Optimal Pharmacovigilance:

Why it is important and how to do it

Steven Black, MD Professor of Pediatrics Division of Infectious Diseases Cincinnati Children’s Hospital Cincinnati, Ohio USA Presentation Overview

• Passive Reporting Systems: VAERS as an example • Broad post licensure evaluations of safety: Phase Four studies, VSD Datalink, and rapid cycle studies • Background rates of events versus vaccine related events • Passive versus Active Surveillance • Global collaborative efforts Passive Reporting: VAERS as an example • VAERS ( Reporting System) relies on passive reporting by physicians or patients. – No real denominator data. – Only cases after vaccine are reported • Statistical techniques have been developed for signal detection using “data mining”). – Relies on comparing expected versus observed numbers of events following vaccine using Bayesian statistics. – In a retrospective analysis evaluating this technique, intussusception following Rotashield vaccine identified early. • EBGM ( Empirical Bayesian Geometric Mean) score elevated for – Intussusception – Diarrhea, GI hemorrhage Passive Reporting: Intussusception Cases Reported to VAERS after Rotashield

45

40 MMWR

35

30

25

20 Cases Cases 15

10

5

0 gen- feb- mar- apr- mag- giu- lug- ago- set- ott- nov- 99 99 99 99 99 99 99 99 99 99 99

4 Empirical Bayesian Geometric Mean Score following Rotashield 20

18

16

14

12

10 EBGM for Intussusception 8

6

4

2

0 Jan Feb Mar April May June July 1 5 Passive Reporting Summary: Intussusception and VAERS

• VAERS was excellent at signal detection for intussusception. • Additional Statistical techniques can lead to earlier detection of vaccine safety signals. • Passive reporting systems, however: – Are prone to bias in that events closest to vaccine are reported preferentially – Underreporting is a major issue. – Misclassification of events can occur. – Information on cases may be limited. – Recognized adverse events are more likely to be reported. – Do not allow assessment of relative or attributable risk. Active Post Marketing Surveillance

Phase Four studies by manufacturers

Datalink Studies

Rapid cycle analyses

Global Collaborations

7 Vaccine Data Link studies vs Passive Reporting Systems.

• Unlike passive reporting systems, data link studies have the potential to – Identify cases of events in an unbiased way. – Allows analysis of unexposed cases – Allow access to medical records to better characterize and understand cases. – Calculate rates of events, relative risk and attributable risk. – Evaluate vaccine impact and changes in disease What Data do you Need?

Computerized Vaccine Data on an individual level Exposure Need a unique identifier to link Outcome or Possible “Adverse Event” Computerized Hospital and/or Clinic Diagnoses NOTE: Can do case series with On an individual level Outcome Data alone to assess risk and hence causality

Demographic Data on a Population Adjust for confounders Evaluation of Uncommon Events Resulting in Medical Utilization: Case Series • Allows calculation of relative risk using only cases of the outcome.

• Probability of being vaccinated within a specified time window prior to is compared to probability of being vaccinated at other times. If there is no association, then these two probabilities should be equal yielding an odds ratio of one. • For self control comparison, exposed and unexposed follow up time are defined for the same individual. • Resolves issue of differences between groups • Issue of “well child effect” bias for pre-exposure time interval • Does not work if event prevents vaccination vax vax

Pre Exposed Post 1 Exposed 2 Post /Pre 2 Developed by Farrington Rapid Cycle Techniques

• Requires rapid access to automated data sources to establish rates of events in baseline and following a new vaccine. Currently in use in VSD with weekly data pulls. • Uses sequential probability testing to test the likelihood of an observed number of cases versus expected. • A “stopping value” is established to either rule in or rule out an association based upon the expected number of cases and risk level. • For intussusception, a risk established after 2589 doses of vaccine given… a similar time frame as VAERS data mining. • Also able to detect a decrease in risk of seizures, fever and other abnormal neurologic events within 12 weeks of introduction of DTaP as compared to DTwP. Rapid Cycle Techniques Sequential Probability Testing: Two Patterns

Intussusception and Rotashield

Facial Paralysis and Men ACWY

12 What do these analyses look like?

An example

Prelicensure Study of MMR-V vs Post Licensure Evaluation in VSD A Case Study: MMR-V Pre-licensure vs Post Licensure Overview • Pre-licensure studies of safety are usually small – Focus on common local and systemic events – Analyses done within predefined windows • Post licensure studies are usually much larger and have ability to look at events more flexibly MMR-V Pre-licensure Safety Black et al. PIDJ 24:8-12, 2005 MMRV MMR & V P-value N=323 N=157 Fever 0-42 39.6% 34.8% ns days s/p vax Fever 5-12 27.7% 18.7% 0.034 days s/p vax

Seizures 1 1 on day 1 ns on day 9

15 VSD Rapid Cycle Evaluation of MMRV Outpatient Visits for Fever by Day after Vaccine at Northern California Kaiser Permanente: 1995-2008

Age 12-23 months 6241 total fever visits after 302,670 MMR+V, 147,762 MMR, 46,390 MMRV, 38,251 VZV

350 MMR 300 MMR+V MMRV 250 V 200 150

100

Events Events / Doses 100,000 50 0 0 5 10 15 20 25 30 35 40

Days after VSD Rapid Cycle Evaluation of MMRV Temporal distribution of seizures after MMRV vaccination vs after simultaneous MMR and varicella vaccination • 25 60

50 20

40 15 30 10 20

5 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42

Days Post-MMRV Vaccine Days Post-MMR + (2/06-9/07) (1/00-8/07)

Nicola Klein as presented at PAS/SPR in Honolulu May 2008 VSD Rapid Cycle Evaluation of MMRV Evaluating for the risk of seizure 7-10 days post- vaccination using chart verified febrile seizures 95% Confidence P-value Odds ratio Interval MMRV versus MMR + 2.3 1.6, 3.2 <0.0001 Varicella* Attributable 95% Confidence Risk Interval MMRV versus MMR + 5.2 / 10,000 2.2, 8.1 Varicella* N= 42 for MMR-V out of 43,353 MMR-V and 124 for 314,599 doses of MMR + V for chart review confirmed seizures. Adjusted for age and influenza season. Increased risk with MMRV cannot be explained by concomitant , temporal trends in seizure, VSD site, age or influenza season. 18 A Comparison of Active versus Passive Surveillance in Burkino Faso • Men A vaccine was introduced into the Sub- Saharan African meningitis belt in December 2010 and almost 12 million people vaccinated in one campaign in Burkino Faso • Two safety surveillance systems established – Nation wide passive surveillance – Active Surveillance in one district

VACCINE in press MenA Introduction in Burkino Faso: Passive Surveillance Data

Per 100,000 MenA Introduction in Burkino Faso: Active Surveillance Data

Per 100,000 Per 100,000

21 MenA Introduction in Burkino Faso: Active Surveillance Data Summary: Active vs Passive Surveillance

• Passive surveillance can be useful for signal detection • Under-reporting can severely limit the usefullness of passive reporting • Active surveillance in a defined sub-population can be more useful than passive surveillance in a larger population. • Both types of surveillance are feasible in developed and developing countries.

23 Summary of Advantages of Population Based Database Studies • Allow calculation of incidence on AEFI and background rates of disease without vaccination • Allow calculation of relative risk • Allow calculation of attributable risk. • Allow adjustment for confounders • Allow assessment of trends including vaccine impact on disease for risk-benefit analyses What to do if you observe a possible signal?

• Evaluation of a possible consistent time association of the event with vaccination • Evaluation in a different analytic framework: self- control analysis or other reference group • Possible associations can serve as a source of hypothesis generation for further studies – ie case- control study conducted for intussusception. The first step in looking at a possible signal:

What happens without vaccines? The importance of knowing background epidemiology

26 What happens without vaccines? “AEFI” without Vaccines: Autoimmune < 30days Outpatient Events in Teens

# events Rate/100,000 Rate Outpatient care Adolescents py Teens Adults Thyroid (…) disorders 859 396 1412.05 Ulcerative 556.x colitis 76 35.4 117.52 Regional 555.x enteritis 68 31.6 97.18 Systemic lupus 7100 erythematosus 63 52.9 120.23 Rheumatoid 7140 arthritis 29 13.5 119.33 37730 Optic neuritis 10 4.7 13.56 Multiple 340 sclerosis 9 4.2 64.18 71659 Polyarthritis 7 3.3 30.74 Claire Anne Seigrist PIDJ Pandemic Flu Safety The importance of background rates of disease in assessment of vaccine safety during mass immunization with H1N1 influenza vaccines Where to from here? • Increasing the sample size and geographic reach – Regional and global consortiums • Maximizes statistical power • Increases local expertise – EXAMPLES: • VAESCO • SOMNIA • WHO GVSI Pilot SOMNIA STUDY: EVALUATING THE RISK OF NARCOLEPSY FOLLOWING ADJUVANTED 2009 H1N1 PANDEMIC VACCINES

BACKGROUND AND RATIONALE

30 Narcolepsy • CNS disorder characterized by excessive daytime sleepiness, abnormal manifestations of REM sleep

o Sleep attacks, disrupted nocturnal sleep, sleep paralysis, hypnagogic hallucinations, cataplexy

o Chronic disease, treated with medication and behavior modification, no cure • Two diagnostic entities

o Narcolepsy with cataplexy, narcolepsy without cataplexy • Brighton Collaboration case definition exists*

o Primarily based on presence of symptoms and an abnormal multiple sleep latency test (MSLT) characteristic of narcolepsy

• Poli et al. Narcolepsy as an adverse event following immunization: case definition and • guidelines for data collection, analysis and presentation. Vaccine. 2013;31(6):994-1007. 31 Narcolepsy pathogenesis

• Narcolepsy with cataplexy thought to be caused by damage to hypocretin-1 secreting neurons in the hypothalamus • Strongly associated with the HLA DQB1*0602 allele

o Present in 5%-38% of people • Thought to be autoimmune and a multi-event process • Suspected infectious triggers include

o Febrile illness

o Influenza

o β hemolytic streptococcal infections 32 Narcolepsy epidemiology*

• Onset can occur at any age, but peaks during the teenage years

o Very rare under age 5 years old

o Rare after age 40 years old • Prevalence estimates vary widely

o Israel: 0.23 per 100,000

o Unites States: 30-56 per 100,000

o Japan: 160 per 100,000 • There is often a long delay from symptom onset to diagnosis (can be many years)

* Silber et al. The epidemiology of narcolepsy in Olmsted County, Minnesota: a population-based study. Sleep 2002;25:197-202; Longstreth et al. Prevalence of narcolepsy in King County, Washington, USA. Sleep Med 2009;10:422-426; Partinen et al. Narcolepsy as an autoimmune disease: the role of H1N1 and vaccination. Lancet neurology 2014;13:600-613. 33 Summary of data on and narcolepsy in children in Europe 2009-2010

34 Hum Vaccin Immunother. 2016 Jan; 12(1): 187–193. Methods: Incidence Rates Study Dynamic retrospective cohort database study

Periods • Before the H1N1 pandemic • During the H1N1 pandemic but pre-vaccination • During/post-H1N1 pandemic and vaccination

Population • 10 databases in 7 countries (540 million person years) Data • Electronic healthcare databases (General Practitioner [GP], source claims) Validation 1. Positive predictive value for the period of case control study 2. All cases in the database (Netherlands, FISABIO)

Analysis • By country, by age group, by period • Join point analysis, IRR between periods

Level of • Differences in the level of granularity that could be submitted detail for privacy reasons

35 Incidence rates (total population) 2003-2013 Site Period Event Person- Incidence* 2003- s time 2013

European Union Denmark 2003-2013 269 17,850,129 1.50 (1.33-1.69) United Kingdom 2003-2013 467 42,897,721 1.09 (0.99-1.19) Spain-Valencia 2009-2013 46 20,458,082 0.22 (0.17-0.28) (validated) Spain-Cataluña 2007-2013 240 34,861,809 0.69 (0.50-0.78) Sweden 2003-2013 1,536 102,027,209 1.52 (1.43-1.59) Netherlands (validated) 2003-2013 14 2,879,712 0.49 (0.29-0.76)

Non-EU Canada-British 2003-2013 278 47,857,684 0.58 (0.32-0.64) Columbia Canada-Alberta 2003-2013 427 51,885,946 0.82 (0.74-0.90) Canada-Manitoba 2003-2010 42 6,405,888 0.66 (0.50-0.86) Taiwan 2003-2012 472 161,407,503 0.29 (0.27-0.32) *Per 100,000 person years 36 Incidence rate in Sweden over time (signaling country)

• An increase in the rate of narcolepsy diagnosis in Sweden • Going up in 2010 and coming down again in 2012

20 Incidence rates in other SOMNIA countries over time

• No change in incidence rates over time in any of the countries or age groups beyond Sweden and in

Taiwan during circulation of wildtype virus 38 SOMNIA

CASE CONTROL STUDY

39 Methods: Retrospective case control study

Index date • Primary: Multiple sleep latency test (MSLT) referral • Secondary: Excessive daytime sleepiness (EDS) or cataplexy onset Case definition • Children: Brighton Classification levels 1-2 • Adults: Brighton Classification levels 1-4 Case finding • Primarily through sleep centers Case • Blinded review locally ascertainment Exposure • AS03, MF59, other H1N109pdm and seasonal influenza vaccines, HPV vaccine Covariates • Infections, comorbidities Analyses • By country, by vaccine, pooled

Other data • Virus circulation 40 Total Follow-up period analysis by vaccine brand in children no)

* Odds ratio (any time versus

• No association for Focetria (MF59), Arepanrix or Pandemrix (AS03) and narcolepsy in children *Pandremix (AS03) (Netherlands): estimated based on a case coverage study in children born between 2004-2009 41 Total followup period analysis by vaccine brand in adults no) Odds ratio (any time versus

• No association for Focetria (MF59), Arepanrix or Pandemrix (AS03) and narcolepsy in adults

42 Conclusions

• Incidence rate data did not show a rise in the rate of narcolepsy following vaccination except in the one signaling country included (Sweden, which used Pandemrix) • Case control analyses for AS03-adjuvanted pandemic H1N1 vaccines (Arepanrix and Pandemrix) did not show evidence of an increased risk of narcolepsy following vaccination in children or adults

o Though data were limited for Pandemrix • Cases control analyses for Focetria (MF59) revealed an non-significant elevated risk estimate in children in the restricted period analysis but not in the total period analysis which included more follow up time 43 The most recent effort: GVSI MCC Pilot

32 countries initially screened for eligibility of 25 hospitals in PAHO Rest of the World 16 countries representing all WHO regions selected Argentina (6 sites) Albania Chile (4 sites) Australia (2 sites) Peru China Uruguay India Costa Rica Iran (2 sites) Honduras Singapore Colombia South Africa Spain (Uganda)

44 GVSI Pilot: Roles and responsibilities

• WHO: – Overall study coordination – Responsible for identification of non-PAHO candidate sites • PAHO: responsible for recruitment & coordination PAHO sites • VACCINE.GRID: responsible for training, quality control, data collection tools and analysis • Funding: US – FDA and GRIP (FP7-EC)

45 Study Goals and Outcomes

• To assess the feasibility, quality and potential for sustainability of a global collaboration using hospital diagnostic data to assess two known vaccine associations – one negative and one positive: – Aseptic meningitis following vaccines (negative) – ITP following measles vaccines (positive) • Using a common protocol and case report forms • Common training tools • Using a central data repository with state of the art protection of patient information • To learn from this effort to inform potential future collaborations – Using a structured survey of participants. – By review of the data quality

*NB: measles is not associated with aseptic meningitis but containing vaccines are to various degrees

46 Training of sites

Non-PAHO PAHO region: – Field trips to Iran, India & – organized several face to face Uganda by WHO & meetings with presentations on the VACCINE.GRID methods & data collection – Training webinars were – Training webinars were organized organized by VACCINE.GRID by Vaccine GRID – Videos (English language) of the – Videos (Spanish language) of the tools and English language tools and Spanish language Manual Manual of Procedures for data of Procedures for data collection collection methods methods – Dummy case abstraction – Close monitoring of the sites by training and submissions and PAHO feedback by Vaccine.GRID – Dummy case abstraction training – Initial data submission & quality and submissions and feedback by control Vaccine.GRID

47 MMR Vaccine types used at Study Sites

48 ITP Risk by type of measles strain

Overall for Any strain SCRI IRR= 5.0 95% CI=2.5-9.7

CCO: OR= 4.7 95% CI= 2.1-10.7 Strain specific SCRI Schwarz: IRR=20.7 95% CI: 2.7-157.6 Edmonston-Zagreb IRR: 11.1 95% CI: 1.4-90.3 Enders´Edmonston IRR: 8.5 95% CI: 1.9-38.1

49 Aseptic Meningitis Risk by type of mumps strain

Any strain SCRI IRR= 10.9 95% CI= 4.2-27.8 Iran CCO: OR= 35.0 95% CI= 4.8-256

Strain specific SCRI Leningrad-Zagreb: IRR=10.8 95% CI= 1.3-87.4

Hoshino/LZ/Urabe (Iran only): IRR= 20.3 95% CI= 4.8-85

50 GVSI Pilot: Conclusions

• The study proved global collaboration to evaluate vaccine safety using a common protocol across diverse countries is feasible • Valid data were generated and known associations were confirmed • The study showed the value of global diversity in that the difference between strains in their risk of AM could be shown in the same study using a common protocol. • Given the rarity of many adverse events, global collaboration is indeed essential.

51 Overall Summary • There are multiple modalities to evaluate safety in the post- licensure setting – Passive reporting signal detection systems such as VAERS – Population based systems such as VSD, HPA. – International collaborations such as VAESCO in Europe – The global vaccine safety data network/GVSI Pilot Network. • Public expectation is for a rapid response to possible concerns. – Without this confidence is undermined and “anti-vaccine” groups get ahead. • Concerns will arise because events occur and occur in clusters in both vaccinees and non vaccinees due to chance alone. • The globalization of manufacturing and new vaccine introductions dictates that vaccine safety assessment is now the responsibility of all countries. • Global collaborations can increase efficiency and capacity

52 Finally….

• It is important to remember that you have already done this once !

• “The hospital surveillance was the key safety monitoring activity, its primary aim being to rapidly identify AEFIs of sufficient severity to require hospital admission and, for some conditions, emergency department (ED)

53 consultation. “