. (1998) [4]. A 1-sided test, significance level of 5% and power of 80% have been used...... 33 Table 8 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rates (IR) for atrial flutter were not available; therefore, we base the calculations only on IR of A-fib obtained from Wilke et al. (2013) [5]. A 1-sided test, significance level of 5% and power of 80% have been used...... 34 Table 3.1 Overview of DL formulations in the Nordic countries. Dates of approval are identical for all EU countries...... 59 Table 6.1 Discussion of potential confounders of the association between DL use and first seizure and first recurrent seizure ...... 64 Table 6.2 Minimum sufficient adjustment sets for the DAG developed (Figure 6.1)...... 70 Table 6.3 Discussion of potential confounders of the association between DL use and A- fib/flu...... 72 Table 6.4 Minimum sufficient adjustment sets for the DAG developed (Figure 6.2)...... 77 Table 6.5 Minimum sufficient adjustment sets for the DAG developed (Figure 6.3)...... 80

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 134

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 7 Epidemiology No.: EP07044.002

LIST OF FIGURES

Figure 1 Example of current exposed periods versus unexposed periods for study subjects included in the association analyses (Substudies 2B, 3B, 4B, and 5B)...... 22 Figure 6.1 DAG for the association between DL use and first seizure and first recurrent seizure...... 69 Figure 6.2 DAG for the association between DL use and A-fib/flu...... 76 Figure 6.3 DAG for the association between DL use and SVT...... 79

EMA/623947/2012 22-Feb-2016

04CN88

05822K P A GE 135

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 8 E pi d e mi ol o g y N o.: E P 0 7044.002

2 LIST OF ABBREVIATI ONS

A E A dverse e v e nt A -fi b Atri al fi brill ati o n A -fl u Atri al fl utt er A -fi b/fl u Atri al fi brill ati o n or atri al fl utt er

P P D

A T C Anato mical Therapeutic Che mical Classification Syste m CH MP Co m mittee for Medicinal Products for Hu man Use C OP D Chronic obstructive pul monary d isease D A G Directe d acyclic gra p h DL Desloratadine DS UR Develop ment safety update report E M A European Medicines Agency E U European Union I C D I nter nati o nal Classificati o n of Diseases I R Incidence rate I R R Incidence rate rati o M A H Mark eti ng a ut h orizati o n h ol d er M S D Merck Sharp & Doh me, a s u bsi di ar y of M erc k & C o. , I nc. P P D

NS AE N o n -serious adverse event PI C P ers o n al id e ntific ati o n c o d e PR AC Phar macovigilance Risk Assess ment Co m mittee PS UR Periodic safety update report O T C Ov er -t h e -counter P P V P ositive pre dictive val ue Q B A Q ua ntitative Bias A nalysis S A E Serious adverse eve nt S V T Supraventricular tachycardia

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K P A GE 136

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 9 E pi d e mi ol o g y N o.: E P 0 7044.002

3 RESP ONSIBLE PARTIES

Pri nci p al i nvestigat or Annette Kj ær Ersb øll, Professor P P D University of Southern Den mark Øster Fari magsgade 5 A D K -1353 Copenhagen K, Den mark P h o n e: P P D E -m ail: P P D

Coordinating investigato rs Den mark : Annette Kj ær Ersb øll, Pr ofess or, P P D , U ni v ersit y of Southern Den mark , Den mark Fi nl a n d : Eero Pukkala, Professor, School of Health Sciences, University of Ta mpere , Fi nl a n d Nor way : Kristian Bolin, Professor, Depart ment of Econo mics, Lund U niv ersity , S weden S weden : Kristian Bolin, Professor, Depart ment of Econo mics, Lund U niv ersity , S weden

S p o ns or co ntact P P D Merck Sharp & Doh me, C or p., a s u bsi di ar y of M erc k & C o., I nc. U G1 D -6 0 P O Box 1000 North Wales, P A 19454 -1099 U nite d States P h o n e: P P D E -m ail: P P D Vendor/collaborator P P D E waldsgade 3 D K -2200 Copenhagen, Den mark P h o n e: P P D w w w.appliedecono mics.dk Expert consultants P P D

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K P A GE 137

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 1 0 E pi d e mi ol o g y N o.: E P 0 7044.002

4 ABSTRACT

Titl e Ass oci ati o n bet ween us e of desl orata di ne a n d ris k of seiz ures , s u prave ntric ular tac hycar dia, a n d atri al fi brill ati o n or fl utt er : A N or dic register -based study Pr ot oc ol V ersi o n 1. 1 D at e 1 6 February 2016 A ut h or Annette Kj ær Ersbøll, Professor, P P D University of Southern D e n m ar k, Copenhagen, Den mark R ati o n al e and background A p ost -auth orizati on safety study is needed to assess the potential risk of desloratadine exposure on seizures , s u prave ntric ular t ac h yc ar di a, a n d atri al fi brill ati o n or fl utt er . Researc h question and T o e x pl or e t h e us e of desl orata di ne i n t h e g e n er al population o bj ecti v e s (Substudy 1) ; t o d escri b e t h e incidence r at e of first s eiz ur e ( S u bst u dy 2 A ); t o e x a mi n e t h e ass ociati o n s bet ween desloratadine exposure a n d ris k of first s eiz ur e ( S ubstud y 2 B); t o d escri b e t h e incidence r at e of s u prave ntric ular tac hycar dia ( S u bst u dy 3 A); t o e x a mi n e t h e ass oci ati o n bet ween desloratadine exposure a n d s u prave ntric ular tachycardia ( S u bst u dy 3 B); t o d escri b e t h e incidence r at e of atri al fi brill ati o n or fl utt er ( S u bst u dy 4 A) ; t o e x a mi n e t h e ass ociati o n s bet ween desloratadine exposure a n d atri al fi brill ati o n or fl utt er ( S u bst u dy 4 B) ; t o d escri b e t h e incidence r at e of first rec urre nt s eiz ur e (Substudy 5 A); a n d t o e x a mi n e t h e ass oci ati o n bet ween desl orata di ne ex p os ur e a n d fir st rec urre nt s eiz ur e (Substudy 5B) . St u d y d esi g n Observational , nation wide , r egist er -b as e d st u dy usi n g pers o n - s p ecific li n k a g e of d at a fr o m t h e n ati o n al p o p ul ati o n r egist ers fr o m D e n m ar k, Fi nl a n d, N or w a y, a n d S weden (“ N or dic c o u ntries”) i ncl u di n g all i n di vi d u als w h o redee med a pr escri pti o n of desl orata di ne a n d all i n di vi d u als wit h a registere d di a g n osis of s eiz ur e , s u prave ntric ular tac hycar dia, or atri al fi brill ati o n or fl utt er . Population T h e p o p ul ati o n c o nsists of a c o h ort of desl orata di ne us ers , a c o h ort of i n di vi d u als wit h seiz ures , a c o h ort of i n di vi d u als wit h s u prave ntric ular tachycardia , a n d a c o h ort of i n di vi d u als wit h atri al fi brill ati o n or fl utt er . T h e g e n er al p o p ul ati o n of t h e f o ur N or dic c o u ntries will b e u s e d t o d eri v e esti mates of t h e ris k ti m e b y a g e, year, and country .

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K PAGE 138

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 11 Epidemiology No.: EP07044.002

Variables The exposure variable of interest is desloratadine for which current use (i.e., exposed period) will be defined as the period after each redeemed prescription equal to the number of days’ supply plus a 4 week grace period to account for intermittent use and a possible wash-out effect. To minimize exposure misclassification, we propose to use person time at least 26 weeks beyond the dispensing date of the last prescription as the “unexposed” reference period.

The outcome variables are first seizure, first supraventricular tachycardia diagnosis, first atrial fibrillation or flutter diagnosis, and first recurrent seizure. Directed acyclic graphs (DAGs) were developed to identify the minimum sufficient adjustment set of confounders to include in the association analysis of each outcome. Data sources Data will be obtained from nationwide population registers, including the national patient registers, the civil registration systems, and the prescription registers. Study size The sample size is fixed, as it will consist of all individuals in four Nordic countries (Denmark, Finland, Norway, and Sweden) who have redeemed at least one prescription for desloratadine or who have received a diagnosis of seizure, supraventricular tachycardia, or atrial fibrillation or flutter. Data analysis A descriptive analysis of desloratadine use in the general population will be performed. Furthermore, the incidence rates of seizure, supraventricular tachycardia, atrial fibrillation or atrial flutter, and first recurrent seizure will be calculated. Among persons ever dispensed desloratadine, the associations between desloratadine exposure and first seizure, supraventricular tachycardia, atrial fibrillation or flutter, and first recurrent seizure will be evaluated using Poisson regression of incidence rates accounting for confounding factors. Additional supplementary analyses will be performed.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 139

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 12 Epidemiology No.: EP07044.002

Milestones Approval of protocol by Committee for Medicinal Products for Human Use (CHMP)/Pharmacovigilance Risk Assessment Committee (PRAC): TBD Submission of application for permission to access national registries: 2 weeks after approval Start of data collection: 4 months after application submitted End of data collection: 12 months after start of data collection Data analysis: 6 months after end of data collection Submission of final report of study results (as a Type II variation): 6 months after completion of data analysis Preparation of manuscripts for scientific publication: First draft 2+ months after submission of the final study report The study milestones are dependent on the timelines for the applications for data access and the timelines for data collection from the registries. We have based the milestones on our best estimates of the time required; however, these timelines are not within control of the responsible parties and recently, the application and data collection processes have been taking longer due to new administrative procedures and requirements that are applicable for some studies many of which are due to additional requirements regarding data privacy. In particular, access to data in Finland requires sequential applications to the various individual registries, each of which has recently taken up to 12 months [Ref. 5.4: 046J05]. The current expectation is that the study report will be finalized 29 months after endorsement of protocol, while the MAH previously worked with an 18 months’ period following protocol endorsement as included in the current version (number 1.1) of the EU Risk Management Plan. If there are further delays due to the application and data collection processes which affect the study timeline or the availability of data in one of the countries, we will inform the PRAC accordingly with a request for extension of the timelines.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 140

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 13 Epidemiology No.: EP07044.002

5 AMENDMENTS AND UPDATES

Amendment or Section of Amendment or Date Reason Update no. Study Protocol Update N/A N/A N/A N/A N/A

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 141

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 14 Epidemiology No.: EP07044.002

6 MILESTONES

Milestone Planned Date Approval of protocol by TBD CHMP/PRAC Submission of application for 2 weeks after approval permission to access national registries Start of data collection 4 months after application submitted End of data collection 12 months after start of data collection Data analysis 6 months after end of data collection Submission of final report of study 6 months after completion of data analysis results (as a Type II variation) Preparation of manuscripts for First draft 2+ months after submission of the final study report scientific publication

The study milestones are dependent on the timelines for the applications for data access and the timelines for data collection from the registries. We have based the milestones on our best estimates of the time required; however, these timelines are not within control of the responsible parties and recently, the application and data collection processes have been taking longer due to new administrative procedures and requirements that are applicable for some studies many of which are due to additional requirements regarding data privacy. In particular, access to data in Finland requires sequential applications to the various individual registries, each of which has recently taken up to 12 months [Ref. 5.4: 046J05]. The current expectation is that the study report will be finalized 29 months after endorsement of protocol, while the MAH previously worked with a 18 months’ period following protocol endorsement as included in the current version (number 1.1) of the EU Risk Management Plan. If there are further delays due to the application and data collection processes which affect the study timeline or the availability of data in one of the countries, we will inform the PRAC accordingly with a request for extension of the timelines.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 142

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 15 Epidemiology No.: EP07044.002

7 RATIONALE AND BACKGROUND

The European Medicines Agency (EMA) has requested that Merck Sharp and Dohme (MSD) consider options for a post-authorization safety study (PASS; Category 3 PV activity) to investigate whether there is an association between desloratadine (DL) use and seizures, supraventricular tachycardia, and atrial fibrillation or flutter in the general population. Since market authorization, there have been a small number of adverse event reports of seizures and supraventricular arrhythmias in patients taking DL, but the case reports do not permit evaluation of the association. Denmark, Finland, Norway, and Sweden (hereafter also referred to as ‘Nordic countries’) offer unique opportunities for such a study by means of the existence of centralized registration of activities in the healthcare sector, covering complete populations over many years.

DL is a prescription oral antihistamine approved in the European Union (EU) for the relief of symptoms associated with allergic rhinitis and urticaria in both adults and children (Annex 3). Data from the Danish National Prescription Register (www.medstat.dk/en; accessed 28 May 2014) suggest that about 270,000 Danish citizens (about 5% of the population) purchase antihistamines with a prescription at least once annually. This figure has been rather stable since the year 2000. The number of Danish citizens who purchase DL by a prescription at least once annually has increased from 7,317 in 2001 to 51,564 in 2012. The number of Danish children aged 0–4 years who purchased DL with a prescription at least once annually has increased from 8 in 2001 to 6,815 in 2012. The number of Danish children aged 5–19 years who purchased DL with a prescription at least once annually has increased from 935 in 2001 to 11,447 in 2012. In 2012, children aged 0–4 years accounted for 37% of all children (0–19 years) who purchased DL at least once. The background incidence of unprovoked seizures is rare; approximately 0.4-0.6 per 1,000 population per year [Ref. 5.4: 00W4CY]. Patients with epilepsy experience seizures; however, not all individuals with seizures have epilepsy and not all seizures evolve into epilepsy [Ref. 5.4: 00W4CY]. A previous study described clinical observations of seizures induced by DL in four children [Ref. 5.4: 045VR5].

Supraventricular arrhythmias include supraventricular tachycardia (SVT), atrial fibrillation (A-fib), and atrial flutter (A-flu). Through review of cardiac events reported since marketing authorization, the MAH determined that SVT and A-fib constituted a potential risk and that the reported cases were distributed across all age groups, not just children. In the present study, new onset SVT and a composite of new onset atrial fibrillation or flutter (A-fib/flu) will be examined as two separate outcomes. The incidence rate (IR) of SVT is reported to be 13/100,000 person-years for persons aged 19 years or younger, 27/100,000 person-years for persons aged 20–64 years, and 122/100,000 person-years for persons aged 65 years or older [Ref. 5.4: 03XNM4]. Wilke et al. (2013) [Ref. 5.4: 03XNL7] reported the incidence of A-fib to increase markedly by age with 0.0016 cases/1,000 person-years in children (<15 years) and approximately 30 cases/1,000 person-years for persons aged 80 years or older. No data on the epidemiology of A-flu were available.

The present protocol proposes a study using registers from the four Nordic countries to describe the association between DL use and seizures, SVT, and A-fib/flu.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 143

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 16 Epidemiology No.: EP07044.002

8 RESEARCH QUESTION AND OBJECTIVES

The research question is to examine the associations between DL use and seizures, SVT, and A- fib/flu in the general population.

Primary hypotheses (stated as null-hypotheses)

 There is no association between current DL use and first seizure.  There is no association between current DL use and SVT.  There is no association between current DL use and A-fib/flu.

Primary objectives

 Describe the use of DL during the study period in the general population overall and stratified by country, age, gender, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status.  Describe the IRs of the following outcomes in the general population:  Incident diagnosis of seizure;  Incident diagnosis of SVT; and  Incident diagnosis of A-fib/flu.  Compare the risk of incident seizure among persons while currently exposed to DL to the risk among the same persons while unexposed to DL for the total DL population and stratified by age, while adjusting for relevant confounding factors.  Compare the risk of incident SVT among persons while currently exposed to DL to the risk among the same persons while unexposed to DL for the total DL population and stratified by age, while adjusting for relevant confounding factors.  Compare the risk of incident A-fib/flu among persons while currently exposed to DL to the risk among the same persons while unexposed to DL for the total DL population and stratified by age, while adjusting for relevant confounding factors.

Secondary hypothesis (stated as a null-hypothesis)

o There is no association between current DL use and first recurrent seizure. Secondary objectives

o Among persons who had a first seizure, describe the IR of first recurrent seizure.

o Among persons who had a first seizure, compare the risk of first recurrent seizure while currently exposed to DL to the risk among the same persons while unexposed to DL, while adjusting for relevant confounding factors.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 144

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 17 Epidemiology No.: EP07044.002

9 RESEARCH METHODS

9.1 Study design

The proposed study will be an observational (non-experimental), nationwide, register-based study using data from the Nordic national population registers. The associations between current DL use and seizures, current DL use and SVT, current DL use and A-fib/flu, and current DL use and first recurrent seizure will be assessed in analyses using person-specific linkage of data (Table 1). The cohort of all DL users will be used for describing DL use (Substudy 1). The cohorts of individuals who have had first seizure, SVT, or A-fib/flu will be used to examine the IRs of first seizure, SVT, and A- fib/flu in the general population (Substudies 2A, 3A, and 4A, respectively). A cohort study design among all DL users (i.e. a risk interval design including the DL-only cohort [Ref. 5.4: 03QPDQ]) will be used for the association between current DL exposure and first seizure, SVT, and A-fib/flu (Substudies 2B, 3B, and 4B, respectively). The cohort of individuals who have had first recurrent seizure will be used for estimating the IR of first recurrent seizure (Substudy 5A). A cohort study design among all DL users who have experienced a first seizure will be used for the association between current DL exposure and first recurrent seizure (Substudy 5B).

Table 1 Overview of Substudies.

Pharmacoepidemiological studies typically use dispensed days’ supply as a surrogate for current drug exposure, assuming they are used every day. However, in contrast to medications for many chronic diseases, antihistamines may be used intermittently, as needed for symptoms. For the main analyses in the present study of the association between current DL use and outcomes of interest (Substudies 2B, 3B, 4B, and 5B), person time exposed to DL will be determined from dispensing records. “Current use” (i.e., “exposed” period) will be defined for each prescription as the sum of days’ supply plus a 4 week grace period to account for intermittent use and a possible wash-out effect. If a new DL prescription redemption occurs during an exposed period (either during the period equal to the sum of the days’ supply or the 4 week grace period), the exposure period extends from that date with a period equal to the sum of days’ supply in the newly redeemed prescription plus a 4 week grace period. Note: Days’ supply will be calculated from the quantity of tablets or amount of solution dispensed and the standard daily dose based on the age of the patient (i.e., 6-11 months: 1 mg/day; 12 months-5 years: 1.25 mg/day; 6-11 years: 2.5 mg/day; ≥12 years: 5 mg/day). Because the drug is used “as needed”, it is quite possible that there may still be exposed days in the period after the latest exposed period (i.e., after the exposed period that includes the EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 145

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 18 Epidemiology No.: EP07044.002

sum of the days’ supply plus a 4 week grace period). To minimize exposure misclassification, we propose to use person time at least 26 weeks beyond the dispensing date of the prior prescription as the “unexposed” reference period. The “unexposed” period is actually a period with remote exposure to DL because dispensing of DL is a condition of entering the study population. We will refer to the “remote exposure” period as “unexposed” (For more details, see section 9.3.1 and Figure 1). The time between the exposed period and the unexposed period is considered neither exposed nor unexposed.

The study is divided into six parts described below (For more details, see section 9.7).

1. Substudy 1

A descriptive analysis of DL use in the general population. To describe DL use in the general population, we will identify both prevalent and incident users of DL and describe the distribution of the number of redeemed DL prescriptions for the total population and stratified by country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status.

2. Substudies 2A and 2B

A descriptive analysis of the IR of first seizure overall in the general population and stratified by country, sex, and age will be conducted (2A).

3. Substudies 3A and 3B

Descriptive analysis of the IR of SVT overall in the general population and stratified by country, sex, and age will be conducted (3A).

4. Substudies 4A and 4B

Descriptive analysis of the IR of A-fib/flu overall in the general population and stratified by country, sex, and age will be conducted (4A).

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 146

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 19 Epidemiology No.: EP07044.002

5. Substudies 5A and 5B

Among persons who have experienced a first seizure, a descriptive analysis of the IR of first recurrent seizure will be conducted (5A).

6. Supplementary analyses

In total, 10 supplementary analyses will be performed to examine the robustness of the results. These are described in further details in section 9.7.6 of the protocol as well as in the SAP.

9.2 Setting

The cohort of individuals with redeemed DL prescriptions will be identified from the four Nordic national prescription registers. Similarly, the cohort of all individuals with seizures (first seizure and first recurrent seizure, the cohort of individuals with SVT, and the cohort of individuals with A-fib/flu will be identified from the four Nordic national patient registers. The population is comprised of all individuals with DL prescriptions and of all individuals with seizures, SVT, or A-fib/flu in the four Nordic countries in the period 2001–2013 in Denmark and Finland, 2008–2013 in Norway, and July 2005–2013 in Sweden (Table 5). Individuals for Substudies 2B, 3B, and 4B become eligible for the study cohort upon first dispensing of DL, and individuals in Substudy 5B become eligible upon first seizure if they have also redeemed at least one DL prescription (See sections 9.2.1 and 9.2.2). Data will be available until and including 2013 for all countries. If possible, data will also be included for 2014.

9.2.1 Inclusion criteria

Substudy 1

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

 General population living in Denmark, Finland, Norway, and Sweden on 01 January of each year.

Substudy 2A

1.0 Individuals who have experienced a seizure during the study period (seizure cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

2.0 General population living in Denmark, Finland, Norway, and Sweden on 01 January of each year.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 147

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 20 Epidemiology No.: EP07044.002

Substudy 2B

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

Substudy 3A

 Individuals who have experienced a SVT during the study period (SVT cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

 General population living in Denmark, Finland, Norway, and Sweden on 01 January of each year.

Substudy 3B

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

Substudy 4A

 Individuals who have had an A-fib/flu diagnosis during the study period (A-fib/flu cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

 General population living in Denmark, Finland, Norway, and Sweden on 01 January of each year.

Substudy 4B

 Individuals who have redeemed at least one prescription of DL during the study period (DL cohort) and have residential location in Denmark, Finland, Norway, or Sweden.

Substudy 5A

 Individuals who have experienced a first recurrent seizure during the study period and have residential location in Denmark, Finland, Norway, or Sweden.

Substudy 5B

9.2.2 Exclusion criteria

Substudy 2A

 Individuals with a diagnosis of seizure, epilepsy, or prescriptions of antiepileptic medicine before entering the study period, as they have prevalent disease.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 148

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 21 Epidemiology No.: EP07044.002

 Individuals with a diagnosis of malignant brain tumor or head trauma before the first seizure, as they are at high risk of seizures due to causes other than DL use.

Substudy 2B

Substudies 3A and 4A

 Individuals with a diagnosis of SVT or A-fib/flu before entering the study period, as they have prevalent disease.

Substudies 3B and 4B

 Individuals with a diagnosis of SVT or A-fib/flu before use of DL, as they have prevalent disease.

 Individuals with a diagnosis of congenital pre-excitation syndrome (e.g., Wolff Parkinson White) before use of DL, as they are at high risk of SVT or A-fib/flu due to causes other than DL.

Substudy 5A

Substudy 5B

 Individuals with a brain tumor (benign and malignant), initiation of treatment with antiepileptic medicine, or head trauma occurring after redemption of first DL prescription

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 149

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 22 Epidemiology No.: EP07044.002

and first seizure will be censored at date of first occurrence, as they are at high risk of recurrent seizures due to causes other than DL use.

9.3 Variables

9.3.1 Exposure

For the association analyses, person time exposed to DL will be determined from dispensing records and period of current use will be defined for each prescription as days’ supply starting from the date of redemption plus a 4 week grace period to account for intermittent use and a possible wash-out effect. To minimize exposure misclassification, we propose to use person time at least 26 weeks beyond the dispensing date of the prior prescription as the “unexposed” reference period (in supplementary analysis 7, an alternative period of 52 weeks beyond the prior prescription is used to defined the unexposed reference period). If a new DL prescription redemption occurs during an exposed period (either during the period equal to the sum of the days’ supply or the 4 week grace period), the exposure period extends from that date with a period equal to the sum of days’ supply in the newly redeemed prescription plus a 4 week grace period. In addition, a prescription redemption before 26 weeks after previous prescription redemption would mark the start of a new exposed period (and would not result in an unexposed period). Figure 1 below provides an example of how persons included in the association analyses enter and exit exposure periods during the study period.

Days’ supply will be calculated from the quantity of tablets or amount of solution dispensed and the standard daily dose based on the age of the patient (i.e., 6-11 months: 1 mg/day; 12 months-5 years: 1.25 mg/day; 6-11 years: 2.5 mg/day; ≥12 years: 5 mg/day).

Figure 1 Example of current exposed periods versus unexposed periods for study subjects included in the association analyses (Substudies 2B, 3B, 4B, and 5B).

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 150

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 23 Epidemiology No.: EP07044.002

Person1

Person2

Person3

Person4

Study period

= days’ supply = unexposed period = Outcome Exposed period { = 4 week grace period = neither exposed nor unexposed Figure 1 illustrates how subjects in the study population (i.e., restricted to persons with at least one redeemed prescription of DL) enter and exit exposed and unexposed periods. An exposed period is a period starting from the day of redemption of a DL prescription and includes the following period equal to the number of days’ supply in the drug packages purchased (illustrated with dark grey boxes) and a 4 week grace period (illustrated with the light grey boxes). Unexposed periods, which start 26 weeks after the dispensing date of the last DL prescription, are depicted with white boxes. The time between the end of the exposed period and the unexposed period is considered neither exposed nor unexposed; these periods are illustrated with shaded boxes. Moreover, the date at which an outcome (i.e., seizure, recurrent seizure, SVT, or A-fib/flu) occurs (if occurring) is marked with a black triangle. For example, person 1 experienced the outcome before redeeming DL for the first time (date of entering the study); and therefore, this person will be excluded. Person 2 experienced the first outcome while in an exposed period and the recurrent outcome occurred in an unexposed period. Person 3 was exposed to DL twice, but did not experience an outcome. Finally, person 4 was exposed to DL once and experienced the outcome in a period where the person was neither exposed nor unexposed; and therefore, this outcome will not count toward the analysis of the association.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 151

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 24 Epidemiology No.: EP07044.002

Loratadine, the parent compound of DL, is also available via prescription and over–the-counter in the Nordic countries. We think it is somewhat unlikely that patients would switch between DL and loratadine; however, it is possible that some of the time counted as unexposed in the analyses could actually be time exposed to loratadine. Moreover, other non-sedating prescription antihistamines are available in the Nordic countries, and if the effect on the outcomes is driven by use of non- sedating antihistamines in general rather than the specific effect of DL exposure, these drugs should be examined to elucidate potential misclassification of exposure. To explore exposure misclassification (i.e., the influence of loratadine and other non-sedating prescription antihistamines), on the associations under study), we will conduct supplementary analyses, which will be specified in the statistical analysis plan.

Regarding missing values in variables from the prescription registers, normally no registration is interpreted as the person does not use the drug (e.g., no registration of DL use is interpreted as the person has not used DL)). Prescription drugs are not available from sources other than the pharmacies, and all drugs purchased at pharmacies are included in the prescription register (the only exception are drugs supplied from hospitals, however less than 1% of DL is supplied by the hospitals). We do not expect missing information for data in the prescription registry that will be used to determine DL exposure (e.g., number of packs per redemption). However, we will examine data to see whether missing information occurs. Missing data handling will be determined after review of the data, but before data from different registers are linked together (e.g. handling of missing values for exposure data is blinded to outcome status and vice versa) and will be described in a data control report, which will be submitted with the study report.

9.3.2 Outcomes

Four outcome variables will be used in the present study: first seizure (Substudy 2B), first SVT (Substudy 3B), first A-fib/flu (Substudy 4B), and first recurrent seizure (Substudy 5B). Except where noted, the case definition of seizure excludes febrile seizure, a condition in infancy and early childhood attributed to fever. However, supplementary analysis 2 will evaluate the outcomes of febrile and non-febrile seizures in children. (The diagnosis of febrile seizure is generally not used in adults.) Information on outcome variables will be obtained from the Nordic national patient registers, including diagnostic and treatment information for patients treated at the secondary and tertiary hospital level in all four Nordic countries using International Classification of Diseases (ICD)- 10 codes. We will exclude seizure cases if they are registered with brain tumor (benign and malignant), stroke, or acute drug intoxication or overdose of drugs during the same hospitalization as the seizure. Table 2 gives an overview of the outcome variables in the study. For all outcome variables, we will include the primary diagnoses from the emergency room or inpatient settings registered in the Nordic national patient registers.

The National Patient Registers are used for reimbursement of services to the hospitals. It is mandatory to enter specific information (such as name, date, main diagnosis) for the record to be

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 152

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 25 Epidemiology No.: EP07044.002

established. Due to the requirements of the national registries, we do not expect missing data on the outcome variables. In Denmark and Finland, it is only possible to register hospitalizations into the Nordic national patient registers if complete information on the primary diagnosis, date, and hospital department is entered into the registration system. Therefore, no missing values should occur for these variables. This procedure is most likely the same for Sweden and Norway. Data in the Danish National Patient Register is automatically checked for missing codes, inconsistencies between diagnosis and gender, incorrect digits and errors in the personal identification code (PIC). If an error is detected, the record is returned to the source hospital for correction [Ref. 5.4: 04C0C2] .No studies have examined the validity of seizures in the total population, but Vestergaard et al (2006) examined the validity of the discharge diagnosis of febrile seizure in children the National Patient Register (ICD10 code R56.0) [Ref. 5.4: 04C0C7]. The positive predictive value (PPV) was 92.8% (95%CI: 88.8-95.7%). The sensitivity (defined as completeness by the authors) is 71.5% (95% CI: 66.3-76.4%) [Ref. 5.4: 04C0C7]. The diagnosis of atrial fibrillation and atrial flutter has been validated in the Danish National Patient Register [Ref. 5.4: 04C0BZ]. The PPV for the combined diagnosis of atrial fibrillation and atrial flutter (I48) was 92.6%. Other studies have found even higher PPVs [Ref. 5.4: 04C0BS] [Ref. 5.4: 04C0BQ]. No studies have validated the diagnosis for supraventricular tachycardia (I47); however, we find it likely that this diagnosis has approximately the same validity in the National Patient Register as seen for atrial fibrillation and atrial flutter.

EMA/623947/2012 22-Feb-2016

04CN88

05822K P A GE 153

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 2 6 E pi d e mi ol o g y N o.: E P 0 7044.002

T a bl e 2 Overvie w of the Outco me Variables in the Study. Description of outco me variable Substud y Substudy 2 B First s eiz ur e: A n i nci de nt case of seiz ure is a pers o n wit h t he first diag n osis of seiz ure ( Substudy 3 B S u prave ntric ular tachycardia ( S V T) : A n i nci d e nt c as e of S V T is a p ers o n wit h first di a g n osis of S V T Substudy 4 B Atri al fi brill ati o n or atri al fl utt er ( A -fi b/fl u) : A n i nci d e nt c as e of A -fi b/fl u is a p ers o n wit h first di a g n osis of A -fi b or first di a g n osis of A -fl u . T h es e t w o di a g n os es will b e c o m bi n e d i nt o a co mposite endpoint Substudy 5B First rec urre nt seiz ure: A first rec urre nt case of seiz ure is a pers o n wit h t he di a g n osis of t h e second seizure I C D -1 0 c o d es ar e list e d i n Annex 4 . Note that the case definition of seizure in these substudies excludes febrile seizure

9. 3. 3 Covariates

I n t h e pr es e nt st u d y, w e h a v e us e d dir ect e d ac yclic gra p hs ( D A Gs) , als o c all e d c a us al diagra ms , f or confounder s el ecti o n. D A Gs ar e a w ell -accepted methodology f or usi n g c a us al kno wledge a n d a s et of f or m al mathe matical pri nci pl es f or s el ecti n g w hic h v ari a bl es t o a dj ust f or w h e n perfor ming ass oci ati o n a nalyses [ R ef. 5. 4: 045RC Q] . T h e y pr o vi d e a syste matic way t o e x pl or e t h e r el ati o ns hi ps bet ween t h e ex p os ures, outco mes, a n d c ovariates ( u ni dir ecti o n al, bi dir ecti o n al, c a us al) a n d f acilit at e d e ali n g wit h a l ar g e nu mber of p ot e nti al confounders. D A Gs h el p m a k e t h e assu mptions u n d erl yi n g a n a n al ysis e x plicit. T h e s el ecti o n of v ari a bl es needed for confounder adjust ment t o o bt ai n a n u n bi as e d esti m at e of t h e ass oci ati o n u n d er st u d y is c all e d t h e mi ni m u m s uffici e nt adjust ment s et of confounders. W e us e d t h e o p e n s o urce a n d fr e el y availa ble s oft ware D A Gitt y f or t h e develop ment of t h e D A Gs [ R ef. 5. 4: 045 WX9] . D A Gitt y h el ps t h e researcher vis u aliz e t h e str uct ur e of r el e v a nt v ari a bl es f or t h e ass oci ati o n u n d er st u d y i n cl u d e d i n t h e D A G, as w ell as t o i d e ntif y t h e mi ni m u m s uffici e nt adjust ment s ets a v ail a bl e f or confounder a dj ust m e nt. I n t h e f oll o wi n g paragraphs , w e pr es e nt a bri ef d escri pti o n of t h e D A G pr ocess ; ho wever , t h e f ull D A G pr oc ess is d escri b e d i n A n n ex 6.

T h e f act ors list e d i n t h e l eft -hand c ol u m n of T a bl e 3 w er e considered potential confounders (i. e., candidate variables) for the association bet ween DL use and seizures, S VT, and/or A -fi b/fl u b as e d o n t h e lit er at ur e, as w ell as c o ns ultati o n wit h a gr o u p of cli nic al e x p erts. T h e p ot e nti al confounders w er e revie wed d uri n g a D A G m e eti n g h el d o n Ja n uary 6, 2015 . At t his m e eti n g, t h e r el ati o ns hi p s a m o n g t h e diff er e nt f act ors w er e disc usse d , a n d D A Gitt y w as us e d t o h el p i d e ntif y t h e mi ni m u m s uffici e nt adjust ment s et of confounders t o i ncl u d e i n t h e ass oci ati o n a n al ysis of e ac h o utc o m e. Three D A Gs i ncl u di n g p ot e nti al confounders of t h e ass oci ati o n bet ween DL us e a n d s eiz ur es, S V T, a n d A- fi b/fl u, r es p ecti v el y, w er e developed b y t h e D a nis h n ati o n al i nvestigat ors, M S D, P P D a n d t w o cli nic al e x p erts. Aft er t h e D A G m e eti n g, w e c o ns ulte d a der matologist t o c o nfir m t h e c orrect ness of t h e developed D A Gs, es p eci all y regar di ng t h e r el ati o ns hi ps bet ween urtic ari a a n d t h e other potential confounders. The D A Gs h el p e d cl arif y whether the potential confounders listed were confounders, colliders, or inter mediate variables and which of the potential confounders constituted a mi ni m u m s uffici e nt adjust ment s et. W e s el ect e d t h e fi n al mi ni m u m s uffici e nt adjust ment s et , w hic h will b e u s e d t o a dj ust f or confounding f act ors i n t h e ass oci ati o n st u di es b as e d on whether w e f o u n d t h at a gi v e n c o m bi n ati o n of confounders w o ul d b e o bt ai n a bl e fr o m t h e r e gist ers. T h e fi n al mi ni m u m s uffici e nt adjust ment s et f or t h e ass oci ati o n st u di es (i. e., 2 B, 3 B , 4 B, a n d 5 B) ar e s e e n i n t h e ri g ht -hand c ol u m ns of T a bl e 3 .

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K PAGE 154

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 27 Epidemiology No.: EP07044.002

Table 3 Information on whether potential confounding factors are included in the DAGs and the minimum sufficient adjustment sets, which will be used for confounder adjustment in the association studies.

Potential confounding factors Potential confounder included in the DAG Confounders included in the selected minimum sufficient adjustment set for each association study Seizures SVT A-fib/flu Seizures SVT A-fib/flu Age (age will be derived as a categorical and a continuous variable: Age groups Yes Yes Yes Yes Yes Yes [0–4 years, 5–19 years, 20–64 years, ≥65 years] and years) Sex (male versus female) Yes Yes Yes Yes Yes Yes Country of residence (Denmark, Finland, Yes Yes Yes Yes Yes Yes Norway, Sweden) Calendar year (years ranging from 2001– Yes Yes Yes Yes Yes Yes 2013) Drug overdose (other than Yes Yes Yes No No No desloratadine) Drug and alcohol abuse Yes Yes Yes No No No Diabetes (both type 1 and type 2) Yes Yes Yes No No No Use of hypoglycemic agents (oral anti- Yes No No No No No diabetics, insulin) Hypertension No Yes Yes No No No Thyroidism (both hypo- and hyperthyroidism, e.g., Grave’s disease, No Yes Yes No No No thyrotoxicosis) Structural heart disease: Left ventricular hypertrophy, left ventricular systolic No Yes Yes No No No dysfunction, chronic heart failure (CHF)

Seasonality (i.e., winter [December- February], spring [March – May], Yes Yes Yes Yes Yes Yes summer [June-August], and autumn [September-November]) Asthmatic status Yes Yes Yes Yes Yes Yes Disease severity of rhinitis Yes Yes Yes Yes Yes Yes Chronic obstructive pulmonary disease Yes Yes Yes No No No (COPD) Smoking Yes Yes Yes No No No Inflammatory disease No Yes Yes No No No Metastatic disease Yes No No No No No Infections No Yes Yes No No No Stroke Yes Yes Yes No No No Chronic urticaria Yes Yes Yes Yes Yes Yes Unspecific autoimmune disease Yes Yes Yes No No No Type 1 allergy Yes Yes Yes No No No Antihypertensive treatment No Yes Yes No No No

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 155

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 28 Epidemiology No.: EP07044.002

Table 4 below shows the variables included in the minimum sufficient adjustment sets in all four association analyses. In addition to being confounders included in the association analyses, the variables in the minimum sufficient adjustment sets will be used to characterize the population using DL in the descriptive study of DL use (Substudy 1). The table describes how each of the variables will be operationalized. ICD-10 codes and ATC codes that will be used for the definition of the variables can be found in Annex 4.

Table 4 Definition of confounders included in the minimum sufficient adjustment sets that will be used for confounder adjustment in the association studies. Confounders included in the Definition Data source minimum sufficient adjustment sets Age For the descriptive analyses of DL use, age will be defined as the age of the Civil registration purchaser at the date of prescription redemption and stratified into age system groups (0–4 years, 5–19 years, 20–64 years, ≥65 years). In the association studies, age is a time varying confounder; and therefore, risk time will be split up in years of age. We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

Sex Sex is a time-independent confounder and will be included as males versus Civil registration females in both the descriptive and association analyses. system

Country Country of residence is a time-independent confounder, as persons will be Civil registration excluded if emigrating the country. Country of residence will be included with system four categories, i.e. Denmark, Finland, Norway, or Sweden in both the descriptive and association analyses.

Calendar year For the descriptive analyses of DL use, calendar year is defined as the year of Prescription prescription redemption. register In the association studies, calendar year is a time varying confounder; and therefore, risk time will be split up in calendar years.

Seasonality For the descriptive analyses of DL use, seasonality is defined as the season Prescription (i.e., winter [December-February], spring [March – May], summer [June- register August], and autumn [September-November]) when the DL prescriptions are redeemed.

In the association studies, seasonality is a time varying confounder; and therefore, risk time will be split up in seasons defined as winter (December– February), spring (March–May), summer (June–August), and autumn (September–November).

Asthma status Both for the descriptive and association analyses, asthmatic status is defined National patient as a binary variable indicating whether or not a person has redeemed register and treatment for asthma defined as at least two prescriptions of inhalant prescription steroids within a six-month period and/or contacts hospitals with a diagnosis register of asthma (including both primary and secondary diagnoses) during a five- year period before first DL exposure. To distinguish persons treated for chronic obstructive pulmonary disease from those treated for asthma, first registered asthma treatment has to be redeemed when the purchaser was 45 years or younger.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 156

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 29 Epidemiology No.: EP07044.002

Confounders included in the Definition Data source minimum sufficient adjustment sets Disease severity of Both for the descriptive and association analyses, severity of rhinitis will be Prescription rhinitis defined as binary variable indicating whether or not a person has received register treatment for severe rhinitis. Persons with severe rhinitis will be identified from the prescription register as persons who have redeemed immunotherapy at least once during a five-year period before first DL exposure.

Chronic urticaria Both for the descriptive and association analyses, chronic urticaria status is National patient defined as a binary variable indicating whether or not a person has a register registered diagnosis of chronic urticaria in the five-year period before first DL exposure.

9.3.4 Other variables

We will also obtain information on immigration and emigration, as well as date of birth and death status from the civil registration systems to be able to calculate the IRs in Substudies 2B, 3B, 4B, and 5B.

The interpretation of register-based variables with regard to missing information was discussed in section 9.3.1 and 9.3.2. For a few of the included register-based variables (e.g., sex and date of birth), all included persons should have information. We do not expect missing information on these variables; however, we will examine data to see whether missing information occurs. Missing data handling will be determined after review of the data, but before data from different registers are linked together (blinded to exposure to DL and outcome status) and will be described in a data control report, which will be submitted along with the study report.

9.4 Data sources

The DL study will include register information from four Nordic countries – Denmark, Finland, Norway, and Sweden. In addition to a long history of collecting high quality information on births, deaths, immigration and emigration, disease incidence, and activities in the healthcare sector [Ref. 5.4: 00W4D0], exceptional opportunities to perform register-based research are driven by the unique PIC introduced in the Nordic countries in the 1960’s and available to all persons with permanent residence in the Nordic countries [Ref. 5.4: 00W4G3]. The PIC makes it possible to link information at the individual level from several registers for scientific research purposes. The national prescription registers and national patient registers within each of the Nordic countries capture all the individual encounters of purchasing prescribed DL and allow sufficient longitudinal data to differentiate between first and recurrent seizures and to identify incident SVT and A-fib/flu cases. Person-specific use of DL will be elucidated from the national prescription registers by obtaining information on redemption of DL prescription for each person [Ref. 5.4: 00W4CX]. Person- specific information on seizures, SVT, and A-fib/flu will be derived from the Nordic national patient registers.

Table 5 presents key information on the population-based health registers in the Nordic countries of relevance for the present study. Data will be available until and including 2013 for all countries. Data

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 157

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 30 Epidemiology No.: EP07044.002

will be applied for including 2014 and used if available. The data extraction period refers to the longest period for which data on exclusion variables (e.g., seizures, SVT, or A-fib/flu before baseline) can be obtained for.

Table 5 Overview of national health registers in the Nordic countries of relevance for the present study. Country Register Denmark Finland Norway Sweden National prescription 1995–2013 1994–2013 2004–2013 2005–2013 register National patient register 1977–2013 (1) 1967–2013 (2) 2008–2013 1987–2013 (3) Civil registration system 1968–2013 1967–2013 1964–2013 1965–2013 2001–2013 2001–2013 2008–2013 July 2005–2013 Study period * * * * (13 years) (13 years) (6 years) (9 years)

Data extraction period 1977–2013 1967–2013 2008–2013 1987–2013 (1) Contacts with outpatient clinics (incl. emergency departments) since 1995. (2) Contacts with outpatient clinics (incl. emergency departments) since 1998. (3) Contacts with outpatient clinics (incl. emergency departments) since 2001. * Data will be available until and including 2013 for all countries. If possible, data will also be included for 2014.

The national prescription registers include information on the date of prescription redemption, information on the purchaser, and information on the drug redeemed (e.g., ATC code, number of pills, daily dose, pack size, and number of packs purchased) [Ref. 5.4: 045TYK]. The Nordic national patient registers include diagnostic and treatment information for patients treated at the secondary and tertiary hospital level [Ref. 5.4: 045TZL] [Ref. 5.4: 045W09]. Clinical experts have been consulted on how to include information from the prescription and patient registers. Information on date of birth, immigration, emigration, and death will be obtained from the civil registration systems [Ref. 5.4: 00W4G3] [Ref. 5.4: 045W09].

9.4.1 Study procedures

9.5 Study size

04CN88

05822K PAGE 158

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 31 Epidemiology No.: EP07044.002

and first A-fib/flu [Substudy 4B]). For Substudies 2B, 3B, and 4B, a cohort study design among DL users will be used. The sample size in the present study will be fixed, as the study population for the association studies 2B, 3B, and 4B consists of all individuals in the four Nordic countries who have redeemed a DL prescription at least once. Hence, the aim of this section is to calculate the minimum detectable incidence rate ratio (IRR). Calculations concern Substudy 2B evaluating the association between DL use and first seizure, Substudy 3B evaluating the association between DL use and first SVT event, and Substudy 4B evaluating the association between DL use and first A-fib/flu. The actual annual IR of seizures, SVT, and A-fib/flu among current DL users is unknown and will be determined in the current study.

Minimal detectable IRR were calculated based on a rearranged formula for sample size calculations developed by Bryant & Morganstein [Ref. 5.4: 045WWJ] (Formula provided in Annex 5). The following are the parameters needed to calculate the minimum detectable IRR: the number of DL users in each age group (i.e., the fixed sample size); the proportion of exposed time (current exposure) and unexposed time (non-current/remote exposure) for DL users; the background event rates of seizures, SVT, and A-fib/flu; and the number of years included in the study along with statistical parameters and assumptions included in the equation (i.e., the significance level, power, and 1-sided versus 2-sided tests).

9.5.1 Fixed sample size

The number of DL users from each country has been estimated based on national drug sale statistics (Annex 1A). The annual unique number of DL users in Denmark, Finland, and Sweden for age groups 0–4, 5–19, 20–64, and ≥65 years is estimated to be 20,000, 86,000, 215,000, and 44,500, respectively. Data for Norway on DL prescriptions and seizures/SVT/A-fib/flu are available for a shorter time period; and therefore, Norway is not included in the calculations of the minimum detectable IRR. The number of annual unique DL users in the four countries is thereby slightly underestimated.

9.5.2 Proportion of current and non-current exposure time among DL users for the purpose of calculation of minimum detectable IRR

DL exposure (i.e., current DL use) is time-varying, and each person might have one or more periods as exposed and one or more periods as unexposed. Individuals in the study population (i.e., individuals who have redeemed at least one prescription of DL) are assumed to be exposed in a period following the date of DL purchase (i.e., current users). An estimate of the number of DL prescriptions redeemed per year in each age group was obtained from the feasibility study performed prior to developing the synopsis and the present protocol. For the minimum detectable IRR calculations, the mean number of prescriptions in each age group per year from the Danish data is used as an estimate of the average annual number of redemptions. The exposure period is then calculated as follows: (mean annual number of prescriptions in each age group*4 weeks (standard days’ supply)) + 4 weeks grace period to account for intermittent use and a possible wash-out effect. Beyond this period, individuals are assumed to be unexposed (i.e., non-current users). The average

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 159

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 32 Epidemiology No.: EP07044.002

number of prescriptions per year was estimated to be 1.50, 1.62, 1.78, and 2.30 for age groups 0–4, 5–19, 20–64, and ≥65 years, respectively1.

9.5.3 Background event rate

The background incidence rate (i.e., IR) of seizures, SVT, and A-fib/flu was obtained from published studies [Ref. 5.4: 00W4CY] [Ref. 5.4: 03XNM4] [Ref. 5.4: 03XNL7]. IRs of seizure among unexposed individuals for age groups 0–4, 5–19, 20–64, and ≥65 years were estimated to be 65, 50, 40, and 40 per 100,000 person-years, respectively [Ref. 5.4: 00W4CY]. The IR of SVT is reported to be 13/100,000 person-years for persons aged 19 years or younger, 27/100,000 person-years for persons aged 20–64 years, and 122/100,000 person-years for persons aged 65 years or older [Ref. 5.4: 03XNM4]. Wilke et al. (2013) [Ref. 5.4: 03XNL7] reported the incidence of A-fib to increase markedly with age with 0.0016 cases/1000 person-years in children (<15 years) and approximately 30 cases/1000 person-years for persons aged 80 years or older. No data on the epidemiology of A- flu were available.

9.5.4 Number of years in the study

9.5.5 Statistical parameters and assumptions

The minimum incidence rate ratio, IRR>1, that can be detected is calculated using the following parameter values and assumptions:

Significance level, α: 0.05 Power, 1-β: 0.80

1 By using Danish figures we might get conservative estimates of the minimal detectable IRR as mean number of prescriptions calculated for the feasibility study showed slightly lower means for Denmark compared to Sweden (figures from Finland and Norway were not available for these calculations). The average number of prescriptions per age group for the 20-64 and ≥65 year’s age groups was calculated as the mean number of annual prescriptions (2001-2012) per age group using data from Denmark for ages ≥16 years. For the 5-19 year age group, the calculation was based on data from Denmark for ages 16-19. For the 0-4 year age group, the calculation was based on Swedish data from the feasibility study, which showed a tendency for a lower number of prescriptions in this age group compared to the older age groups. EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 160

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 33 Epidemiology No.: EP07044.002

1-sided test

9.5.6 Example of calculating the minimum detectable IRR

Mean number of Annual IR of seizures Estimated number of individuals with at Minimum detectable IRR Age group among unexposed weeks exposed based least one annual DL t=2 and 6 years (for age 0-4 (years) individuals per on mean number of DL prescription in the years and age ≥5 years) 100,000 person-years prescriptions per person Nordic countries 0–4 65 20,000 10.00 3.0 5–19 50 86,000 10.48 1.5 20–64 40 215,000 11.12 1.3 ≥65 40 44,500 13.20 1.7

Table 7 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rate (IR) of SVT is based on Orejarena et al. (1998) [Ref. 5.4: 03XNM4]. A 1-sided test, significance level of 5% and power of 80% have been used. Mean number of Annual IR of SVT among Estimated number of individuals with at Minimum detectable IRR Age group unexposed individuals weeks exposed based least one annual DL t=2 and 6 years (for age 0- (years) per 100,000 person- on mean number of DL prescription in the 4 years and age ≥5 years) years prescriptions per person Nordic countries 0–4 13 20,000 10.00 8.2 5–19 13 86,000 10.48 2.1 20–64 27 215,000 11.12 1.4 ≥65 122 44,500 13.20 1.4

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 161

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 34 Epidemiology No.: EP07044.002

Table 8 The minimum detectable incidence rate ratio (IRR>1) was calculated based on the Nordic population of DL users with at least one prescription of DL per year. Incidence rates (IR) for atrial flutter were not available; therefore, we base the calculations only on IR of A-fib obtained from Wilke et al. (2013) [Ref. 5.4: 03XNL7]. A 1-sided test, significance level of 5% and power of 80% have been used. Mean number of Annual IR of A-fib Estimated number of individuals with at Minimum detectable IRR Age group among unexposed weeks exposed based least one annual DL t=2 and 6 years (for age 0- (years) individuals per 100,000 on mean number of DL prescription in the 4 years and age ≥5 years) person-years prescriptions per person Nordic countries 0–4 1.6 20,000 10.00 54.0 5–19 3.8 86,000 10.48 3.4 20–64 168.9 215,000 11.12 1.6 ≥65 2154.3 44,500 13.20 1.1

9.6 Data management

 All national scientific coordinators will facilitate the construction of the study populations:

 Study population consisting of all DL users during the study period.

 Study population consisting of all individuals with seizure during the study period and the years prior to the study period.

 Study population consisting of all individuals with SVT during the study period and years prior to the study period.

 Study population consisting of all individuals with A-fib/flu during the study period and years prior to the study period.

 All national scientific coordinators are responsible for acquiring and validating the datasets and will explore how the data can be combined from the registers. Data quality control includes but is not restricted to check for legal values for each categorical variable, check of consistency between dates (at least date of birth before all other dates and date of death after all dates), and check and advise on the handling of missing data. All national scientific coordinators will produce a data control report describing the checks performed. All

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 162

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 35 Epidemiology No.: EP07044.002

national scientific coordinators will derive the final dataset from the data obtained from the registers by combining the registers according to the study designs. All national scientific coordinators will produce a data control report describing the checks performed of the final dataset, including reasons for modifications and exclusions. In this process, all national coordinators have to agree on the reasons for exclusion (e.g., missing value on crucial variables, chronological errors in the relation between dates, non-legal values of categorical variables, and extreme values of continuous variables). The national scientific coordinators should send the final data set and the data control reports that describe the data and provide suggestions on how to handle missing values and invalid codes to the Danish scientific coordinator. A template for the data control report will be provided by the Danish scientific coordinator and will include the following requirements:

1. Information on known misclassification of each variable (e.g., underreporting, low sensitivity or specificity, categorization with obvious invalid values, etc.). 2. Check for legal values for each categorical variable. Check for reasonable distribution of variables. Include advice on how to handle unexpected observations. 3. Check for reasonable minimum, maximum, and central tendency (median, mean) for each continuous variable. Check for outliers must be performed (e.g., exploratory plots, such as box-whiskers plot). Include advice on how to handle unexpected observations. 4. Check of chronological relation between date variables: At least date of birth before all other dates and date of death after all dates, but also reasonable relation of dates of diagnoses and prescriptions. Unexpected patterns should be described and solutions for handling such observations should be included. 5. Check of missing information on variables and include advice on how to handle such observations (e.g., delete observations with missing information, put missing observations into a specific category, etc.).  The final datasets from Finland, Norway, and Sweden are transferred to Statistics Denmark where all subsequent data handling is done by the Danish scientific coordinator.  The Danish scientific coordinator combines data as described by the document developed by all national scientific coordinators and the datasets from all countries will be joined into a combined analysis dataset. Relevant variables will be derived.  The Danish scientific coordinator will assess the data validity of all countries by logical checks, examination of extreme values, and missing data. It is important that identification numbers are maintained to facilitate linkage back to the original datasets to be able to check the data and for the sake of transparency. 9.7 Data analysis

Prior to conducting the data analyses, we will perform data management (as described in section 9.6) to ensure data quality and to correct inconsistencies and errors in the data. The data analysis will include the following five steps listed below. Note that specifications concerning all pre-defined supplementary analyses will be detailed in a separate statistical analysis plan before database lock and start of data analysis. The assumptions of the statistical models performed will be evaluated for each model in the analysis.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 163

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 36 Epidemiology No.: EP07044.002

9.7.1 Substudy 1: Descriptive analysis of exposure of DL

A cohort study describing DL use in the general population will be performed. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to avoid left-truncation bias. In addition, to avoid analogous misclassification of prevalent with incident DL dispensings among immigrants, we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. Prevalent users will be defined as persons who have at least one prescription of DL in the period of interest (e.g., the entire study period or each year). Incident users are first time users of DL in the period of interest. The following descriptive analyses will be performed:

9.7.2 Substudy 2: First seizure

Substudy 2A

A cohort study estimating the incidence of first seizures in the general population will be conducted. The cohort of individuals with seizures and the distribution of the population by age, year , and country will be used to derive estimates for the general population. The number of individuals in the general population on 01 January of each year will be obtained from the NORDCAN database and used as an estimate of the number of person years at risk assuming each person in the population on 01 January of each year contributes one person year at risk [Ref. 5.4: 045VPW]. The IRs of seizure will be shown for the total population and stratified by country, sex, and age. The 95% confidence interval for IRs will be calculated as:

where IR is the incidence rate, exp is the exponential function, and N new users is the number of new users [Ref. 5.4: 045VXZ].

Substudy 2B

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 164

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 37 Epidemiology No.: EP07044.002

DL is included in the cohort and enters the cohort at the date of first prescription redemption. Study individuals are followed until occurrence of a seizure; occurrence of one of the conditions described in section 9.2.2; 31 December 2013; emigration; or death, whichever comes first. The association between exposure to DL and first seizure will be evaluated using Poisson regression of the IR of first seizure, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to account for truncation bias. In addition we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) (Table 4). We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

9.7.3 Substudy 3: SVT

Substudy 3A

A cohort study estimating the incidence of SVT in the general population will be performed. The cohort of DL users, the cohort of individuals with SVT, and the distribution of the population by age, year, and country will be used to derive estimates in the general population. The number of individuals in the general population on 01 January of each year will be obtained from the NORDCAN database and used as an estimate of the number of person years at risk assuming each person in the population on 01 January of each year contributes one person year at risk [Ref. 5.4: 045VPW]. The IRs of SVT will be shown for the total population and stratified by country, sex, and age. The confidence interval of IRs is calculated as outlined above under Substudy 2A.

Substudy 3B

A cohort study among ever-DL users analyzing the association between exposed and unexposed time (as defined in Substudy 2B) and first SVT will be performed for the total population and stratified by age. A person who has redeemed at least one prescription of DL is included in the cohort and enters the cohort at the date of first prescription redemption. Study individuals are followed until occurrence of a SVT; 31 December 2013; emigration; or death, whichever comes first. The association between exposure to DL use and first SVT will be evaluated using Poisson regression of the IR of first SVT, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to account for truncation bias. In addition we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) (Table 4). We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

9.7.4 Substudy 4: Atrial fibrillation/flutter

Substudy 4A

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 165

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 38 Epidemiology No.: EP07044.002

A cohort study estimating the incidence of A-fib/flu in the general population will be performed. The cohort of individuals with A-fib/flu and the distribution of the population by age, year, and country will be used to derive estimates in the general population. The number of individuals in the general population on 01 January of each year will be obtained from the NORDCAN database and used as an estimate of the number of person years at risk assuming each person in the population on 01 January of each year contributes one person year at risk [Ref. 5.4: 045VPW]. The IRs of A-fib/flu will be shown for the total population and for each country, sex, and age. The confidence interval of IRs is calculated as outlined above under Substudy 2A.

Substudy 4B

A cohort study among ever-DL users analyzing the association between exposed and unexposed time (as defined in Substudy 2B) and first A-fib/flu diagnosis will be performed for the total population and stratified by age. A person who has redeemed at least one prescription of DL is included in the cohort and enters the cohort at the date of first prescription redemption. Study individuals are followed until occurrence of A-fib/flu; 31December 2013; emigration or death, whichever comes first. The association between exposure to DL use and first A-fib/flu will be evaluated using Poisson regression of the IR of first A-fib/flu, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to account for truncation bias. In addition we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar years, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) (Table 4). We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

9.7.5 Substudy 5: First recurrent seizure

Substudy 5A

Substudy 5B

04CN88

05822K PAGE 166

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 39 Epidemiology No.: EP07044.002

exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar years, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) (Table 4). We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

9.7.6 Supplementary analyses

The following supplementary analyses will be conducted. Unless otherwise specified, supplementary analyses will be performed on the total population (i.e. not stratified by age) and using the primary exposure definition.

(1) alternative definitions of exposure based on time since last DL dispensing (periods 0-4, 5-8, 9-16, and 17-26 weeks each compared with >26 week since last DL dispensing) for Substudies 2B, 3B, 4B, and 5B. Because as-needed medications, such as antihistamines may not be taken daily, this approach reflects the clinical expectation that the probability of actual exposure on a given day is highest shortly after filling a prescription and diminishes with increasing time. Each time a DL prescription is refilled, the time since last dispensing will reset to 0.

(2) for Substudies 2B and 5B, differentiating between febrile and non-febrile seizures for children aged 0–4 years.

(4) the association analyses in Substudies 2B, 3B, and 4B will be stratified by countries to examine potential differences across countries.

(5) additional supplementary analyses, which will be specified in the SAP, will evaluate the potential effect of exposure misclassification in Substudies 2B, 3B, and 4B (i.e. evaluate use of loratadine and other non-sedating prescription antihistamines).

(7) a supplementary analysis for Substudies 2B, 3B, 4B, and 5B for which non-exposed periods start 52 weeks following the last prescription redemption.

(8) a supplementary analysis using an alternative adjustment set to examine the robustness of the study results. This adjustment set consists of age, sex, country, calendar year, seasonality, severity of rhinitis, asthmatic status, diabetes, hypo-/hyperthyroidism, inflammatory disease, infections, type 1 allergy. For the outcomes supraventricular tachycardia and atrial fibrillation or flutter, antihypertensive treatment will also be added.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 167

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 40 Epidemiology No.: EP07044.002

(10) a supplementary analysis restricted to calendar time where misclassification due to over-the- counter (OTC) use does not exist (i.e., prior to OTC availability of DL in Denmark and Finland (only OTC sale in 2013). No OTC sale of DL has occurred in Norway and Sweden in the study period).

9.8 Quality control

By signing this protocol, the Sponsor agrees to be responsible for implementing and maintaining a quality management system with written development procedures and functional area standard operating procedures to ensure that studies are conducted and data are generated, documented, and reported in compliance with the protocol, accepted standards of Good Clinical and Pharmacoepidemiology Practice, and all applicable federal, state, and local laws, rules and regulations relating to the conduct of the study.

The study is register-based. As a result, data quality is difficult to ascertain directly. However, previous studies have examined the validity and quality of information in the Nordic registers. The Danish National Prescription Registry contains data of high quality, including detailed information on dispensed drugs, and as the register covers all prescription dispensed in Danish pharmacies, loss to follow-up is unlikely for individuals with permanent residence in Denmark [Ref. 5.4: 045TYK]. The completeness of the Norwegian and Swedish prescription registers is characterized as good [Ref. 5.4: 00W4CX] [Ref. 5.4: 046B3C]. Completeness of registration in the Norwegian prescription register is ensured by law and quality checks are carried out monthly to identify possible errors or inconsistencies [Ref. 5.4: 00W4CX]. The patient identity data are only missing for approximately 0.3% of all items in the Swedish prescription register [Ref. 5.4: 046B3C]. The Finnish prescription register has been described in detail and considered as excellent [Ref. 5.4: 0469YG]. Data quality in the Danish National Patient Register is overall assumed to be of high quality; however validity of data depends on the diagnosis under consideration; the positive predictive values for diagnoses similar to those in this study are >90%. [Ref. 5.4: 04C0C2]. The Norwegian Patient Register has relatively good agreement with the Norwegian Cancer Register [Ref. 5.4: 0469WT] and the completeness and accuracy of the Finnish Patient Register has been evaluated as varying from satisfactory to very good for common diagnoses with positive predictive values ranging from 75%– 99% [Ref. 5.4: 046B36]. The validity of the Swedish Patient Register is high for many, but not all diagnoses. The positive predictive values of most diagnoses in the Swedish Patient Register compared to medical records ranges from 85–95% [Ref. 5.4: 03RSH6].

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 168

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 41 Epidemiology No.: EP07044.002

9.9 Limitations of the research methods

Exposure: Use of DL

 The prescription registers capture information on purchasing of drugs from pharmacies. The actual adherence of prescribed drugs and consumption cannot be established in these databases. This is a major limitation because the drugs concerned are taken on an as needed basis to treat symptomatic disease. Thus, it is difficult to establish when persons are truly exposed to the drug. Therefore, supplementary analyses will be performed to characterize the relationship of prescribing and timing of events of interest by categorizing time since DL prescription to the outcomes of interest and by using 52 instead of 26 weeks beyond previous prescription redemption as start of an unexposed period.

 Left-truncation is a potential bias in the study meaning that we have no information on DL use before the start of registration of prescriptions in each country. Left-truncation bias in relation to DL use is only a relevant issue for Sweden and Norway since DL was available from 2001, but the prescription registers in Norway and Sweden were established in 2004 and 2005, respectively (Table 4). Therefore, we do not know whether a DL user in the first year of registration in Norway and Sweden is a long-term user (prevalent user) or a first- time user (incident user). To account for this potential bias, a person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included. This is not a limitation in Denmark and Finland since DL was first approved in 2001 and the prescription registers in Denmark and Finland were established in 1995 and 1994, respectively. However, left-truncation bias might also occur when persons are immigrating into the study population. Therefore we also exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden.

 There is a possibility that patients could purchase DL over-the-counter; however, because of reimbursement systems in the Nordic countries, it is actually less expensive to purchase prescription DL. In Norway and Sweden DL has only been allowed for prescription sale; However, in Denmark, DL has been available for over-the-counter sale since 2013, but over- the-counter sales only counted for less than 4 % of the total amount of DL purchased in 2013. A supplementary analysis will examine the associations under consideration restricting the time period to include only years where DL was not allowed for OTC sale.

 Loratadine, the parent compound of DL, is also available via prescription and over-the- counter in the Nordic countries. We think it is somewhat unlikely that patients would switch between DL and loratadine. However, it is possible that some of the time counted as

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 169

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 42 Epidemiology No.: EP07044.002

unexposed in the analyses could actually be time exposed to loratadine or over-the-counter DL. Moreover, other non-sedating prescription antihistamines are available in the Nordic countries, and if the effect on the outcomes is driven by non-sedating prescription antihistamines in general, rather than the specific effect of desloratadine exposure only, these drugs should be examined to elucidate potential misclassification of exposure. To explore exposure misclassification (i.e., the influence of loratadine and other non-sedating prescription antihistamines on the associations under study), we will conduct supplementary analyses, which will be specified in the statistical analysis plan.

Outcome: Episodes of seizures, SVT, and A-fib/flu

 The coverage of calendar years in the national patient registers differ between the Nordic countries and may be insufficient to capture the complete history of seizures, SVT, or A- fib/flu at the individual level. Regarding seizures, recurrent seizures would be studied as primary seizures, if first seizure happened before start of registration.

 Truncation is a potential bias in the study meaning that we have no information on seizures, SVT, and A-fib/flu before the start of registration of each of these variables. For Denmark, Finland, and Sweden, information on seizures, SVT, and A-fib/flu is available from 1977, 1967, and 1987, respectively. Thus, the impact of truncation bias on results from these countries is limited. However, information on seizures, SVT, and A-fib/flu in Norway is only available from 2008, which limits the study period of the Norwegian data further and increases the risk of truncation bias.

 The present study aims at studying incident outcomes occurring in the community setting. To differentiate between incident and prevalent disease, we exclude persons with registered disease before baseline. In addition, we used surrogates of prevalence (i.e., redeemed prescriptions of antiepileptic medicine) to exclude persons with prevalent disease. To capture disease cases emerging in the community setting, we limit ascertainment to those sites of care where incident presentation would be present and use only primary diagnoses of hospitalization, as secondary diagnoses often represent conditions emerging during hospitalization.

Confounders

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 170

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 43 Epidemiology No.: EP07044.002

Information on the three confounders of severity of rhinitis, asthmatic status, and chronic urticaria status is challenging to obtain from the nationwide registers for many reasons. First of all, these conditions in general emerge in the primary health care sector where patients normally seek care from their general practitioner for treatment of symptoms. Second, truncation bias may, as with regard to the measures of exposure and outcome, also occur in relation to these three confounders in Norway as the patient register was established in 2008, making the follow-up period rather short. Third, different issues regarding codes and use of drugs for the three conditions have to be realized and discussed to understand the potential impact of residual confounders in the present study. We have discussed these issues with clinical experts and the essence of the discussion is given below.

 Severity of rhinitis is measured as use of immunotherapy and the binary measure of severity of rhinitis will not capture different levels of disease severity. Therefore, we expect the sensitivity of the measure to be rather low and severity of rhinitis to be underreported in the present study. However, we do not expect misclassification of those persons who received immunotherapy, as it is assumed that persons receiving immunotherapy also have severe rhinitis. A quantitative bias analysis examining the impact of misclassification of severity of rhinitis on the results will be performed.

 Identifying persons with asthma by use of register data is challenging due to the fact that the drugs used for treatment of asthma overlap with the treatment of chronic obstructive pulmonary disease (COPD). In order not to misclassify persons with COPD into the group of persons with asthma, we identify persons with asthma as those initiating asthma treatment before the age of 45 years. However, this means that for the part of the population aged 45 years or older, we will only have information on asthmatic status from the patient register in which hospitalized cases are registered. These cases are more likely to have severe asthma than those identified through the prescription register meaning that residual confounders may be more likely to occur among the population older than 45 years than those younger. A quantitative bias analysis examining the impact of misclassification of asthma on the results will be performed.

 Chronic urticaria is a very rare diagnosis given at highly specialized hospital units. This means that we expect a delay from onset of symptoms of chronic urticaria and the diagnosis of chronic urticaria of approximately 2-5 years. In the period between onset of disease and diagnosis, persons may receive antihistamines, including DL prescribed by the general practitioner, for their symptoms. Therefore, chronic urticaria may be underreported in the hospital register and the sensitivity of the measure is expected to be somewhat low. To get an impression of how important this misclassification will be in the present study, we will look at how many of the persons with a diagnosis of chronic urticaria have redeemed prescriptions of DL during the five years prior to the date of diagnosis. A quantitative bias analysis examining the impact of misclassification of chronic urticaria on the results will be performed.

 As in any association study, there is always the possibility of unmeasured confounders or residual confounding resulting from roughly categorized confounders that could affect the results of the study.

Variables used for inclusion and exclusion criteria

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 171

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 44 Epidemiology No.: EP07044.002

In Substudies 2B and 5B, seizure outcomes for which a drug overdose is also registered at the same hospitalization are excluded. In general, drug overdoses are likely to be underreported in the registries, as the registries often do not have sufficient detail to differentiate between overdoses from different types of drugs. Therefore, it is not possible to distinguish between overdoses from DL and overdoses from other drugs. Because overdoses of DL are rare compared to other drug overdoses, persons with any drug overdose registered during the same hospitalization as the seizure are excluded from analyses 2B and 5B. As a result, this study cannot determine whether the seizure outcomes are specifically associated with DL overdoses.

Sample size

Based on the assumptions made, the sample size is not adequate in the youngest age group for the assessment of the potential association between DL use and A-fib. Also, in this age group only relatively strong associations (IRR=8.2) can be properly assessed for the association between DL use and SVT.

9.10 Other aspects

N/A

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 172

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 45 Epidemiology No.: EP07044.002

10 PROTECTION OF HUMAN SUBJECTS

This is an observational study with no administration of any therapeutic or prophylactic agent. Subjects observed in this study will continue with the normal standard of care as provided by their personal physician. Pre-existing national register data will be the sole data source.

According to Danish, Finnish, Norwegian, and Swedish law, register-based studies can be carried out without consent from the data subjects where the processing takes place for the sole purpose of carrying out statistical or scientific studies of significant public importance and where such processing is necessary in order to carry out these studies. It is an absolute requirement that the publication of statistical or scientific results may never reveal the identity of individuals or otherwise compromise data subjects. We will obtain approval by the data agencies in the four countries before data management and data analyses will be performed.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 173

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 46 Epidemiology No.: EP07044.002

11 MANAGEMENT AND REPORTING OF ADVERSE EVENTS/ADVERSE REACTIONS

11.1 Definition of adverse event

An adverse event (AE) is defined as any untoward medical occurrence in a patient or clinical investigation subject administered a pharmaceutical product or who undergoes a protocol-specified procedure and which does not necessarily have to have a causal relationship with this treatment or procedure. Therefore, an AE can be any unfavorable and unintended sign (e.g., including an abnormal laboratory finding), symptom, or disease temporally associated with the use of a medicinal product or protocol-specified procedure, whether or not considered related to the medicinal product or protocol-specified procedure. Any worsening (i.e., any clinically significant adverse change in frequency and/or intensity) of a pre-existing condition that is temporally associated with the use of the Sponsor’s product, is also an AE.

Changes resulting from normal growth and development that do not vary significantly in frequency or severity from expected levels are not to be considered AEs. Examples of this may include, but are not limited to, teething, typical crying in infants and children, and onset of menses or menopause occurring at a physiologically appropriate time.

Sponsor's product includes any pharmaceutical product, biological product, device, diagnostic agent, or protocol-specified procedure, whether investigational (including placebo or active comparator product) or marketed, manufactured by, licensed by, provided by, or distributed by the Sponsor for human use.

Adverse events may occur during the course of the use of the Sponsor's product in studies or within the follow-up period specified by the protocol, or prescribed in clinical practice, from overdose (whether accidental or intentional), from abuse, and from withdrawal.

11.2 Definition of serious adverse event

Serious adverse event (SAE) means an adverse event which is fatal or life threatening, results in persistent or significant disability/incapacity, requires inpatient hospitalization, prolongation of existing inpatient hospitalization, a congenital anomaly/birth defect, or is another important medical event. Other important medical events that may not result in death, may not be life- threatening, or may not require hospitalization may be considered a SAE when, based upon appropriate medical judgment, they may jeopardize the patient or subject and may require medical or surgical intervention to prevent one of the other outcomes listed previously. Examples of such medical events include allergic bronchospasm requiring intensive treatment in an emergency room or at home, and blood dyscrasias or convulsions that do not result in inpatient hospitalization.

11.3 Other relevant safety information

The following events are considered important safety information and should be collected/reported using the same timeframes and reporting methods as SAEs:

 Exposure to product during pregnancy or lactation

EMA/623947/2012 22-Feb-2016

04CN88

05822K P A GE 174

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 4 7 E pi d e mi ol o g y N o.: E P 0 7044.002

 Lack of eff ect  Overdose

1 1. 4 Causality assess ment

A c a us alit y assess ment (attri b uti o n) m ust b e perfor med a n d recorded f or e ac h S A E a n d n o n -s er i o us A E ( NS AE) i n r el ati o ns hi p t o a S p o ns or's pr o d uct. D uri n g st u di es wit h dir ect p ati e nt c o ntact ( visits), t h e assess ment of c a us alit y will b e deter mined b y a n i nvestigat or w h o is a q u alifi e d p h ysici a n acc or di ng t o his/ h er b est cli nic al judg ment. Us e t h e f oll o wi n g crit eri a as guidance , as n ot all crit eri a m ust b e pr es e nt t o b e i n dic ati v e of c a us alit y t o a S p o ns or's pr o d uct: T h er e is e vi d e nc e of exposure t o t h e S p o ns or's pr o d uct; t h e t e m p or al sequence of t h e A E o ns et r el ati v e t o t h e a d mi nistr ati o n of t h e S p o ns or's pr o d uct is r e as o n a bl e; a n d t h e A E is m or e li k el y e x pl ai n e d b y t h e S p o ns or's pr o d uct t h a n b y a n ot her c a us e. I n st u di es wit h o ut dir ect p ati e nt c o nt act, t h e assess ment of c a us alit y w o ul d b e deter mined b y a n ot ati o n of attri b uti o n i n m e dic al r ec or ds. C a us alit y c a n b e assi g n e d b y t h e i nvestigat or or t h e S p o ns or. E x a m pl es i ncl u d e a dr ug -induced r as h t h at a n i nvestigat or attri b ut es t o a specific product, or a clinical notation that a product was discontinued because it caused inso mnia.

1 1. 5 Adverse event repo rti n g

T his is a n o n -i nterve nti o nal r egist er -b as e d st u d y. N o re p orti ng of i n di vi d u al c as es t o reg ulat ory agencies is planned because there is no access to individual patient/subject charts for this study, and n o s p ecific attri b uti o n of c as es is p ossi bl e. H o w e v er, if through t h e conduct of t his st u d y, a n investigator beco mes a ware of any S AE or NS AE that is attributed to any investigational or marketed pr o d uct manufactured b y M erck , t h e i nvestigat or will c o m pl et e a n A dverse E v e nt R e p ort F or m (A n n e x 1 B ) i n E n glis h a n d s u b mit S A Es wit hi n 2 4 h o urs a n d n o n -s eri o us A Es wit hi n 1 0 c al e n d ar d a ys vi a F ax t o M erc k Gl o b al S af ety at P P D ( U S), or t oll -fr e e f ax P P D ( e x -U S a n d U S a v ail a bilit y).

T h e e n d of st u dy re p ort, a n d a n y i nt eri m a n al ysis, will i ncl u d e aggregate listi n gs of all S A Es a n d N S A Es attri b uta bl e t o DL a n d will b e pr o vi d e d t o reg ulat ory a g e nci es as r e q uir e d. All i nt eri m a n d fi n al st u d y r e p orts will b e i ncl u d e d i n P eri o dic S af ety U p date Re p orts ( P S U Rs) a n d/ or Develop ment S af ety U p dat e Re p orts ( DS URs) u ntil c o m pl eti o n of t h e st u d y as r e q uir e d. S A Es a n d spontaneously reported NS AEs attributable to OT HER investigational or marketed products manufactured by Merck will be collected and reported to regulatory agencies as individual cases as r e q uir e d.

Beca use t his st u d y r eli es e xcl usi v el y o n auto mated d at a, it is v er y u nli k el y t h at a n i nvestigat or will beco me a ware of a suspected A E .

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K PAGE 175

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 48 Epidemiology No.: EP07044.002

12 PLANS FOR DISSEMINATING AND COMMUNICATING STUDY RESULTS

The results will be disseminated in working reports. In addition, the results will be submitted for publication in international peer-reviewed journals.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 176

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 49 Epidemiology No.: EP07044.002

13 REFERENCES

[Ref. 5.4: 00W4CX] Furu K, Wettermark B, Andersen M, Martikainen JE, Almarsdottir AB, Sorensen HT. The nordic countries as a cohort for pharmacoepidemiological research. Basic Clin Pharmacol 2009;106:86- 94. [Ref. 5.4: 00W4CY] Hauser WA, Beghi E. First seizure definitions and worldwide incidence and mortality. Epilepsia 2008;49(Suppl 1):8-12. [Ref. 5.4: 00W4D0] Thygesen L, Daasnes C, Thaulow I, Bronnum-Hansen H. Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving. Scan J Public Health 2011;39(Suppl 7):12-6. [Ref. 5.4: 00W4G3] Pedersen CB. The Danish Civil Registration System. Scan J of Public Health 2011;39(7 suppl):22-5. [Ref. 5.4: 03QPDQ] Glanz JM, McClure DL, Xu S, Hambidge SJ, Lee M, Kolczak MS, et al. Four different study designs to evaluate vaccine safety were equally validated with contrasting limitations. J Clin Epidemiol 2006;59:808-18. [Ref. 5.4: 03RSH6] Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim J-L, Reuterwall C, et al. External review and validation of the Swedish national inpatient register. BMC Public Health 2011;11:1-16. [Ref. 5.4: 03XNL7] Wilke T, Groth A, Mueller S, Pfannkuche M, Verheyen F, Linder R, et al. Incidence and prevalence of atrial fibrillation: an analysis based on 8.3 million patients. Europace 2013;15:486-93. [Ref. 5.4: 03XNM4] Orejarena LA, Vidaillet H Jr, DeStefano F, Nordstrom DL, Vierkant RA, Smith PN, et al. Paroxysmal supraventricular tachycardia in the general population. J Am Coll Cardiol 1998;31(1):150-7. [Ref. 5.4: 045RCQ] Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999 Jan;10(1):37-48. [Ref. 5.4: 045TYK] Kildemoes HW, Sorensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health. 2011 Jul;39(7 Suppl):38-41. [Ref. 5.4: 045TZL] Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health. 2011 Jul;39(7 Suppl):30-3. [Ref. 5.4: 045VCJ] Woodward M. Epidemiology: study design and data analysis. 2nd ed. Boca Raton (United States): Chapman & Hall/CRC; c1999. Chapter 8, Sample size determination; p. 381-426. [Ref. 5.4: 045VPW] The NORDCAN project [Internet]. Copenhagen: Association of Nordic Cancer Registries; c2009 [updated 2014 Dec 17]. Available from: http://www-dep.iarc.fr/NORDCAN/english/frame.asp. [Ref. 5.4: 045VR5] Cerminara C, El-Malhany N, Roberto D, Lo Castro A, Curatolo P. Seizures induced by desloratadine, a second-generation antihistamine: clinical observations. Neuropediatrics. 2013 Aug;44(4):222-4. [Ref. 5.4: 045VXZ] Kirkwood BR, Sterne JA. Essential Medical Statistics. 2nd ed. Malden (MA): Blackwell Publishing; 2003. [Ref. 5.4: 045W09] Pukkala E. Nordic biological specimen bank cohorts as basis for studies of cancer causes and control: quality control tools for study cohorts with more than two million sample donors and 130,000 prospective cancers. In: Dillner J, editor. Methods in biobanking. New York: Springer; 2011. p. 61-112. [Ref. 5.4: 045WWJ] Bryant E, Morganstein DR. Sample size determination for longitudinal surveys. Survey Research Methods Section, American Statistical Association Meeting; 1987 Aug 17-20; San Francisco, CA: 1987. [Ref. 5.4: 045WX9] Textor J, Hardt J, Knuppel S. DAGitty: a graphical tool for analyzing causal diagrams. Epidemiology. 2011 Sep;22(5):745. [Ref. 5.4: 0469WT] Bakken IJ, Gystad SO, Christensen OO, Huse UE, Laronningen S, Nygard J, et al. Comparison of data from the Norwegian Patient Register and the Cancer Registry of Norway. Tidsskr Nor Laegeforen. 2012 Jun 12;132(11):1336-40.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 177

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 50 Epidemiology No.: EP07044.002

[Ref. 5.4: 0469YG] Klaukka T. The Finnish database on drug utilization. Nor Epidemiol. 2001;11(1):19-22. [Ref. 5.4: 046B36] Sund R. Quality of the Finnish Hospital Discharge Register: a systematic review. Scand J Public Health. 2012 Aug;40(6):505-15. [Ref. 5.4: 046B3C] Wettermark B, Hammar N, Fored CM, Leimanis A, Otterblad Olausson P, Bergman U, et al. The new Swedish Prescribed Drug Register-- opportunities for pharmacoepidemiological research and experience from the first six months. Pharmacoepidemiol Drug Saf. 2007 Jul;16(7):726- 35. [Ref. 5.4: 046J05] Sulamaa A, Prami T, Sipila R, Linna M, Hahl J, Miettinen T, et al. [Description of the permit process for a national registry study - implementation of the current care guideline for diabetes mellitus as an example]. Suom Laakaril. 2014;46:3085-9a. Finnish. [Ref. 5.4: 04C0BQ] Frost L, Vestergaard P. Alcohol and risk of atrial fibrillation or flutter: a cohort study. Arch Intern Med. 2004 Oct 11;164(18):1993-8. [Ref. 5.4: 04C0BS] Frost L, Andersen LV, Vestergaard P, Husted S, Mortensen LS. Trend in mortality after stroke with atrial fibrillation. Am J Med. 2007 Jan;120(1):47-53. [Ref. 5.4: 04C0BZ] Rix TA, Riahi S, Overvad K, Lundbye-Christensen S, Schmidt EB, Joensen AM. Validity of the diagnoses atrial fibrillation and atrial flutter in a Danish patient registry. Scand Cardiovasc J. 2012 Jun;46(3):149- 53. [Ref. 5.4: 04C0C2] Schmidt M, Schmidt SA, Sandegaard JL, Ehrenstein V, Pedersen L, Sorensen HT. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol. 2015 Nov 17;7:449- 90. [Ref. 5.4: 04C0C7] Vestergaard M, Obel C, Henriksen TB, Christensen J, Madsen KM, Ostergaard JR, et al. The Danish National Hospital Register is a valuable study base for epidemiologic research in febrile seizures. J Clin Epidemiol. 2006 Jan;59(1):61-6.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 178

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 51 Epidemiology No.: EP07044.002

ANNEX 1. LIST OF STAND-ALONE DOCUMENTS

No. Document Reference No. Date Title 1A. N/A May 2014 Number of Desloratadine Users 1B. N/A July 2015 AE Report Form

Documents are available upon request.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 179

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 52 Epidemiology No.: EP07044.002

ANNEX 2. ENCEPP CHECKLIST FOR STUDY PROTOCOLS

Doc.Ref. EMA/540136/2009

ENCePP Checklist for Study Protocols (Revision 2, amended)

Adopted by the ENCePP Steering Group on 14/01/2013

The European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP) welcomes innovative designs and new methods of research. This Checklist has been developed by ENCePP to stimulate consideration of important principles when designing and writing a pharmacoepidemiological or pharmacovigilance study protocol. The Checklist is intended to promote the quality of such studies, not their uniformity. The user is also referred to the ENCePP Guide on Methodological Standards in Pharmacoepidemiology which reviews and gives direct electronic access to guidance for research in pharmacoepidemiology and pharmacovigilance.

For each question of the Checklist, the investigator should indicate whether or not it has been addressed in the study protocol. If the answer is “Yes”, the page number(s) of the protocol where this issue has been discussed should be specified. It is possible that some questions do not apply to a particular study (for example in the case of an innovative study design). In this case, the answer ‘N/A’ (Not Applicable) can be checked and the “Comments” field included for each section should be used to explain why. The “Comments” field can also be used to elaborate on a “No” answer.

This Checklist should be included as an Annex by marketing authorisation holders when submitting the protocol of a non-interventional post-authorisation safety study (PASS) to a regulatory authority (see the Guidance on the format and content of the protocol of non-interventional post- authorisation safety studies). Note, the Checklist is a supporting document and does not replace the format of the protocol for PASS as recommended in the Guidance and Module VIII of the Good pharmacovigilance practices (GVP).

Study title:

Association between use of desloratadine and risk of seizures, supraventricular tachycardia, and atrial fibrillation or flutter: A Nordic register-based study

Study reference number:

Study not registered

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 180

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 53 Epidemiology No.: EP07044.002

Section 1: Milestones Yes No N/A Page Number(s) 1.1 Does the protocol specify timelines for 1.1.1 Start of data collection2 14 1.1.2 End of data collection3 14 1.1.3 Study progress report(s) 1.1.4 Interim progress report(s) 1.1.5 Registration in the EU PAS register 1.1.6 Final report of study results. 14 Comments: This study does not include a study progress report or an interim report

Section 2: Research question Yes No N/A Page Number(s) 2.1 Does the formulation of the research question and objectives clearly explain: 15 2.1.1 Why the study is conducted? (e.g. to address an important public health concern, a risk identified in the risk management plan, an emerging safety issue) 2.1.2 The objective(s) of the study? 16 2.1.3 The target population? (i.e. population or subgroup to whom the study results are intended to be 19-22 generalised) 2.1.4 Which formal hypothesis(-es) is (are) to be 16 tested? 2.1.5 If applicable, that there is no a priori - hypothesis? Comments:

Section 3: Study design Yes No N/A Page Number(s)

3.1 Is the study design described? (e.g. cohort, case- control, randomised controlled trial, new or alternative design)

3.2 Does the protocol specify the primary and secondary (if applicable) endpoint(s) to be investigated?

3.3 Does the protocol describe the measure(s) of effect? (e.g. relative risk, odds ratio, deaths per 1000 person-years, absolute risk, excess risk, incidence rate ratio, hazard ratio, number needed to harm (NNH) per year)

Comments:

2 Date from which information on the first study is first recorded in the study dataset or, in the case of secondary use of data, the date from which data extraction starts. 3 Date from which the analytical dataset is completely available. EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 181

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 54 Epidemiology No.: EP07044.002

Section 4: Source and study populations Yes No N/A Page Number(s) 4.1 Is the source population described? 19 4.2 Is the planned study population defined in terms of: 19 4.2.1 Study time period? 28 4.2.2 Age and sex? 28 4.2.3 Country of origin? 24-25 4.2.4 Disease/indication? 29 4.2.5 Co-morbidity? 28 4.2.6 Seasonality? 4.3 Does the protocol define how the study population will be sampled from the source population? (e.g. 19-22 event or inclusion/exclusion criteria)

Comments:

Section 5: Exposure definition and measurement Yes No N/A Page Number(s) 5.1 Does the protocol describe how exposure is 22-24 defined and measured? (e.g. operational details for defining and categorising exposure) 5.2 Does the protocol discuss the validity of exposure 40-42 measurement? (e.g. precision, accuracy, prospective ascertainment, exposure information recorded before the outcome occurred, use of validation sub-study) 5.3 Is exposure classified according to time windows? 22-24 (e.g. current user, former user, non-use) 5.4 Is exposure classified based on biological 22-24 mechanism of action and taking into account the pharmacokinetics and pharmacodynamics of the drug? 5.5 Does the protocol specify whether a dose- dependent or duration-dependent response is 22-24 measured? Comments:

Section 6: Endpoint definition and measurement Yes No N/A Page Number(s) 6.1 Does the protocol describe how the endpoints are 24-26 defined and measured? 40-42 6.2 Does the protocol discuss the validity of endpoint measurement? (e.g. precision, accuracy, sensitivity, EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 182

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 55 Epidemiology No.: EP07044.002

Section 6: Endpoint definition and measurement Yes No N/A Page Number(s) specificity, positive predictive value, prospective or retrospective ascertainment, use of validation sub-study) Comments:

Section 7: Confounders and effect modifiers Yes No N/A Page Number(s) 7.1 Does the protocol address known confounders? 26-29 (e.g. collection of data on known confounders, methods of controlling for known confounders) 7.2 Does the protocol address known effect 26-29 modifiers? (e.g. collection of data on known effect modifiers, anticipated direction of effect) Comments: Regarding 7.2: we stratify the analysis by age and in a sensitivity analysis by country

Section 8: Data sources Yes No N/A Page Number(s) 8.1 Does the protocol describe the data source(s) used in the study for the ascertainment of: 8.1.1 Exposure? (e.g. pharmacy dispensing, general 29-30 practice prescribing, claims data, self-report, face-to-face interview, etc.) 8.1.2 Endpoints? (e.g. clinical records, laboratory markers 29-30 or values, claims data, self-report, patient interview including scales and questionnaires, vital statistics, etc.) 29-30 8.1.3 Covariates? 8.2 Does the protocol describe the information available from the data source(s) on: 8.2.1 Exposure? (e.g. date of dispensing, drug quantity, dose, number of days of supply prescription, daily dosage, 22-24 prescriber) 8.2.2 Endpoints? (e.g. date of occurrence, multiple event, 24-26 severity measures related to event) 8.2.3 Covariates? (e.g. age, sex, clinical and drug use 26-29 history, co-morbidity, co-medications, life style, etc.) 8.3 Is a coding system described for: 8.3.1 Diseases? (e.g. International Classification of 61 Diseases (ICD)-10) 8.3.2 Endpoints? (e.g. Medical Dictionary for Regulatory 61 Activities (MedDRA) for adverse events) 8.3.3 Exposure? (e.g. WHO Drug Dictionary, Anatomical Therapeutic Chemical (ATC)Classification System) 61 8.4 Is the linkage method between data sources described? (e.g. based on a unique identifier or other) 29-30 Comments:

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 183

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 56 Epidemiology No.: EP07044.002

Section 9: Study size and power Yes No N/A Page Number(s)

9.1 Is sample size and/or statistical power calculated? 30-34

Comments:

Section 10: Analysis plan Yes No N/A Page Number(s) 10.1 Does the plan include measurement of excess 35-40 risks?

10.2 Is the choice of statistical techniques described? 35-40

10.3 Are descriptive analyses included? 35-40

10.4 Are stratified analyses included? 35-40 35-40 10.5 Does the plan describe methods for adjusting for confounding?

10.6 Does the plan describe methods addressing 35-40 effect modification? Comments: Effect Modification will be evaluated using stratified analyses by age. Different methods for evaluating effect modification have not been discussed

Section 11: Data management and quality Yes No N/A Page control Number(s)

11.1 Is information provided on the management of 34-35 missing data? 11.2 Does the protocol provide information on data 34-35 storage? (e.g. software and IT environment, database maintenance and anti-fraud protection, archiving) 11.3 Are methods of quality assurance described? 34-35

11.4 Does the protocol describe possible quality 34-35 issues related to the data source(s)? 11.5 Is there a system in place for independent review of study results? Comments:

Section 12: Limitations Yes No N/A Page Number(s) EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 184

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 57 Epidemiology No.: EP07044.002

Section 12: Limitations Yes No N/A Page Number(s) 12.1 Does the protocol discuss: 12.1.1 Selection biases? 41-44 12.1.2 Information biases? (e.g. anticipated direction and magnitude of such biases, 41-44 validation sub-study, use of validation and external data, analytical methods)

12.2 Does the protocol discuss study feasibility? (e.g. 41-44 sample size, anticipated exposure, duration of follow-up in a cohort study, patient recruitment) 12.3 Does the protocol address other limitations? 41-44 Comments: The protocol also addresses the limitations in relation to register data and period of available data in the different countries potentially causing truncation bias

Section 13: Ethical issues Yes No N/A Page Number(s) 13.1 Have requirements of Ethics 45 Committee/Institutional Review Board approval been described? 13.2 Has any outcome of an ethical review procedure been addressed? 13.3 Have data protection requirements been 45 described? Comments:

Section 14: Amendments and deviations Yes No N/A Page Number(s) 14.1 Does the protocol include a section to document 13 future amendments and deviations? Comments:

Section 15: Plans for communication of study Yes No N/A Page results Number(s)

15.1 Are plans described for communicating study 48 results (e.g. to regulatory authorities)? 15.2 Are plans described for disseminating study 48 results externally, including publication?

Comments:

EMA/623947/2012 22-Feb-2016

04CN88

05822K P A GE 185

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 5 8 E pi d e mi ol o g y N o.: E P 0 7044.002

P P D

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K PAGE 186

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 59 Epidemiology No.: EP07044.002

ANNEX 3. ABOUT DESLORATADINE

Desloratadine (DL) is a drug used to treat allergies. It is marketed under several trade names, such as Aerius, Azomyr, and NeoClarityn in the EU, and in the rest of the world as, Claramax, Clarinex, Larinex, Dazit, Deselex, and Delot. It is an active metabolite of Loratadine, which is also on the market. DL is available as tablets (including orally disintegrating and extended release) and as syrup.

DL is a second-generation H1-antagonist. It is a tricyclic antihistamine, which has a selective and peripheral H1-antagonist action. It is an antagonist at histamine H1 receptors, and an antagonist at all subtypes of the muscarinic acetylcholine receptor. It has a long-lasting effect, and in moderate and low doses, does not cause drowsiness because it does not readily enter the central nervous system. Unlike other antihistamines, DL is also effective in relieving nasal congestion, particularly in patients with allergic rhinitis. Most common side-effects are fatigue, dry mouth, headache, and gastrointestinal disturbances. In the EU, DL is a prescription drug approved for the relief of symptoms associated with allergic rhinitis and urticaria in both adults and children. The Anatomical Therapeutic Classification (ATC) code for DL is R06AX27.

Table 3.1 Overview of DL formulations in the Nordic countries. Dates of approval are identical for all EU countries.

Formulation Date of approval

Film coated tablet, 5 mg 15 January 2001

Orodispersible tablet, 2.5 mg 23 April 2007

Orodispersible tablet, 5.0 mg 23 April 2007

Oral solution, 0.5mg/ml 23 April 2007

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 187

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 60 Epidemiology No.: EP07044.002

ANNEX 4. ICD-10 CODES AND ATC CODES

Diagnosis/drugs to be identified in Proposed classification system Proposed codes by clinicians registers

Acute drug intoxication and overdose ICD-10 T88.6 of drugs T88.7 T36 T50

Antiepileptic medicine ATC N03A N05BA

Asthma ICD-10 J45

ATC R03A and R03B

Atrial fibrillation or flutter ICD-10 I48

Brain Tumor (both malignant and ICD-10 C70 and C71 benign) D33

Chronic urticaria ICD-10 L50 (excluding acute urticaria: L50.8D) L56.3

Congenital pre-excitation syndrome ICD-10 I45.6

Desloratadine ATC R06AX27

Epilepsy ICD-10 G40

Head trauma ICD-10 S06

Loratadine ATC R06AX13

Malignant brain tumor ICD-10 C70 and C71

Seizure ICD-10 R56

Severe rhinitis (immunotherapy) ATC V01AA (excluding V01AA07)

Stroke (all types) ICD-10 I61 I62 I63 I64.9

Supraventricular arrhythmias ICD-10 I47.1

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 188

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 61 Epidemiology No.: EP07044.002

ANNEX 5. MINIMUM DETECTABLE INCIDENCE RATE RATIO CALCULATION

Using an alternative sample size formula from Woodward, we obtained similar results [Ref. 5.4: 045VCJ].

EMA/623947/2012 22-Feb-2016

04CN88

05822K P A GE 189

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 6 2 E pi d e mi ol o g y N o.: E P 0 7044.002

A N N E X 6 . DIRECTED ACYCLIC G R AP HS ( D A GS) F OR T HE STUDY ON DESL ORATADINE AND RI S K OF FI RST SEIZ U RE, FI RS T R ECURRENT SEIZ U RE, AT RI AL FIB R ILLATI ON OR FLUTTER, AND SUPRAVENTRICULAR TACHYCARDIA

Pri mary D A G meeting D at e 06 January 2015 Orga nizer P P D Attendees P P D

Annette Kj ær Ersbøll

P P D

M S D P P D

Cli nic al e x p erts P P D

Purpose To choose which confounding factors should be included and adjusted for in the study on desloratadine ( DL) and risk of seizures and supraventricular arrhyth mias. Supple mentary DA G meeting D at e 1 1 Marc h 2015 Attendees P P D

Annette Kj ær Ersbøll

P P D

Cli nic al e x p ert P P D Purpose To further discuss and clarify the relationship bet ween der matological factors such as chronic urticaria and the other variables included in the D A Gs developed at the pri mary D A G meeting. Soft ware D A Gitt y (http:// w w w.dagitty.net/)

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K PAGE 190

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 63 Epidemiology No.: EP07044.002

In the present study, we have used directed acyclic graphs (DAGs), also called causal diagrams for confounder selection. DAGs are a well-accepted methodology for using causal knowledge and a set of formal mathematical principles for selecting which variables to adjust for when performing association analyses [Ref. 5.4: 045RCQ]. They provide a systematic way to explore the relationships between the exposures, outcomes, and covariates (unidirectional, bidirectional, causal) and facilitate dealing with a large number of potential confounders. DAGs help make the assumptions underlying an analysis explicit. The selection of variables needed for confounder adjustment to obtain an unbiased estimate of the association under study is called the minimum sufficient adjustment set of confounders. We used the open source and freely available software DAGitty for the development of the DAGs [Ref. 5.4: 045WX9]. DAGitty helps the researcher visualize the structure of relevant variables for the association under study included in the DAG, as well as to identify the minimum sufficient adjustment sets available for confounder adjustment.

This document outlines the process used to develop the DAGs for the current study.

The purpose of the DAG Workshop on 06 January 2015 was to choose which confounding factors should be included and adjusted for in the study on DL and risk of seizures and supraventricular arrhythmias. This was done by going through all proposed potential confounders; and furthermore, adding potential confounders proposed by the external experts during the meeting. When discussing the previously proposed supraventricular arrhythmias outcome, it was brought up by the clinical experts whether it is reasonable to analyze this as a combined outcome (i.e., combining the diagnoses of A-fib/flu and SVT). The clinical experts indicated that pooling the diagnoses together as one outcome would not be clinically optimal and that it would be more appropriate to separate the diagnoses into two outcomes (i.e., A-fib/flu as one outcome and SVT as another outcome). Due to this discussion, we have chosen to separate SVT and A-fib/flu; and therefore, we developed three DAGs: one for seizures and recurrent seizures, one for A-fib/flu, and one for SVT.

During the DAG development process, we discussed the relevant potential confounders in the left columns of Tables 1 and 3 in the order in which they are listed. Key issues discussed are described in the “Discussion” column. It should be noted that this may have influenced the discussion of the relation between two potential confounders meaning that an association between two potential confounders was discussed when the second potential confounder was added to the DAG (e.g., the association between chronic obstructive pulmonary disease (COPD) and smoking was not only discussed when COPD was included in the DAG, but also at the time when smoking was added). After the DAG meeting, the minutes of the meeting were circulated to all participants for further comments and/or suggestions. This may also have influenced the order in which the confounders were listed in Tables 1 and 3. Furthermore, we arranged a meeting with a dermatologist to get his suggestions for the DAG. We were especially interested in how to include chronic urticaria (one of the indications for desloratadine), how chronic urticaria was related to other factors already in the DAGs, and whether we needed to include additional factors or associations in the DAGs, which might be different than for the allergic rhinitis indication.

We used the open source and freely available software DAGitty for the development of the DAGs [Ref. 5.4: 045WX9]. DAGitty helps the researcher visualize the structure of relevant factors for the association under study included in the DAG, as well as to identify the minimum sufficient adjustment sets available for confounder adjustment to provide an unbiased estimate of the association of interest.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 191

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 64 Epidemiology No.: EP07044.002

After the DAG meeting and discussions, the final directed paths that were drawn from each factor to the other factors in the DAGs were listed in the right column. Furthermore, we listed the different minimum sufficient adjustment sets identified by use of DAGitty for the three DAGs developed. We selected the final minimum sufficient adjustment set that will be used to adjust for confounding factors in the association studies based on whether we found that a given combination of confounders would be obtainable from the registers. Since there is no information on smoking or type 1 allergy available in the population registers, minimum sufficient adjustment sets including these two potential confounders were not eligible for selection. In cases where multiple minimum sufficient adjustment sets were candidates for confounder adjustment, we chose the minimum sufficient adjustment set where we found that the combination of confounders had the highest possible validity and quality in the population registers. Operational definitions of the variables in the minimum sufficient adjustment set are provided in the study protocol.

Afterwards, the DAGs were updated according to the final confirmed DAG meeting minutes and the final results were added to the protocol.

DAG for the association between DL use and first seizure and first recurrent seizure

It was decided that the DAG for the first seizure and the first recurrent seizure should be the same.

Table 6.1, Figure 6.1, and Table 6.2 provide information on the results of the discussion of the DAG concerning the association between DL use and first seizure and first recurrent seizure. Table 6.1 lists the potential confounders discussed during the development of the DAG, Figure 6.1 shows the final developed DAG, and Table 6.2 provides the results of the DAG in terms of the minimum sufficient adjustment sets of confounders to include in the association analyses to obtain a confounder adjusted estimate of the association between DL use and first seizure and first recurrent seizure. Finally, the specific minimum sufficient adjustment set that will be used to adjust for confounding of the association between DL and seizures was identified and included the following confounders: age, sex, country, calendar year, disease severity of rhinitis, asthmatic status, and chronic urticaria.

Table 6.1 Discussion of potential confounders of the association between DL use and first seizure and first recurrent seizure

Potential Discussion List of directed paths (head-to-tail arrows) confounders Age Age affects the use of DL. There is a directed path from age to: Inflammatory disease Disease severity of rhinitis Infections Asthmatic status COPD DL Smoking Metastatic cancer Drug and alcohol abuse Drug overdose of drugs other than DL Seizures

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 192

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 65 Epidemiology No.: EP07044.002

Potential Discussion List of directed paths (head-to-tail arrows) confounders Stroke Thyroidism Diabetes Sex Sex is associated with disease severity of rhinitis, There is a directed path from sex to: as the prevalence of rhinitis is higher in males Smoking than females. Drug and alcohol abuse There are sex differences in the incidence of Stroke stroke and smoking. Inflammatory disease There are sex differences in use of DL. Diabetes Thyroidism Chronic urticaria Disease severity of rhinitis DL Country of There are probably country differences in use of There is a directed path from country of residence DL. residence to: (Denmark, Country differences in smoking. DL Finland, Norway, Smoking Sweden Calendar year Preliminary data from the prescription registers There is a directed path from calendar year to: demonstrate that the number of unique patients Inflammatory disease treated with DL varies by calendar year in each Diabetes country, initially increasing and then showing some variation. Some, but not all of the variation Chronic urticaria may be due to differences in seasonality/pollen. Type 1 allergy Year is also associated with seasonality in that Seasonality the influence of pollen may differ from year to DL year. Smoking There is a temporal trend in smoking, diabetes, Stroke inflammatory disease, type 1 allergy, and stroke. Seasonality There is a seasonal trend in DL use. There is a directed path from seasonality to: The effect of seasonality could change from year DL to year. Infections It was suggested that pollen level could be used Disease severity of rhinitis as a proxy for the effect of seasonality. One of Type 1 allergy the experts explained that this is not simple because pollen level may vary from area to area Asthmatic status and day to day. Pollen may also have a different influence from year to year for the same individuals. Thus, information on pollen level will not be included. Seasonality will be used as a proxy measure. There is a seasonal trend in disease severity of rhinitis and asthmatic status. Disease severity Disease severity of rhinitis affects the use of DL. There is a directed path from disease severity of of rhinitis Very strong association. DL is a prescription drug rhinitis to: and if you have severe rhinitis and use DL antihistamines regularly, then you might choose Asthmatic status a prescription drug as DL, as this drug is subsidized. The severity of rhinitis therefore influences whether you prefer DL over over-the-

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 193

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 66 Epidemiology No.: EP07044.002

Potential Discussion List of directed paths (head-to-tail arrows) confounders counter antihistamines. Disease severity of rhinitis affects asthmatic status. Asthmatic status Asthmatic status affects DL use. There is a directed path from asthmatic status to: DL Infections Chronic Age affects risk of COPD. There is a directed path from COPD to: obstructive Inflections pulmonary disease (COPD) Comorbidity Could be measured as history of hospital Not relevant admissions or as an index (e.g., the Charlson index). Comment (MSD): This is a non-specific indicator. Really a proxy for something else. Should be specified what part of comorbidity on top of the other indicators discussed should be included. Specific comorbidities are better. Comment (expert): Agree that Charlson should not be used. Conclusion: Do not include comorbidity index, number of admissions, or number of hospital contacts. Smoking Smoking has an effect on seizures, COPD, There is a directed path from smoking to: asthmatic status, stroke, and metastatic cancer. Metastatic cancer Seizures Stroke Asthmatic status COPD Metastatic Metastatic cancer can increase risk of seizures. There is a directed path from metastatic cancer cancer to: Seizures Use of other Comment (MSD): Most drugs are not associated Not relevant drugs with seizures at normal doses. In addition, this would be extremely difficult to operationalize. Drug-drug interactions have been shown but these interactions are generally very small. It is more probable that high drug doses may influence high use of other drugs. Conclusion: Do not include use of other drugs. Drug overdose Overdose of drugs other than DL affects the risk There is a directed path from drug overdose of (of drugs other of seizures. drugs other than DL to: than DL) Comment (MSD): One of the most common Seizures reasons of new seizures is overdose of alcohol and cocaine. The coding of overdose and the reason for overdose is not easily found in register studies. May be too complicated to operationalize. Comment (expert): Not specific enough codes for what kind of drug overdoses. Suggests using all EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 194

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 67 Epidemiology No.: EP07044.002

Potential Discussion List of directed paths (head-to-tail arrows) confounders drug overdoses as one variable (binary). Comment (expert): Very unreliable information in the registers. Persons with code for overdose (alcohol, drug, others) will probably be reliable, but there will be underreporting and it is difficult to differentiate between different types of overdose. Comment (MSD): The problem gets very messy, because drug overdose due to DL is part of this group and we may not be able to separate this specific group. Conclusion: Combine any overdose as one binary variable. Drug and alcohol Drug and alcohol abuse affects risk of seizures There is a directed path from drug and alcohol abuse and overdose of drugs other than DL. abuse to: Seizures Drug over dose of other factors than DL Diabetes Diabetes affects use of hypoglycemic agents and There is a directed path from diabetes to: the risk of stroke Stroke Both diabetes type 1 and type 2 should be Hypoglycemic agents included.

Hypoglycemic Use of hypoglycemic agents affects the risk of There is a directed path from hypoglycemic agents (oral anti- seizures. agents to: diabetics, insulin) Seizures Stroke Stroke affects the risk of seizures. We will not There is a directed path from stroke to: differentiate between hemorrhagic or ischemic Seizures stroke as both increase the risk of seizures. Inflammatory disease Chronic urticaria Chronic urticaria is associated with use of very There is a directed path from chronic urticaria to: high doses (much higher than the usual dose) of DL DL (Guidelines recommend up to 4 times the standard dose). Associated with sex. Prevalence of urticaria/chronic urticaria is higher in females than males. There is a temporal trend in prevalence of diagnosed chronic urticaria (increases) (i.e., calendar year affects chronic urticaria). Thyroidism Thyroidism is associated with risk of stroke. There is a directed path from thyroidism to: (Thyroidism Age affects risk of thyroidism. Stroke includes both There are sex differences in risk of thyroidism. hypo- and hyperthyroidism). Unspecific Unspecific autoimmune disease is associated There is a directed path from unspecific autoimmune with risk of chronic urticaria, thyroidism, and autoimmune disease to: disease diabetes. Diabetes Thyroidism Chronic urticaria Inflammatory Inflammatory disease is associated with risk of There is a directed path from inflammatory disease chronic urticaria. disease to:

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 195

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 68 Epidemiology No.: EP07044.002

Potential Discussion List of directed paths (head-to-tail arrows) confounders Stroke increases risk of inflammatory disease. Chronic urticaria Sex and age are associated with risk of inflammatory disease. Infections Infections can affect risk of chronic urticaria. There is a directed path from infections to: There are sex differences in infections. Chronic urticaria COPD, asthmatic disease, and seasonality are associated with risk of infections. Type 1 allergy Type 1 allergy can affect risk of chronic urticaria There is a directed path from type 1 allergy to: and rhinitis. Disease severity of rhinitis Seasonality affects risk of type 1 allergy. Asthmatic status Chronic urticaria

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 196

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 69 Epidemiology No.: EP07044.002

Figure 6.1 DAG for the association between DL use and first seizure and first recurrent seizure.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 197

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 70 Epidemiology No.: EP07044.002

Table 6.2 Minimum sufficient adjustment sets for the DAG developed (Figure 6.1). Potential confounders Minimum sufficient adjustment sets 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Age X X X X X X X X X X X XXXXXXXX Sex X XX X X XX X X X X XXXXXX Country X X X X X X X X X Calendar year X X X X X X X X X X X XXXX Seasonality X X X X X X X X X X X Disease severity of rhinitis X X X X X X X X X X Asthmatic status X X X X X X X X X X X COPD X X X X Smoking X X X X X X X X X X Metastatic disease Drug overdose Drug and alcohol abuse X X Diabetes X X X X X X X Hypoglycemic agents X X Stroke X X X X X X XXX Chronic urticaria X X X X Thyroidism (hypo- X X X X X /hyperthyroidism) Unspecific autoimmune disease X X X X X X Inflammatory disease X X X X X X Infections X X X XX Type 1 allergy X X X X X X X X X X Light grey background = confounder information not available in population registers (e.g., self-reported lifestyle factors). Dark grey background = selected minimum sufficient adjustment set (number 5) and the alternative set (number 8).

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 198

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 71 Epidemiology No.: EP07044.002

The minimum sufficient adjustment set chosen for analysis

Table 6.2 shows the possible minimum sufficient sets available. The set used to adjust for confounding of the association between DL and seizures included the following confounders: age, sex, country, calendar year, disease severity of rhinitis, asthmatic status, and chronic urticaria. The selected set was chosen as it was the only one not including smoking and type 1 allergy.

Some rare diseases are treated with very high doses of desloratadine

It was brought up during the discussion that some rare diseases are treated with very high doses of DL (e.g., chronic urticaria). It is assumed that dermatologists prescribe DL in high doses for these rare diseases. A dermatologist was consulted on this matter to be sure we included the relevant diseases and relationships with other potential confounders in the DAG. We should consider how to take the high use of DL among patients with chronic urticaria into account when analyzing the data (e.g., in a supplementary analysis where we exclude persons who have redeemed high doses of DL or have a diagnosis of chronic urticaria).

Recurrent seizure

The same confounders should be included as for first seizure. Change in drug use due to seizure could influence the drug prescriptions.

In conclusion, it would complicate things to include other potential confounders also in light of the low number of outcomes in this setting.

DAG for the association between DL use and A-fib/flu

Table 6.3, Figure 6.2, and Table 6.4 provide information on the results of the discussion of the DAG concerning the association between DL use and A-fib/flu. Table 6.3 lists the potential confounders discussed during the development of the DAG, Figure 6.2 shows the final developed DAG, and Table 6.3 provides the results of the DAG in terms of the minimum sufficient adjustment sets of confounders to include in the association analyzes to obtain a confounder adjusted estimate of the association between DL use and first diagnosis of A-fib/flu. Finally, the specific minimum sufficient adjustment set that will be used to adjust for confounding of the association between DL and A- fib/flu was identified and included the following confounders: age, sex, country, calendar year, disease severity of rhinitis, asthmatic status, and chronic urticaria.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 199

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 72 Epidemiology No.: EP07044.002

Table 6.3 Discussion of potential confounders of the association between DL use and A-fib/flu.

Potential Discussion List of directed paths (head-to-tail arrows) confounders Age Age affects the use of DL. There is a directed path from age to: Inflammatory disease Drug and alcohol abuse Hypertension A-fib/flu Pre-stroke Diabetes Structural heart disease Thyroidism Disease severity of rhinitis COPD DL Infections Smoking Asthmatic status Drug overdose of drugs other than DL Sex There are sex differences in risk of A-fib/flu. There is a directed path from sex to: Sex is associated with disease severity of rhinitis, Smoking as prevalence of rhinitis is higher in males than Inflammatory disease females. Hypertension There are sex differences in smoking and the A-fib/flu incidence of stroke, inflammatory disease, hypertension, structural heart disease and Pre-stroke thyroidism. Structural heart disease There are sex differences in DL use. Thyroidism Chronic urticaria Disease severity of rhinitis DL Drug and alcohol abuse Diabetes Country There are probably country differences in use of There is a directed path from country to: DL. Smoking Country differences in smoking. DL Calendar year Preliminary data from the prescription registries There is a directed path from calendar year to: demonstrate that the number of unique patients Smoking treated with DL varies by calendar year in each Inflammatory disease country, initially increasing and then showing some variation. Some, but not all of the A-fib/flu variation may be due to differences in Pre-stroke seasonality/pollen. Diabetes Year is also associated with seasonality in that Chronic urticaria the influence of pollen may differ from year to Type 1 allergy year. Seasonality There is a temporal trend in A-fib/flu, smoking, DL diabetes, inflammatory disease, type 1 allergy, and stroke. EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 200

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 73 Epidemiology No.: EP07044.002

Potential Discussion List of directed paths (head-to-tail arrows) confounders Seasonality There is a seasonal trend in DL use. There is a directed path from seasonality to: There are seasonal trends in A-fib/flu, disease DL severity of rhinitis, asthmatic status, infections, Infections inflammatory disease, type 1 allergy, and stroke. A-fib/flu The effect of seasonality could change from year Disease severity of rhinitis to year. Type 1 allergy It was suggested that pollen level could be used as a proxy for the effect of seasonality. One of Asthmatic status the experts explained that this is not simple in that pollen level may vary from area to area and day to day. Pollen may also have a different influence from year to year for the same individuals. Thus, information on pollen level will not be included. Seasonality will be used as a proxy measure. Disease severity Disease severity of rhinitis affects the use of DL. There is a directed path from disease severity of of rhinitis Very strong association. DL is a prescription drug rhinitis to: and if you have severe rhinitis and use DL antihistamines regularly, then you might choose Asthmatic status a prescription drug as DL, as this drug is subsidized. The severity of rhinitis therefore influences whether you prefer DL over over-the- counter antihistamines. Disease severity of rhinitis affects asthmatic status. Asthmatic status Asthmatic status affects DL use. There is a directed path from asthmatic status to: DL Infections COPD Age affects risk of COPD. There is a directed path from COPD to: COPD affects risk of A-fib/flu and infections. A-fib/flu Infections Comorbidity Could be measured as history of hospital Not relevant admissions or as an index (e.g., the Charlson index). Comment (MSD): A non-specific indicator. Really a proxy for something else. Should be specified what part of comorbidity on top of the other indicators discussed should be included. Specific comorbidities are better. Comment (expert): Agree that Charlson should not be used. Conclusion: Do not include comorbidity index, number of admissions, or number of hospital contacts. Smoking Smoking has an effect on COPD, asthmatic There is a directed path from smoking to: status, hypertension, stroke, and structural heart Hypertension disease. Pre-stroke Structural heart disease Asthmatic status COPD

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 201

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 74 Epidemiology No.: EP07044.002

Potential Discussion List of directed paths (head-to-tail arrows) confounders Inflammatory Inflammatory disease is associated with risk of A- There is a directed path from inflammatory disease fib/flu, structural heart disease, and chronic disease to: urticaria. A-fib/flu Stroke increases risk of inflammatory disease. Structural heart disease Sex and age are associated with risk of Chronic urticaria inflammatory disease. Infections Infections can affect risk of A-fib/flu and chronic There is a directed path from infections to: urticaria. A-fib/flu There are sex differences in infections. Chronic urticaria COPD, asthmatic disease, and seasonality can affect risk of infections. Hypertension Hypertension affects risk of A-fib/flu and There is a directed path from hypertension to: increases risk of structural heart disease and A-fib/flu stroke. Pre-stroke Structural heart disease Antihypertensive treatment Drug overdose (of Overdose of drugs other than DL may induce A- There is a directed path from drug overdose of drugs other than fib/flu. drugs other than DL to: DL) Conclusion: Include any overdose as one binary A-fib/flu variable. Drug and alcohol Drug and alcohol abuse affects risk of A-fib/flu There is a directed path from drug and alcohol abuse and overdose of drugs other than DL. abuse to: A-fib/flu Drug overdose of drugs other than DL Thyroidism (hypo- Thyroidism is associated with of A-fib/flu, stroke, There is a directed path from thyroidism to: /hyperthyroidism) and structural heart disease. A-fib/flu and synonymous Age affects risk of thyroidism. Pre-stroke diagnoses (e.g., There are sex differences in risk of thyroidism. Grave’s disease, Structural heart disease thyrotoxicosis) Diabetes Diabetes affects risk of A-fib/flu, hypertension, There is a directed path from diabetes to: stroke, and structural heart disease. Hypertension Both diabetes type 1 and type 2 should be A-fib/flu included. Pre-stroke Structural heart disease Structural heart Structural heart disease affects risk of A-fib/flu There is a directed path from structural heart disease: Left and stroke. disease to: ventricular A-fib/flu hypertrophy, left Pre-stroke ventricular systolic dysfunction, CHF Stroke Stroke does not increase the risk of any factors There is a directed path from pre-stroke to: included in the DAG, but is the effect of factors Inflammatory disease included. Stroke is divided into two types: 1) pre-stroke that occurs independently of A-fib/flu and 2) post-stroke that is caused by A-fib/flu. OBS: A-fib/flu affects risk of post-stroke Chronic urticaria Chronic urticaria is associated with use of very There is a directed path from chronic urticaria to:

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 202

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 75 Epidemiology No.: EP07044.002

Potential Discussion List of directed paths (head-to-tail arrows) confounders high doses (much higher than the recommended DL dose) of DL. Associated with sex. Prevalence of urticaria/chronic urticaria is higher in females than males. There is a temporal trend in prevalence of diagnosed chronic urticaria (increases) (i.e., calendar year affects chronic urticaria). Inflammatory disease and infections can increase the risk of chronic urticaria. Unspecific Unspecific autoimmune disease is associated There is a directed path from unspecific autoimmune with risk of chronic urticaria, thyroidism, and autoimmune disease to: disease diabetes. Diabetes Thyroidism Chronic urticaria Type 1 allergy Type 1 allergy can increase the risk of chronic There is a directed path from type 1 allergy to: urticaria and rhinitis. Disease severity of rhinitis Seasonality affects risk of type 1 allergy. Asthmatic status Chronic urticaria Antihypertensive Hypertension affects the use of antihypertensive There is a directed path from antihypertensive treatment treatment. treatment to: Use of antihypertensive treatment increases the Chronic urticaria risk of chronic urticaria.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 203

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 76 Epidemiology No.: EP07044.002

Figure 6.2 DAG for the association between DL use and A-fib/flu.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 204

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 77 Epidemiology No.: EP07044.002

Table 6.4 Minimum sufficient adjustment sets for the DAG developed (Figure 6.2). Potential Minimum sufficient adjustment sets confounders 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Age X X X XXXX X X X XXXXXXXX Sex X X X XXXX X X X XXXXXXXX Country X X X X X Calendar year X X X XXXX X X X XXXXXXXX Seasonality X XX XXXX X XX XXXXXXXX Disease severity of X X X X X X rhinitis Asthmatic status X X X X X X X X XXXX COPD X X X X X X X Smoking X X X X X X X X X X X Inflammatory X X X XXXX X X X X XXX disease Infections X X X XXX X X X XXXXX Hypertension X X X X X XXX Drug overdose Drug and alcohol abuse Thyroidism (hypo- X X X X X X X X /hyperthyroidism) Diabetes X X X X X X X X Structural heart X X disease Stroke Chronic urticaria X X X X X Unspecific X X X X X X autoimmune disease Type 1 allergy X X X X X X X Antihypertensive X X X XXX treatment Light grey background = confounder information not available in population registers (e.g., self-reported lifestyle factors). Dark grey background = selected minimum sufficient adjustment set (number 8) and the alternative set (number 1).

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 205

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 78 Epidemiology No.: EP07044.002

The minimum sufficient adjustment set chosen for analysis

Table 6.4 shows the possible minimum sufficient sets available. The set used to adjust for confounding of the association between DL and A-fib/flu included the following confounders: age, sex, country, calendar year, disease severity of rhinitis, asthmatic status, and chronic urticaria. The selection of which minimum sufficient adjustment set to use for confounder adjustment was based on our evaluation of the validity and quality of confounder information available from the registers. The combined set of confounders in the final selected minimum adjustment set was found to have the highest possible validity and quality.

DAG for the association between DL use and SVT

The DAG was almost similar for this outcome as for the A-fib/flu outcome. The following were the only changes:

No arrows from the following variables to SVT:

 Inflammatory diseases  Hypertension  COPD  Calendar year  Seasonality  Diabetes

In this DAG, we do not differentiate between pre and post stroke, as this is not relevant for the SVT outcome and since the arrow points from stroke to SVT.

Figure 6.3 shows the final developed DAG and Table 6.3 provides the results of the DAG in terms of the minimum sufficient adjustment set of confounders to include in the association analyzes to obtain a confounder adjusted estimate of the association between DL use and first diagnosis of SVT. In Table 6.5, the minimum sufficient sets in the DAG of the association between DL use and SVT are listed. Finally, the specific minimum sufficient adjustment set that will be used to adjust for confounding of the association between DL and SVT was identified and included the following confounders: age, sex, country, calendar year, disease severity of rhinitis, asthmatic status, and chronic urticaria.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 206

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 79 Epidemiology No.: EP07044.002

Figure 6.3 DAG for the association between DL use and SVT.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 207

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 80 Epidemiology No.: EP07044.002

Table 6.5 Minimum sufficient adjustment sets for the DAG developed (Figure 6.3). Potential confounders Minimum sufficient adjustment sets 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Age X X X X X XXX X X X XXXXXX Sex X X X X X XXX X X X XXXXXX Country X X X X X Calendar year X X X X X XXX X X X XXXXX Seasonality X X X X X X X X X XX Disease severity of rhinitis X X X X XX Asthmatic status X X X X X X X X X XX COPD X X X X Smoking X X X X X X X Inflammatory disease X X X X X XXX X X X X Infections X X X X X X X XXX Hypertension X X X X X X Drug overdose Drug and alcohol abuse Thyroidism (hypo-/hyperthyroidism) X X X X X X X X X Diabetes X X X X X X Structural heart disease X X X X X Stroke X X X X X Chronic urticaria X X X X X X X X Unspecific autoimmune disease X X X X X X Type 1 allergy X X X X X X X X X X Antihypertensive treatment X X X X X X Light grey background = confounder information not available in population registers (e.g., self-reported lifestyle factors). Dark grey background = selected minimum sufficient adjustment set (number 9) and the alternative set (number 3).

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 208

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 81 Epidemiology No.: EP07044.002

The minimum sufficient adjustment set chosen for analysis

Table 6.5 shows the possible minimum sufficient sets available. The set used to adjust for confounding of the association between DL and SVT included the following confounders: age, sex, country, calendar year, disease severity of rhinitis, asthmatic status, and chronic urticaria. The selection of which minimum sufficient adjustment set to use for confounder adjustment was based on our evaluation of the validity and quality of confounder information available from the registers. The combined set of confounders in the final selected minimum adjustment set was found to have the highest possible validity and quality.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 209

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 82 Epidemiology No.: EP07044.002

ANNEX 7. ADMINISTRATIVE AND REGULATORY DETAILS

Confidentiality:

Confidentiality of Data

By signing this protocol, the investigator affirms to the Sponsor that information furnished to the investigator by the Sponsor will be maintained in confidence, and such information will be divulged to the Institutional Review Board, Ethics Review Committee or similar or expert committee; affiliated institution and employees, only under an appropriate understanding of confidentiality with such board or committee, affiliated institution and employees. Data generated by this study will be considered confidential by the investigator, except to the extent that it is included in a publication as provided in the Publications section of this protocol.

Confidentiality of Subject Records

By signing this protocol, the investigator agrees that the Sponsor (or Sponsor representative), Institutional Review Board/Independent Ethics Committee (IRB/IEC), or Regulatory Agency representatives may consult and/or copy study documents in order to verify worksheet/case report form data. By signing the consent form, the subject agrees to this process. If study documents will be photocopied during the process of verifying worksheet/case report form information, the subject will be identified by unique code only; full names/initials will be masked prior to transmission to the Sponsor.

By signing this protocol, the investigator agrees to treat all subject data used and disclosed in connection with this study in accordance with all applicable privacy laws, rules, and regulations.

Confidentiality of Investigator Information

By signing this protocol, the investigator recognizes that certain personal identifying information with respect to the investigator, and all sub-investigators and study site personnel, may be used and disclosed for study management purposes, as part of a regulatory submissions, and as required by law. This information may include:

 Name, address, telephone number and e-mail address;  Hospital or clinic address and telephone number;  Curriculum vitae or other summary of qualifications and credentials; and  Other professional documentation.

Consistent with the purposes described above, this information may be transmitted to the Sponsor, and subsidiaries, affiliates and agents of the Sponsor, in your country and other countries, including countries that do not have laws protecting such information. Additionally, the investigator’s name and business contact information may be included when reporting certain serious adverse events to regulatory agencies or to other investigators. By signing this protocol, the investigator expressly consents to these uses and disclosures.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 210

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 83 Epidemiology No.: EP07044.002

If this is a multicenter study, in order to facilitate contact between investigators, the Sponsor may share an investigator’s name and contact information with other participating investigators upon request.

Administrative:

Compliance with Financial Disclosure Requirements

Financial Disclosure requirements are outlined in the US Food and Drug Administration Regulations, Financial Disclosure by Clinical Investigators (21 CFR Part 54). It is the Sponsor's responsibility to determine, based on these regulations, whether a request for Financial Disclosure information is required. It is the investigator's/subinvestigator's responsibility to comply with any such request.

The investigator/subinvestigator(s) agree, if requested by the Sponsor in accordance with 21 CFR Part 54, to provide his/her financial interests in and/or arrangements with the Sponsor to allow for the submission of complete and accurate certification and disclosure statements. The investigator/subinvestigator(s) further agree to provide this information on a Certification/Disclosure Form, commonly known as a financial disclosure form, provided by the Sponsor or through a secure password-protected electronic portal provided by the Sponsor. The investigator/subinvestigator(s) also consent to the transmission of this information to the Sponsor in the United States for these purposes. This may involve the transmission of information to countries that do not have laws protecting personal data.

Compliance with Law, Audit and Debarment

By signing this protocol, the investigator agrees to conduct the study in an efficient and diligent manner and in conformance with this protocol; generally accepted standards of Good Pharmacoepidemiology Practice and all applicable federal, state and local laws, rules and regulations relating to the conduct of the study.

The investigator also agrees to allow monitoring, audits, Institutional Review Board/Independent Ethics Committee review and regulatory agency inspection of study-related documents and procedures and provide for direct access to all study-related source data and documents.

The investigator agrees not to seek reimbursement from subjects, their insurance providers or from government programs for procedures included as part of the study reimbursed to the investigator by the Sponsor.

The Investigator shall prepare and maintain complete and accurate study documentation in compliance with Good Pharmacoepidemiology Practice, standards and applicable federal, state and local laws, rules and regulations; and, for each subject participating in the study, provide all data, and, upon completion or termination of the study, submit any other reports to the Sponsor as required by this protocol or as otherwise required pursuant to any agreement with the Sponsor.

Study documentation will be promptly and fully disclosed to the Sponsor by the investigator upon request and also shall be made available at the investigator’s site upon request for inspection, copying, review and audit at reasonable times by representatives of the Sponsor or any regulatory agencies. The investigator agrees to promptly take any reasonable steps that are requested by the

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 211

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 84 Epidemiology No.: EP07044.002

Sponsor as a result of an audit to cure deficiencies in the study documentation and worksheets/case report forms.

The investigator must maintain copies of all documentation and records relating to the conduct of the study in compliance with all applicable legal and regulatory requirements. This documentation includes, but is not limited to, the protocol, worksheets/case report forms, advertising for subject participation, adverse event reports, subject source data, correspondence with health authorities and IRBs/ERCs, consent forms, investigator’s curricula vitae, monitor visit logs, laboratory reference ranges, laboratory certification or quality control procedures and laboratory director curriculum vitae. All study documents shall be made available if required by relevant health authorities. The investigator must consult with the Sponsor prior to discarding study and/or subject files.

The investigator will promptly inform the Sponsor of any regulatory agency inspection conducted for this study.

Persons debarred from conducting or working on studies by any court or regulatory agency will not be allowed to conduct or work on this Sponsor’s studies. The investigator will immediately disclose in writing to the Sponsor if any person who is involved in conducting the study is debarred or if any proceeding for debarment is pending or, to the best of the investigator’s knowledge, threatened.

In the event the Sponsor prematurely terminates a particular study site, the Sponsor will promptly notify that site’s IRB/IEC.

According to European legislation, a Sponsor must designate an overall coordinating investigator for a multi-center study (including multinational). When more than one study site is open in an EU country, Merck, as the Sponsor, will designate, per country, a national principal coordinator (Protocol CI), responsible for coordinating the work of the principal investigators at the different sites in that Member State, according to national regulations. For a single-center study, the Protocol CI is the principal investigator. In addition, the Sponsor must designate a principal or coordinating investigator to review the study report that summarizes the study results and confirm that, to the best of his/her knowledge, the report accurately describes the conduct and results of the study in the study’s final report. The Sponsor may consider one or more factors in the selection of the individual to serve as the Protocol CI and or CSR CI (e.g., availability of the CI during the anticipated review process, thorough understanding of study methods, appropriate enrolment of subject cohort, timely achievement of study milestones). The Protocol CI must be a participating study investigator.

Compliance with Study Registration and Results Posting Requirements

Guidance: Registration is only required for PASS studies (safety and/or efficacy).

Under the terms of the Food and Drug Administration Modernization Act (FDAMA) and the Food and Drug Administration Amendments Act (FDAAA), the Sponsor of the study is solely responsible for determining whether the study and its results are subject to the requirements for submission to the Clinical Trials Data Bank, www.clinicaltrials.gov. Merck, as Sponsor of this study, will review this protocol and submit the information necessary to fulfil these requirements. Merck entries are not limited to FDAMA/FDAAA mandated studies. Information posted will allow subjects to identify potentially appropriate studies for their disease conditions and pursue participation by calling a EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 212

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 85 Epidemiology No.: EP07044.002

central contact number for further information on appropriate study locations and site contact information.

By signing this protocol, the investigator acknowledges that the statutory obligations under FDAMA/FDAAA are that of the Sponsor and agrees not to submit any information about this study or its results to the Clinical Trials Data Bank.

EMA/623947/2012 22-Feb-2016

04CN88

05822K PAGE 213

Product: MK-4117 Protocol/Amendment No.: 203-01 Page 86 Epidemiology No.: EP07044.002

ANNEX 8. PRAC ENDORSEMENT

Placeholder for PRAC Endorsement Letter.

EMA/623947/2012 22-Feb-2016

04CN88

05822K P A GE 214

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 8 7 E pi d e mi ol o g y N o.: E P 0 7044.002

A N N E X 9 . QUALIFIED PERS ON F OR PHAR MACOVIGILANCE ( QPP V)

P P D E x ec uti v e Dir ect or, E U Q u alifi e d Pers o n f or Risk Manage ment & Phar macovigilance T el: P P D ; G S M: P P D ; F a x: P P D E m ail: P P D

P P D Director, E U Phar macovigilance (and Deputy QPP V) T el: P P D G S M: P P D F a x: P P D E m ail: P P D

E mergency/ Out of Hours: GS M nu mbers above or via P P D

Hertford Road Hoddesdon Hertfordshire E N11 9B U Telephone 01992 467272

D e ar Sir/ Mada m ,

Re: E U QPP V Signature Page for P ASS

I N N: Desloratadine Pr o d uct: M K -4117 Pr ot oc ol N o.: 2 0 3 E pi d e mi ol o g y N o.: E P 0 7044.002 Pr ot oc ol D at e : 1 3 J ul y 2015 Pr ot oc ol Titl e: Association bet ween use of desloratadine and risk of seizures, s u prave ntric ular tac hycar dia, a n d atrial fi brillati o n or fl utter: A N or dic register -based study M A H: M erc k S h ar p & Doh me Lt d .

I n li n e wit h t h e G ui d eli n e o n G o o d Phar maco v i gil a nc e Pr actic e ( G V P), M o d ul e VIII – P ost - A ut h ori zati o n S af ety St u di es ( P ASS) a n d acc or di ng t o M S D i nt er n al S O Ps, t his st u d y h as b e e n revie wed and approved by the European Qualified Person for Phar macovigilance.

Y o urs fait hf ully , P P D

Executive Director, E U Qualified Person f or Risk Manage ment & Phar macovigilance

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K P A GE 215

Product: MK -4117 Protocol/A mend ment No.: 2 0 3 -0 1 P a g e 8 8 E pi d e mi ol o g y N o.: E P 0 7044.002

A N N E X 1 0. SI GNATURE PA GE

Merck Representative

TYPED NA ME SI G N A T U R E D ATE P P D

Pri nci pal I nvestigat or

I a gr e e t o conduct t his st u d y i n accordance wit h t h e d esi g n o utli n e d i n t his pr ot oc ol a n d t o a bi d e b y all pr o visi o ns of t his pr ot oc ol (i ncl u di n g ot h er m a n u als a n d docu ments referenced fr o m t his pr ot oc ol) ; d e vi ati o ns fr o m t h e pr ot oc ol ar e acce pta ble o nl y wi t h a m ut u all y agree d u p o n pr ot oc ol a mend ment. I a gr e e t o conduct the st u d y i n accordance wit h g e n er all y accepted standards of G o o d Phar macoepide miology Pr actic e. I als o a gr e e t o r e p ort all infor mation or d at a i n accordance wit h t h e pr ot oc ol. I understand t h at infor mation t h at i d e ntifi es m e will b e us e d a n d discl os e d as descri be d i n t he pr ot oc ol.

P P D

P P D P P D

E MA/623947/2012 2 2 -F e b -2016

04 C N88

05822 K DESLORATADINE PAGE 216 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

ANNEX 5 Statistical Analysis Plan

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K P A GE 217

Product: MK -4117 Pr ot oc ol N o.: 4 1 1 7 -2 0 3 St atistic al A n al ysis Pl a n Epide miology No.: EP O7044.002

S p o ns or: M erc k S h ar p & D o h m e C or p., a s u bsi di ar y of M erc k & C o., I nc. (hereafter referred to as the Sponsor or Merck) One Merck Drive P. O. B o x 1 0 0 Whitehouse Station, NJ 08889 -0 1 0 0, U. S. A.

Titl e: Association bet ween use of desloratadine and risk of seizures, s u prave ntric ular tac hycar dia, a n d atrial fi brillati o n or fl utter: A N or dic register - based study Statistical A nalysis Pla n (S A P)

A d diti o nal s pecificati o ns of t he a nalyses described in the desloratadine pr ot oc ol

Versi o n 1: 2 9 S e pt 2016

I nvestigat or

Institute of Applied Econo mics and Health Research  Martha E mneus, Director, MSc (econ)  M ary Rosenz weig, project manager N ati o n al I nstit ut e of P u blic H e alt h, U ni v ersit y of Southern Den mark  Annette Kj ær Ersbøll, Professor and principal investigator  Lau Caspar Thygesen, Associate Professor  Thora Majlund Kj ærulff, Research Assistant Nati o nal scie ntific c o or di nat ors  Den mark: Annette Kj ær Ersbøll, Professor, National Institute of Public Health, University of Southern Den mark, Copenhagen, Den mark  Fi nla n d: Eer o P ukkala, Direct or of Statistics, Fi n nis h Ca ncer Registry, I nstit ute f or Statistical a n d Epide miological Cancer Research, Helsinki. Professor in Epide miology, School of Health Sci e nc es, University of Ta mpere  Nor way: Kristian Bolin, Professor, Institute of Econo mics and Center for Health Econo mics, Gothenburg University  S weden: Kristian Bolin, Professor, Institute of Econo mics and Center for Health Econo mics, Gothenburg Universi t y Expert group  P P D

)

04J CZ V

05822 K P A GE 218

Product: MK -4117 Pr ot oc ol N o.: 4 1 1 7 -2 0 3 St atistic al A n al ysis Pl a n Epide miology No.: EP O7044.002

Table of contents

List of a b breviatio ns 6

Introduction 7

Additio nal specificatio ns for sectio n 9.4 i n the protocol ( Data Sources / Study Procedures) 7 Study procedures (i. e., s ecti o n 9. 4. 1 i n t h e pr ot oc ol ) 7 T a bl e I. E sti m at e d f oll o w -up periods for data extraction when data are accessed in 3 Q2016 – 3 Q2017 8

A d ditio nal s pecificatio ns for sectio n 9. 7 i n t he protocol ( Data a nalysis) 9 S ubst udy 1: Descriptive a nalysis of expos ure of desloratadi ne (i.e., sectio n 9.7.1 i n the protocol) 9 Substudies 2 -5: First s ei z ur e, s u pr a v e ntri c ul ar t a c h y c ar di a, atri al fi brill ati o n or fl utt e r a n d first rec urre nt s ei z ur e (i.e., s e cti o n 9. 7. 2 – 5 i n t h e pr ot o c ol) 1 0 S ubst u dy 2: First seiz ure (i. e., s e cti o n 9. 7. 2 i n t h e pr ot o c ol ) 1 0 Substudy 3: S V T (i. e., s e cti o n 9. 7. 3 i n t h e pr ot o c ol ) 1 1 S u bst u d y 4: Atri al fi brill ati o n /fl utt er (i. e., s e cti o n 9. 7. 4 i n t h e pr ot o c ol ) 1 1 S ubst udy 5: First rec urre nt seiz ure (i. e., s e cti o n 9. 7. 5 i n t h e pr ot o c ol ) 1 2 Additional infor mation for Substudies 2 A, 3 A, 4 A, and 5 A 1 2 Additional infor mation for Substudies 2 B, 3 B, 4 B, and 5 B 1 3 Supple mentary analyses 1 -4 (s e cti o n 9. 7 . 6 i n t he protocol) 1 4 Supple mentary analysis 5: Exposure misclassification (section 9.7.6 in the protocol) 1 4 Table II. Use of lorata di ne, ru pata di ne, a n d DL i n the Nor dic cou ntries 1 5 T a bl e III. N o n -sedative antihista mines marketed in the Nordic countries 1 6 Supple mentary analysis 6: Higher risk a mong first -ever prescription rede mption (section 9.7.6 in the protocol) 1 7 Supple mentary analysis 7: Non -exposed period starts 52 weeks follo wing last prescription rede mption (sectio n 9. 7. 6 i n t he protocol) 1 7 Supple mentary analysis 8: Alternative confounder adjust men t set (sectio n 9. 7. 6 i n t he protocol) 1 8 Table I V. Definition of confounders included in the pri mary and alternative mini mu m sufficient adjust ment set that will be used for confounder adjust ment in the association studies. 1 8 Supple mentary analysis 9: Quantitative bias analysis (section 9.7.6 in the protocol) 2 0 Supple mentary analysis 10: Only periods with no O TC sales (section 9.7.6 in the protocol) 2 2

S pecificatio n of data extractio n 2 2 Study population 2 2 Table V. Definition of pri mary study populations* 2 3 T a bl e VI. D efi nit ion of the secondary study populations* 2 3 Data extraction periods 2 3 Table VII. Operationalization and data extraction period of variables included in the study 2 4 V ari a bl es 2 6 Table VIII: Overvie w of when the IC D7, IC D8, IC D9 and IC D10 codes were used in each of the Nordic c o u ntries. 2 6

Refere nces 2 7

04J CZ V

05822 K P A GE 219

Product: MK -4117 Pr ot oc ol N o.: 4 1 1 7 -2 0 3 St atistic al A n al ysis Pl a n Epide miology No.: EP O7044.002

Ta ble s hells for t he deslorata di ne st u dy 2 9 Overvie w of the analyses in the desloratadine study 2 9 Table 0. Overvie w of the analyses in the desloratadine study 2 9 Substudy 1: Descriptive analysis of the desloratadine use 3 0 Table 1.1 Incidence rates and prevalence proportions of desloratadine ( DL) users in the general population bet ween 2001 and 2015 for the total po p ulatio n a n d stratifie d by age gro u ps, co u ntry, sex, cale n dar year, seaso nality, ast h matic stat us, severity of r hi nitis a n d c hro nic urticaria stat us. 3 0 Table 1.2 Distribution of nu mber of D D D of desloratadine ( DL) in the total po p ulatio n a n d stratifie d by age, sex, co u ntry, cale n dar year, seaso nality, ast h matic stat us, severity of r hi nitis a n d c hro nic urticaria status a mong DL users 3 2 Substudy 2 A: First seizure 3 3 Table 2 A. Nu mber of persons with a first ti me diagnoses of seizures, risk ti me and incidence rate of seizures overall and stratified by age groups, country and sex 3 3 Substudy 2 B: Desloratadine and first seizure 3 4 Table 2 B1. Nu mber of first seizure cases, risk ti me and incidence rate across persons currently and not currently exposed to desloratadine ( DL) 3 4 Table 2 B2. Analysis of association bet ween desloratadine ( DL) exposure and first seizure 3 4 Substudy 3 A: First supraventricular tachycardia 3 5 Table 3 A. Nu mber of persons diagnosed with supraventricular tachycardia (S V T), risk ti me a n d incidence rate of supraventricular tachycardia overall and stratified by age groups, country and sex 3 5 Substudy 3 B: Desloratadine and supraventricular tachycardia 3 6 Table 3 B1. Nu mber o f supraventricular tachycardia (S V T) cases, risk ti me and incidence rate across persons currently and not currently exposed to desloratadine ( DL) 3 6 Table 3 B2. Analysis of association bet ween desloratadine ( DL) exposure and first s u prave ntric ular tachycardia (S V T) 3 6 S u bst u d y 4 A: Atri al fi brill ati o n a n d fl utt er 3 7 Table 4 A. Nu mber of perso ns diag nosed with atrial fibrillatio n or flutter ( A -fi b /fl u), ri s k ti m e a n d i nci de nce rate of atrial fi brillatio n of fl utters overall a n d stratifie d by age gro u ps, co u ntry a n d sex 3 7 S u bst u d y 4 B: D esl or at a di n e a n d atri al fi brill ati o n or fl utt er 3 8 T a bl e 4 B 1. N u m b er of atri al fi brill ati o n or fl utt er ( A -fi b/fl u) cases, risk ti me a n d i nci de nce rate across persons currently and not currently exposed to desloratadine ( DL) 3 8 Table 4 B2. Analysis of association bet ween desloratadine ( DL) exposure and first atrial fibrillation or fl utter ( A -fi b /fl u) 3 8 Substudy 5 A: First recurrent seizure 3 9 Table 5 A. Nu mber of perso ns with a first recurre nt diag nosis of seizures, risk ti me a n d i nci de nce rate of first rec urre nt seiz ures 3 9 Substudy 5 B: Desloratadine and first recurrent seizure 4 0 Table 5 B1. Nu mber of first recurrent seizure cases, risk ti me and incidence rate a cross perso ns currently and not currently exposed to desloratadine ( DL) 4 0 Table 5 B2. Analysis of association bet ween desloratadine ( DL) exposure and first recurrent seizure 4 0 Supple mentary a n al ysis 1 4 1 Table S1.1 Analysis of association bet ween desloratadine ( DL) exposure and first seizure when using an alternative exposure categorization 4 1 T a bl e S 1. 2 A n al y si s of association bet ween desloratadine ( DL) exposure and supraventricular tachycardia (S V T) when using an alternative exposure categorization 4 1

04J CZ V

05822 K P A GE 220

Product: MK -4117 Pr ot oc ol N o.: 4 1 1 7 -2 0 3 St atistic al A n al ysis Pl a n Epide miology No.: EP O7044.002

Table S1.3 Analysis of association bet ween desloratadine ( DL) exposure and atrial f i brill ati o n or fl utter ( A -fib/flu) when using an alternative exposure categorization 4 1 Table S1.4 Analysis of association bet ween desloratadine ( DL) exposure and first recurrent seizure when using an alternative exposure c ategorizatio n 4 2 Supple mentary analysis 2 4 3 Table S2.1 Analysis of association bet ween desloratadine ( DL) exposure and first febrile seizure and fir st n o n -febrile seizure a mong children age 0 – 4 years 4 3 Table S2.2 Analysis of association bet ween desloratadine ( DL) exposure and first recurrent febrile seizure a nd first recurre nt no n -febrile seizure a mong children age 0 – 4 years 4 3 Supple mentary analysis 3 4 4 Table S3.1 Analysis of association bet ween desloratadine ( DL) exposure and first seizure when excluding persons with a diagnosis of chronic urticaria or high use of DL 4 4 Table S3.2 Analysis of association bet ween desloratadine ( DL) exposure and first supraventricular tachycardia (S V T) when excluding persons with a diagnosis of chronic urticaria or high use of DL 4 4 Table S3.3 A nalysis of associatio n bet wee n desloratadi ne ( DL) ex posure a nd first atrial fibrillatio n or fl utter ( A -fib/flu) when excluding persons with a diagnosis of chronic urticaria or high use of DL 4 4 Supple menta r y a n al ysis 4 4 5 Table S4.1 Analysis of association bet ween desloratadine ( DL) exposure and first seizure for the total population and stratified by country 4 5 T a bl e S 4. 2 A n al y si s of a s s o ci a tion bet ween desloratadine ( DL) exposure and supraventricular tachycardia (S V T) for the total population and stratified by country 4 5 Table S4.3 Analysis of association bet ween desloratadine ( DL) exposure and atrial fibrill ati o n of fl utter ( A -fib/fl u) for t he total po p ulatio n a n d stratifie d by co u ntry 4 6 Supple mentary analysis 5 4 7 Table S5.1. Analysis of association bet ween alternative exposure definitions (d esloratadine ( DL) and loratadine or non -sedating antihista mines) and first seizure 4 7 Table S5.2. Analysis of association bet ween alternative exposure definitions (desloratadine ( DL) and loratadine or non -s e d ati n g antihista mines) and first supraventricular tachycardia (S V T) 4 7 Table S5.3. Analysis of association bet ween alternative exposure definitions (desloratadine ( DL) and loratadine or non -se dati ng a nti hista mi nes) a n d first atri al fi brill ati o n or fl utt er ( A -fi b /fl u) 4 7 Supple mentary analysis 6 4 8 Table S6.1. Analysis of association bet ween desloratadine ( DL) exposure and first seizure stratified by nu mber of pr escription rede mptions 4 8 Table S6.2. Analysis of association bet ween desloratadine ( DL) exposure and first supraventricular tachycardia (S V T) stratified by nu mber of prescription rede mptions 4 8 Table S 6. 3. A nalysis of associatio n bet wee n deslorata di ne ( DL) ex posure a n d first atrial fibrillatio n or fl utter ( A -fib/flu) stratified by nu mber of prescription rede mptions 4 8 Supple mentary analysis 7 4 9 Table S7.1 Analysis of association bet ween desloratadine ( DL) exposure and first seizure when non - exposed periods starts 52 weeks fro m last prescription rede mption 4 9 Table S7.2 Analysis of association bet ween desloratadine ( DL) exposure and first supraventricular tachycardia (SVT) when non -exposed periods starts 52 weeks fro m last prescription rede mption 4 9 Table S7.3 Analysis of association bet ween desloratadine ( DL) ex p o s ur e a n d fir st atri al fi brill ati o n or fl utter ( A -fib/flu) when non -exposed periods starts 52 weeks fro m last prescription rede mption 4 9 Supple mentary analysis 8 5 0 T a bl e S 8. 1 A n al y si s of a ssociation bet ween desloratadine ( DL) exposure and first seizure when adjusting for an alternative confounder adjust ment set 5 0 Table S8.2 Analysis of association bet ween desloratadine ( DL) exposure and first supraventricul ar tachycardia (S V T) when adjusting for an alternative confounder adjust ment set 5 0

04J CZ V

05822 K P A GE 221

Product: MK -4117 Pr ot oc ol N o.: 4 1 1 7 -2 0 3 St atistic al A n al ysis Pl a n Epide miology No.: EP O7044.002

Table S8.3 A nalysis of associatio n bet wee n desloratadi ne ( DL) ex posure a nd first atrial fibrillatio n or fl utter ( A -fi b /fl u) w h e n a dj u sti n g for an alternative confounder adjust ment set 5 0 Supple mentary analysis 9 5 1 Table S9.1. Qua ntitative Bias A nalysis (si mple approach): misclassificatio n of asth matic status, s e v erit y of r hi ni tis, a n d c hro nic urticaria stat us 5 1 T a bl e S 9. 2. Q u a ntit ati v e Bi a s A n al y si s (i n di vi d u al r e c or d l e v el a n al y si s): mi s cl a s sifi c ati o n of c hr o ni c urticaria stat us, severity of r hi nitis a n d ast h matic stat us* 5 1 Supple mentary analysis 10 5 2 Table S10.1 Analysis of association bet ween desloratadine ( DL) exposure and first seizure in periods wit ho ut over -t h e -co u nter sales 5 2 Table S10.2 Analysis of association bet ween desloratadine ( DL) exposure and first supraventricular tachycardia (S V T) in periods without over -t h e -co u nter sales 5 2 Table S10.3 Analysis of association bet ween desloratadin e ( D L) e x p o s ur e a n d fir st atri al fi brill ati o n or fl utter ( A -fib/flu) i n periods without over -t h e -co u nter sales 5 2

04J CZ V

05822 K P A GE 222

Product: MK -4117 Pr ot oc ol N o.: 4 1 1 7 -2 0 3 St atistic al A n al ysis Pl a n Epide miology No.: EP O7044.002

List of a b breviati o ns

A- fi b Atri al fi brill ati o n A- fl u Atri al fl utt er A- fi b/fl u Atri al fi brill ati o n or atri al fl utt er

P P D

A T C A n at o mi c al T h er a p e uti c C h e mi c al Cl a s sifi c ati o n S y st e m D A G Dir e ct e d a c y cli c gr a p h D L Desloratadine I C D I nt er n ati o n al Cl a s sifi c ati o n of Diseases I R Incidence r at e I R R Incidence rate r ati o PI C P er s o n al i d e ntifi c ati o n code O T C O v er -t h e -c o u nt er Q B A Q u a ntit ati v e Bi a s A n al y si s S V T Supraventricular tachy c ar di a

04J CZ V

05822 K PAGE 223

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Introduction

The statistical analyses that will be performed in the project titled: “Association between use of desloratadine and risk of seizures, supraventricular tachycardia, and atrial fibrillation or flutter: A Nordic register-based” are in general described in the protocol.The present statistical analysis plan (SAP) provides additional specification, describes the analyses in more detail where needed, and operationalizes variables not yet specified in the protocol. Section 9.4 “Data sources” and section 9.7 “Data analysis” from the protocol are inserted marked in italics and highlighted in gray to facilitate the reader’s understanding of the SAP. The final report will also include a description of data control performed which will describe any differences in variables across countries, missing data strategies, variable derivations, etc. Any minor modifications to the statistical analysis plan, including refinements of the ICD and ATC codes in Tables III, V, VI, and VII, will be documented in the methods section of the final report.

Additional specifications for section 9.4 in the protocol (Data Sources / Study Procedures)

The DL study will include register information from four Nordic countries – Denmark, Finland, Norway, and Sweden. In addition to a long history of collecting high quality information on births, deaths, immigration and emigration, disease incidence, and activities in the healthcare sector [Ref. 5.4: 00W4D0], exceptional opportunities to perform register-based research are driven by the unique PIC introduced in the Nordic countries in the 1960’s and available to all persons with permanent residence in the Nordic countries [Ref. 5.4: 00W4G3]. The PIC makes it possible to link information at the individual level from several registers for scientific research purposes. The national prescription registers and national patient registers within each of the Nordic countries capture all the individual encounters of purchasing prescribed DL and allow sufficient longitudinal data to differentiate between first and recurrent seizures and to identify incident SVT and A-fib/flu cases. Person-specific use of DL will be elucidated from the national prescription registers by obtaining information on redemption of DL prescription for each person [Ref. 5.4: 00W4CX]. Person- specific information on seizures, SVT, and A-fib/flu will be derived from the Nordic national patient registers.

The national prescription registers include information on the date of prescription redemption, information on the purchaser, and information on the drug redeemed (e.g., ATC code, number of pills, daily dose, pack size, and number of packs purchased) [Ref. 5.4: 045TYK]. The Nordic national patient registers include diagnostic and treatment information for patients treated at the secondary and tertiary hospital level [Ref. 5.4: 045TZL, 045W09]. Clinical experts have been consulted on how to include information from the prescription and patient registers. Information on date of birth, immigration, emigration, and death will be obtained from the civil registration systems [Ref. 5.4: 00W4G3, 045W09].

Study procedures (i.e., section 9.4.1 in the protocol) This is an observational, register-based study and pre-existing health-related national register data will be the sole data source. According to Danish, Finnish, Norwegian, and Swedish law, register-

7 04JCZV

05822K PAGE 224

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

based studies can be carried out without consent from the individual subjects where the processing takes place for the sole purpose of carrying out statistical or scientific studies of significant public importance and where such processing is necessary in order to carry out these studies. It is an absolute requirement that the publication of statistical or scientific results may never reveal the identity of individuals or otherwise compromise data subjects. We will obtain approval by the data agencies in the four countries before data management and data analyses will be performed.

Table 5 in section 9.4 of the protocol provided an overview of the periods of available data, the study period and the data extraction period. Table I below provides updated information on data availability. If data from 2016 are available, the study period will be extended to include the most recent data. The actual study period for each country is influenced by the periods of available data in the relevant registers as well as disease and drug-free periods applied when analyzing the data. To identify incident disease events we apply a disease-free run-in period as long as possible for the registers in Denmark, Sweden and Finland. However, as the national patient register was established in 2008 in Norway we apply a two-year disease-free run-in period in Norway. Furthermore, since the prescription registers were established in 2004 and July 2005 in Norway and Sweden, respectively, we will apply a drug-free run-in period to account for the potential bias from left-truncation. This means that a person with prescription redemption of desloratadine (DL) in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in the analysis. In summary and with the aim to clarify the study periods, the study period for all substudies (unless explicitly specified in the SAP) is 2001–2015 for Denmark and Finland, 2006–2015 for Sweden (given the drug-free period of 6 months) and 2010–2015 for Norway (given the two-year disease-free run-in period). The data extraction period refers to the longest period for which data on exclusion variables (e.g., seizures, SVT, or A-fib/flu before baseline) can be obtained.

Table I. Estimated follow-up periods for data extraction when data are accessed in 3Q2016– 3Q2017 Country Register Denmark Finland Norway Sweden

National prescription 1995–2015 1994–2015 2004–2015 2005–2015 register National patient register 1977–2015 (1) 1967–2015 (2) 2008–2015 1987–2015 (3)

Civil registration system 1968–2015 1967–2015 1964–2015 1965–2015

2001–2015 2001–2015 2010–2015 2006–2015 Estimated study period (15 years) (15 years) (6 years) (11 years)

Data extraction period 1977–2015 (or 1967–2015 (or 2008–2015 (or most 1987–2015 (or most recent most recent data recent data available) most recent data data available) available) available)

8 04JCZV

05822K PAGE 225

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Additional specifications for section 9.7 in the protocol (Data analysis) Prior to conducting the data analyses, we will perform data management (as described in section 9.6) to ensure data quality and to correct inconsistencies and errors in the data. The data analysis will include the following five steps listed below. Note that specifications concerning all pre-defined supplementary analyses will be detailed in a separate statistical analysis plan (i.e., this document) before database lock and start of data analysis. The assumptions of the statistical models performed will be evaluated for each model in the analysis.

Substudy 1: Descriptive analysis of exposure of desloratadine (i.e., section 9.7.1 in the protocol)

A cohort study describing DL use in the general population will be performed. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to avoid left-truncation bias. In addition, to avoid analogous misclassification of prevalent dispensings with incident DL dispensings among immigrants, we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. Prevalent users will be defined as persons who have at least one prescription of DL in the period of interest (e.g., the entire study period or each year). Incident users are first time users of DL in the period of interest. The following descriptive analyses will be performed:

In accordance with the remaining analyses, the number of individuals in the general population each year will be obtained from the NORDCAN database and used as an estimate of the number of person-years at risk assuming each person in the population each year contributes one person-year at risk [Ref. 5.4: 045VPW]. When calculating the incidence rates, the population at risk is defined as the population on July 1 of the given year for Denmark and Norway, the average of the population on January 1 and December 31 of the given year for Sweden and the average of the population December 31 that year and December 31 the previous year for Finland (i.e., the population at risk in 2001 is the average of the population at risk December 31, 2001 and December 31, 2000). When calculating the prevalence proportion, the denominator will be the population on January 1 of the given year. The 95 % confidence intervals (CI) will be calculated for both incidence rates (IR) and prevalence proportions.

The table shells for Substudy 1 are Table 1.1 and Table 1.2 (see section “Table shells for the desloratadine study”).

9 04JCZV

05822K PAGE 226

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudies 2-5: First seizure, supraventricular tachycardia, atrial fibrillation or flutter and first recurrent seizure (i.e., section 9.7.2–5 in the protocol)

The descriptions of the analyses of the substudies below are taken from the protocol. To reduce duplication of text, the next section includes further details that apply to a number of the substudies.

Substudy 2: First seizure (i.e., section 9.7.2 in the protocol)

Substudy 2A A cohort study estimating the incidence of first seizures in the general population will be conducted. The cohort of individuals with seizures and the distribution of the population by age, year, and country will be used to derive estimates for the general population. The number of individuals in the general population on 01 January of each year will be obtained from the NORDCAN database and used as an estimate of the number of person years at risk assuming each person in the population on 01 January of each year contributes one person year at risk [Ref. 5.4: 045VPW]. The IRs of seizure will be shown for the total population and stratified by country, sex, and age. The 95% confidence interval for IRs will be calculated as:

where IR is the incidence rate, exp is the exponential function, and N new users is the number of new users [Ref. 5.4: 045VXZ].

Substudy 2B

A cohort study among ever-DL users analyzing the association between exposed and unexposed time and first seizures will be performed for the total population and stratified by age. Current use (i.e., “exposed” period) will be defined for each prescription as days’ supply plus a 4 week grace period to account for intermittent use and a possible wash-out effect. Unexposed time will be defined as the period starting 26 weeks from the dispensing date of the prior DL prescription until the next DL prescription redemption, if any. A person who has redeemed at least one prescription of DL is included in the cohort and enters the cohort at the date of first prescription redemption. Study individuals are followed until occurrence of a seizure; occurrence of one of the conditions described in section 9.2.2; 31 December 2013; emigration; or death, whichever comes first. The association between exposure to DL and first seizure will be evaluated using Poisson regression of the IR of first seizure, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to account for truncation bias. In addition we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) (Table 4). We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

10 04JCZV

05822K PAGE 227

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 3: SVT (i.e., section 9.7.3 in the protocol)

Substudy 3A

Substudy 3B

A cohort study among ever-DL users analyzing the association between exposed and unexposed time (as defined in Substudy 2B) and first SVT will be performed for the total population and stratified by age. A person who has redeemed at least one prescription of DL is included in the cohort and enters the cohort at the date of first prescription redemption. Study individuals are followed until occurrence of a SVT; 31 December 2013; emigration; or death, whichever comes first. The association between exposure to DL use and first SVT will be evaluated using Poisson regression of the IR of first SVT, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to account for truncation bias. In addition we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar year, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) (Table 4). We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

Substudy 4: Atrial fibrillation/flutter (i.e., section 9.7.4 in the protocol)

Substudy 4A A cohort study estimating the incidence of A-fib/flu in the general population will be performed. The cohort of individuals with A-fib/flu and the distribution of the population by age, year, and country will be used to derive estimates in the general population. The number of individuals in the general population on 01 January of each year will be obtained from the NORDCAN database and used as an estimate of the number of person years at risk assuming each person in the population on 01 January of each year contributes one person year at risk [Ref. 5.4: 045VPW]. The IRs of A-fib/flu will be shown for the total population and for each country, sex, and age. The confidence interval of IRs is calculated as outlined above under Substudy 2A.

Substudy 4B

A cohort study among ever-DL users analyzing the association between exposed and unexposed time (as defined in Substudy 2B) and first A-fib/flu diagnosis will be performed for the total population

11 04JCZV

05822K PAGE 228

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

and stratified by age. A person who has redeemed at least one prescription of DL is included in the cohort and enters the cohort at the date of first prescription redemption. Study individuals are followed until occurrence of A-fib/flu; 31December 2013; emigration or death, whichever comes first. The association between exposure to DL use and first A-fib/flu will be evaluated using Poisson regression of the IR of first A-fib/flu, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to account for truncation bias. In addition we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar years, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) (Table 4). We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

Substudy 5: First recurrent seizure (i.e., section 9.7.5 in the protocol)

Substudy 5A

Substudy 5B

A cohort study among ever-DL users with a previous seizure analyzing the association between exposed and unexposed time (as defined in Substudy 2B) and first recurrence of seizures (excluding febrile seizure) will be performed. A person who has redeemed at least one prescription of DL and has experienced a seizure is included in the cohort and enters the cohort at the date of first seizure. Study individuals are followed until first recurrent seizure; occurrence of any condition mentioned in section 9.2.2; 31 December 2013; emigration; or death, whichever comes first. The association between exposure to DL use and first recurrent seizure will be evaluated using Poisson regression of the IR of first recurrent seizure, and the measure of effect is the IRR. A person with prescription redemption of DL in Norway and Sweden within the first 6 months after establishment of the prescriptions registers will not be included in order to account for truncation bias. In addition we exclude persons with DL prescription redemptions within 6 months after immigrating to Denmark, Finland, Norway, or Sweden. The analyses will be adjusted for confounders identified from the DAG (i.e., country, age, sex, calendar years, seasonality, severity of rhinitis, asthmatic status, and chronic urticaria status) (Table 4). We will evaluate whether age can be included as a continuous variable by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

Additional information for Substudies 2A, 3A, 4A, and 5A

A specification has to be made regarding the number of persons at risk, which is not correctly stated in the protocol, where it is specified to be general population on January 1 of each year. More correctly when calculating the incidence rates, the population at risk is defined as the population on July 1 of the given year for Denmark and Norway, the average of the population on January 1 and

12 04JCZV

05822K PAGE 229

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

December 31 of the given year for Sweden and the average of the population December 31 that year and December 31 the previous year for Finland (i.e., the population at risk in 2001 is the average of the population at risk December 31, 2001 and December 31, 2000).

The table shells for Substudies 2A, 3A, 4A and 5A are Table 2A, Table 3A, Table 4A and Table 5A, respectively (see section “Table shells for the desloratadine study”).

Additional information for Substudies 2B, 3B, 4B, and 5B

Rationale for the choice of study design: Substudy 2B, 3B, 4B and 5B are historical cohort study using prospectively collected data from nationwide population registers with a cohort defined by first use of DL. During the follow-up period, DL will be treated as a time-varying exposure. Using a variety of exposure definitions, all person-time will be categorized as exposed, unexposed, or unclear (period immediately following a course of current treatment). Incidence rates will be calculated for each outcome of interest for each exposure category, and incidence rate ratios will be estimated for exposed person time compared to unexposed person-time aggregated across each cohort. In this sense, it is a risk interval design, but it differs from the type of self-controlled risk interval design which has frequently been used in vaccine safety research, in that it does not condition the comparison within individuals [Ref. 5.4: 03QPDQ]. A design similar to the one we propose was used to study the effect of cisapride on ventricular arrhythmias [Ref. 5.4: 04CD8S].

A self-controlled risk interval design (SCRID) analyzed by conditional Poisson regression might better control for time-constant confounding within individuals than the design we propose. However, our design avoids an important potential bias that may occur in a SCRID. In that approach, in order to be included in the conditional Poisson analysis, an individual needs to contribute both an exposed and a non-exposed period. This would exclude persons with only an exposed period, an analytic choice which could bias toward the null. If in fact there is a drug-associated risk, we would expect the outcome to occur during the first-ever exposure period for some persons included in the study. These persons would then be censored from the analysis at the time they experience the outcome and hence would only have had an exposed period. An analysis based only on those patients who have had at least one exposed and one unexposed period could therefore potentially underestimate the true risk. Underestimation of the true risk is likely to be a more important bias for this study than confounding by time constant factors; therefore we have maintained our proposal to adjust for both time-constant and time-varying confounders in the analysis, rather than performing a conditional Poisson regression.

The table shells for Substudy 2B are Table 2B1 and Table 2B2; the table shells for Substudy 3B are Table 3B1 and 3B2; the table shells for Substudy 4B are Table 4B1 and Table 4B2; and the tables shells for Substudy 5B are Table 5B1 and Table 5B2 (see section “Table shells for the desloratadine study”).

13 04JCZV

05822K PAGE 230

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analyses 1-4 (section 9.7.6 in the protocol)

(1) alternative definitions of exposure based on time since last DL dispensing (periods 0-4, 5- 8, 9-16, and 17-26 weeks each compared with >26 week since last DL dispensing) for Substudies 2B, 3B, 4B, and 5B. Because as-needed medications, such as antihistamines may not be taken daily, this approach reflects the clinical expectation that the probability of actual exposure on a given day is highest shortly after filling a prescription and diminishes with increasing time. Each time a DL prescription is refilled, the time since last dispensing will reset to 0.

(2) for Substudies 2B and 5B, differentiating between febrile and non-febrile seizures for children aged 0–4 years.

(4) the association analyses in Substudies 2B, 3B, and 4B will be stratified by countries to examine potential differences across countries.

Note: No additional detail was considered necessary for the above analyses.

The table shells for Supplementary analysis 1 are Tables S1.1–S1.4, the tables shells for Supplementary analysis 2 are S2.1–S2.2, the tables shells for Supplementary analysis 3 are Tables S3.1–S3.3 and the tables shells for Supplementary analysis 4 are Tables S4.1–S4.3 (see section “Table shells for the desloratadine study”).

Supplementary analysis 5: Exposure misclassification (section 9.7.6 in the protocol)

Loratadine and rupatadine are two drugs metabolized to DL in the liver. Loratadine was marketed in Denmark in 1988, Norway and Sweden in 1989, and Finland in 1995. As of 2013, when the most current data are available, rupatadine had not been marketed in the Nordic countries. However, if in future years of the study, which may extend to 2015 depending on data availability, rupatadine dispensing does appear in the registries, it will be included in supplementary analysis 5 (Table II).

14 04JCZV

05822K PAGE 231

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Loratadine is available both by prescription and over-the-counter (OTC) without a prescription. An average of 63% of loratadine purchased in the Danish population, 93% in Norway, 41% in Sweden and 72% in Finland from 2001 to 2013 was purchased by prescription and therefore also registered at the individual-level in the prescription register. The reason why some persons choose to purchase OTC drugs by a prescription is that prescription drugs are often subsidized and will therefore be less expensive for the purchaser. Non-sedative antihistamines include a range of different drugs (including DL and loratadine) used in different time periods across the Nordic countries, but in this supplementary analysis, they will be considered as one group (Table III).

Table II. Use of loratadine, rupatadine, and DL in the Nordic countries Country Drug Loratadine (R06AX13) Rupatadine (R06AX28) Desloratadine (R06AX27) Denmark Dispensing group OTC medicine also sale Only prescription. Only prescription 2001- in non-pharmacy outlets. 2012. In 2012 and 2013 parallel-imported products are launched Date for first recorded April 28th 1988 June 8th 2015 January 15th 2001 use or launch date Sales by prescription 60,965 (63 %) 0 39,157 (99.5%) 2001-13 (DDD x 1000) Total sales 2001-13 97,265 (100 %) 0 39,347 (100 %) (DDD x 1000) Norway Dispensing group “Unntatt fra Only prescription. Only prescription 2001- reseptplikten” OTC Dispensing group 2013. Dispensing group medicine sale. “Prescription C” “Prescription C”*

Date for first recorded July 10th 1989 Date for first launch January 15th 2001 use or launch date permission October 21st 2013. No sale in 2013 and 2014 (not on the market). Sales by prescription 116,440 (93 %) 0 152,514 (100 %) 2004-13 (DDD x 1000) Total sales 2004-13 125,743 (100 %) 0 152,514 (100 %) (DDD x 1000) Sweden Dispensing group OTC medicine also sale Only prescription. Only prescription 2001- in non-pharmacy outlets. 2013 except two brands, Desloratadine Sandoz and Flynise from Actavis (launched August 16th 2012 and April 16th 2015). Date for first recorded February 3rd 1989 Date for first launch January 15th 2001. use or launch date permission September 5th 2013. No sale in 2013 and 2014 (not on the market). Sales by prescription 106,074 (41 %) 0 162,661 (100 %) 2006-13 (DDD x 1000)

Total sales 2006-13 258,640 (100 %) 0 162,661 (100 %) (DDD x 1000)

15 04JCZV

05822K PAGE 232

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Country Drug Loratadine (R06AX13) Rupatadine (R06AX28) Desloratadine (R06AX27) Finland Only prescription from Dispensing group Prescription and OTC Only prescription 2001-2012. From 2013 onwards, also OTC sale Date for first recorded May 8th 1995 --- January 15th 2001 use or launch date Sales by prescription 67,937 (72 %) 0 93,076 (98%) 2001-13 (DDD x 1000) Total sales 2001-13 94,637 (100 %) 0 94,662 (100 %) (DDD x 1000) Data source Sweden: https://www.socialstyrelsen.se/statistik/statistikdatabas/lakemedel; https://lakemedelsverket.se/LMF/?type=product Data source Denmark: Medstat.dk Data source Norway: http://www.norpd.no/; http://www.legemiddelforbruk.no/; http://www.msd.no/Documents/Regulatory/pdf/Clarityn_tbl_SPC.pdf http://www.legemiddelverket.no/_layouts/Preparatomtaler/Spc/13-9498.pdf http://www.ema.europa.eu/docs/no_NO/document_library/EPAR_- _Product_Information/human/000313/WC500025540.pdf Data source Finland: http://www.kela.fi/web/en/medicine-reimbursement-statistics_finnish-statistics-on-medicines; http://www.fimea.fi/web/en/databases_and_registeries/consumption/finnish_statistics_on_medicines

Two sensitivity analyses will be performed for which a wider exposure definition will be used to evaluate the potential effect of exposure misclassification in Substudies 2B, 3B, and 4B. First, the exposure measure for this analysis will be defined as redeemed prescriptions of DL and loratadine grouped together. Second, the exposure measure will be further expanded to include all non- sedative antihistamines purchased by prescription in the Nordic countries during the study period. The prescription registers in Denmark, Norway and Sweden include all prescriptions purchased regardless of whether these are subsidized by the state, whereas the Finnish register only includes purchased prescriptions of subsidized drugs. Neither of these analyses will be able to capture OTC use of DL, loratadine, or other non-sedative antihistamines.

Table III. Non-sedative antihistamines marketed in the Nordic countries Denmark (2001-2015) Finland (2001-2015) Norway (2001-2014) Sweden (2001-2015) Cetirizine R06AE07 Cetirizine R06AE07 Cetirizine R06AE07 Cetirizine R06AE07 Levocetirizine R06AE09 Levocetirizine R06AE09 Levocetirizine R06AE09 Terfenadine R06AX12 Terfenadine R06AX12 Terfenadine R06AX12 Loratadine R06AX13 Loratadine R06AX13 Loratadine R06AX13 Loratadine R06AX13 Acrivastine R06AX18 Acrivastine R06AX18 Acrivastine R06AX18 Ebastine R06AX22 Ebastine R06AX22 Ebastine R06AX22 Ebastine R06AX22 Mizolastine R06AX25 Mizolastine R06AX25 Mizolastine R06AX25 Fexofenadine R06AX26 Fexofenadine R06AX26 Fexofenadine R06AX26 Fexofenadine R06AX26 Desloratadine R06AX27 Desloratadine R06AX27 Desloratadine R06AX27 Desloratadine R06AX27 Rupatadine R06AX28 Rupatadine R06AX28 Bilastine R06AX29 (2013 - Bilastine R06AX29 (2014) 2014) Data source Sweden: https://www.ehalsomyndigheten.se/ Data source Denmark: Medstat.dk Data source Norway: Norwegian Drug Wholesale Statistics, Norwegian Institute of Public Health Data source Finland: Kela.fi

16 04JCZV

05822K PAGE 233

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

The table shells for Supplementary analysis 5 are Tables S5.1–S5.3 (see section “Table shells for the desloratadine study”).

Supplementary analysis 6: Higher risk among first-ever prescription redemption (section 9.7.6 in the protocol)

We will perform a supplementary analysis for Substudies 2B, 3B, and 4B in which we will evaluate whether the potential risk of the outcomes is higher following the first ever-DL prescription redemption compared to the risk following the second or the third (or more) prescription redemptions. In this analysis exposed risk time will be divided into groups according to numerical order of prescription redemptions for each person. For this supplementary analysis a cumulative exposure variable (first, second, third or more versus non-exposed) will be analyzed. The exposure variable (Exposure status) will be categorized as: currently DL unexposed; currently DL exposed following first prescription redemption; currently DL exposed following second prescription redemption; and currently DL exposed following third or more prescription redemption. The exposure variable will be analyzed as a fixed effect (as done in Substudies 2B, 3B and 4B) with 4 levels in the supplementary analyses (in Substudies 2B, 3B and 4B, the exposure variable has 2 categories corresponding to currently DL exposed and currently DL unexposed). If the exposure status variable is significant, pairwise comparisons between the 4 levels will be performed. If a pairwise comparison is performed, IRRs will be calculated. Pairwise comparisons with 1) the non- exposed category as reference and 2) with the currently DL exposed following first prescription redemption as reference are both of interest. Correspondingly, two sets of IRRs will be calculated, 1) one set of IRRs with the non-exposed category as reference and 2) one set of IRRs with the category currently DL exposed following first prescription redemption as the reference.

The table shells for Supplementary analysis 6 are Tables S6.1–S6.3 (see section “Table shells for the desloratadine study”).

Supplementary analysis 7: Non-exposed period starts 52 weeks following last prescription redemption (section 9.7.6 in the protocol)

(7) a supplementary analysis for Substudies 2B, 3B, 4B, and 5B for which non-exposed periods start 52 weeks following the last prescription redemption.

The table shells for Supplementary analysis 7 are Tables S7.1–S7.3 (see section “Table shells for the desloratadine study”).

17 04JCZV

05822K PAGE 234

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 8: Alternative confounder adjustment set (section 9.7.6 in the protocol)

(8) a supplementary analysis using an alternative adjustment set to examine the robustness of the study results. This adjustment set consists of age, sex, country, calendar year, seasonality, severity of rhinitis, asthmatic status, diabetes, hypo-/hyperthyroidism, inflammatory disease, infections, type 1 allergy. For the outcomes supraventricular tachycardia and atrial fibrillation or flutter, antihypertensive treatment will also be added.

The operationalization of these variables is presented in Table IV below. The alternative confounder adjustment set was chosen because it was assumed that the variables will be available in the registers and because this set could be used in all four analyses. There is an overlap between the operationalization of two of the confounders: severity of rhinitis and type 1 allergy, and if we find a strong correlation when examining the frequency distributions of these two variables, we will merge the variables together as one single variable and include it in the analysis.

Table IV. Definition of confounders included in the primary and alternative minimum sufficient adjustment set that will be used for confounder adjustment in the association studies. Confounders Confounders included in the included in the primary minimum alternative minimum Definition Data source sufficient sufficient adjustment adjustment set set

Age Age Age is a time varying confounder; and therefore, risk Civil time will be split up in years of age. We will evaluate registration whether age can be included as a continuous variable system by examining the assumption of linearity. Alternatively, age will be included as a categorical variable.

Sex Sex Sex is a time-independent confounder and will be Civil included as males versus females in both the registration descriptive and association analyses. system

Country Country Country of residence is a time-independent Civil confounder, as persons will be censored if emigrating registration from the country. Country of residence will be system included with four categories, i.e. Denmark, Finland, Norway, or Sweden in both the descriptive and association analyses.

Calendar year Calendar year Calendar year is a time varying confounder; and Prescription therefore, risk time will be split up in calendar years. register

Seasonality Seasonality In the association studies, seasonality is a time Prescription varying confounder; and therefore, risk time will be register split up in seasons defined as winter (December– February), spring (March–May), summer (June– August), and autumn (September–November).

18 04JCZV

05822K PAGE 235

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Confounders Confounders included in the included in the primary minimum alternative minimum Definition Data source sufficient sufficient adjustment adjustment set set

Asthma status Asthma status Asthmatic status is defined as a binary variable National indicating whether or not a person has redeemed patient register treatment for asthma defined as at least two and prescriptions of inhalant steroids within a six-month prescription period and/or contacts to hospitals with a diagnosis register of asthma (including both primary and secondary diagnoses) during a five-year period before first DL exposure. To distinguish persons treated for chronic obstructive pulmonary disease from those treated for asthma, first registered asthma prescription has to be redeemed when the purchaser was 45 years or younger.

Disease severity of Disease severity of Severity of rhinitis will be defined as a binary variable Prescription rhinitis rhinitis indicating whether or not a person has received register treatment for severe rhinitis. Persons with severe rhinitis will be identified from the prescription register as persons who have redeemed immunotherapy at least once during a five-year period before first DL exposure.

Chronic urticaria --- Both for the descriptive and association analyses, chronic urticaria status is defined as a binary variable indicating whether or not a person has a registered diagnosis of chronic urticaria in the five-year period before first DL exposure.

--- Diabetes Diabetes is defined as a binary variable indicating National whether or not a person has a registered diagnosis of patient register diabetes or has at least two prescription redemptions and of glucose-lowering drugs registered in the prescription prescription register in the five-year period before register first DL exposure.

--- Hypo- Hypo-/hyperthyroidism is defined as a binary variable National /hyperthyroidism indicating whether or not a person has a registered patient register diagnosis of hypo-/hyperthyroidism or has at least and two prescription redemptions of drugs for treatment prescription of hypo-/hyperthyroidism (i.e., Thycapsol and register Eltroxin) registered in the prescription register in the five-year period before first DL exposure.

--- Inflammatory disease Inflammatory disease is defined as a binary variable National indicating whether or not a person has a registered patient register diagnosis of inflammatory bowel disease, psoriasis, and rheumatic diseases, vasculitis, or sarcoidosis or has prescription redeemed at least two prescriptions of Daivonex register (drug against psoriasis) in the five-year period before first DL exposure. --- Infections Infection is defined as a binary variable indicating National whether or not a person has a registered diagnosis of patient register lung infection (pneumonia) or sinusitis in the five-year period before first DL exposure.

19 04JCZV

05822K PAGE 236

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Confounders Confounders included in the included in the primary minimum alternative minimum Definition Data source sufficient sufficient adjustment adjustment set set

--- Type 1 allergy Type 1 allergy is defined as a binary variable indicating National whether or not a person has a registered diagnosis of patient register acute urticaria, anaphylaxis, quinckes oedema or has and at least one prescription redemptions of prescription immunotherapy drugs registered in the prescription register register in the five-year period before first DL exposure. --- Antihypertensive Use of antihypertensive treatment will be defined as National treatment binary variable indicating whether or not a person has prescription received treatment for hypertension (ACE inhibitor, register angiotensin II receptor antagonist, calcium channel blocker, beta-blockers, alpha-blockers, thiazide (diuretic treatment), methyldopa, or moxonidine). Persons treated with antihypertensive drugs will be identified from the prescription register as persons who have redeemed antihypertensive prescription at least twice during a five-year period before first DL exposure.

The table shells for Supplementary analysis 8 are Tables S8.1–S8.3 (see section “Table shells for the desloratadine study”).

Supplementary analysis 9: Quantitative bias analysis (section 9.7.6 in the protocol)

Quantitative bias analysis (QBA) is used to quantify the amount of systematic error that may occur in a study [Ref. 5.4: 04C0BP, 04C0BT, 04CH6J]. In our study, adequate information on some of the potential confounders (i.e., severity of rhinitis, asthma status, and chronic urticaria), will be challenging to obtain from the sources available (i.e., national patient registers). The national patient registers contain data for contacts to the secondary and tertiary hospitals in the Nordic countries, but we do not have access to data on treatment of these conditions in the primary health care setting. We therefore propose to perform a QBA to try to estimate the effect of misclassification of these confounders on the results. QBA can be performed at either the summarized level or individual record level [Ref. 5.4: 04C0BT]. The advantage of performing QBA at the summarized level is that it is fast and simple to run (https://sites.google.com/site/biasanalysis/); however, QBA of summarized data is performed on crude estimates of the association under consideration, and therefore does not take into account the complexity of the statistical regression model. In contrast, using individual record level QBA methods make it possible to include information on other confounders while examining the impact of misclassification in the regression models [Ref. 5.4: 04CH6J]. The disadvantage of QBA performed at the individual record level is that it requires substantial computational power to run these analyses on huge datasets.

Strategy of QBA analyses We propose to select the type of QBA to be used based on the extent of confounding identified in the empiric data, as well as feasibility considerations. The QBAs will be applied for substudies 2B, 3B

20 04JCZV

05822K PAGE 237

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

and 4B. If the adjusted estimate is within 10% of the crude estimate, we will conduct a simple QBA at the summary level; otherwise we will conduct the QBA at the individual record level to examine the potential impact of misclassification of some of the covariates (i.e., severity of rhinitis, asthma status, and chronic urticaria) on the main study results. If it is not possible to run this analysis due to computational limitations, we propose an analysis restricted to use of data from Denmark only. If none of these individual record level analyses are feasible, we propose to conduct an alternative QBA based on summarized data on the crude association between exposure and outcome. We acknowledge that the interpretation of results from the latter analysis may be challenging due to simplification of the model (e.g., the lack of adjustment for the remaining confounders in the model). However, we judge this to be our best option with regard to quantifying the potential impact of misclassified covariates on the associations under study.

Sensitivity and specificity Regarding the validity of the observed information on the three confounders, we probably have low sensitivity (i.e., the true positive rate or the proportion of persons with the diseases that are correctly identified as such). However, when persons have received the diagnosis for these conditions in the hospital setting, it is likely that they also have the condition, which means that the specificity is high (i.e., the true negative rate or the proportion of healthy people who are correctly identified as not having the condition). We find it unlikely that the potential misclassification of the three confounders is differential and dependent on either use of DL or future development of seizure, SVT or A-fib/flu.

Summary level approach When performing the summary level analysis we will obtain the observed crude association between DL and the three outcomes and specify 12 different scenarios of bias (i.e., define combinations of assumed sensitivity and specificity) to see how these scenarios would have affected the study results. The specificity will be set to 0.985, 0.990, 0.995, 0.999, and 1.0 and the sensitivity to 0.2, 0.4, 0.6, 0.8, and 1.0.

Individual record level approach For the individual record level QBA, we will perform a probabilistic bias analysis in which the biased measure of association will be corrected for misclassification. Overall the principle is to estimate the positive and negative predictive values based on the observed data and the assumed distribution of the sensitivity and specificity. The predictive values are used to estimate the probabilities of the study population being correctly classified. Based on these probabilities a distribution of the variable is generated. By sampling from this distribution, we construct a new variable assigning a new value of the covariate under consideration for each individual in the study population. We rerun the analysis using the new version of the variable to get an estimate corrected for the misclassification bias. The steps of this procedure are as follows:

21 04JCZV

05822K PAGE 238

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

max=0.8. Monte Carlo simulations are used to realize the distributions. A random number generator is used to sample values for sensitivity and specificity from the distributions. After iterating the analysis 999 times, we get a distribution of the results. From this distribution, the point estimate corrected for systematic error is the median and the 2.5th and 97.5th percentile can be interpreted as the 95 % CI interval. Furthermore, we will calculate the total error of the model presented as the sum of the systematic error and the random error in the model.

The table shells for Supplementary analysis 9 are Tables S9.1 for the summery level approach and Table S9.2 for the individual record level approach. Only one of the two tables will be used depending on which analysis we perform (see section “Table shells for the desloratadine study”).

Supplementary analysis 10: Only periods with no OTC sales (section 9.7.6 in the protocol)

The table shells for Supplementary analysis 10 are Tables S10.1–S10.3 (see section “Table shells for the desloratadine study”).

Specification of data extraction

Study population

The four primary study populations will consist of all persons in the Nordic countries who have redeemed at least one prescription of DL (ATC code: R06AX27), persons with a diagnosis of either seizure (ICD10 code: R56), SVT (ICD10 code: I47), or A-fib/flu (ICD10 codes I48) in the study period.

The study period varies across the Nordic countries due to the availability of data in the population registers (See section “Additional specifications for Section 9.4 in the protocol” for specification of the study period). In the following specification of the data extraction, the study period for Denmark and Finland is used to exemplify the specification of the data extraction given in this section. However, for Norway and Sweden changes will be made with regard to the periods of data availability in the prescription and national patient registers. Only primary diagnoses are needed (Table V)

22 04JCZV

05822K PAGE 239

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Table V. Definition of primary study populations* Populations ICD-10 codes ATC codes Study period 1.1.2001 to 31.12.2015 (or most Desloratadine users --- R06AX27 recent available data) 1.1.2001 to 31.12 2015 (or most Seizure R56** --- recent available data) Supraventricular 1.1.2001 to 31.12.2015 (or most I47** --- tachycardia recent available data) Atrial fibrillation or 1.1.2001 to 31.12.2015 (or most I48** --- flutter recent available data) * For persons included in the study populations between 2001-2015 we need information on the diseases from the register start to be able to differentiate between prevalent and incident disease (see also section on “Data extraction periods”). **All subcategories included

To describe the number of DL users and non-DL users according to asthmatic status, chronic urticaria status and severity of rhinitis, we need data on the total populations with these conditions. Furthermore, we need the population of loratadine and rupatadine (two other antihistamines) users and the population of non-sedating antihistamine users (including DL) in order to perform sensitivity analyses on misclassification of exposure in which we will use two broader definitions of antihistamine use. These secondary study populations are described below in Table VI.

Table VI. Definition of the secondary study populations* Populations ICD-10 codes ATC codes Study period 1.1.2001 to 31.12.2015 (or most Asthma J45** R03B**, R03A** recent available data) L50** (excluding acute 1.1.2001 to 31.12 2015 (or most Chronic urticaria --- urticaria (L50.8D), L56.3 recent available data V01AA**(excluding 1.1.2001 to 31.12.2015 (or most Severe rhinitis --- V01AA07) recent available data) Loratadine and 1.1.2001 to 31.12.2015 (or most --- R06AX13, R06AX28 rupatadine users recent available data) non-sedating 1.1.2001 to 31.12.2015 (or most --- R06A** antihistamine users*** recent available data) * For persons included in the study populations between 2001-2015 we need information on the diseases from the register start to be able to differentiate between prevalent and incident disease (see also section on “Data extraction periods”). **All subcategories included ***The subcodes to differentiate between sedative and non-sedative antihistamines will be determined

Data extraction periods

For all persons included in the four primary populations (Table V) and five secondary populations (Table VI), we need their history of diagnosis and treatment (from the start of registration of the registers) to be able to differentiate between incident and prevalent disease. E.g., for a person with a diagnosis of seizure between January 1, 2001 and December 31, 2015, we need information on any prior diagnosis of seizure. We also need full information on treatment, disease, migration,

23 04JCZV

05822K PAGE 240

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

demographics and death during the study period for persons in the nine study populations (see Table VII).

Furthermore for all four primary populations (Table V), we need the full history of treatment (from 1996 onwards) and redeemed prescription drugs (from 1996 onwards) regardless of diagnosis/drug in order to describe the study population and take confounding factors into account in the analyses. The variables needed from the registers are specified below in Table VII. From all registers we need the personal identification number (PIN) to be able to link information.

Table VII. Operationalization and data extraction period of variables included in the study Registers Information needed* Data extraction Variable period** Civil Demographics (age, sex and vital status) At the time of data PIN Registration extraction Date of birth System Sex Vital status and date of death Migration 1.1.2001 to Immigration/emigration 31.12.2015 (or most status recent available data) Date of immigration/emigration Prescription Use of drugs: From register start – PIN register Desloratadine (ATC code: R06AX27) 31.12.2015 (or most ATC code Asthma treatment (ATC codes:R03A, R03B) recent available data) Date of purchase Severe rhinitis, i.e. use of immunotherapy (ATC Dose codes:V01AA) Pack size Antiepilepsy medicine (ATC code: N03A, N05BA) Number of packs redeemed Volume Use of drugs: 1.1.1996 to Glucose-lowering drug (insulin, metformin etc.): 31.12.2015 (or most (ATC: A10) recent available data) Thycapzol: (ATC: H03BB02) Eltroxin: (ATC: H03AA01) Daivonex: (ATC: D05AX02) ACE inhibitor: (ATC: C09A) Angiotensin II receptor antagonist: (ATC: C09C) Calcium channel blocker: (ATC: C08CA) Beta-blockers: (ATC: C07AB) Alpha-blockers: (ATC: C02CA) Thiazide (diuretic treatment): (ATC: C03A - C03E) Methyldopa: (ATC: C02AB) Moxonidine: (ATC: C02AC05) Loratadine (ATC:R06AX13 ) 1.1.2001 to Rupatadine ( ATC: R06AX28) 31.12.2015 (or most Non-sedating antihistamines (ATC: R06A) recent available data)

24 04JCZV

05822K PAGE 241

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Table VII. Operationalization and data extraction period of variables included in the study Registers Information needed* Data extraction Variable period** National Seizures (ICD8: 34518-34511, ICD9: 7803, ICD10: From register start – PIN Patient R56) 31.12.2015 (or most Admission identification Register SVT (ICD8:42790, ICD9: 3457, ICD10: I47) recent available data) number A-fib/flu (ICD8: 42793+42794, ICD9: 4273, ICD10: Primary diagnosis I48) Secondary diagnoses Chronic urticaria (ICD8: 90891, ICD9:7080-7098 + Date of admission 6927, ICD10 code: L50 (excluding L50.8D), L56.3) Type of patient (inpatient, Asthma (ICD8: 49300-49309, ICD9: 4930-4939, emergency, outpatient) ICD10 code: J45) Type of admission (acute, Epilepsy (ICD8: 34509-34511, ICD9: 3457, ICD10 elective) code: G40) Congenital pre-excitation syndrome (ICD8: 42721-42723, ICD9: 4267-4268, ICD10: I45.6) Head trauma (ICD8: 85119, ICD9: 850-854, ICD10: S06) Malignant brain tumor (ICD8: 19100-19109, ICD9: 1921-1923+1910-1919, ICD10 code:C70, C71) Benign brain tumor (ICD8: 22500-22509, ICD9: 2250-2259, ICD10 code: D33) Type 1 allergy (ICD10 code: L50.8D, T78.2, T78.3, 1.1.1996 to T78.4) 31.12.2015 (or most Diabetes (ICD10 code: E10-E14) recent available data) Hypo- and hyperthyroidism (ICD10 code: E00-03, E05-07) Inflammatory Bowel Disease (ICD10 code: K50- K52) Psoriasis (ICD10 code: L40) Rheumatic diseases (ICD10 code: I01-I09 Rheumatic heart diseases) Rheumatoid arthritis (ICD10: M05-08) Vasculitis (ICD10 code: L95 (Vasculitis in skin); M05 (Rheumatoid vasculitis) Sarcoidosis (ICD10 code: D86) Lung infection (pneumonia) (ICD10 code: J13- J18) Sinusitis (ICD10 code: J01(acute), J32 (chronic)) Stroke (ICD8: 43601-43700, ICD9: 1.1.2001 to 431+432+434+436 (All codes are included in 430- 31.12.2015 (or most 438), ICD10 codes: I61, I62, I63, I64.9) recent available data) Acute drug intoxication and overdose of drugs (ICD8: 96009-97959, ICD9: 191+192+960- 979+7088, ICD10 code:T88.6, T88.7, T36, T50) Cancer Malignant brain tumor (ICD7:193, ICD10 From register start – PIN register code:C70, C71) 31.12.2015 (or most Diagnosis Benign brain tumor (ICD7: ICD7:223, ICD10 code: recent available data) Date of diagnosis D33) Tumor number Register on Cause of death 1.1.2001 to Date of death Causes of 31.12.2015 (or most Underlying cause of death Deaths recent available data) (ICD10 code) Contributory cause(s) of

25 04JCZV

05822K PAGE 242

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Table VII. Operationalization and data extraction period of variables included in the study Registers Information needed* Data extraction Variable period** death (ICD10 code)

* For all ICD8, ICD9, ICD10 and ATC codes all subcategories are needed. ** The study period in the Table is set to end December 31, 2015; however, the most recent available data at the time of data collection will be included.

Variables

Beside the list of ICD10 codes presented in Annex 4 of the protocol, we also need ICD8 and ICD9 codes for the diseases used to exclude prevalent cases or diseases used to exclude persons from the analyses. ICD8, ICD9 and ICD10 codes have been used in different time periods in the four Nordic countries [Ref. 5.4: 03RS26] (see Table VIII below).

Table VIII: Overview of when the ICD7, ICD8, ICD9 and ICD10 codes were used in each of the Nordic countries. ICD8 ICD9 ICD10 Denmark 1969–1993 ICD9 never used in Denmark 1994–present Finland 1969–1986 1987–1995 1996–present Norway 1969–1985 1986–1995 1996–present Sweden 1969–1986 1987–1996 1997–present

26 04JCZV

05822K PAGE 243

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

References

[Ref. 5.4: 00W4CX] Furu K, Wettermark B, Andersen M, Martikainen JE, Almarsdottir AB, Sorensen HT. The nordic countries as a cohort for pharmacoepidemiological research. Basic Clin Pharmacol 2009;106:86-94. [Ref. 5.4: 00W4D0] Thygesen L, Daasnes C, Thaulow I, Bronnum- Hansen H. Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving. Scan J Public Health 2011;39(Suppl 7):12-6. [Ref. 5.4: 00W4G3] Pedersen CB. The Danish Civil Registration System. Scan J of Public Health 2011;39(7 suppl):22-5. [Ref. 5.4: 03QPDQ] Glanz JM, McClure DL, Xu S, Hambidge SJ, Lee M, Kolczak MS, et al. Four different study designs to evaluate vaccine safety were equally validated with contrasting limitations. J Clin Epidemiol 2006;59:808-18. [Ref. 5.4: 03RS26] Engholm G, Ferlay J, Christensen N, Bray F, Gjerstorff ML, Klint Å, et al. NORDCAN - a Nordic tool for cancer information, planning, quality control and research. Acta Oncologica 2010;49:725-36. [Ref. 5.4: 045TYK] Kildemoes HW, Sorensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health. 2011 Jul;39(7 Suppl):38-41. [Ref. 5.4: 045TZL] Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health. 2011 Jul;39(7 Suppl):30-3. [Ref. 5.4: 045VPW] The NORDCAN project [Internet]. Copenhagen: Association of Nordic Cancer Registries; c2009 [updated 2014 Dec 17]. Available from: http://www- dep.iarc.fr/NORDCAN/english/frame.asp. [Ref. 5.4: 045VXZ] Kirkwood BR, Sterne JA. Essential Medical Statistics. 2nd ed. Malden (MA): Blackwell Publishing; 2003. [Ref. 5.4: 045W09] Pukkala E. Nordic biological specimen bank cohorts as basis for studies of cancer causes and control: quality control tools for study cohorts with more than two million sample donors and 130,000 prospective cancers. In: Dillner J, editor. Methods in biobanking. New York: Springer; 2011. p. 61- 112. [Ref. 5.4: 04C0BP] Fox MP, Lash TL, Greenland S. A method to automate probabilistic sensitivity analyses of

27 04JCZV

05822K PAGE 244

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

misclassified binary variables. Int J Epidemiol. 2005 Dec;34(6):1370-6. [Ref. 5.4: 04C0BT] Lash TL, Fox MP, Fink AK. Statistics for biology and health. Gail M, Krickeberg K, Samet J, Tsiatis A, Wong W, editors. Dordrecht: Springer; Applying quantitative bias analysis to epidemiologic data; 2009. [Ref. 5.4: 04CD8S] Enger C, Cali C, Walker AM. Serious ventricular arrhythmias among users of cisapride and other QT-prolonging agents in the United States. Pharmacoepidemiol Drug Saf. 2002 Sep;11(6):477- 86. [Ref. 5.4: 04CH6J] Lash TL, Fink AK. Semi-automated sensitivity analysis to assess systematic errors in observational data. Epidemiology. 2003 Jul;14(4):451-8.

28 04JCZV

05822K PAGE 245

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Table shells for the desloratadine study

Overview of the analyses in the desloratadine study

Table 0. Overview of the analyses in the desloratadine study Analyses Description Substudy 1: DL use Prevalent users and incident users in the total population and stratified by age groups, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis and chronic urticaria status DDD of DL for the total population and stratified by age groups, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis and chronic urticaria status Substudy 2 A Incidence rates of seizure B DL and seizures Substudy 3 A Incidence rates of SVT B DL and SVT Substudy 4 A Incidence rates of A-fib/flu B DL and A-fib/flu Substudy 5 A Incidence rates of first recurrent seizure B DL and first recurrent seizures Supplementary analyses 1 Alternative definition of DL exposure 2 Stratified by febrile and non-febrile seizures among children aged 0–4 3 Exclusion of persons diagnosed with chronic urticaria or have used high doses of DL 4 Stratified by country and restricted to Denmark and Finland merged together 5 Analysis of the potential effect of exposure misclassification 6 Risk after first-ever DL prescription redemption (RX) compared to second, third, fourth (RX)etc. 7 Definition of non-exposed periods changed to 52 weeks following the last prescription redemption 8 Use of an alternative adjustment set of confounders 9 Quantitative bias analysis of confounder misclassification 10 Restricting analyses to period with no over-the-counter sale of DL A-fib/flu= atrial fibrillation or flutter; DL=desloratadine; SVT=supraventricular tachycardia

29 04JCZV

05822K PAGE 246

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 1: Descriptive analysis of the desloratadine use

Table 1.1 Incidence rates and prevalence proportions of desloratadine (DL) users in the general population between 2001 and 2015 for the total population and stratified by age groups, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis and chronic urticaria status. Number of Incidence Prevalence

Number of Person-years at IR (n/100,000 95 % CI Number of General PP (n/100,000 95 % CI incident risk* persons) prevalent population* persons) users users Total population Age groups 0–4 years 5–19 years 20–64 years 65+ years Country Denmark Finland Norway Sweden Sex Males Females Calendar year 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Seasonality Winter

30 04JCZV

05822K PAGE 247

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Number of Person-years at IR (n/100,000 95 % CI Number of General PP (n/100,000 95 % CI incident risk* persons) prevalent population* persons) users users Spring Summer Autumn Asthmatic No registration ------status Asthma ------Severity of No registration ------rhinitis Sever rhinitis ------Chronic No registration ------urticaria Chronic urticaria ------IR=incidence rate; PP=prevalence proportion * Person-years at risk are estimated by use of the population size January 1 each year in 5-year age groups.

31 04JCZV

05822K PAGE 248

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Table 1.2 Distribution of number of DDD of desloratadine (DL) in the total population and stratified by age, sex, country, calendar year, seasonality, asthmatic status, severity of rhinitis and chronic urticaria status among DL users Per DL user Mean SD Median Min-max Total DL population DDD of DL Age groups 0–4 years 5–19 years 20–64 years 65+ years Country Denmark Finland Norway Finland Sex Males Females Calendar years 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Seasonality Winter Spring Summer Autumn Asthmatic status No registration Asthma

32 04JCZV

05822K PAGE 249

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Severity of rhinitis No registration Sever rhinitis Chronic urticaria No registration Chronic urticaria

Substudy 2A: First seizure

Table 2A. Number of persons with a first time diagnoses of seizures, risk time and incidence rate of seizures overall and stratified by age groups, country and sex Variable Level First ever seizure N Risk time (person-years) IR 95% CI Overall Age groups 0–4 5–19 20–64 ≥65 Country Denmark Finland Norway Sweden Sex Males Females

33 04JCZV

05822K PAGE 250

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 2B: Desloratadine and first seizure

Table 2B1. Number of first seizure cases, risk time and incidence rate across persons currently and not currently exposed to desloratadine (DL) Age groups Exposure status level Persons included (N) First seizures (N) Risk time Incidence rate 95% CI All Currently DL exposed Yes Not currently DL exposed No 0–4 years Currently DL exposed Yes Not currently DL exposed No 5–19 years Currently DL exposed Yes Not currently DL exposed No 20–64 years Currently DL exposed Yes Not currently DL exposed No ≥65 years Currently DL exposed Yes Not currently DL exposed No

Table 2B2. Analysis of association between desloratadine (DL) exposure and first seizure Crude model* Fully adjusted model** Age groups Exposure status level IRR 95% CI p-value IRR* 95% CI p-value All Currently DL exposed Yes Not currently DL exposed No 1 Ref 0–4 years Currently DL exposed Yes Not currently DL exposed No 1 Ref 5–19 years Currently DL exposed Yes Not currently DL exposed No 1 Ref 20–64 years Currently DL exposed Yes Not currently DL exposed No 1 Ref ≥65 years Currently DL exposed Yes Not currently DL exposed No 1 Ref * adjusted for age, sex, country and calendar year **adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

34 04JCZV

05822K PAGE 251

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 3A: First supraventricular tachycardia

Table 3A. Number of persons diagnosed with supraventricular tachycardia (SVT), risk time and incidence rate of supraventricular tachycardia overall and stratified by age groups, country and sex Variable Level First ever SVT N Risk time (person-years) IR 95% CI Overall Age groups 0–4 5–19 20–64 ≥65 Country Denmark Finland Norway Sweden Sex Males Females

35 04JCZV

05822K PAGE 252

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 3B: Desloratadine and supraventricular tachycardia

Table 3B1. Number of supraventricular tachycardia (SVT) cases, risk time and incidence rate across persons currently and not currently exposed to desloratadine (DL) Age groups Exposure status level Persons included (N) SVT (N) Risk time Incidence rate 95% CI All Currently DL exposed Yes Not currently DL exposed No 0–4 years Currently DL exposed Yes Not currently DL exposed No 5–19 years Currently DL exposed Yes Not currently DL exposed No 20–64 years Currently DL exposed Yes Not currently DL exposed No ≥65 years Currently DL exposed Yes Not currently DL exposed No

Table 3B2. Analysis of association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) Crude model* Fully adjusted model** Age groups Exposure status level IRR 95% CI p-value IRR* 95% CI p-value All Currently DL exposed Yes Not currently DL exposed No 1 Ref 0–4 years Currently DL exposed Yes Not currently DL exposed No 1 Ref 5–19 years Currently DL exposed Yes Not currently DL exposed No 1 Ref 20–64 years Currently DL exposed Yes Not currently DL exposed No 1 Ref ≥65 years Currently DL exposed Yes Not currently DL exposed No 1 Ref * adjusted for age, sex, country and calendar year **adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmaticstatus, severity of rhinitis, chronic urticaria status.

36 04JCZV

05822K PAGE 253

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 4A: Atrial fibrillation and flutter

Table 4A. Number of persons diagnosed with atrial fibrillation or flutter (A-fib/flu), risk time and incidence rate of atrial fibrillation of flutters overall and stratified by age groups, country and sex Variable Level First ever A-fib/flu N Risk time (person-years) IR 95% CI Overall Age groups 0–4 5–19 20–64 ≥65 Country Denmark Finland Norway Sweden Sex Males Females

37 04JCZV

05822K PAGE 254

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 4B: Desloratadine and atrial fibrillation or flutter

Table 4B1. Number of atrial fibrillation or flutter (A-fib/flu) cases, risk time and incidence rate across persons currently and not currently exposed to desloratadine (DL) Age groups Exposure status level Persons included (N) A-fib/flu (N) Risk time Incidence rate 95% CI All Currently DL exposed Yes Not currently DL exposed No 0–4 years Currently DL exposed Yes Not currently DL exposed No 5–19 years Currently DL exposed Yes Not currently DL exposed No 20–64 years Currently DL exposed Yes Not currently DL exposed No ≥65 years Currently DL exposed Yes Not currently DL exposed No

Table 4B2. Analysis of association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) Crude model* Fully adjusted model** Age groups Exposure status level IRR 95% CI p-value IRR* 95% CI p-value All Currently DL exposed Yes Not currently DL exposed No 1 Ref 0–4 years Currently DL exposed Yes Not currently DL exposed No 1 Ref 5–19 years Currently DL exposed Yes Not currently DL exposed No 1 Ref 20–64 years Currently DL exposed Yes Not currently DL exposed No 1 Ref ≥65 years Currently DL exposed Yes Not currently DL exposed No 1 Ref * adjusted for age, sex, country and calendar year **adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmaticstatus, severity of rhinitis, chronic urticaria status.

38 04JCZV

05822K PAGE 255

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 5A: First recurrent seizure

Table 5A. Number of persons with a first recurrent diagnosis of seizures, risk time and incidence rate of first recurrent seizures First recurrent seizure N Risk time (person-years) IR 95% CI

39 04JCZV

05822K PAGE 256

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Substudy 5B: Desloratadine and first recurrent seizure

Table 5B1. Number of first recurrent seizure cases, risk time and incidence rate across persons currently and not currently exposed to desloratadine (DL) Exposure status level Persons included (N) First recurrent seizure (N) Risk time Incidence rate 95% CI Currently DL exposed Yes Not currently DL exposed No

Table 5B2. Analysis of association between desloratadine (DL) exposure and first recurrent seizure Crude model* Fully adjusted model** Exposure status level IRR 95% CI p-value IRR* 95% CI p-value Currently DL exposed Yes Not currently DL exposed No 1 Ref * adjusted for age, sex, country and calendar year **adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmaticstatus, severity of rhinitis, chronic urticaria status.

40 04JCZV

05822K PAGE 257

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 1

Table S1.1 Analysis of association between desloratadine (DL) exposure and first seizure when using an alternative exposure categorization Age groups Variable First seizures (N) Risk time Overall p-value IRR* 95% CI All 0–4 weeks 5–8 weeks 9–16 weeks 17–26 weeks ≥26 weeks 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S1.2 Analysis of association between desloratadine (DL) exposure and supraventricular tachycardia (SVT) when using an alternative exposure categorization Age groups Variable SVT (N) Risk time Overall p-value IRR* 95% CI All 0–4 weeks 5–8 weeks 9–16 weeks 17–26 weeks ≥26 weeks 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S1.3 Analysis of association between desloratadine (DL) exposure and atrial fibrillation or flutter (A-fib/flu) when using an alternative exposure categorization Age groups Variable A-fib/flu (N) Risk time Overall p-value IRR* 95% CI All 0–4 weeks 5–8 weeks 9–16 weeks 17–26 weeks ≥26 weeks 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

41 04JCZV

05822K PAGE 258

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Table S1.4 Analysis of association between desloratadine (DL) exposure and first recurrent seizure when using an alternative exposure categorization Age groups Variable First recurrent seizures Risk time Overall p-value IRR* 95% CI (N) All 0–4 weeks 5–8 weeks 9–16 weeks 17–26 weeks ≥26 weeks 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

42 04JCZV

05822K PAGE 259

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 2

Table S2.1 Analysis of association between desloratadine (DL) exposure and first febrile seizure and first non-febrile seizure among children age 0–4 years Febrile status Exposure status level First seizures (N) Risk time p-value IRR* 95% CI All Currently DL exposed Yes Not currently DL exposed No 1 Ref Febrile seizures Currently DL exposed Yes Not currently DL exposed No 1 Ref Non-febrile seizures Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S2.2 Analysis of association between desloratadine (DL) exposure and first recurrent febrile seizure and first recurrent non-febrile seizure among children age 0–4 years Febrile status Exposure status level First recurrent seizures (N) Risk time p-value IRR* 95% CI All Currently DL exposed Yes Not currently DL exposed No 1 Ref Febrile seizures Currently DL exposed Yes Not currently DL exposed No 1 Ref Non-febrile seizures Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

43 04JCZV

05822K PAGE 260

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 3

Table S3.1 Analysis of association between desloratadine (DL) exposure and first seizure when excluding persons with a diagnosis of chronic urticaria or high use of DL Exposure status level First seizures (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis.

Table S3.2 Analysis of association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) when excluding persons with a diagnosis of chronic urticaria or high use of DL Exposure status level SVT (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis.

Table S3.3 Analysis of association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) when excluding persons with a diagnosis of chronic urticaria or high use of DL Exposure status level A-fib/flu (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis.

44 04JCZV

05822K PAGE 261

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 4

Table S4.1 Analysis of association between desloratadine (DL) exposure and first seizure for the total population and stratified by country Country Exposure status level First seizures (N) Risk time p-value IRR* 95% CI All Currently DL exposed Yes Not currently DL exposed No 1 Ref Denmark Currently DL exposed Yes Not currently DL exposed No 1 Ref Finland Currently DL exposed Yes Not currently DL exposed No 1 Ref Norway Currently DL exposed Yes Not currently DL exposed No 1 Ref Sweden Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country (only for the analysis of the total population), sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S4.2 Analysis of association between desloratadine (DL) exposure and supraventricular tachycardia (SVT) for the total population and stratified by country Country Exposure status level SVT (N) Risk time p-value IRR* 95% CI All Currently DL exposed Yes Not currently DL exposed No 1 Ref Denmark Currently DL exposed Yes Not currently DL exposed No 1 Ref Finland Currently DL exposed Yes Not currently DL exposed No 1 Ref Norway Currently DL exposed Yes Not currently DL exposed No 1 Ref Sweden Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country (only for the analysis of the total population), sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

45 04JCZV

05822K PAGE 262

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Table S4.3 Analysis of association between desloratadine (DL) exposure and atrial fibrillation of flutter (A-fib/flu) for the total population and stratified by country Country Exposure status level A-fib/flu (N) Risk time p-value IRR* 95% CI All Currently DL exposed Yes Not currently DL exposed No 1 Ref Denmark Currently DL exposed Yes Not currently DL exposed No 1 Ref Finland Currently DL exposed Yes Not currently DL exposed No 1 Ref Norway Currently DL exposed Yes Not currently DL exposed No 1 Ref Sweden Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country (only for the analysis of the total population), sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

46 04JCZV

05822K PAGE 263

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 5

Table S5.1. Analysis of association between alternative exposure definitions (desloratadine (DL) and loratadine or non-sedating antihistamines) and first seizure Exposure definition Exposure status level First seizures (N) Risk time p-value IRR* 95% CI DL and loratadine Currently exposed Yes Not currently exposed No 1 Ref non-sedating antihistamines Currently exposed Yes Not currently exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S5.2. Analysis of association between alternative exposure definitions (desloratadine (DL) and loratadine or non-sedating antihistamines) and first supraventricular tachycardia (SVT) Exposure definition Exposure status level SVT (N) Risk time p-value IRR* 95% CI DL and loratadine Currently exposed Yes Not currently exposed No 1 Ref non-sedating antihistamines Currently exposed Yes Not currently exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S5.3. Analysis of association between alternative exposure definitions (desloratadine (DL) and loratadine or non- sedating antihistamines) and first atrial fibrillation or flutter (A-fib/flu) Exposure definition Exposure status level A-fib/flu (N) Risk time p-value IRR* 95% CI DL and loratadine Currently exposed Yes Not currently exposed No 1 Ref non-sedating antihistamines Currently exposed Yes Not currently exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

47 04JCZV

05822K PAGE 264

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 6

Table S6.1. Analysis of association between desloratadine (DL) exposure and first seizure stratified by number of prescription redemptions Exposure status First seizures (N) Risk time p-value IRR* 95% CI IRR* 95% CI Currently DL exposed following first prescription redemption 1 Ref Currently DL exposed following second prescription redemption Currently DL exposed following third or more prescription redemption Not currently DL exposed 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S6.2. Analysis of association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) stratified by number of prescription redemptions Exposure status SVT (N) Risk time p-value IRR* 95% CI IRR* 95% CI Currently DL exposed following first prescription redemption 1 Ref Currently DL exposed following second prescription redemption Currently DL exposed following third or more prescription redemption Not currently DL exposed 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S6.3. Analysis of association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) stratified by number of prescription redemptions Exposure status A-fib/flu (N) Risk time p-value IRR* 95% CI IRR* 95% CI Currently DL exposed following first prescription redemption 1 Ref Currently DL exposed following second prescription redemption Currently DL exposed following third or more prescription redemption Not currently DL exposed 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

48 04JCZV

05822K PAGE 265

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 7

Table S7.1 Analysis of association between desloratadine (DL) exposure and first seizure when non-exposed periods starts 52 weeks from last prescription redemption Exposure status level First seizures (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S7.2 Analysis of association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) when non-exposed periods starts 52 weeks from last prescription redemption Exposure status level SVT (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S7.3 Analysis of association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) when non- exposed periods starts 52 weeks from last prescription redemption Exposure status level A-fib/flu (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

49 04JCZV

05822K PAGE 266

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 8

Table S8.1 Analysis of association between desloratadine (DL) exposure and first seizure when adjusting for an alternative confounder adjustment set Exposure status level First seizures (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, sex, country, calendar year, seasonality, severity of rhinitis, asthma status, diabetes, hypo/hyper-thyroidism, inflammatory disease, infections, and type 1 allergy

Table S8.2 Analysis of association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) when adjusting for an alternative confounder adjustment set Exposure status level SVT (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, sex, country, calendar year, seasonality, severity of rhinitis, asthma status, diabetes, hypo/hyper-thyroidism, inflammatory disease, infections, type 1 allergy, and antihypertensive treatment

Table S8.3 Analysis of association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) when adjusting for an alternative confounder adjustment set Exposure status level A-fib/flu (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, sex, country, calendar year, seasonality, severity of rhinitis, asthma status,diabetes, hypo/hyper-thyroidism, inflammatory disease, infections, type 1 allergy, and antihypertensive treatment

50 04JCZV

05822K PAGE 267

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 9

As mentioned earlier in the description of Supplementary analysis 9, Table S9.1. will show the analysis if data fit the summary approach and Table S9.2 will show the analysis if data fit the individual record level approach, hence only one of the two tables will be used.

Table S9.1. Quantitative Bias Analysis (simple approach): misclassification of asthmatic status, severity of rhinitis, and chronic urticaria status Scenario Assumed specificity Assumed sensitivity IRR 0 1 1 (uncorrected) 1 0.9995 0.8 2 0.9995 0.6 3 0.9995 0.4 4 0.9995 0.2 5 0.9950 0.8 6 0.9950 0.6 7 0.9950 0.4 8 0.9950 0.2 9 0.9900 0.8 10 0.9900 0.6 11 0.9900 0.4 12 0.9900 0.2

Table S9.2. Quantitative Bias Analysis (individual record level analysis): misclassification of chronic urticaria status, severity of rhinitis and asthmatic status* Analysis Point estimate (median)** 95 % simulation interval (2.5th and 97.5th percentiles) Conventional (random error only) Systematic error (misclassification of confounders) Total error * Specificity: min=0.985, mode1=0.990, mode2=0.995, max=1.00; sensitivity: min=0.2, mode1=0.4, mode2=0.6, max=0.8 **All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

51 04JCZV

05822K PAGE 268

Product: MK-4117 Protocol No.: 4117-203 Statistical Analysis Plan Epidemiology No.: EPO7044.002

Supplementary analysis 10

Table S10.1 Analysis of association between desloratadine (DL) exposure and first seizure in periods without over-the- counter sales Exposure status level First seizures (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S10.2 Analysis of association between desloratadine (DL) exposure and first supraventricular tachycardia (SVT) in periods without over-the-counter sales Exposure status level SVT (N) Risk time p-value IRR* 95% CI

Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

Table S10.3 Analysis of association between desloratadine (DL) exposure and first atrial fibrillation or flutter (A-fib/flu) in periods without over-the-counter sales Exposure status level A-fib/flu (N) Risk time p-value IRR* 95% CI Currently DL exposed Yes Not currently DL exposed No 1 Ref *All analyses adjusted for confounders in the minimum sufficient adjustment set: age, country, sex, calendar year, seasonality, asthmatic status, severity of rhinitis, chronic urticaria status.

52 04JCZV

05822K DESLORATADINE PAGE 269 EPIDEMIOLOGY NO.: EP07044.002 PROTOCOL NO/AMENDMENT NO.: MK-4117-203 EU PAS REGISTER NO./EUDRACT NO.: 15038

ANNEX 6 Variable Definitions

(EU GUIDANCE: 23 JANUARY 2013 EMA/738724/2012)

05822K PAGE 270

Annex 6 Variable Definitions

Variable Definition Data source ATC ICD-8 ICD-9 ICD-10

EXPOSURE

Desloratadine R06AX27

Loratadine R06AX13

Non-sedating R06AE07, antihistamines R06AE09, R06AX12, Prescription Exposure status R06AX13, register R06AX18, R06AX22, R06AX25, R06AX26, R06AX27, R06AX28, R06AX29 OUTCOME VARIABLES

Non-febrile seizure - - R568*

Febrile seizure - - R560*

Supraventricular 42790 4270 I47* tachycardia A binary variable with categories yes and no Patient register Atrial fibrillation or 42793, 42794 4273 I48* flutter

Recurrent non-febrile - - R568* seizure

05822K PAGE 271

Variable Definition Data source ATC ICD-8 ICD-9 ICD-10

VARIABLES USED FOR ADJUSTMENT Age A time varying variable categorized as 0-4, 5-19, 20- Civil 64, ≥65 years registration system Sex Male versus female Civil registration system Country Country of residence with categories Denmark, Civil Finland, Norway, and Sweden registration system Calendar year A time-varying variable Calendar years 2001, 2002, 2003, …, 2015 for Prescription prescriptions except for Tables 2A, 3A, 4A and 5B, register; where calendar year is defined for incident Patient diagnoses (seizure, SVT, atrial fibrillation and register recurrent seizure, respectively) Seasonality A time-varying variable Prescription register Winter (December–February), spring (March–May), summer (June–August), and autumn (September– November). Asthma status A binary variable with categories yes and no, based on prescriptions and/or contacts to the hospital Redeemed treatment for asthma defined as at least Prescription R03A* two prescriptions of inhalant steroids within a six- register R03B* month period during a five-year period before first DL exposure. First registered asthma prescription has to be redeemed when the purchaser was 45 years or younger

05822K PAGE 272

Variable Definition Data source ATC ICD-8 ICD-9 ICD-10

VARIABLES USED FOR ADJUSTMENT Contacts to hospitals with a diagnosis of asthma Patient - - J45* (including both primary and secondary diagnoses) register during a five-year period before first DL exposure Severe rhinitis A binary variable with categories yes and no, based on prescriptions Redeemed immunotherapy at least once during a Prescription V01AA* five-year period before first DL exposure register (excluding V01AA07) Chronic urticaria A binary variable with categories yes and no, based on hospitals contacts Registered diagnosis of chronic urticaria in the five- Patient L50* year period before first DL exposure. register (excluding L50.8D), L56.3 Diabetes A binary variable with categories yes and no, based on prescriptions and/or contacts to the hospital Redeemed at least two prescriptions of glucose- Prescription A10* lowering drugs in the five-year period before first register DL exposure Registered diagnosis of diabetes in the five-year Patient E10*-E14* period before first DL exposure register Hypo- A binary variable with categories yes and no, based /hyperthyroidism on prescriptions and/or contacts to the hospital Redeemed at least two prescriptions of drugs for Prescription H03BB02 treatment of hypo-/hyperthyroidism in the five-year register H03AA01 period before first DL exposure Registered diagnosis of hypo-/hyperthyroidism Patient E00*-E07* register

05822K PAGE 273

Variable Definition Data source ATC ICD-8 ICD-9 ICD-10

VARIABLES USED FOR ADJUSTMENT Inflammatory A binary variable with categories yes and no, based disease on prescriptions and/or contacts to the hospital Redeemed at least two prescriptions of Daivonex Prescription D05AX02 (drug against psoriasis) in the five-year period register before first DL exposure Registered diagnosis of inflammatory bowel Patient K50*-K52* disease, psoriasis, rheumatic diseases, vasculitis, or register L40* sarcoidosis in the five-year period before first DL M05*-M08* exposure L95* D86* Infection A binary variable with categories yes and no, based on hospitals contacts Registered diagnosis of lung infection (pneumonia) Patient J12*-J18*, or sinusitis in the five-year period before first DL register J21*-J22* exposure. J01*, J32* Type 1 allergy A binary variable with categories yes and no, based on prescriptions and/or contacts to the hospital Redeemed at least one prescription of Prescription V01AA immunotherapy drugs in the five-year period before register first DL exposure Registered diagnosis of acute urticaria, anaphylaxis, Patient L50.8D* quinckes oedema in the five-year period before first register T78* DL exposure A binary variable with categories yes and no, based on hospitals contacts

05822K PAGE 274

Variable Definition Data source ATC ICD-8 ICD-9 ICD-10

VARIABLES USED FOR ADJUSTMENT Antihypertensive Redeemed antihypertensive prescription (ACE Prescription C09A* treatment inhibitor, angiotensin II receptor antagonist, calcium register C09C* channel blocker, beta-blockers, alpha-blockers, C08CA* thiazide (diuretic treatment), methyldopa, or C07AB* moxonidine) at least twice during a five-year period C02CA* before first DL exposure C03A*-C03E* C02AB* C02AC05 * Subcodes

05822K