I4V-MC-B011 Non-interventional PASS Protocol Page 1
Post-Authorisation Safety Study (PASS) Information
Study I4V-MC-B011: A Retrospective Cohort Study to Assess the Safety of Baricitinib Title Compared with Other Therapies Used in the Treatment of Rheumatoid Arthritis in Nordic Countries Version identifier Version 1.0 Date of last version 23 Jul 2018 EU PAS Register No.: EUPAS25151 Active substance Baricitinib; ATC code L04AA37 Medicinal product(s): Olumiant (baricitinib) 2-mg and 4-mg film-coated tablets Product reference: EU/1/16/1170 Procedure number: EMEA/H/C/004085 Marketing authorisation Eli Lilly Nederland BV, Papendorpseweg 83, 3528BJ Utrecht, The Netherlands holder(s) Joint PASS No Research question and This study aims to evaluate the safety of baricitinib among rheumatoid arthritis (RA) objectives patients treated in routine clinical care. Primary objectives: (1) to compare the incidence rates and profiles of: serious infections overall (including herpes zoster) and opportunistic infections (including tuberculosis, Candida infections, and progressive multifocal leukoencephalopathy); major adverse cardiovascular events; malignancies overall (including lymphoma and typically virus-induced malignancies such as cervical and many oropharyngeal cancers); and venous thromboembolism, among RA patients with long-term exposure to baricitinib compared to similar patients with long-term exposure to other indicated medications; (2) to describe the incidence rates of the following individual outcomes: lymphoma; herpes zoster; opportunistic infections such as tuberculosis, Candida, and progressive multifocal leukoencephalopathy; rhabdomyolysis; agranulocytosis; hyperlipidaemia(hypercholesterolaemia, hypertriglyceridaemia); gastrointestinal perforations; and liver injury. Secondary objectives: (3) to monitor the incidence rates of the aggregate outcomes of serious infections overall, MACE, malignancies overall, and VTE in very elderly patients, that is, those ≥75 years of age; (4) to assess the effectiveness of risk minimisation activities by describing the pattern of use of baricitinib and the occurrence of pregnancy, active tuberculosis or active viral hepatitis, and monitoring and treatment of lipid levels in relation to such use in routine clinical care. Countries of study Denmark, Finland, Norway, and Sweden Author PPD PPD
E-mail: PPD Phone: PPD
Approval Date: 29-Nov-2018 GMT I4V-MC-B011 Non-interventional PASS Protocol Page 2
Marketing Authorisation Holder
Marketing authorisation Eli Lilly Nederland BV, Papendorpseweg 83, 3528BJ Utrecht, The Netherlands holder (MAH) MAH contact person PPD Eli Lilly and Company Lilly Corporate Center Indianapolis, IN 46285 United States Email: PPD Telephone: PPD I4V-MC-B011 Non-interventional PASS Protocol Page 3
1. Table of Contents
Section Page 1. Table of Contents...... 3 2. List of Abbreviations ...... 9 3. Responsible Parties ...... 11 4. Abstract ...... 12 5. Amendments and Updates...... 14 6. Milestones...... 15 7. Rationale and Background ...... 16 8. Research Question and Objectives...... 18 9. Research Methods...... 19 9.1. Study Design...... 19 9.2. Setting...... 19 9.2.1. Study Population...... 20 9.2.2. Eligibility...... 20 9.2.3. Patient Subgroups of Special Interest ...... 21 9.3. Variables...... 21 9.3.1. Drug Exposure...... 21 9.3.1.1. Drug Exposure and Cohort Identification for Malignancy Analyses...... 23 9.3.1.2. Drug Exposure and Cohort Identification for Nonmalignancy Analyses ...... 23 9.3.1.3. Example of Exposure Classification ...... 24 9.3.1.3.1. Medication Restarts ...... 26 9.3.2. Outcomes...... 28 9.3.2.1. Malignancy, Excluding Nonmelanoma Skin Cancer ...... 28 9.3.2.2. Nonmelanoma Skin Cancer ...... 28 9.3.2.3. Serious Infections ...... 29 9.3.2.4. Opportunistic Infections ...... 29 9.3.2.5. Major Adverse Cardiovascular Events...... 30 9.3.2.6. Venous Thromboembolism...... 30 9.3.2.7. Gastrointestinal Perforations...... 30 9.3.2.8. Liver Injury ...... 30 9.3.2.9. Rhabdomyolysis ...... 31 9.3.2.10. Hyperlipidaemia (Including Hypercholesterolaemia and Hypertriglyceridaemia)...... 31 I4V-MC-B011 Non-interventional PASS Protocol Page 4
9.3.2.11. Myelosuppresion (Agranulocytosis) ...... 31 9.3.3. Covariates...... 31 9.4. Data Sources ...... 32 9.4.1. The Nordic Health System ...... 32 9.4.2. National Clinical Rheumatology Registries ...... 33 9.4.3. National Hospital Registries...... 35 9.4.4. Prescription Data...... 36 9.4.5. The Medical Birth Registers...... 37 9.4.6. Education and Income Data...... 37 9.4.7. Quality Assurance and Quality Control ...... 37 9.4.8. Data Period ...... 38 9.5. Study Size ...... 38 9.6. Data Management ...... 40 9.6.1. Data to Be Collected ...... 40 9.6.1.1. Missing Data ...... 40 9.7. Data Analysis...... 41 9.7.1. Analysis Population ...... 41 9.7.2. Background and Rationale for Propensity Scores ...... 42 9.7.3. Propensity-Score Definition and Estimation ...... 42 9.7.4. Using the Propensity Score to Address Channelling Bias ...... 42 9.7.5. Evaluation of the Propensity-Score Model and Stratification...... 43 9.7.6. Malignancy Analysis...... 43 9.7.6.1. Malignancy Excluding Nonmelanoma Skin Cancer ...... 43 9.7.6.2. Nonmelanoma Skin Cancer ...... 45 9.7.7. Nonmalignancy Analyses...... 46 9.7.7.1. Serious Infections ...... 46 9.7.7.2. Opportunistic Infections ...... 47 9.7.7.3. Major Adverse Cardiovascular Events...... 48 9.7.7.4. Venous Thromboembolic Events...... 48 9.7.7.5. Analysis of Individual Outcomes (Objective 2)...... 49 9.7.7.6. Patient Subgroups of Special Interest...... 50 9.7.8. Sensitivity Analysis...... 50 9.7.8.1. Malignancy, Excluding Nonmelanoma Skin Cancer ...... 50 9.7.8.1.1. Assessment of the Association between Duration of Baricitinib Exposure and Malignancy Incidence ...... 50 9.7.8.1.2. Assessment of the Association between JAK Inhibitors and Malignancy ...... 51 I4V-MC-B011 Non-interventional PASS Protocol Page 5
9.7.8.1.3. Assessment of the Association between Baricitinib Exposure and Malignancy, within Periods of Time after Initiating Medication...... 51 9.7.8.1.4. Assessment of the Association between Baricitinib Exposure and Malignancy, Allowing for Various Latency Periods ...... 51 9.7.8.2. Nonmelanoma Skin Cancer ...... 51 9.7.8.3. Nonmalignancy Outcomes, All...... 52 9.7.8.4. Serious Infections ...... 52 9.7.8.4.1. Serious Infections Including Recurrent Infections ...... 52 9.7.8.4.2. Including Patients with a History of Serious Infection...... 52 9.7.8.5. Opportunistic Infections ...... 52 9.7.8.5.1. Including Recurrent Opportunistic Infections...... 52 9.7.8.5.2. Including Patients with History of Opportunistic Infection ...... 53 9.7.8.6. Major Adverse Cardiovascular Events...... 53 9.7.8.6.1. Intent-to-Treat Analysis of the Association between Baricitinib Exposure and MACE...... 53 9.7.8.6.2. Assessment of the Association between Duration of Baricitinib Exposure and MACE Incidence...... 53 9.7.8.7. Venous Thromboembolic Events ...... 53 9.7.9. Assessment of Effectiveness of Risk Minimisation Activities...... 53 9.8. Quality Control ...... 54 9.9. Limitations of the Research Methods...... 54 9.9.1. Channelling Bias in Observational Studies ...... 55 9.9.2. Assessing Malignancy Risk in a Real-World Setting ...... 55 9.9.3. Evaluation of Patient Outcomes ...... 56 9.9.4. Generalisability...... 56 9.10. Other Aspects...... 57 9.10.1. File Retention and Archiving ...... 57 10. Protection of Human Subjects ...... 58 11. Management and Reporting of Adverse Events/Adverse Reactions ...... 59 12. Plans for Disseminating and Communicating Study Results...... 60 13. References ...... 61 14. Annex ...... 68 I4V-MC-B011 Non-interventional PASS Protocol Page 6
List of Tables Table Page Table 9.1. Classification of Medications for the Treatment of Rheumatoid Arthritis in Nordic Countries ...... 22 Table 9.2. Malignancy Diagnoses and Their Corresponding ICD-10 Codes and Positive Predictive Value (Where Available)...... 28 Table 9.3. Sample List of Serious Infections and Their Corresponding ICD-10 Codes ...... 29 Table 9.4. Opportunistic Infections ...... 29 Table 9.5. Major Adverse Cardiovascular Events and Their Corresponding ICD-10 Codes and Positive Predictive Value (Where Available) ...... 30 Table 9.6. Baseline Covariates for Consideration in Each Outcome-Specific Analysis ...... 32 Table 9.7. National Rheumatology Registries in Nordic Countries – Number of Patients and Inclusion Criteria ...... 34 Table 9.8. Brief Description of the National Clinical Rheumatology Registries Included in the Study and Their Variables of Interest ...... 35 Table 9.9. Brief Description of the National Hospital Registries Included in the Study and Their Variables of Interest ...... 36 Table 9.10. Sample Size to Detect a Hazard Ratio of 1.5...... 39 Table 9.11. Sample Size to Detect a Given Effect Size, by Average Length of Follow-up...... 40 Table 14.1. Description of Data Sources and Their Respective Variables of Interest Used in the Present Study...... 79 Table 14.2. Regulatory Requirement for Access to Registries and Databases ...... 83 I4V-MC-B011 Non-interventional PASS Protocol Page 7
List of Figures Figure Page Figure 9.1. Example of exposure classification...... 25 I4V-MC-B011 Non-interventional PASS Protocol Page 8
List of Annexes Annex Page Annex 1. List of Standalone Documents ...... 69 Annex 2. ENCePP Checklist for Study Protocols...... 70 Annex 3. Additional Information ...... 79 I4V-MC-B011 Non-interventional PASS Protocol Page 9
2. List of Abbreviations
Term Definition
AE adverse event
ApEHR The Institute of Applied Economics and Health Research
ATC Anatomical Therapeutic Chemical Classification System bDMARD biologic disease-modifying anti-rheumatic drug
BMI body mass index cDMARD conventional disease-modifying anti-rheumatic drug
CI confidence interval
DANBIO The Danish Rheumatology Database
DMARD disease-modifying anti-rheumatic drug
ENCePP European Network of Centres for Pharmacoepidemiology and Pharmacovigilance
EPN Regional ethics review board, Sweden
EU-RMP European Union–Risk Management Plan
HR hazard ratio
ICD International Classification of Diseases
JAK Janus kinase
MACE major adverse cardiovascular events
MAH marketing authorisation holder
MI myocardial infarction
NMSC nonmelanoma skin cancer
NOR-DMARD The Norwegian Antirheumatic Drug Register
OTC over-the-counter
PML Progressive multifocal leukoencephalopathy
PPV positive predictive value
RA rheumatoid arthritis
ROB-FIN The National Register for Biologic Treatment in Finland I4V-MC-B011 Non-interventional PASS Protocol Page 10
SAP statistical analysis plan
SD Statistics Denmark
SRQ The Swedish Rheumatology Quality Register
TB tuberculosis
THL National Institute for Health and Welfare tsDMARD targeted synthetic disease-modifying anti-rheumatic drug
VTE venous thromboembolism I4V-MC-B011 Non-interventional PASS Protocol Page 11
3. Responsible Parties
PPD
Phone: PPD E-mail: PPD
Eli Lilly Principal Investigator PPD Eli Lilly and Company Lilly Corporate Center Indianapolis, IN 46285 United States Telephone: PPD Email: PPD I4V-MC-B011 Non-interventional PASS Protocol Page 12
4. Abstract
Study I4V-MC-B011: A Retrospective Cohort Study to Assess the Safety of Baricitinib Compared with Other Therapies Used in the Treatment of Rheumatoid Arthritis in Nordic Countries Version: 1.0 Main author: PPD
Rationale and Background Baricitinib is a Janus kinase (JAK) 1/JAK2 selective inhibitor recently approved in Europe and countries in other regions for the treatment of moderate-to-severe rheumatoid arthritis (RA). Data from clinical studies in patients with RA have been evaluated and demonstrate that baricitinib is effective and generally well tolerated; however, the long-term safety profile among patients with RA in routine clinical practice has not been characterised. The purpose of this study is to assess the long-term safety of baricitinib compared with other systemic therapies used in the treatment of patients with RA in the course of routine clinical care. Research Question and Objectives This study aims to evaluate the safety of baricitinib among patients with RA receiving routine clinical care through the following specific objectives: Primary objectives: 1. To compare the incidence rates and profiles of the following aggregate outcomes: • serious infections overall (including herpes zoster) and opportunistic infections (including tuberculosis, Candida infections, and progressive multifocal leukoencephalopathy [PML]), • major adverse cardiovascular events (MACE), • malignancies overall (including lymphoma and typically virus-induced malignancies such as cervical and many oropharyngeal cancers), and • venous thromboembolism (VTE), among RA patients treated with baricitinib versus similar patients treated with other medications indicated for RA. 2. To describe the incidence rates of the following individual outcomes: lymphoma; herpes zoster; opportunistic infections such as tuberculosis, Candida, and PML; rhabdomyolysis; agranulocytosis; hyperlipidaemia (hypercholesterolaemia, hypertriglyceridaemia); gastrointestinal perforations; and liver injury.
Secondary Objectives: 3. To monitor the incidence rates of the aggregate outcomes of serious infections overall, MACE, malignancies overall, and VTE in very elderly patients, that is, ≥75 years of age. 4. To assess the effectiveness of risk minimisation activities by describing the pattern of use of baricitinib and the occurrence of pregnancy, active tuberculosis or active viral I4V-MC-B011 Non-interventional PASS Protocol Page 13
hepatitis, and the monitoring of lipid levels in relation to baricitinib use in routine clinical care. (This objective complements the aims of Study I4V-MC-B010, which aims to assess the effectiveness of risk minimisation activities.) Study design This study will use a retrospective cohort design to evaluate patients with RA from 4 Nordic countries (Denmark, Finland, Norway, and Sweden). Population Patients diagnosed with RA by a rheumatologist and identified from Nordic rheumatoid arthritis registries during the period 2017–2025 will be included in the study. Variables Data on selected clinical characteristics related to RA activity and disease severity, comorbidities, concomitant medications, and targeted adverse events will be collected directly from Nordic nationwide health registers. Data Sources Data for this study will come from various Nordic healthcare registries in Denmark, Finland, Norway, and Sweden: rheumatoid arthritis registries, prescription registers, patient registers, cause of death registries, and Medical birth registers, education and income registers. Study Size The final study size will depend on the total eligible patients present within the Nordic Rheumatoid Arthritis Registries between 2017 and 2025. A cohort of at least 4000 baricitinib-exposed patients and at least 4000 disease-modifying anti-rheumatic drug (DMARD) comparator patients will provide sufficient statistical power to estimate the risk of the chosen outcomes. Data Analysis Baricitinib will be the treatment of interest for all analyses. Comparisons will be made with 2 cohorts: a biologic cohort consisting of patients treated with biologic DMARDs (bDMARDs) and a cohort consisting of patients receiving treatment with conventional DMARDs (cDMARDs). Other JAK inhibitors will be excluded from the main analyses but will be considered in sensitivity analyses to evaluate potential class effects. I4V-MC-B011 Non-interventional PASS Protocol Page 14
5. Amendments and Updates
Not applicable. I4V-MC-B011 Non-interventional PASS Protocol Page 15
6. Milestones
Milestone Planned date
Start of data collection/extractiona Estimated Q4 2018
Study Progress Reportsb Included annually in baricitinib PBRER/PSUR
To be determined based on at least 24 months of data Final Report for Objective 4 in at least 50% of the discrete healthcare databases
End of data collectionc Estimated Q4 2026
Approximately 1 year after the end of data collection; Final Study Report (Objectives 1, 2, 3) estimated 2027 a For secondary data sources, the start of data collection corresponds to the date when data extraction is initiated. b Progress reports will include descriptive information on participants and their outcomes in each report. Because of the data lag and the lengthy process for data access, the first submission may include data only from Sweden and Denmark. c Date at which the complete analytic dataset is available. I4V-MC-B011 Non-interventional PASS Protocol Page 16
7. Rationale and Background
Rheumatoid arthritis (RA) is a chronic autoimmune inflammatory disease characterised by progressive joint destruction, systemic complications, and reduced survival (Smolen and Steiner 2003; Colmegna et al. 2012). It has a profoundly negative impact on the quality of life of those affected, particularly among those with moderate-to-severe disease (Choy and Panayi 2001; Allaire et al. 2009; Wasserman 2011). Baricitinib is a Janus kinase (JAK) 1/JAK2 selective inhibitor recently approved in Europe for the treatment of moderate-to-severe rheumatoid arthritis (RA). In clinical studies of patients with RA, baricitinib produced clinically meaningful improvements across all relevant domains of efficacy, including signs and symptoms, low disease activity and remission rates, physical function, and patient-reported outcomes, as well as inhibiting progressive radiographic joint damage. Data from clinical studies in patients with RA have been evaluated and demonstrate that baricitinib is effective and generally well tolerated; however, the long-term safety profile among patients with RA in routine clinical practice has not been characterised. Rheumatoid arthritis is associated with a number of serious comorbidities (CDC 2015 [WWW]). Obesity and smoking are risk factors for RA; consequently, patients with RA have a higher prevalence of these risk factors than age-matched controls (Crowson et al. 2013; Chang et al. 2014). In addition, among patients with RA, there is a high prevalence of comorbidities including myocardial infarction (MI), stroke, malignancy, venous thromboembolism (VTE), infections, hypertension, and gastrointestinal ulcer. The EU-RMP (version 6.0) for baricitinib describes 1 important identified risk (herpes zoster) and 9 important potential risks (malignancies [including lymphoma and typically virus-induced malignancies such as cervical and many oropharyngeal cancers], serious and opportunistic infections [including tuberculosis (TB), Candida infections, progressive multifocal leukoencephalopathy (PML)], myelosuppression [agranulocytosis], myopathy including rhabdomyolysis, potential for drug-induced liver injury, gastrointestinal perforations, major adverse cardiovascular events [MACE], VTE, and foetal malformation following exposure in utero). During the last 2 decades, biologic therapies have emerged that target cytokines (e.g., tumour necrosis factor-α and interleukin-6) and T or B cells and have improved outcomes for patients not responding adequately to conventional disease-modifying anti-rheumatic drugs (cDMARDs) (Smolen et al. 2007). However, there are still many patients who do not respond adequately to biologic treatment options. Janus kinase inhibitors comprise a new treatment option and act to inhibit cytokine signalling and other inflammation mediators (Leonard and O’Shea 1998) with differing degrees of specificity for 1 or more members of the JAK family (O’Shea et al. 2013). Baricitinib is a selective and reversible inhibitor of JAK1 and JAK2 that has shown efficacy across the spectrum of patients with active RA. Hypercholesterolemia and upper respiratory tract infections are common adverse reactions (ARs) (Olumiant SmPC, 2017). I4V-MC-B011 Non-interventional PASS Protocol Page 17
The long-term safety of biologics and new therapies entering the market is of interest to rheumatologists, regulators, and healthcare professionals (Ramiro et al. 2017). With a total population of 26.6 million and comprehensive national healthcare systems, Nordic countries provide a unique setting for the study of the safety of new medications. Rheumatoid arthritis has a prevalence of 0.5% to 1.0% among adults in these countries, with an incidence of 25 to 45 per 100,000 person-years (Puolakka et al. 2010; Neovius et al. 2011; Eriksson et al. 2013; Antti Kuuliala [WWW]; Leddgikt [WWW]). Detailed information on these patients is available through dedicated national RA clinical databases: The Danish Rheumatology Database (DANBIO), The National Register for Biologic Treatment in Finland (ROB-FIN) The Norwegian Disease-Modifying Antirheumatic Drug Register (NOR-DMARD) The Swedish Rheumatology Quality Register (SRQ) These registries collect data from annual mandatory reporting on the treatment of patients with RA using disease-modifying anti-rheumatic drugs (DMARDs). In combination with the availability of additional data from recognised national health registers and long-term follow-up of patients in the healthcare systems, this creates an ideal environment to study the safety of new medications. This protocol summarises the analyses that will be conducted to evaluate the safety profile of baricitinib. In particular, the important risks described in the baricitinib RMP will be investigated in comparison with similar patients receiving other (biologic DMARD [bDMARD], cDMARD) treatments for RA. This protocol also contributes to the assessment of risk minimisation activities (Objective 4) by describing the pattern of use of baricitinib and specific outcomes from the risk minimisation activities among patient populations who will have received the patient alert card and who will have been treated by healthcare providers who received the communication. I4V-MC-B011 Non-interventional PASS Protocol Page 18
8. Research Question and Objectives
This study aims to evaluate the safety of baricitinib among patients with RA receiving routine clinical care, through the following specific objectives: Primary objectives: 1. To compare the incidence rates and profiles of: serious infections overall (including herpes zoster) and opportunistic infections (including TB, Candida infections, and PML), MACE, malignancies overall (including lymphoma and typically virus-induced malignancies such as cervical and many oropharyngeal cancers), and VTE among patients with RA treated with baricitinib versus similar patients treated with other medications indicated for RA. 2. To describe the incidence rates of the following individual outcomes: lymphoma; herpes zoster; opportunistic infections such as TB, Candida, and PML; rhabdomyolysis; agranulocytosis; hyperlipidaemia (hypercholesterolaemia, hypertriglyceridaemia); gastrointestinal perforations; and liver injury. Secondary objectives: 3. To monitor the incidence rates of the aggregate outcomes of serious infections overall, MACE, malignancies overall, and VTE in very elderly patients, that is, those ≥75 years of age. 4. To assess the effectiveness of risk minimisation activities by describing the pattern of use of baricitinib and the occurrence of pregnancy, active tuberculosis or active viral hepatitis, and changes in lipid levels in relation to baricitinib use in routine clinical care. (This objective complements the aims of Study I4V-MC-B010, which aims to assess the effectiveness of risk minimisation activities.) I4V-MC-B011 Non-interventional PASS Protocol Page 19
9. Research Methods
9.1. Study Design This study will be based on a retrospective national cohort of patients with RA in 4 Nordic countries: Denmark, Finland, Norway, and Sweden. The basis for the study or analytic cohort will be patients included in the national rheumatology registries, and therefore secondary use of data: Denmark: the Danish Rheumatology Database (DANBIO) Finland: the National Register for Biologic Treatment in Finland (ROB-FIN) Norway: the Norwegian Disease-Modifying Anti-Rheumatic Drugs Registry (NOR-DMARD) Sweden: the Swedish Rheumatology Quality Register (SRQ). The registries listed above will be enriched with clinical, prescription, socioeconomic, and demographic information from other data sources available in each Nordic country: national patient registries: information on diagnoses given at hospitals prescription registries: information on comedication medical birth registries: information on pregnancy and pregnancy outcomes socioeconomic registers: important confounder information; education and income Due to the universal healthcare available in the 4 countries, the affordability of medication is not a limiting issue and essentially all Nordic patients with RA receive treatment and are captured in these registries, allowing for observation of the effects of treatment in a real-life setting. Patients are included in the rheumatology registries based solely on treatment and there is a nearly complete record of all treated patients with RA. The combination of the registry data with the other healthcare data available in each country will create the enriched cohort that will be used for investigating the safety of baricitinib use.
9.2. Setting Residents of Nordic countries (that is, Denmark, Finland, Norway, and Sweden) are served by national universal healthcare. Data related to the clinical treatment of residents are collected in several national registers and databases. All prescriptions (over 99%) redeemed at community pharmacies are registered in the national prescription registries, with the exception of drugs administered at hospitals, over-the-counter (OTC) and vaccines. Patient data are also gathered in national hospital registries (e.g., discharge diagnoses, procedures), pregnancy registries (i.e., comprehensive information on all live births), and clinical databases (e.g., national rheumatology registries). All relevant registries in each country are listed in Table 14.1 of Annex 3. The Nordic rheumatology registers include information about treatment with any DMARD. This will include information on treatment with baricitinib and other targeted synthetic DMARDs (tsDMARDs) (although this is not currently reflected in the registries’ operating protocols or names). In 2015, the Danish registry (DANBIO) included approximately 26,000 patients with prevalent RA and captured 96% of the >10,000 patients with RA who received treatment with I4V-MC-B011 Non-interventional PASS Protocol Page 20
biologics. In the same year, the Finnish ROB-FIN included >8,000 patients with RA, the Norwegian NOR-DMARD >12,000 patients with RA, and finally the Swedish SRQ database included almost 40,000 patients with RA, of which 12,000 are treated with bDMARDs.
9.2.1. Study Population This study will be based on data from patients with RA included in the public healthcare registries in the Nordic Countries: Denmark, Finland, Norway, and Sweden, over the calendar period of interest for the study. All patients with RA treated with a DMARD are to be registered in the clinical databases, regardless of their specific treatment. This will include, but is not limited to, patients with RA treated with biologic drugs and patients with moderate to severe RA. Patients will be included after the RA diagnosis has been confirmed by a rheumatologist and the completeness of registration is over 90%. Diagnostic criteria for RA follow the American College of Rheumatology/European League Against Rheumatism classification criteria (Ibfelt et al. 2016). Although the 4 Nordic countries have the same criteria for diagnosing RA, the threshold for initiating DMARD treatment might differ between countries, hospitals, and individual physicians. In general, the prevalence of RA in the Nordic Countries is 0.5% to 1.0% among adults, with an incidence of 25 to 45 per 100,000 person-years (Puolakka et al. 2010; Neovius et al. 2011; Eriksson et al. 2013; Antti Kuuliala [WWW]; Leddgikt [WWW]). With a total population of 26.6 million, the Nordic RA population consists of approximately 175,000 patients. As approximately 10 per 100,000 adults initiate treatment with bDMARDs each year, this will correspond to approximately 2600 new patients in treatment each year. Patients with RA receiving treatment with baricitinib or other tsDMARDs will also be included in these registers, as will patients treated with cDMARDs. Patients receiving treatment with cDMARDs will be available from the rheumatology registers and/or from other data sources (for example, national patient registries).
9.2.2. Eligibility Inclusion criteria All patients with RA who are included in 1 of the Nordic rheumatology registries over the study period (2017–2025). Exclusion criteria The patient is unable or unwilling to provide informed consent to participate in the national rheumatology databases, which require informed consent from the patient. The remaining national registries used in the study include the national population, with no requirement for patient consent. The patient has a history of JAK inhibitor use prior to initiation of drug treatment (cDMARD, biologic, or baricitinib). Sensitivity analyses will be performed in which the inclusion/exclusion criteria, exposure ascertainment, and outcome identification will vary from that in the primary I4V-MC-B011 Non-interventional PASS Protocol Page 21
analyses (e.g., patients exposed to baricitinib and tofacitinib, both JAK inhibitors, will be evaluated to assess potential class effects). It is worth noting that patients may be removed from a register if they request, but this is rare. The observation period for each patient will continue until the occurrence of an event of interest, the end of the study period, death, or emigration, whichever comes first. As these are national registries that include almost the entirety of the eligible population, the total number of patients available for the current study will depend on the size of the RA population in the Nordic region and the market share of the respective tsDMARD, bDMARD, and cDMARD groups in each country.
9.2.3. Patient Subgroups of Special Interest This study will describe the characteristics of, and estimate the proportion of, patients with RA who are diagnosed with severe hepatic impairment, evidence of hepatitis B or C infection, or a history of lymphoproliferative or malignant disease within the previous 5 years, and who also receive treatment with baricitinib. These data will also be used to monitor the incidence of primary aggregate outcomes in very elderly patients, i.e., those ≥75 years of age. Use of baricitinib during pregnancy, which is contraindicated, will also be evaluated, as will use in combination with bDMARDs or other JAK inhibitors. Pregnant women will be identified from the National Medical Birth registries (Section 9.4.5) if the pregnancy ended in a live birth or stillbirth. Pregnancies ending in early termination (spontaneous or elective abortions) will be identified through the National Hospital registries (Section 9.4.3), International Classification of Diseases (ICD)-10 diagnosis codes DO02-DO06. Breastfeeding mothers, however, cannot be identified through national registries.
9.3. Variables
9.3.1. Drug Exposure In Nordic countries, most bDMARDs are administered in hospital and it is anticipated for tsDMARDs, including baricitinib, that prescriptions will be dispensed at the hospital. Regardless, prescriptions dispensed outside the hospital setting can be identified through national prescription registries that cover all prescriptions redeemed at community pharmacies. Identification of exposures to the medications of interest will rely on information from the dedicated national clinical rheumatology registries that include information on tsDMARD, bDMARD, and cDMARD prescriptions. All bDMARDs and tsDMARDs are prescribed solely at the hospitals and therefore captured in the RA registries. Some cDMARDs (e.g., methotrexate) are prescribed in the primary care sector, but nonetheless captured in the RA registries in connection with outpatient visits to rheumatologists, who are required to report the information to the national RA registries. Exposure will be defined as treatment with a tsDMARD, in this case baricitinib (Anatomical Therapeutic Chemical Classification System [ATC] L04AA37), or a bDMARD or cDMARD in I4V-MC-B011 Non-interventional PASS Protocol Page 22
a rheumatology database in one of the Nordic countries (see data sources in Table 14.1 of Annex 3). Length of treatment, switches between or within cDMARDs or bDMARDs (or biosimilars), and addition to existing therapies, will be identified from the same registries. In particular, exposure to DMARDs will be collected in the registries as “Time since treatment initiation” and as “Cumulative duration of treatment”. The high proportion of the RA population that is included in the register, for example, greater than 95% of the region’s patients treated with a bDMARD, and the high completeness of treatment data in these registers will allow for robust classification of exposures into treatment groups of interest for analysis. Concomitant exposure to other medications (non-DMARD) will be obtained from the national prescription registries (see data sources in Table 14.1 of Annex 3). Medications used to treat RA will be classified as described below and in Table 9.1.
Table 9.1. Classification of Medications for the Treatment of Rheumatoid Arthritis in Nordic Countries cDMARDs bDMARDs tsDMARD Sulfasalazine Abatacept Certolizumab pegol Tofacitinib (A07EC01) (L04AA24) (L04AB05) (L04AA29) Methotrexate Adalimumab Golimumab (L04AX03) (L04AB04) (L04AB06) Hydroxychloroquine Etanercept Rituximab (P01BA02) (L04AB01) (L01XC02) Leflunomide Infliximab Tocilizumab (L04AA13) (L04AB02) (L04AC07) Sarilumab* (L04AC14) Abbreviations: bDMARDs = biologic disease-modifying anti-rheumatic drugs; cDMARDs = conventional disease- modifying anti-rheumatic drugs; tsDMARD = targeted synthetic disease-modifying anti-rheumatic drug. * This medication is not marketed in some Nordic countries.
New DMARDs and biosimilars for the treatment of RA will be added as they are approved. Tofacitinib, a JAK inhibitor used to treat patients with moderate-to-severe RA, will be excluded from all analyses except some sensitivity analyses that evaluate the potential existence of class effects. Simultaneous use of biologic medications will be enumerated, although this pattern of use is expected to be rare as it contravenes current Nordic treatment guidelines, which do not recommend the use of multiple biologic medications (Singh et al. 2015). The classification of drug exposure within this study for the evaluation of malignancy differs from the classification used for other outcomes. This is to accommodate the long latency of malignant outcomes even after a causal exposure (see Sections 9.3.1.1 and 9.3.1.2, respectively). For malignancy, assignment to exposure groups will be based on an intent-to-treat approach and selection will be hierarchical. For other, non-malignancy outcomes, an “as-treated” approach will be used instead. I4V-MC-B011 Non-interventional PASS Protocol Page 23
9.3.1.1. Drug Exposure and Cohort Identification for Malignancy Analyses For the primary outcome of malignancy, follow-up time will be categorised into 3 hierarchical and mutually exclusive exposure groups or cohorts: cDMARDs, biologic medications (bDMARDs), and baricitinib. To accommodate the long latency that would be expected for malignant outcomes that occur after an exposure, exposure assignment for malignancies will be hierarchical. Once exposure to a biologic medication occurs, time may not be attributed to the cDMARD cohort, and once exposure to baricitinib occurs, time may not be attributed to the other cohorts. This approach is conservative because it will tend to attribute malignant events to baricitinib, regardless of subsequent exposures and of latent effects of past exposures. cDMARDs cohort: Biologic-naive cDMARD users, with no previous exposure to a JAK inhibitor. Follow-up will begin at treatment initiation and continue until initiation of a biologic medication or baricitinib, end of the study period, death, emigration, or incident malignancy; a new index date will be assigned if a patient initiates a biologic medication or baricitinib; new and continuing user status will be updated at each time point to ensure that patients who re-initiate use of a previously used cDMARD are censored. bDMARD or biologic cohort: Patients initiating a biologic medication with no previous exposure to a JAK inhibitor. Follow-up will begin at treatment initiation and will continue until the initiation of a JAK inhibitor, end of the study period, death, emigration, or incident malignancy; a new index date will be assigned if a patient initiates baricitinib; new and continuing user status will be updated at each time point to ensure that patients who re-initiate use of a previously used bDMARD are censored. Baricitinib cohort: Patients initiating baricitinib with no prior exposure to another JAK inhibitor. Follow-up will begin at treatment initiation and will continue until initiation of another JAK inhibitor, end of the study period, death, emigration, or incident malignancy.
9.3.1.2. Drug Exposure and Cohort Identification for Nonmalignancy Analyses Unlike malignancy, nonmalignant outcomes are expected to occur in closer temporal proximity to a putatively causal exposure, so it is reasonable for the classification of exposures to reflect changes in treatment regimens. Thus, three cohorts similar to those described in Section 9.3.1.1 will be created, but exposure will be classified using an as-treated approach. Using this approach, person–time will accrue based on the treatment received and will reflect actual use during each “medication episode”. cDMARDs cohort: cDMARD users with no prior exposure to a biologic DMARD or a JAK inhibitor. Follow-up will begin at treatment initiation and continue until initiation of a biologic DMARD or a JAK inhibitor , discontinuation of all cDMARDs, end of the study period, death, emigration, or occurrence of an incident event; a new index date will be assigned when a patient initiates a new cDMARD; new and continuing user status will be updated at each time point to ensure that patients who re-initiate use of a previously used cDMARD are censored (Table 9.1). bDMARDs or biologic cohort: Patients initiating a biologic medication with no prior exposure to a JAK inhibitor. Follow-up will begin at treatment initiation and will I4V-MC-B011 Non-interventional PASS Protocol Page 24
continue until the initiation of a JAK inhibitor/tsDMARD, medication discontinuation, end of study period, death, emigration, or occurrence of an incident event; a new index date will be assigned when a patient initiates a new bDMARD (Table 9.1); new and continuing user status will be updated at each time point to ensure that patients who re-initiate use of a previously used bDMARD are censored; concomitant use of a cDMARD will be assessed and included in the analysis as a time-dependent covariate. Baricitinib cohort: Patients initiating baricitinib with no prior exposure to another JAK inhibitor. Follow-up will begin at treatment initiation and will continue until initiation of a bDMARD or other JAK inhibitor/tsDMARD, medication discontinuation, end of study period, death, emigration, or occurrence of an incident event; concomitant use of a cDMARD will be assessed and included in the analysis as a time-dependent covariate. A window corresponding to five half-lives or 30 days, whichever is longer, will be added to the end of each treatment period. This window is the extension period. In case the standard interval between medication administration is longer than five half-lives (e.g., 183 days between rituximab infusions), the standard treatment interval will be used. During this window, patients will continue to accrue time “at risk” for a brief period after the medication is stopped.
9.3.1.3. Example of Exposure Classification Patients in the national rheumatology registries are under routine clinical care and may add, discontinue, or switch medications during the course of follow-up. This results in switching within the bDMARD cohort and among the 3 medication cohorts. Detailed examples of the attribution of at-risk time are described below. Each patient will be assigned an index date, when the patient begins contributing person–time for a time-to-event analysis. Information on relevant potential confounding factors is collected at this time, for example, history of infection at baseline. This information is used in the propensity-score matching described in Section 9.7.3 to help ensure that confounding factors are evenly distributed across the groups being compared. When a patient switches medication, a new index date will be assigned, and that patient will be re-matched with another initiator from the comparison group. The alternative, for the original patient who switched medication to remain matched to a continuing user of the previous medication, is not appropriate as patients who switch therapies may have different baseline risk with regard to the study outcomes than patients who continue their treatment. The use of concomitant cDMARDs will be included as a time-dependent covariate and patients who start or stop concomitant use will not have their cohort observation time censored. Follow-up time will be measured from the index date to initiation of a medication in another exposure cohort, medication discontinuation, end of study period, death, emigration, or occurrence of an incident event (for time-to-event analysis). As described earlier, an extension period corresponding to 5 half-lives will be added to the treatment period to extend the time at risk for that medication. An event occurring during use of a medication or during the subsequent extension period would be assigned to the discontinued medication. However, if a new medication is initiated within the extension period, the time at risk for the previous medication will end when the new medication is initiated. To account for ambiguities in the attribution of I4V-MC-B011 Non-interventional PASS Protocol Page 25 events that occur shortly after a switch, a sensitivity analysis will be conducted in which such events are attributed to the previous medication. Detailed description of the sensitivity analysis can be found in Section 9.7.8. An event occurring during use of a medication or during the subsequent risk window would be assigned to the discontinued medication. An example of the follow-up time calculation for a patient who switches biologic medications is provided in Figure 9.1.
Figure 9.1. Example of exposure classification.
(a) Switching within medication cohorts The example in Figure 9.1(a) illustrates a patient who: At the start of the registry, initiates adalimumab treatment and continues until Month 12. o Adalimumab time: 12 months + 5 half-lives (70 days or ~2 months for the purpose of this example) = 14 months o Biologic exposure cohort total time: 14 months At Month 18, the patient initiates etanercept and continues until Month 24. o Etanercept time: 6 months + 30 days (5 half-lives = 21 days < 30 days) = 7 months o Biologic exposure cohort total time: 14 + 7 = 21 months At Month 25, the patient initiates infliximab and continues until Month 36. o Infliximab time: 11 months + 5 half-lives (45 days or 1.5 months) = 12.5 months o Biologic exposure cohort total time: 21 + 12.5 = 33.5 months By the end of this observation period, this patient will contribute three treatment episodes to bDMARDs and 33.2 months of person-time in total to the biologic group. (b) Switching between medication cohorts I4V-MC-B011 Non-interventional PASS Protocol Page 26
Switching between medication cohorts may also occur and will be managed similarly to the within-medication cohort switch described above. This scenario is illustrated in Figure 9.1(b). For this example, assume an analysis for a non-malignancy outcome so exposures are assigned “as treated” rather than following a hierarchy. Just as for switching within a cohort, this patient: At the start of the registry initiates treatment with adalimumab and continues treatment until Month 12. o Adalimumab time: 12 months + 5 half-lives (70 days or ~2 months for the purpose of this example) = 14 months o Biologic exposure cohort: 14 months o However, at month 6 the patient initiates concomitant methotrexate and continues treatment until Month 15. Concomitant methotrexate will be included in the analysis as a time-dependent covariate. At Month 15, the patient switches to baricitinib and is now included in the baricitinib cohort until Month 24. o Baricitinib exposure cohort: 9 months + 30 days (5 half-lives = 2.7 days << 30 days) = 10 months At Month 25, the patient initiates etanercept and continues on treatment until the end of the observation period at Month 42. o Etanercept time: 17 months + 30 days (5 half-lives = 21 days < 30 days) = 18 o Biologic exposure cohort: 14 + 18 = 32 months By the end of this observation period, this patient will contribute two treatment episodes to bDMARDs and one treatment episode to baricitinib. Had this been an intent-to-treat analysis for malignancy, the patient would have remained in the baricitinib exposure group and would not have been eligible to contribute person-time to the biologic exposure cohort upon switching to etanercept at Month 25. After initiating baricitinib in Month 15, all subsequent person-time would have been contributed to the baricitinib exposure cohort (i.e., ‘ever exposed’ to baricitinib for malignancy outcomes).
9.3.1.3.1. Medication Restarts In accord with the new user study design (Lund et al. 2015), only patients who newly initiate treatment will be included in analyses. However, if the number of new users is not sufficient to adequately power the analyses, patients who restart a previous medication (that is, prevalent users) may be included in sensitivity analyses. Sufficient numbers will be evaluated with regard to 80% power to detect at least a 1.5-fold increase in risk among the baricitinib exposure cohort compared to the comparator cohort. If prevalent users need to be included, additional details will need to be considered, as described below. Patients in routine care may stop and restart medications at the discretion of their physician. For instance, patients may be required to discontinue their biologic DMARD temporarily to receive a live vaccine (e.g., herpes zoster vaccine) or patients may appear to have discontinued while they are hospitalised, since dispensings/administrations of medications are not observed in hospital. I4V-MC-B011 Non-interventional PASS Protocol Page 27
Medication restarts will not affect the analysis of malignancy outcomes because the intent-to- treat approach for malignancies considers “ever exposure” versus “never exposure” to baricitinib. However, medication restarters will be excluded from analysis of non-malignancy outcomes which employs the new user design to avoid selection bias related to discontinuation and prevalent use of study medications. Patients who restart medications may have a different baseline risk for the study outcomes than patients who continue treatment or initiate a new treatment. Those who restart a previous treatment may be at reduced risk of AE’s compared to others, especially AE’s likely to lead to discontinuation, such as serious AE’s. Including patients who restart medications is not consistent with the new user study design; however, restarting a medication may be correlated with the outcome of interest so such restarts should not be ignored. Therefore, at a minimum a description of this group of patients by exposure cohort will be provided which includes the distribution of baseline covariates especially risk factors for the outcome of interest. If the number of new users is not sufficient to adequately power the analyses, patients who restart a previous medication (i.e., prevalent users) may be included. To address the potential differences in baseline risk between restarters and incident users, new index dates would be assigned for patients who restart any specific drug within a treatment exposure cohort which also addresses those who restart a treatment exposure. They would then be matched to restarters in the comparison cohort. Patients who restart and incident users of medications will be compared to further evaluate the appropriateness of including both in an analysis. In particular, the proportion of restarters between the groups being compared and the distribution of risk factors will need to be evaluated to determine whether it is feasible to proceed with an analysis that includes both. In real-world practice, non-adherence may be a problem in that patients may not take the prescribed amount of medication at the desired intervals. To distinguish patients who continue their treatment despite gaps from patients who discontinue their medication and restart, the following approach will be applied: Definition of discontinuation: If the gap between the end of the days supply of one dispensing and the next dispensing is more than 60 days, the next dispensing will not be incorporated into the treatment period and the patient will be considered to have discontinued treatment. For medications that are administered as infusions, the recommended interval will be used instead of days supply. For example, the recommended interval between infliximab infusions is 56 days. When constructing the treatment period, if the gap between visits for infusions is longer than 56+60=116 days, the second visit will not be incorporated into the treatment period. A treatment episode may only be classified as a medication restart if the patients have previously discontinued that medication, and patients who discontinue their medication may not have their exposure time extended to allow a restart to be considered part of their original treatment. Finally, the proportion of patients who have additional dispensings after discontinuation will be assessed and the distribution of the time interval between the date of the last dispensing of a I4V-MC-B011 Non-interventional PASS Protocol Page 28
treatment period and the subsequent dispensing of the same medication will be examined. Prior to initiation of comparative analyses, information from this assessment will be used to confirm the appropriateness of the above definition. Sensitivity analyses, such as those in Section 9.7.8.3, will not include treatment episodes of medication restarts.
9.3.2. Outcomes The primary outcomes in this study are those that will be examined in comparative analyses versus the bDMARD and cDMARD exposure groups: serious infections overall, MACE, malignancy overall excluding nonmelanoma skin cancer (NMSC), and VTE. Secondary outcomes will be described, but may not be evaluated separately in comparative analyses if the number of events would not provide sufficient statistical power (i.e., 80%) to detect a difference between comparison groups: lymphoma; herpes zoster; opportunistic infections such as TB, Candida infections, and PML; rhabdomyolysis; agranulocytosis; hyperlipidaemia; gastrointestinal perforations; and evidence of liver injury. The following sections describe each outcome, including a sample of ICD-10 codes, with the final list for each outcome to be included in the statistical analysis plan (SAP).
9.3.2.1. Malignancy, Excluding Nonmelanoma Skin Cancer The occurrence of malignancy is captured through the National Patient registries that include discharge diagnoses for all patients in contact with a hospital. Table 9.2 shows the primary malignancy diagnostic codes and their corresponding positive predictive values (PPVs). The final list of diagnoses to be included and relevant ICD-10 codes will be included in the SAP.
Table 9.2. Malignancy Diagnoses and Their Corresponding ICD-10 Codes and Positive Predictive Value (Where Available) Diagnosis ICD-10 code PPV (95% CI) Malignant Neoplasm C00–C75 98.0 (89.4–99.9) (Malignant neoplasms, stated or presumed (Thygesen et al. 2011) to be primary, of specified sites) C43 – Malignant melanoma of skin NA Malignant Neoplasm C76–C80 100 (92.9–100) (Malignant neoplasms of ill-defined, (Thygesen et al. 2011) secondary, and unspecified sites) Malignant neoplasm C81–C96 84.5 (82.2–86.5) (Malignant neoplasms, stated or presumed (Nørgaard et al. 2005) to be primary, of lymphoid, haematopoietic and related tissue) Abbreviations: ICD-10 = International Classification of Diseases (ICD)-10 diagnosis codes; PPV = positive predictive value.
9.3.2.2. Nonmelanoma Skin Cancer Cases of NMSC will be included from the national patient registries, which include discharge diagnoses for all patients in contact with a hospital. The outcome examined in this study is NMSC, ICD-10 codes C44.X, but the final codes to be evaluated will be included in the SAP. I4V-MC-B011 Non-interventional PASS Protocol Page 29
9.3.2.3. Serious Infections This study will include all serious infections treated at hospitals. Only infections requiring contact with a hospital will be included, given that this information is available in the national patient registries. Minor infections that do not require hospitalisation will not be included in the analyses. A sample of serious infections of interest, which will be included for any patient with contact with a hospital who receives these diagnostic codes, is shown in Table 9.3. The final list of diagnoses to be included and relevant ICD-10 codes will be included in the SAP.
Table 9.3. Sample List of Serious Infections and Their Corresponding ICD-10 Codes Diagnosis ICD-10 Code Bursitis (Other bursopathies) M71 Cellulitis L03 Diverticulitis (Diverticular disease of intestine) K57 Septicaemia A40 - Streptococcal sepsis (PPV 21.7 [Madsen et al. 1998]) A41 – Other sepsis Influenza and pneumonia J09–J18 Bronchitis (Other acute lower respiratory J20–J22 infections) Gastroenteritis A00–A09 (Intestinal infectious diseases) Meningitis/Encephalitis G00–G09 (Inflammatory diseases of the central nervous system) Herpes zoster B02 Abbreviations: International Classification of Diseases (ICD)-10 diagnosis codes; PPV = positive predictive value.
Description on how these outcomes will be analysed is provided in Section 9.7.7.1.
9.3.2.4. Opportunistic Infections This study will include serious opportunistic infections that required a patient to contact a hospital. Nonserious opportunistic infections that do not require a Nordic patient to contact a hospital will not be included in the analyses. Opportunistic infections of interest based on those included in the RMP are presented in Table 9.4. Others may be added subsequently; a complete list of opportunistic infections and diagnostic codes will be included in the SAP.
Table 9.4. Opportunistic Infections Diagnosis ICD-10 Code Progressive multifocal leukoencephalopathy (PML) A81.2 Tuberculosis A15–A19 Candida (Candidiasis) B37 I4V-MC-B011 Non-interventional PASS Protocol Page 30
Abbreviation: ICD-10 = International Classification of Diseases (ICD)-10 diagnosis codes.
Discussion of the analysis is provided in Section 9.7.7.2.
9.3.2.5. Major Adverse Cardiovascular Events Major adverse cardiovascular events are composite cardiovascular endpoints that include fatal and nonfatal MI, fatal and nonfatal ischaemic stroke, and cardiovascular death. Incident MACE, and acute MI as a separate outcome, will be captured based on ICD diagnostic and procedure codes available in the national patient registers. ICD-10 diagnostic codes and the corresponding PPV are shown in Table 9.5, with the final list to be included in the SAP.
Table 9.5. Major Adverse Cardiovascular Events and Their Corresponding ICD-10 Codes and Positive Predictive Value (Where Available) Diagnosis ICD-10 code PPV (95% CI) MI (Acute myocardial infarction) I21 100 (97.5–100) Ischaemic stroke (Cerebral infarction) 97.0 (84.7–99.5) (Krarup et al. I63 2007) Cardiovascular death (Sudden cardiac death) 50.0 (35.5– 64.5) (Joensen et I46.1 al. 2009) Abbreviations: ICD-10 = International Classification of Diseases (ICD)-10 diagnosis codes; MI = myocardial infarction; PPV = positive predictive value.
9.3.2.6. Venous Thromboembolism The occurrence of venous thromboembolic events is captured based on ICD diagnostic and procedure codes available in national patient registers as discharge diagnoses and includes both pulmonary embolism and deep vein thrombosis. VTE will be identified, in inpatient diagnosis through the use of diagnostic codes and, in outpatient diagnoses, through the use of diagnostic codes and a requirement for an anticoagulant dispensing within 31 days. Confirmation of VTEs will be explored based on additional clinical information. The final list of diagnoses to be included and relevant ICD-10 codes will be included in the SAP.
9.3.2.7. Gastrointestinal Perforations The definition of this outcome will include, but not be limited to, diverticulitis with perforation and gastrointestinal ulcers with perforation. The final codes used to define this outcome will be included in the SAP.
9.3.2.8. Liver Injury Through the national patient registers we will capture all discharge diagnoses of toxic liver disease (ICD-10 code K71), which include drug-induced liver injury. Furthermore, we will include any occurrence of events indicative of hepatic injury (i.e., any hepatic event that requires liver biopsy), procedure codes KJJA20 and KJJA21. Analyses of this outcome will only consider hepatic events that resulted in a hospital contact. The ICD-10 diagnostic codes used to define this outcome will be included in the SAP. I4V-MC-B011 Non-interventional PASS Protocol Page 31
9.3.2.9. Rhabdomyolysis The occurrence of rhabdomyolysis as a discharge diagnosis will be captured through the national patient registers, with ICD-10 code M62.8 (rhabdomyolysis). Confirmation of rhabdomyolysis will be through review of additional clinical information from available medical charts.
9.3.2.10. Hyperlipidaemia (Including Hypercholesterolaemia and Hypertriglyceridaemia) The occurrences of hypercholesterolaemia (ICD-10 code E78.0) and hypertriglyceridaemia (ICD-10 code E78.1) are captured through the national patient registers. Only cases with a hospital contact can be identified through the registers.
9.3.2.11. Myelosuppresion (Agranulocytosis) The occurrence of discharge diagnoses indicating agranulocytosis, for example, ICD-10 code D70.2 (Other drug-induced agranulocytosis), will be captured through the national patient registers, with the final code list included in the SAP. Only cases with a hospital contact can be caught through the registers.
9.3.3. Covariates The Nordic rheumatoid arthritis registries collect information about a variety of disease measures related to disease duration, severity, and treatment. Race, alcohol consumption, smoking, and body mass index (BMI) are important potential confounders of the association between drug exposure and the safety outcomes of interest. However, information on race and alcohol use is not included in the Nordic RA registries, smoking is not included in the Finnish RA clinical database, and BMI is not included in the Swedish or Norwegian databases. Smoking and BMI will be evaluated in a restricted analysis to investigate its predictive value and impact on the estimate using data from the countries where they are available. The covariates listed in Table 9.6 will be considered in the analyses for their potential to confound the association between exposure to a medication indicated for RA and the outcomes under investigation. Further explanation is provided in Section 9.7. I4V-MC-B011 Non-interventional PASS Protocol Page 32
Table 9.6. Baseline Covariates for Consideration in Each Outcome-Specific Analysis Outcome Baseline Covariates for Consideration in the Statistical Model Malignancy excluding NMSC Age, sex, education, income, RA disease activity (DAS28), functional status (HAQ), RA disease duration, modified Charlson Comorbidity score, personal history of cancer (excluding NMSC), personal history of NMSC, previous biologic medication use, healthcare resource utilisation Serious and opportunistic Age, sex, education, income, RA disease activity (DAS28), RA disease infections duration, diabetes mellitus, chronic lung disease, liver disorder, ischaemic heart disease, previous serious infections or antibiotic use, glucocorticoid use, previous cDMARD or bDMARD use, healthcare resource utilisation Major adverse cardiovascular Age, sex, education, income, RA disease activity (DAS28), RA disease event duration, history of cardiovascular disease (MI, stroke, unstable angina, hospitalised congestive heart failure, ventricular arrhythmia, cardiovascular revascularisation procedure, coronary artery disease, and transient ischaemic attack), diabetes mellitus, current or history of hypertension (redemption of antihypertension agents, dyslipidaemia (redemption of lipid modifying agents, ATC C10), use of prescription aspirin, use of lipid-lowering agents or anti- platelet agents, healthcare resource utilisation Venous thromboembolism Age, sex, education, income, RA disease activity (DAS28), functional status (HAQ), RA disease duration, history of cancer, cardiovascular disease (hospitalised congestive heart failure, ventricular or atrial arrhythmia), diabetes mellitus, previous VTE, recent pregnancy; recent hospitalisation, surgery or trauma; use of prescription aspirin, anticoagulants, systemic glucocorticoids, methotrexate, oral contraceptives or hormone replacement therapy, previous biologic medication use, healthcare resource utilisation Abbreviations: ATC = Anatomical Therapeutic Chemical Classification System; DAS28 = Disease Activity Score modified to include the 28 diarthrodial joint count; cDMARD = conventional disease-modifying anti-rheumatic drug; HAQ= Health Assessment Questionnaire; MI = myocardial infarction; NMSC = nonmelanoma skin cancer; RA = rheumatoid arthritis; VTE = venous thromboembolism.
9.4. Data Sources
9.4.1. The Nordic Health System In the Nordic countries, about 80% of the funding of the healthcare comes from public sources (Kristiansen et al. 2000). County councils provide most of the healthcare services in Denmark, Norway, and Sweden. In Finland, the municipalities provide most of the healthcare. Payment according to the number of patients recruited, in combination with service fees, is used for all Danish general practitioners and the majority of the Finnish, whereas various fee-for-service systems are used in the other countries. All Nordic countries had global hospital budgeting in the 1980s; since then, Finland, Norway, and Sweden have implemented other systems, predominantly combinations of diagnosis-related group financing and global budgets. The amounts of resources devoted to healthcare are about the same in all 4 countries whether measured by the proportion of gross domestic product devoted to healthcare, or by hospital beds or doctor/patient ratios. In monetary terms, Denmark I4V-MC-B011 Non-interventional PASS Protocol Page 33
and Norway spend more than Finland and Sweden. Despite similar amounts of resources, they are quite differently used across the countries. Differences of a factor of 2 or more are observed, for example, for pharmaceuticals. All the Nordic countries have increased patient co-payments during the 1990s. Finland also has the greatest number of private hospitals. Despite all these differences, the Nordic healthcare systems are quite similar when seen in a global perspective. The Nordic countries’ healthcare data are well-suited to assess long-term effects of drug exposure. Through the 10- or 11-digit code assigned to each citizen in the Nordic countries, included in the national registers, it is possible to link information from different registers and thereby follow each individual from the beginning of life until death. The national registers in the Nordic countries, which are constructed in a similar way and with similar contents, have been used for numerous studies and contributed to important scientific works (see Table 14.1 and Table 14.2 of Annex 3). Rare exposures and rare outcomes require large databases, and as each of the Nordic countries are small countries with a population ranging from 5.3 million in Norway to 10 million in Sweden, the data in each country are probably too sparse to evaluate associations between specific drugs and rare outcomes. Combining the 4 Nordic countries gives a cohort of 26.6 million persons.
9.4.2. National Clinical Rheumatology Registries Identification of patients with RA and their DMARD treatment will be gathered from the national clinical rheumatology registries. The criteria for being included in the registries slightly differs between the countries, but they all have in common the criterion that they include all patients with RA being treated with a DMARD. However, registration in the databases is voluntary and patients may ask not to be included. This is, however, rare and estimated to be less than 5%. In Table 9.7 and Table 9.8, the 4 national registries are described in summary. I4V-MC-B011 Non-interventional PASS Protocol Page 34
Table 9.7. National Rheumatology Registries in Nordic Countries –Number of Patients and Inclusion Criteria Total Patients Registry Country Registry Inclusion Criteria (n) The Danish Rheumatology Database Patients with inflammatory RA; Denmark 30,000 (DANBIO) (Ibfelt et al. 2016) gout; SLE The Swedish Rheumatology Quality Biologic DMARD initiation in any Sweden 55,000 Register (SRQ) (Eriksson et al. 2014) rheumatic patient The Norwegian Antirheumatic Drug Inflammatory joint disease; use of Register (NOR-DMARD) (Kvien et al. Norway 8000 biologic DMARD 2005) The National Register for Biologic Rheumatic disease using biologic Treatment in Finland (ROB-FIN) Finland 6000 DMARD (Konttinen et al. 2006) Abbreviations: DMARD = disease-modifying anti-rheumatic drug; RA = rheumatoid arthritis; SLE = systemic lupus erythematosus. Note: Current inclusion criteria, and some registry names, have not been updated to reflect that the registries will include information on patients treated with targeted synthetic disease-modifying anti-rheumatic medications, including baricitinib. I4V-MC-B011 Non-interventional PASS Protocol Page 35
Table 9.8. Brief Description of the National Clinical Rheumatology Registries Included in the Study and Their Variables of Interest Sample Variables of Interest Register Brief Description Available for This Study The Danish Rheumatology Research register and a data source DMARD treatment, RA disease Database (DANBIO) (Ibfelt et al. for rheumatologic diseases for activity (DAS28), functional status 2016) monitoring clinical quality. (HAQ), RA disease duration The Swedish Rheumatology Research register that collects DMARD treatment, RA disease Quality Register (SRQ) (Eriksson et clinical data on patients with RA, as activity (DAS28), functional status al. 2014) well as other rheumatic diseases, (HAQ), RA disease duration and may be enriched with data on comorbid conditions, prescription drug dispensings, and mortality from national data sources. The Norwegian Antirheumatic Register-based longitudinal DMARD treatment, RA disease Drug Register (NOR-DMARD) observational study of which the activity (DAS28), functional status (Kvien et al. 2005) main objectives are to study the (HAQ), RA disease duration effectiveness of treatment of inflammatory joint diseases with biological DMARDs in clinical practice. The National Register for Biologic ROB-FIN was designed to monitor DMARD treatment, RA disease Treatment in Finland (ROB-FIN) the safety, effectiveness, and activity (DAS28), functional status (Konttinen et al. 2006) cost-effectiveness of biological (HAQ), RA disease duration treatments in rheumatic diseases. In addition to the data collected by rheumatologists at routine care visits, supplementary information is retrieved from national healthcare registers. Abbreviations: DAS28 = Disease Activity Score modified to include the 28 diarthrodial joint count; DMARD = disease-modifying anti-rheumatic drug; HAQ= Health Assessment Questionnaire; RA = rheumatoid arthritis.
9.4.3. National Hospital Registries Information on the chosen outcomes (Section 9.3.2) will be gathered from the national hospital registries (Table 9.9). They include discharge diagnoses of all patients in contact with a hospital. Personal 10- or 11-digit number allows linkage of information from different registers whereby each individual and their diagnoses can be followed up from the beginning of life until death. Table 9.9 summarises the 4 registries. I4V-MC-B011 Non-interventional PASS Protocol Page 36
Table 9.9. Brief Description of the National Hospital Registries Included in the Study and Their Variables of Interest Sample Variables of Year of Register Brief Description Interest Available for Establishment This Study The Danish National Patient Information on all patients in Discharge diagnoses and Registry (Andersen et al. 1977 contact with a Danish hospital. their date. 1999) National Inpatient Registry Information on all completed in- Hospital admission and (IPR) (Sweden) (Ludvigsson 1964 and out-patient admissions at public discharge, diagnoses, 2011 [WWW]) hospitals. surgery, including dates. Norwegian Patient Registry Contains a combination of data Hospital admission and (NPR) (Nilssen et al. 2014) recorded at treatment sites, when discharge, diagnoses, 2008 patients have received referral or surgery, including dates. treatment in a hospital, an outpatient clinic or a contract specialist. Care Register for Health The purpose of the register is to Discharge diagnoses and Care (Finland) (THL collect data on the activities of their date. [WWW]) 1969 health centres, hospitals, and other institutions providing inpatient care and on the patients treated in them.
9.4.4. Prescription Data Each Nordic country has a nationwide prescription database containing electronically submitted information on prescriptions dispensed by community pharmacies. In total, the databases cover the countries’ 26.6 million inhabitants. In addition, data from the autonomic region Åland Islands are included in the Finnish data, but the data from the autonomous regions of the Faroe Islands and Greenland are not included in the Danish data. The data collected are determined by country-specific regulations but all include information on the prescriptions together with information from different administrative registries. In most countries, data are transferred electronically monthly from pharmacies to the prescription database. All individuals/patients included in the prescription databases have a unique personal identifier based on their person identification number, permitting linkage between various population-based data sources. Some prescription databases routinely include the date of death and migration, whereas others need to be linked to this information. Regarding drug exposure, the Nordic article number is a unique identifier for each drug formulation of a medicinal product used in the Nordic countries. This number constitutes the link to other registries providing detailed information on dispensed drugs. The drugs are classified according to the global ATC system. Numbers of World Health Organization’s defined daily doses dispensed are recorded, as well as the number of packages and the reimbursement code. There are several challenges in using these data. First, the reimbursement system differs between the countries. Second, the indication for the prescription is not yet available in the data. The dispensing (redemption) date and retail price are included in all the registries, but the prescription date is at present not included in the Norwegian and Danish prescription databases. I4V-MC-B011 Non-interventional PASS Protocol Page 37
9.4.5. The Medical Birth Registers Each of the Nordic countries has kept medical birth registers for decades, all with compulsory notification (Irgens 2000; NBHW 2003 [WWW]; Langhoff-Roos et al. 2014). All live births as well as stillbirths from varying gestational ages in the different countries are notified to the registers. All registers contain basic information on the mother, the neonate, and the father. Linkage to other registers and national databases using the personal identity numbers can provide additional data on diseases and medical conditions of the mother, the father, and the neonate, as well as on social conditions, education, prescribed medications, and social security/insurance data. Linkages require study-specific permission from national authorities, which is usually obtained. Thus, it is possible to conduct longitudinal and intergenerational studies and even in some instances include information on relatives and offspring within the period of registration. Diagnoses are registered as ICD-10 codes. The international origin of the codes for some main groups created through the registers allows for cross-country research on large populations within the Nordic countries. However, codes for each individual case are assigned on national platforms and this may involve minor differences between the countries. Birth notification forms are linked to, or part of, the system and thus to population census offices.
9.4.6. Education and Income Data All Nordic countries have high-quality education and income data on each citizen in the country (Education Register–Statistics Norway [WWW]; Epland [WWW]; The Finish Register of Completed Education and Degrees and Student Flow Statistics [WWW]; The Swedish Register of Education [WWW]; Haider et al. 2008; Johnell et al. 2008; Ringbäck Weitoft et al. 2008; Selmer et al. 2009; Baadsgaard et al. 2011; Jensen et al. 2011; Webster 2014; Andersen et al. 2016). Registers that have all been used for research purposes several times before are based on national statistics on education and tax reports. Not all education is comparable and therefore an adjusted variable on educational length will be made and used as a proxy for level of education. Income data will be indexed by a fixed year and indexed to a currency rate and categorised in quartiles.
9.4.7. Quality Assurance and Quality Control All aspects of data analysis will be conducted according to standard procedures of the Research Group at the department of Clinical Pharmacology, Copenhagen University Hospital. All statistics and programming procedures will be conducted by 1 analyst and validated by another. For all data processing and analysis steps, the validation analyst will review the programme along with input and output datasets, and for select steps of the project will employ double-programming techniques to reduce the potential for programming errors. The data in health registries and clinical databases may be used for research purposes if approval has been obtained from the authorities. There are strict rules for what data can be accessed, and only data that are deemed necessary for the study and acceptable from a societal perspective can be obtained. I4V-MC-B011 Non-interventional PASS Protocol Page 38
The data collected from the nation-specific databases will, if possible, be pooled into 1 dataset at Statistics Denmark (SD). Statistics Denmark is a secure facility that hosts a secure computing environment. Access to data will be through a secure and encrypted virtual private network connection. Individual access requires clearance by SD. The Institute of Applied Economics and Health Research (ApEHR) will use the following model developed from several years of experience with acquiring and managing data from multiple Nordic countries. 1. ApEHR, assisted by the scientific country partner identified for each country, will seek regulatory permission to get data from national/local registries. 2. After obtaining appropriate permission, relevant national registries will be accessed to obtain study data. The following procedures will be followed in each participating country: a. ApEHR will determine the necessary definition/formats for data extracts based on the data request outlined in the protocol. Specific rules for extracting and interpreting data from the registries will be developed by ApEHR for each country based on expert advice prior to extracting any study data. b. ApEHR’s scientific country partner will submit specific variable lists to the registries. c. Data from the individual national registries will be sent to the scientific country partner who will perform a quality control and then submit the data to be entered into a consolidated database by ApEHR to establish a pooled multi-country dataset.
9.4.8. Data Period The study period is described in Section 6. It is important to note that the study will access data that lags by approximately 1 year, on average, so that data that is entered today is not available until one year later. This is the result of the structure of the Nordic data and is independent of this study. The number of patients included in each exposure cohort will be monitored periodically to ensure that the duration of the study data period is appropriate for the size of the analytic data required to address the primary comparative objectives.
9.5. Study Size Sample size calculations were performed for the outcomes and background incidence rates given in Table 9.10. Calculations were based on the following assumptions: 80% power 1-sided type I error rate of 0.025 5% loss of patients per year in the registry 3 years of patient accrual followed by up to 8 years of follow-up. I4V-MC-B011 Non-interventional PASS Protocol Page 39
The sample size calculations were calculated with nQuery Advisor 7.0 (survival analysis, two- group test based on exponential survival accrual period and dropouts). These results give appropriate power calculations for a Cox proportional hazards model.
Table 9.10. Sample Size to Detect a Hazard Ratio of 1.5 Sample Size per Group Incidence Rate Outcome Reference to Detect 1.5-fold (per 100 PY) Increase in Risk MACE Lindhardsen et al. 2014 1.71 896 Serious infection Van Dartel et al. 2013 2.59 611 Malignancy Askling et al. 2009 0.934 1594 VTE Holmqvist et al. 2012 0.59 2491 Abbreviations: MACE=Major adverse cardiovascular events; PY=person-years; VTE=venous thromboembolism. With the anticipated study size of 4000 baricitinib-treated patients and 4000 comparator patients, the 8-year study has adequate power to detect at least a hazard ratio of 1.5 after completion. The proportion of subjects who achieve 8 years of observation will depend primarily on the market uptake of baricitinib, as the comprehensive nature of Nordic national healthcare ensures that the large majority of patients will be observed from the time of their entry into the study to the end of the study. At the time of this protocol, over 800 patients are estimated to have received treatment with baricitinib. These patients will be eligible for the maximum observation period. For varying lengths of average follow-up, Table 9.11 provides an overview of the impact on the statistical power to detect a 1.5-fold increase in risk of MACE and malignancy. Serious infections are acute events and there is ample study power to detect any associations that may be present. The incidence rate of VTE has demonstrated evidence of constant hazard throughout the clinical programme, with no evidence of an increase in risk over time. Statistical power to detect an association with a 1.5-fold increase in risk of VTE is therefore dependent only on identifying sufficient events. I4V-MC-B011 Non-interventional PASS Protocol Page 40
Table 9.11. Sample Size to Detect a Given Effect Size, by Average Length of Follow-up Hazard Ratio = 1.5 Hazard Ratio = 2.0 (Baricitinib versus (Baricitinib versus Average Length of Follow-up Comparator) Comparator) Outcome Time Baricitinib Comparator Baricitinib Comparator (years) Patients Patients Patients Patients (n) (n) (n) (n) 1 13,535 13,535 4170 4170 a VTE 2 6792 6792 2094 2094 3 4544 4544 1401 1401 1 4702 4702 1450 1450 b MACE 2 2375 2375 733 733 3 1600 1600 495 495 1 3121 3121 963 963 Serious Infectionc 2 1585 1585 490 490 3 1073 1073 333 333 1 8568 8568 2640 2640 d Malignancy 2 4308 4308 1329 1329 3 2888 2888 891 891 a Based on background incidence rate of VTE of 0.59 cases per 100 PY (Holmqvist et al. 2012). b Based on background incidence rate of MACE of 1.71 cases per 100 PY (Lindhardsen et al. 2014). c Based on background incidence rate of serious infection of 2.59 per 100 PY (Van Dartel et al. 2013). d Based on background incidence rate of malignancy of 0.934 cases per 100 PY (Askling et al. 2009).
9.6. Data Management Datasets and analytic programmes will be created using SAS and will be managed and stored as required by relevant national laws and regulations and consistent with Applied Economics record retention procedures. All data management will be based on pseudo-anonymised data and performed at a secure server system for each country separately. All statistical analyses will be validated through double programming. Data from the 4 Nordic countries will be gathered at Statistics Denmark (SD), the national statistics center, as Denmark is the only country that does not allow their data to cross their border. Statistics Denmark is a secure facility that hosts a secure computing environment. Access to data will be through a secure and encrypted virtual private network connection. Individual access requires clearance by SD.
9.6.1. Data to Be Collected Refer to Section 9.3 for information about data to be extracted.
9.6.1.1. Missing Data Imputation will be considered only for variables that would be used to adjust for potential confounding (“adjusting variables”) or for generating propensity scores. Imputation will not be used to account for missing information about exposure to DMARDs, including baricitinib, or I4V-MC-B011 Non-interventional PASS Protocol Page 41
about the safety outcomes of interest as this would not be appropriate for the main dependent and independent variables in the Cox regression models. In general, the completeness of data is very high and the number of missing data low. The majority of adjustment variables have less than 1% missing values. If missing data for a particular variable exceed 15% of the cohort size, imputation of the missing values for the adjusting variables will be considered before modelling the data. If imputation is deemed necessary, multiple imputation by chained equations (MICE) will be considered (Royston 2004) This is a multiple multivariate imputation method that is described by van Buuren and colleagues (1999). Other methods may be considered as needed.
9.7. Data Analysis Analyses will be conducted separately for each outcome and will include descriptive analyses, comparative analyses (where appropriate), and any relevant sensitivity analyses. Propensity scores will be used to address imbalances in the potential confounding factors across the comparison groups that may confound the association between treatment and study outcomes. Prior to any comparative analyses, an assessment of baseline demographic characteristics will be conducted to investigate the possibility of heterogeneity or unmeasured confounding that would impact the ability to create balanced cohorts for inclusion in the comparative analyses. For all analyses, baricitinib will be the treatment of interest. The cDMARD and biologic cohorts will serve as the reference groups. Comparison with cDMARD users is intended to permit evaluation of any potential risks associated with baricitinib that might not otherwise be detected by comparison to biologic medications. Patients present in the data from the date of market availability in 2017 to the end of the study period will be included in the analysis. Patients will be followed for as long as they remain in the data between those calendar periods until they are censored as a result of: diagnosis of one of the events of interest, death, emigration, or the end of study period, whichever comes first. A hazard ratio of the outcome among patients with RA ever receiving baricitinib treatment compared with patients treated with cDMARDs or bDMARDs will be calculated using Cox proportional hazards models with further adjustment as described in Section 9.3.3. For patients who do not start treatment with a bDMARD at the date of the first registration in one of the rheumatology databases, the exposure status changes from never exposed to ever exposed at the date of start of first bDMARD treatment. Cases and person-years of observation before bDMARD treatment are allocated to the non-exposed group whereas cases and person-years after treatment are allocated to the exposed group. Risk ratios of outcomes among patients with RA compared with the general population will be calculated as standardised incidence ratios (i.e., the ratio between observed and expected cases during follow-up). Results will be presented unadjusted and adjusted for relevant confounders as described in Section 9.3.3. A 1-sided p- value <0.05 will be considered to be statistically significant. Hazard ratios will be presented with 95% confidence intervals (CIs).
9.7.1. Analysis Population The analysis population for all outcomes is patients included in the Nordic rheumatoid arthritis registries, who are members of the drug-exposure groups defined in Section 9.3.1. Subgroup I4V-MC-B011 Non-interventional PASS Protocol Page 42
analyses will be performed on patients of special interest (e.g., those aged 75 years) if there is sufficient sample size. Sample size and statistical power for subgroup analyses may be limited.
9.7.2. Background and Rationale for Propensity Scores Drug exposure in pharmacoepidemiological studies does not occur at random and is the result of patient-, physician-, and healthcare system–related factors. When these factors are associated with the outcome of interest, comparisons of different drug-exposure cohorts will be confounded due to channelling bias. Propensity scores address the imbalance across drug-exposure cohorts by providing a mechanism to compare patients with concordant baseline risk but discordant exposure (Schneeweiss 2007). For clarity, covariates included in the propensity-score models are also referred to as confounders because they confound the association between exposure and outcome.
9.7.3. Propensity-Score Definition and Estimation A propensity score is an estimate of the probability that a patient receives a particular treatment, conditional on measured characteristics at the time a treatment decision is made (Rosenbaum and Rubin 1983). For this study, a patient’s propensity score will reflect the predicted probability of exposure to a medication, given a patient’s characteristics at the index date. Propensity scores will be estimated using logistic regression models predicting the probability of baricitinib exposure compared with exposures in the other groups (cDMARD and bDMARD users). These models will be constructed separately for each primary outcome (Section 9.3.2). The models will include variables that are known risk factors for the outcomes of interest and that are also associated with systemic treatments for RA. Covariates considered for inclusion in the propensity-score models are provided in Table 9.6. The inclusion of interaction and nonlinear terms will be guided by clinical judgement. Evaluation of the propensity-score models is discussed in Section 9.7.5. Newly marketed drugs often experience changes in prescribing patterns over time so that a patient characteristic that was once associated with treatment selection becomes more or less relevant over the drug’s life cycle. The time between marketing authorisation and the stabilisation of use patterns and market share is a particularly dynamic time in the life cycle of a drug (Schneeweiss et al. 2011). To account for this, given that the cohort size allows for it, a calendar-specific propensity score will be employed as described by Mack et al. (2013) and Seeger et al. (2003). These methods estimate the propensity-score models and match patients within blocks of calendar time to account for temporal variation in prescribing patterns.
9.7.4. Using the Propensity Score to Address Channelling Bias Each patient in the study will have at least 1 estimated propensity score that represents the probability of exposure at the index date, given baseline characteristics (new medication starts will be considered a new index date). Matching and stratification on the propensity score is relatively straightforward with 2 exposure groups, but becomes increasingly complex as the number of exposure groups increases. Multiple exposure groups and the possibly limited registry sample available for this study means that matching may result in a high number of I4V-MC-B011 Non-interventional PASS Protocol Page 43
unmatched patients and stratification may result in strata with few or no patients. Therefore, this study will examine pairwise comparisons of exposure cohorts: the baricitinib cohort versus the biologic cohort and the baricitinib cohort versus the cDMARDs cohort. For each comparison, the baricitinib exposure cohorts will be matched separately to the comparator exposure cohorts. Matching will be performed using an objective algorithm and will be discussed further in the SAP. The effectiveness of the matching will be evaluated and the propensity-score model will be adjusted as appropriate. More information on the evaluation of the propensity score is provided in Section 9.7.5. The propensity-score model and matching will be finalised before initiating any safety outcome analyses. If the number of matched patients is prohibitively small, limiting the possibility of conducting a comparative analysis, other applications of the propensity score, such as matching within strata of propensity scores, will be considered. One-to-many matching will also be considered if deemed necessary to increase sample size.
9.7.5. Evaluation of the Propensity-Score Model and Stratification Before initiating any outcome analyses, the ability of the propensity-score stratification to balance the distribution of baseline confounders and reduce channelling bias will be evaluated. The appropriateness of the propensity-score modelling is judged on whether balance on pretreatment characteristics is achieved between the treatment and reference groups (Rubin 2006; D’Agostino 2007; Spreeuwenberg et al. 2010). Standardised differences will be used to assess differences between the cohorts across all measured baseline covariates before and after propensity matching. As a general rule, standardised differences greater than 0.10 indicate an imbalance that may require further investigation (Austin and Mamdani 2006; Austin 2011). Higher-level terms or interactions may be considered when a variable is unbalanced across the baricitinib and reference cohorts or when informed by clinical judgement (e.g., an interaction between age and sex for MACE outcomes).
9.7.6. Malignancy Analysis Analyses will be performed for malignancy, excluding non-melanoma skin cancer, and non- melanoma skin cancer. Malignancies have a long latency period; consequently, they are not easily attributed to a specific exposure. To account for this ambiguity, malignancy analyses will consider the risk of malignancy associated with ever use of baricitinib. These analyses will include only patients initiating treatment, that is, “new users.” The incidence of malignancies will be monitored in very elderly patients (aged ≥75 years). Proportional hazards of the exposure groups will be evaluated using Kaplan–Meier survival curves. If the proportional hazards assumption is violated, other models that relax the proportional hazards assumption will be considered.
9.7.6.1. Malignancy Excluding Nonmelanoma Skin Cancer The outcome of interest for malignancy analysis will include all diagnostic codes for malignancies excluding NMSC. All patients with an active malignancy (diagnosed within 12 months) at enrolment will be excluded from these analyses. Within the propensity score- I4V-MC-B011 Non-interventional PASS Protocol Page 44
matched population(s) a number of descriptive statistics and rates will be generated to understand the data before comparative analyses begin: number of people with past baricitinib or biologic use at baseline baseline demographic and clinical characteristics and standardised differences for the baricitinib, biologic, and cDMARDs cohorts (all patients, matched patients, and unmatched patients) description of RA treatments in each cohort and median duration of exposure of each treatment history of malignancy at registry enrolment, based on available data prevalence of active malignancy at baseline baseline demographic and clinical characteristics for all patients excluded due to active malignancy distribution of survival time post-index date for the baricitinib, biologic, and cDMARDs cohorts (all patients, matched patients, and unmatched patients after propensity-score models are achieved) pattern of medication use post-index date for the baricitinib, biologic, and cDMARDs cohorts (all patients, matched patients, and unmatched patients) distribution of survival time for all malignancy outcomes excluding NMSC and malignancy by type, among the 3 most common malignancies, for the baricitinib, biologic, and cDMARDs cohorts (all patients, matched patients, and unmatched patients) temporal pattern of incidence over repeated time windows since index date After the specified descriptive statistics are calculated, calendar-specific propensity-score matching will be used to match patients between cohorts as described in Section 9.7.4. Descriptive statistics requiring matched cohorts will then be conducted. Comparative analyses will not begin until finalisation of the exposure cohorts and propensity-score models are achieved. Cox proportional hazards regression models will be used to estimate the hazard ratios (HRs) and 95% CIs of incident malignancies, excluding NMSC, among patients in the baricitinib cohort versus the biologic cohort and the baricitinib cohort versus the cDMARDs cohort (null hypothesis [H0]: HR = 1). The model will contain the exposure cohort variable, any variables that remain unbalanced after propensity-score matching (Table 9.6), and a time-dependent variable for any concomitant RA treatment. Patients will be censored upon development of a malignancy excluding NMSC; switch to a medication in a different exposure cohort; five half- lives following discontinuation of a medication; or at the end of the study period, withdrawal from the registry, emigration, or death. Model diagnostics will be performed to identify any influential observations. These analyses may be modified to focus on the class of JAK inhibitors, rather than baricitinib only, if capturing “ever exposure” to baricitinib without including other JAK inhibitor use is not feasible. However, in the scenario where few patients exist with exposure to baricitinib and without exposure to other JAK inhibitors, descriptive statistics will be provided separately for patients exposed singly to each JAK inhibitor (i.e., those I4V-MC-B011 Non-interventional PASS Protocol Page 45
exposed only to baricitinib and those exposed only to each other JAK inhibitor). Sensitivity analyses will be performed accordingly and are discussed in Section 9.7.8.1.
9.7.6.2. Nonmelanoma Skin Cancer The outcome for this analysis is definite NMSC as defined in Section 9.3.2.2. All patients with an active malignancy (diagnosed within 12 months) at enrolment will be excluded from these analyses. Within the propensity score-matched population(s) a number of descriptive statistics and rates will be generated before comparative analyses begin: number of people with past baricitinib or biologic use at baseline baseline demographic and clinical characteristics and standardised differences for the baricitinib, biologic, and cDMARDs cohorts (all patients, matched patients, and unmatched patients) description of RA treatments in each cohort and median duration of exposure of each treatment history of NMSC at registry enrolment, based on available data prevalence of active NMSC at baseline baseline demographic and clinical characteristics for all patients excluded due to active malignancy distribution of survival time post-index date for the baricitinib, biologic, and cDMARDs cohorts (all patients, matched patients, and unmatched patients) pattern of medication use post-index date for the baricitinib, biologic, and cDMARDs cohorts (all patients, matched patients, and unmatched patients) temporal pattern of incidence over repeated time windows since index date After the specified descriptive statistics are calculated, calendar-specific propensity scores will be used to match patients between cohorts as described in Section 9.7.4. Descriptive statistics requiring matched cohorts will then be conducted. Comparative analyses will not begin until finalisation of the exposure cohorts and propensity-score models are achieved. Cox proportional hazards regression models will be used to estimate HRs and 95% CIs of incident NMSCs among patients in the baricitinib cohort versus the biologic medications cohort and the baricitinib cohort versus the cDMARDs cohort (null hypothesis [H0]: HR = 1). The model will contain exposure cohort variable, any variables that remain unbalanced after propensity-score matching (Table 9.6), and a time-dependent variable for any concomitant RA treatment. Patients will be censored upon development of NMSC; switch to a medication in an alternate exposure cohort; five half-lives following discontinuation of a medication; or at the end of the study period, withdrawal from the registry, emigration, or death. Model diagnostics will be performed to identify any influential observations. These analyses may be modified to focus on the class of JAK inhibitors rather than baricitinib if capturing “ever exposure” to baricitinib without including other JAK inhibitor use is not feasible. However, in the scenario where few patients exist with exposure to baricitinib and without exposure to other JAK inhibitors, descriptive statistics will be provided separately for patients exposed singly to each JAK inhibitor (i.e., those exposed only to baricitinib and those exposed only to each other JAK I4V-MC-B011 Non-interventional PASS Protocol Page 46
inhibitor). A sensitivity analyses will be performed accordingly and is discussed in Section 9.7.8.2.
9.7.7. Nonmalignancy Analyses Analyses will be performed for serious infections, opportunistic infections (including TB), MACE, and VTE. These analyses will only consider treatment initiations. Before beginning comparative analyses, a number of descriptive statistics and crude rates will be generated to understand the registry data: baseline demographic and clinical characteristics and standardised differences for the cDMARDs cohort, biologic cohort, and baricitinib cohort (all patients, matched patients, and unmatched patients) prevalence of the outcomes at registry enrolment baseline demographic and clinical characteristics for all patients excluded due to prevalent secondary outcome distribution of follow-up time for the cDMARDs cohort, biologic cohort, and baricitinib cohort (all patients, matched patients, and unmatched patients) distribution of baseline demographic and clinical characteristics and standardised differences for matched patients in the cDMARDs cohort, biologic cohort, and baricitinib cohort by exposure duration the number of new medication starts for matched patients within the cDMARDs cohort, biologic cohort, and baricitinib cohort Comparative analyses will be implemented using calendar-specific propensity-score matching to control for confounding. Sensitivity analyses will be performed accordingly (Section 9.7.8.1). No comparative analyses will begin until finalisation of the exposure cohorts and propensity-score models are achieved. Details of outcome-specific analyses are presented below. Proportional hazards of the exposure groups will be evaluated using Kaplan–Meier survival curves. If the proportional hazards assumption is violated, other models that relax the proportional hazards assumption will be considered.
9.7.7.1. Serious Infections The outcome for this analysis is first serious infection as defined in Section 9.3.2.3. In addition to the descriptive statistics and crude rates described in Section 9.7.7, analyses of serious infections will include the following: distribution of survival time until first serious infection for the cDMARDs cohort, biologic cohort, and baricitinib cohort (all patients, matched patients, and unmatched patients) crude rate of first serious infection and first serious infection by site of infection and for the prespecified infections (Section 9.3.2.3) per 100 patient-years for the cDMARDs cohort, biologic cohort, and baricitinib cohort (all patients, matched patients, and unmatched patients); within cohorts of matched patients, stratified by concomitant DMARD use I4V-MC-B011 Non-interventional PASS Protocol Page 47
the distribution of the number of serious infections per patient the incidence of serious infections prior to use of baricitinib and after commencing baricitinib treatment. Any patient with an existing serious infection at baseline (diagnosed within 3 months) will be excluded from all analyses, including baseline descriptive statistics, crude rates, and comparative analyses. Propensity scores will be used to match patients from different exposure cohorts as described in Section 9.7.4. Cox proportional hazards regression will be used to compare the hazards of first serious infection of the baricitinib and cDMARDs cohorts and the baricitinib and biologic cohorts. All models will include the exposure cohort, concomitant cDMARD use, concomitant glucocorticoid use, and any variables that remain unbalanced after propensity-score matching. Patients will be censored at the first serious infection, switch to a medication in an alternate exposure cohort, five half-lives following discontinuation of a medication, end of the study period, withdrawal from the registry, emigration, or death. Another analysis that includes all serious infections will also be performed and is described in Section 9.7.8.4.
9.7.7.2. Opportunistic Infections The outcomes for this analysis are first opportunistic infection as defined in Section 9.3.2.4. In addition to the descriptive statistics and crude rates described in Section 9.7.7, analyses for opportunistic infections will include the following for first opportunistic infection: pattern of medication use post-index date for the baricitinib, biologic, and cDMARDs cohorts (all patients, matched patients, and unmatched patients) distribution of survival time until first opportunistic infection for the cDMARDs cohort, biologic cohort, and baricitinib cohort (all patients, matched patients, and unmatched patients) crude rate of the first opportunistic infection, the first opportunistic infection by type, and the first prespecified opportunistic infection (Section 9.3.2.4) per 100 patient-years for the cDMARDs cohort, biologic number cohort, and baricitinib cohort (all patients, matched patients, and unmatched patients), stratified by concomitant DMARD use the distribution of opportunistic infections per patient the distribution of serious and non-serious opportunistic infections the incidence of serious infections, including first and subsequent infections, both prior to use of baricitinib and after commencing baricitinib treatment. Any patient with an existing opportunistic infection (diagnosed within 3 months) at baseline will be excluded from all analyses, including baseline descriptive statistics and crude rates. Propensity scores will be used to match patients as described in Section 9.7.4. Cox proportional hazards regression will be used to compare the hazards of first opportunistic infection of the baricitinib and cDMARDs cohorts and the baricitinib and nonbiologic cohorts. All models will include the exposure cohort, concomitant cDMARD use, and any variables that remain unbalanced after propensity-score matching. Patients will be censored at the first opportunistic infection, switch to a medication in an alternate exposure cohort, five half-lives following I4V-MC-B011 Non-interventional PASS Protocol Page 48
discontinuation of a medication, at the end of the study period, withdrawal from the registry, emigration, or death. These analyses will be performed for first opportunistic infections. Proportional hazards of the exposure groups will be evaluated using Kaplan–Meier survival curves. If the proportional hazards assumption is violated, other models that relax the proportional hazards assumption will be considered.
9.7.7.3. Major Adverse Cardiovascular Events The outcome for this analysis is all incident MACE (Section 9.3.2.5). In addition to the descriptive statistics and crude rates described in Section 9.7.7, analyses for MACE will include the following: distribution of survival time until first MACE for the cDMARDs cohort, biologic cohort, and baricitinib cohort (all patients, matched patients, and unmatched patients) crude rate per 100 patient-years of first MACE as a component outcome and by individual event for the biologic cohort and the baricitinib cohort (all patients, matched patients, and unmatched patients) and within cohorts of matched patients, stratified by concomitant DMARD use. the incidence of nonfatal MACE, both prior to use of baricitinib and after commencing baricitinib treatment. Propensity scores will be used to match patients from different exposure cohorts as described in Section 9.7.4. Cox proportional hazards regression will be used to compare the hazards of MACE between the baricitinib and cDMARDs cohorts and the baricitinib and biologic cohorts. All models will include the exposure cohort, concomitant cDMARD use, and any variables that remain unbalanced after propensity-score matching. Patients will be censored at the occurrence of MACE, switch to a medication in an alternate exposure cohort, five half-lives following discontinuation of a medication, the end of the study period, withdrawal from the registry, emigration, or death.
9.7.7.4. Venous Thromboembolic Events The outcome for this analysis is all incident VTEs (Section 9.3.2.6). Because of the recognised risk of VTEs associated with some medications, e.g., methotrexate, analyses may be conducted with inclusion of a time-dependent covariate to account for concomitant treatment with such drugs. In addition, because other specific medications (e.g., certolizumab pegol) may also have different baseline risk of VTEs, an assessment of incidence among users of individual, commonly used medications will also be conducted. Overall analyses, however, will evaluate the comparative risk of VTEs among baricitinib users versus patients treated with biologic medications, as noted in the rheumatology data. In addition to the descriptive statistics and crude rates described in Section 9.7.7, analyses for VTEs will include the following: Pattern of medication use post-index date for the baricitinib, biologic, and cDMARD cohorts (all patients, matched patients, and unmatched patients) I4V-MC-B011 Non-interventional PASS Protocol Page 49
distribution of survival time until first VTE for the cDMARDs cohort, biologic cohort, and baricitinib cohort (all patients, matched patients, and unmatched patients) crude rate of first VTE per 100 patient-years for the biologic cohort and the baricitinib cohort (all patients, matched patients, and unmatched patients) and within cohorts of matched patients, stratified by concomitant methotrexate use. Propensity scores will be used to match patients from different exposure cohorts as described in Section 9.7.4. Cox proportional hazards regression will be used to compare the hazards of VTEs of the baricitinib and cDMARDs cohorts (as feasible given number of events observed) and the baricitinib and biologic cohorts. All models will include the exposure cohort, concomitant cDMARD use, and any variables that remain unbalanced after propensity-score matching. Patients will be censored at the occurrence of VTE, switch to a medication in an alternate exposure cohort, five half-lives following discontinuation of a medication, the end of the study period, withdrawal from the registry, emigration, or death.
9.7.7.5. Analysis of Individual Outcomes (Objective 2) Several primary outcomes include aggregated events that are of interest on their own, but that are not expected to occur with sufficient frequency to permit comparative analyses to be conducted. Other infrequent outcomes are not part of the aggregate categories, but are also of interest. These outcomes are those listed in the second objective: lymphoma; herpes zoster; opportunistic infections such as TB, Candida, and PML; rhabdomyolysis; agranulocytosis; hyperlipidaemia; gastrointestinal perforations; and evidence of drug-induced liver injury. Descriptive analyses will be conducted for these outcomes and will include the following: distribution of survival time until first occurrence of the event of interest in the cDMARDs, biologic, and baricitinib cohorts (all patients, matched patients, and unmatched patients) crude rate of the first occurrence of the safety event per 100 patient-years for the cDMARDs, biologic, and baricitinib cohorts (all patients, matched patients, and unmatched patients) Distribution of number of safety events per patient Incidence of nonlethal, potentially recurrent events of interest prior to use of baricitinib and after commencing baricitinib treatment. For specific opportunistic infections, herpes zoster, and gastrointestinal perforation, it may be useful to examine crude incidence rates stratifying by concomitant use of glucocorticoids. Other stratifications may also be informative and may be included as appropriate. In addition, for certain outcomes, incidence rates may vary by use of specific medications or classes of medications, for example, gastrointestinal perforations among IL-6 inhibitors such as tocilizumab. Thus, it may be useful to estimate crude incidence rates for each bDMARD or subgroup of bDMARDs as feasible considering the number of patients exposed to each medication and the suitability of each outcome. Ascertainment of some events is likely to be subject to detection bias, and the ability to determine whether differences in rates are due to use of a particular medication or due to differences in clinical surveillance may be limited. Finally, I4V-MC-B011 Non-interventional PASS Protocol Page 50
if the incidence of any outcome in this category occurs more frequently than anticipated, comparative analyses may be considered, contingent upon having sufficient statistical power, that is, ≥80% power to detect a 2-fold difference in risk between the baricitinib and the comparator cohort.
9.7.7.6. Patient Subgroups of Special Interest Demographic information from patients with severe hepatic impairment, evidence of hepatitis B or C infection, or a history of lymphoproliferative or malignant disease within the previous 5 years will be summarised and their proportion of the baricitinib-treated RA patient population estimated. The incidence of the aggregate outcomes of interest from Objective 1 will be estimated among the very elderly (aged ≥75 years) by subcohorts of bDMARD, cDMARD, and baricitinib users when possible, based on the population size and exposure time. Although preliminary investigation suggests that the use of JAK inhibitors during pregnancy may be uncommon, information about the safety of baricitinib among pregnant women is important. Therefore, periodic monitoring of the data will evaluate use during pregnancy. The frequency of congenital malformations will also be estimated following exposure, but only for the offspring of women in treatment during pregnancy. For each of these outcomes, counts and incidence will be provided as a measure of the frequency of these events.
9.7.8. Sensitivity Analysis Sensitivity analyses will be performed to examine the impact of assumptions on study conclusions. An underlying assumption for all the analyses presented in this protocol is the absence of unmeasured confounding. It is possible that some of the potential confounding variables may not be available within the national health registers. To address this issue, formal quantitative bias analysis methods, such as 1) a rule-out approach, as presented by Delaney and Seeger (2013), and/or 2) probabilistic bias analysis which uses estimates that are corrected based on plausible distributions of the unmeasured confounder through assessment of simulated datasets, as presented by Fox and Lash (2014), will be used for comparative analyses to quantify the effect that an unmeasured confounder would have on study results. Additional sensitivity analyses are presented below, by outcome.
9.7.8.1. Malignancy, Excluding Nonmelanoma Skin Cancer
9.7.8.1.1. Assessment of the Association between Duration of Baricitinib Exposure and Malignancy Incidence This sensitivity analysis will consider the risk of malignancy associated with cumulative duration of baricitinib exposure and will be conducted regardless of results from the main malignancy analysis. Cumulative duration of baricitinib exposure will be captured from the national rheumatology databases and will be calculated for each person in the baricitinib cohort. Exposure time will commence upon baricitinib initiation and will continue until the drug discontinuation date, initiation of another JAK inhibitor, the development of a malignancy, emigration, death, withdrawal from the study, or the end of the study follow-up period. Analyses I4V-MC-B011 Non-interventional PASS Protocol Page 51
will include crude incidence rates of malignancies excluding NMSC among quintiles of baricitinib exposure. Additionally, a logistic regression will be performed to assess the association of duration of baricitinib exposure with malignancy. The model will include duration of exposure and important confounding variables (Table 9.6). This analysis may be modified to focus on the class of JAK inhibitors rather than specifically baricitinib, if evaluating “ever exposure” to baricitinib without also including other JAK inhibitor use is not feasible due to small numbers. However, in the scenario where few patients exist with exposure to baricitinib and no exposure to other JAK inhibitors, descriptive statistics will be provided for patients exposed only to baricitinib.
9.7.8.1.2. Assessment of the Association between JAK Inhibitors and Malignancy If the primary analysis reveals a significant result in favour of either baricitinib or the comparator, a sensitivity analysis that considers JAK inhibitors (e.g., baricitinib, tofacitinib) as a class will be performed. Exposure cohorts will be constructed as described in Section 9.3.1.1 and analysis will proceed as outlined in Section 9.3.1.2. For this analysis, patients treated with JAK inhibitors other than baricitinib will be included. Descriptive statistics will be provided for both baricitinib and other JAK inhibitors, based on “ever” use of either medication group (or on exclusive use should there be >5 exclusive users in either group).
9.7.8.1.3. Assessment of the Association between Baricitinib Exposure and Malignancy, within Periods of Time after Initiating Medication A sensitivity analysis will be conducted to evaluate the malignancy events that occur only within the first 12 months after initiating medication. This analysis, and others such as within the first 24 months, will include only those malignancies that occur within the specified periods after the index date. The pattern of results from these analyses will aid in the identification of detection bias and potential clustering of events after initiating therapy. Analyses will proceed as described in Section 9.7.6.
9.7.8.1.4. Assessment of the Association between Baricitinib Exposure and Malignancy, Allowing for Various Latency Periods A sensitivity analysis will be conducted, regardless of the results of the primary analysis, to explore latency periods for the development of malignancies. This analysis will exclude malignancies that occur within 12 months after treatment initiation. Cancer develops over a lengthy period and the inclusion of events that cannot plausibly be related to a medication exposure will tend to bias results towards the null and reduce the ability to detect an effect, if one were truly to exist. By evaluating results of analyses with multiple exclusion windows (first 12 months, first 24 months, first 48 months or longer after treatment initiation, depending on the sample size available for analysis), these analyses will aid in the identification of events with longer latency.
9.7.8.2. Nonmelanoma Skin Cancer If the primary NMSC analysis reveals a significant result in favour of either baricitinib or the comparator, a sensitivity analysis that considers the JAK inhibitors baricitinib, tofacitinib, and I4V-MC-B011 Non-interventional PASS Protocol Page 52
any others as a class will be performed. Exposure cohorts will be constructed as described in Section 9.3.1.1, and analysis will proceed as outlined in Section 9.7.6.2.
9.7.8.3. Nonmalignancy Outcomes, All An analysis will be conducted in which a different rule will be used to assign events among patients who switch to a new medication before the at-risk window of the previous medication had ended or within a 30-day window after the end of the at-risk window (whichever is longer). In this sensitivity analysis, any event that occurs during the at-risk window of the previous medication, or within at least 30 days of switching to a new medication, will be attributed to the previous medication. This analysis will apply to patients switching within the bDMARD cohort or between the bDMARD and baricitinib cohorts. These analysis will include treatment episodes reflecting incident use and will not include episodes of medication restarts.
9.7.8.4. Serious Infections
9.7.8.4.1. Serious Infections Including Recurrent Infections An analysis that considers all serious infections, including any recurrent infections, will also be conducted as a sensitivity analysis. Propensity scores will be calculated, and patients will be matched on propensity score as described in Section 9.7.4. Generalised estimating equation negative binomial regression models with a log link will be used to estimate the relative rate and 95% CI for all serious infections between the baricitinib and cDMARDs cohorts and the baricitinib and biologic cohorts. The within-patient association will be accounted for by assuming a first-order autoregressive correlation structure. Any variables that remain unbalanced after propensity-score matching will be included in the model.
9.7.8.4.2. Including Patients with a History of Serious Infection A sensitivity analysis will be conducted including patients with a history of serious infection, defined as having 1 or more infections that required hospitalisation during the baseline period.
9.7.8.5. Opportunistic Infections
9.7.8.5.1. Including Recurrent Opportunistic Infections An analysis that considers all serious opportunistic infections, including recurrent opportunistic infections, will also be conducted as a sensitivity analysis. Propensity scores will be calculated, and patients will be matched on propensity score as described in Section 9.7.4. Generalised estimating equation negative binomial regression models with a log link will be used to estimate the relative rate and 95% CI for all opportunistic infections between the baricitinib and cDMARDs cohorts and the baricitinib and biologic cohorts. The within-patient association will be accounted for by assuming a first-order autoregressive correlation structure. Any variables that remain unbalanced after propensity-score matching will be included in the model. I4V-MC-B011 Non-interventional PASS Protocol Page 53
9.7.8.5.2. Including Patients with History of Opportunistic Infection A sensitivity analysis will be conducted including patients with history of opportunistic infection, defined as having one or more hospitalised opportunistic infections during the baseline period.
9.7.8.6. Major Adverse Cardiovascular Events
9.7.8.6.1. Intent-to-Treat Analysis of the Association between Baricitinib Exposure and MACE A sensitivity analysis to evaluate the occurrence of MACE as a result of atherosclerosis will be conducted. Similar to malignancy, these events are likely to have a long period of clinical latency before detection and will not be easily attributable to a specific exposure. To account for this ambiguity, this analysis will extend the at-risk period beyond the treatment window. In other words, this will estimate the risk associated with ever use of baricitinib using an intent-to-treat assignment of exposure, as described in Section 9.3.1.1. Analyses will otherwise proceed as described in Section 9.7.7.3.
9.7.8.6.2. Assessment of the Association between Duration of Baricitinib Exposure and MACE Incidence This sensitivity analysis will consider the risk of MACE associated with cumulative duration of baricitinib exposure and will be conducted regardless of results from the main MACE analysis. Cumulative duration of baricitinib exposure will be captured from physician enrolment or follow-up forms and will be calculated for each person in the baricitinib cohort. Exposure time will commence upon baricitinib initiation and will continue until the drug discontinuation date, initiation of another JAK inhibitor, the occurrence of MACE, death, disenrolment from the database, or the end of the study follow-up period. Analyses will include crude incidence rates of MACE among quintiles of baricitinib exposure. Additionally, a logistic regression will be performed to assess the association of duration of baricitinib exposure with MACE. The model will include duration of exposure and important confounding variables.
9.7.8.7. Venous Thromboembolic Events An analysis that excludes medications with recognised risk of VTEs according to the summary of product characteristics will also be conducted as a sensitivity analysis. Analyses will proceed as described in Section 9.7.7.4.
9.7.9. Assessment of Effectiveness of Risk Minimisation Activities Outcomes will be described for those patients with RA who are treated with baricitinib. Detailed information to describe the identification of the events of interest, that is, diagnostic codes and relevant algorithms, will be included in the SAP. These analyses will provide information to address important safety information, shown in italics, from the healthcare professional educational material and patient alert card (RMP v2.0, Annex 11 Mock-up of Proposed Additional EU Risk Minimisation Measures): 1. Infections I4V-MC-B011 Non-interventional PASS Protocol Page 54
Screen patients to rule out active tuberculosis and active viral hepatitis before starting Olumiant therapy o Examine the occurrence of first prescriptions for baricitinib among patients diagnosed with active tuberculosis or active viral hepatitis who have prescriptions consistent with such treatment. 2. Pregnancy Olumiant must not be used during pregnancy o Assess the proportion of women with diagnostic codes consistent with pregnancy who refill prescriptions for baricitinib during the period of time they might be pregnant. 3. Changes in lipid parameters Assess lipid parameters approximately 12 weeks following initiation of Olumiant therapy o Evaluate the initiation or escalation of lipid-lowering therapy in relation to use of baricitinib o Describe the temporal pattern of diagnostic codes for hyperlipidemia relative to a first prescription for baricitinib Data are not available for identifying lipid testing. The current analyses are intended as a proxy measure for increased monitoring of lipids within the recommended period.
9.8. Quality Control All data gathering and analyses will be overseen by 2 pharmacoepidemiologists experienced in the field of register-based research. Programming for this project will be conducted by a primary analyst and validated by a separate analyst (validation analyst). For all data processing steps, the validation analyst will review the programme along with input and output datasets. For the analysis steps of the project, double-programming techniques to reduce the potential for programming errors will be employed.
9.9. Limitations of the Research Methods The study makes use of existing data from central databases and health registries that serve clinical and administrative purposes not related to the present study. Because features of RA may themselves influence the frequency of some or all of the outcomes of interest for this study, such as disease severity or duration, patients treated for RA with biologic medications are the most appropriate comparison group since RA patients treated with biologics are likely those with moderate-to-severe RA. The use of concomitant cDMARD and other medication received (e.g., intramuscular glucocorticoids), proxies for disease severity or treatment response, will also be adjusted for in the analyses. The following points should be considered: The quality of the data (i.e., completeness, PPV) is generally high, but coding errors and misclassifications may exist due to differences in coding practices across institutions and countries. I4V-MC-B011 Non-interventional PASS Protocol Page 55
The majority of sales of non-prescription OTC medicines are not in the prescription databases. Only OTC medicines prescribed and dispensed to individual patients, for example, for obtaining reimbursement in chronic diseases, are included. The indication for use and the prescribed dose are to some extent included, but only as free text and thus not easily available for research purposes. Patient-level data on drug use in hospitals and other institutions are not routinely collected. The performance of an algorithm to identify cases is a balance between the sensitivity of the case definition and its specificity. When identifying all possible cases is prioritised, this tends to increase the proportion of false positives and includes some events that are not true cases. Conversely, when a very specific algorithm is the goal, this comes at the cost of a higher proportion of false negatives or missed diagnoses. Misclassification of the outcome, i.e, false positives and missed diagnoses, can bias results towards or away from the null, depending on whether misclassification is present to the same extent in the groups being compared or disproportionately affects one of the treatment exposure cohorts. In general, poor performance of a selected outcome definition should affect the comparison groups (patients treated with baricitinib and those treated with bDMARDs) equally. Such non-differential misclassification of the outcome would reduce the ability of the study to detect a positive association, should one exist, but any results obtained would be unbiased estimates of the true relative risk. When available, definitions for specific outcomes will be based on validated claims-based algorithms and which will be evaluated based on the result of literature reviews. Not all outcomes will have well-defined diagnostic codes and procedures to identify them, however, and this will be a limitation of the study.
9.9.1. Channelling Bias in Observational Studies Drug exposure in pharmacoepidemiological studies does not occur completely at random and is a result of patient, physician, and system-related factors. When these factors are associated with the outcome of interest, comparisons of different drug-exposure cohorts will be confounded due to channelling bias. The present study addresses this limitation by applying propensity-score matching, as appropriate for each outcome. Propensity scores address this imbalance by providing a mechanism to compare patients with concordant baseline risk but discordant exposure (Schneeweiss 2007). Calendar-specific matching will be implemented to account for changes in treatment patterns that commonly occur after a new drug enters the marketplace; however, propensity scores are only able to adjust for measured confounders. The possibility of unmeasured confounding and the possible influence on study results will be considered in the final report.
9.9.2. Assessing Malignancy Risk in a Real-World Setting Malignancy outcomes have a long latency period and as a result are not easily attributed to a particular drug exposure. To account for this ambiguity, the analysis for malignancy considers the risk associated with the use of baricitinib, regardless of subsequent medication changes. Although this approach is considered to be conservative from the standpoint that attribution of a I4V-MC-B011 Non-interventional PASS Protocol Page 56 malignancy to baricitinib will not be missed, it ignores the duration of baricitinib exposure and exposure to other systemic medications. Typically, the risk of malignancy increases with increasing exposure to an identified risk factor. If treatment with baricitinib were a risk factor for malignancy, combining patients with varying durations of exposure would have the potential to bias the measure of association towards the null. To address this issue, an analysis that examines the duration of baricitinib use will be performed. Another challenge of studying malignancy is the effect of screening on the incidence rate. If the analysis reveals an association between exposure to baricitinib and malignancy or exposure to a JAK inhibitor and malignancy, a sensitivity analysis will be conducted that eliminates malignancy cases within the first 12 months of follow-up due to the low likelihood of biologic plausibility and the possibility of detection bias resulting from increased surveillance upon switching to a new medication. Only new initiators at baseline will be considered for the analysis in order to identify malignancy cases that occur within 12 months of drug initiation. Finally, the class of JAK inhibitors is fairly new, and questions remain regarding the risk of baricitinib relative to other JAK medications. If the malignancy analysis is able to consider the risk of malignancy associated with baricitinib, rather than the class of JAK inhibitors, a sensitivity analysis that compares baricitinib to other JAK inhibitors will be conducted. This sensitivity analysis will test the hypothesis that the effects of baricitinib and other JAK inhibitors on the incidence of malignancy are similar.
9.9.3. Evaluation of Patient Outcomes A survey of healthcare professionals, that is, rheumatologists (I4V-MC-B010) will assess physicians’ understanding of the important safety messages included in the healthcare communication and in the patient alert card (included in the EU-RMP v2.0). Data from the Nordic registries included in this study will contribute to the assessment of risk minimisation activities in Europe for baricitinib by providing information on the actual outcomes observed among patients who would have received the patient alert card and would have been treated by physicians who received the healthcare communication. However, it is important to recognise that patient and prescriber behaviours cannot be easily evaluated from patient clinical information. In particular, the monitoring of lipids cannot be directly assessed as no data are available to identify if laboratory tests have been ordered. The proposed analyses that evaluate changes in lipid parameters instead, serve as proxies.
9.9.4. Generalisability The national health registers from the Nordic countries include longitudinal data from each entire nation; there is minimal risk of selection bias. Therefore, the findings from this study should be generalisable, as the limitations in the present study are not expected to affect the biologic relations studied. The national health registers from the Nordic countries include longitudinal data from each entire nation. The patients with RA included in this study represent all, or essentially all, patients I4V-MC-B011 Non-interventional PASS Protocol Page 57 receiving RA care so there is minimal risk of selection bias. The comprehensive nature of these data will ensure that any results from this study will apply to all patients in Nordic countries. Patients with RA in Nordic countries may not necessarily represent all adults with RA in Europe. Examination of baseline characteristics and comparison with other data sources that include patients with RA may help to clarify the extent to which results from the registry have external validity. Regardless, findings from this study are expected to be internally valid and provide valuable information about the long-term safety of baricitinib.
9.10. Other Aspects Not applicable.
9.10.1. File Retention and Archiving All datasets from the 4 Nordic countries will be gathered and analysed at Statistics Denmark. Only aggregated anonymised data containing information with more than 3 individuals will be extracted. Data will be stored according to the strict guidelines by Statistics Denmark. Datasets will be archived and stored at Statistics Denmark until the end of the study, when all analyses have been performed. I4V-MC-B011 Non-interventional PASS Protocol Page 58
10. Protection of Human Subjects
For each country, ethical approval will be obtained if required by national rules and regulations. The approvals will be required before the start of the study. In Denmark, secondary data do not require ethical permission, but approval from the Danish Data Protection Agency according to the Act on Processing of Personal Data. In Finland, an approval from the ethical board is not necessary when applying for register data. Instead, an application needs to be completed and sent to the to the controllers of the registers. An example of a register controller is National Institute for Health and Welfare (THL) or Statistics Finland. This application process also applies to ROB-FIN register. In Norway, an approval is needed from both the Norwegian National Research Ethics Committees (REK) and the Norwegian Data Protection Agency (NDPA) according to the Personal Data Act. Some of the registers require an additional approval from REK if they assess that the level of detailed information or number of variables can be indirectly identifiable from 1 or more health registers. The approval from REK is a dispensation from secrecy. In Sweden, all studies involving living individuals must be approved by a Regional Ethical Review Board (EPN) when the research involves the handling of personal data. Personal data refers to any information that is directly or indirectly attributable to a living individual. I4V-MC-B011 Non-interventional PASS Protocol Page 59
11. Management and Reporting of Adverse Events/Adverse Reactions
This is a non-interventional study based on secondary use of data and, therefore, all protocol-defined adverse events (AEs) will be collected and summarised in the final study report. Protocol-defined AEs: see Section 8. No other AEs will be collected. I4V-MC-B011 Non-interventional PASS Protocol Page 60
12. Plans for Disseminating and Communicating Study Results
The final report will be submitted to the European Medicines Agency. The study will also be registered in the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP) Registry, and the study findings may be submitted to a scientific congress and/or to a peer-reviewed journal. I4V-MC-B011 Non-interventional PASS Protocol Page 61
13. References
Allaire S, Wolfe F, Niu J, LaValley MP, Zhang B, Reisine S. Current risk factors for work disability associated with rheumatoid arthritis: Recent data from a US national cohort. Arthritis Rheum. 2009;61(3):321-328. Andersen JT, Mastrogiannis D, Andersen NL, Petersen M, Broedbaek K, Cejvanovic V, Nielsen TK, Poulsen HE, Jimenez-Solem E. Diclofenac/misoprostol during early pregnancy and the risk of miscarriage: a Danish nationwide cohort study. Arch Gynecol Obstet. 2016;294(2):245-250. Andersen TF, Madsen M, Jorgensen J, Mellemkjoer L, Olsen JH. The Danish National Hospital Register. A valuable source of data for modern health sciences. Dan Med Bull. 1999;46(3):263-268. Antti Kuuliala: Studies on immune activation in rheumatoid arthritis and reactive arthritis [FINNISH]. [cited 2017 Jul 24]. Available at: http://ethesis.helsinki.fi/julkaisut/laa/haart/vk/kuuliala/tiiviste.html Askling J, van Vollenhoven RF, Granath F, Raaschou P, Fored CM, Baecklund E, Dackhammar C, Feltelius N, Coster L, Geborek P, Jacobsson LT, Lindblad S, Rantapaa-Dahlqvist S, Saxne T, Klareskog L. Cancer risk in patients with rheumatoid arthritis treated with anti-tumor necrosis factor alpha therapies: does the risk change with the time since start of treatment? Arthritis Rheum. 2009;60(11):3180-9. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivar Behav Res. 2011;46(3):399-424. Austin PC, Mamdani MM. A comparison of propensity score methods: a case-study estimating the effectiveness of post-AMI statin use. Stat Med. 2006;25(12):2084-2106. Baadsgaard M, Quitzau J. Danish registers on personal income and transfer payments. Scand J Public Health. 2011;39(7 suppl):103-105. [CDC] Centers for Disease Control and Prevention. Rheumatoid arthritis (RA). 2015. Available at: http://www.cdc.gov/arthritis/basics/rheumatoid.htm. Accessed June 7, 2017. Chang K, Yang S, Kim S, Han K, Park S, Shin J. Smoking and Rheumatoid Arthritis. Int J Mol Sci. 2014;15(12):22279-22295. Centre for Epidemiology, National Board of Health and Welfare. The Swedish Medical Birth Register: A Summary of Content and Quality. 2003. Available at: http://www.socialstyrelsen.se/publikationer2003/2003-112-3. Choy EH, Panayi GS. Cytokine pathways and joint inflammation in rheumatoid arthritis. N Engl J Med. 2001;344(12):907-916. Colmegna I, Ohata BR, Menard HA. Current understanding of rheumatoid arthritis therapy. Clin Pharmacol Ther. 2012;91(4):607-620. Crowson CS, Matteson EL, Davis JM, Gabriel SE. Contribution of obesity to the rise in incidence of rheumatoid arthritis. Arthritis Care Res. 2013;65(1):71-77. D’Agostino RB. Estimating treatment effects using observational data. JAMA. 2007;297(3):314. I4V-MC-B011 Non-interventional PASS Protocol Page 62
Delaney JA, Seeger JD. Sensitivity Analysis. Agency for Healthcare Research and Quality (US); 2013 [cited 2017 Jul 23]. Available at: https://www-ncbi-nlm-nih- gov.ep.fjernadgang.kb.dk/books/NBK126178/. Education Register–Statistics Norway. Available at: https://www.ssb.no/a/english/publikasjoner/pdf/nos_d361_en/nos_d361_en.pdf. Epland J. Towards a Register-Based Income Statistics – the Construction of the Norwegian Income Register. Available at: https://www.ssb.no/a/histstat/doc/doc_199805.pdf. Eriksson JK, Askling J, Arkema EV. The Swedish Rheumatology Quality Register: optimisation of rheumatic disease assessments using register-enriched data. Clin Exp Rheumatol. 2014;32(5 suppl 85):S-147-149. Eriksson JK, Neovius M, Ernestam S, Lindblad S, Simard JF, Askling J. Incidence of rheumatoid arthritis in Sweden: a nationwide population-based assessment of incidence, its determinants, and treatment penetration. Arthritis Care Res. 2013;65(6):870-878. Frontpage – Väestörekisterikeskus. [cited 2017 Jul 25]. Available at: http://vrk.fi/en/frontpage. Furu K. Establishment of the nationwide Norwegian Prescription Database (NorPD) – new opportunities for research in pharmacoepidemiology in Norway. Nor Epidemiol. 2008;18(2):129-136. Furu K, Wettermark B, Andersen M, Martikainen JE, Almarsdottir AB, Sørensen HT. The Nordic countries as a cohort for pharmacoepidemiological research. Basic Clin Pharmacol Toxicol. 2010;106(2):86-94. Gaist D, Sorensen HT, Hallas J. The Danish prescription registries. Dan Med Bull. 1997;44(4):445-448. Haider SI, Johnell K, Ringbäck Weitoft G, Thorslund M, Fastbom J. Patient educational level and use of newly marketed drugs: a register-based study of over 600,000 older people. Eur J Clin Pharmacol. 2008;64(12):1215-1222. Helweg-Larsen K. The Danish register of causes of death. Scand J Public Health. 2011;39(7 suppl):26-29. Holmqvist ME, Neovius M, Eriksson J, Mantel A, Wallberg-Jonsson S, Jacobsson LT, Askling J. Risk of venous thromboembolism in patients with rheumatoid arthritis and association with disease duration and hospitalization. JAMA. 2012;308(13):1350-6. Ibfelt EH, Jensen D, Hetland ML. The Danish nationwide clinical register for patients with rheumatoid arthritis: DANBIO. Clin Epidemiol. 2016;8:737-742. Irgens LM. The Medical Birth Registry of Norway. Epidemiological research and surveillance throughout 30 years. Acta Obstet Gynecol Scand. 2000;79(6):435-439. Jensen VM, Rasmussen AW. Danish Education Registers. Scand J Public Health. 2011;39(7 suppl):91-94. Joensen AM, Jensen MK, Overvad K, Dethlefsen C, Schmidt E, Rasmussen L, Tjønneland A, Johnsen S. Predictive values of acute coronary syndrome discharge diagnoses differed in the Danish National Patient Registry. J Clin Epidemiol. 2009;62(2):188-194. I4V-MC-B011 Non-interventional PASS Protocol Page 63
Johnell K, Weitoft GR, Fastbom J. Education and use of dementia drugs: a register-based study of over 600,000 older people. Dement Geriatr Cogn Disord. 2008;25(1):54-59. Kildemoes HW, Sorensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health. 2011;39(7 suppl):38-41. Knudsen LB, Olsen J. The Danish Medical Birth Registry. Dan Med Bull. 1998;45(3):320-323. Konttinen L, Honkanen V, Uotila T, Pöllänen J, Waahtera M, Romu M, Puolakka K, Vasala M, Karjalainen A, Luukkainen R, Nordström DC; ROB-FIN study group. Biological treatment in rheumatic diseases: results from a longitudinal surveillance: adverse events. Rheumatol Int. 2006;26(10):916-922. Krarup L-H, Boysen G, Janjua H, Prescott E, Truelsen T. Validity of stroke diagnoses in a National Register of Patients. Neuroepidemiology. 2007;28(3):150-154. Kristiansen IS, Pedersen KM. [Health care systems in the Nordic countries – more similarities than differences?]. Tidsskr Den Nor Laegeforening Tidsskr Prakt Med Ny Raekke. 2000;120(17):2023-2029. Kvien TK, Heiberg, Lie E, Kaufmann C, Mikkelsen K, Nordvåg BY, Rødevand E. A Norwegian DMARD register: prescriptions of DMARDs and biological agents to patients with inflammatory rheumatic diseases. Clin Exp Rheumatol. 2005;23(5 suppl 39):S188-194. Langhoff-Roos J, Krebs L, Klungsøyr K, Bjarnadottir RI, Källén K, Tapper A-M, Jakobsson M, Børdahl PE, Lindqvist PG, Gottvall K, Colmorn LB, Gissler M. The Nordic medical birth registers – a potential goldmine for clinical research. Acta Obstet Gynecol Scand. 2014;93(2):132-137. Lash TL, Fox MP, MacLehose RF, Maldonado G, McCandless LC, Greenland S. Good practices for quantitative bias analysis. Int J Epidemiol. 2014;43(6):1969-1985. Leddgikt (Revmatoid artritt). Norsk Revmatikerforbund. Available at: https://www.revmatiker.no/diagnose/leddgikt/. Leonard WJ, O’Shea JJ. JAKS AND STATS: Biological Implications. Annu Rev Immunol. 1998;16(1):293-322. Lindhardsen J, Gislason GH, Jacobsen S, Ahlehoff O, Olsen AM, Madsen OR, Torp-Pedersen C, Hansen PR. Non-steroidal anti-inflammatory drugs and risk of cardiovascular disease in patients with rheumatoid arthritis: a nationwide cohort study. Ann Rheum Dis. 2014;73(8):1515-21. Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim J-L, Reuterwall C, Heurgren M, Olausson PO. External review and validation of the Swedish national inpatient register. BMC Public Health. 2011;11:450. Ludvigsson JF, Almqvist C, Bonamy A-KE, Ljung R, Michaëlsson K, Neovius M, et al. Stephansson O, Ye W. Registers of the Swedish total population and their use in medical research. Eur J Epidemiol. 2016 Feb;31(2):125–-136. Mack CD, Glynn RJ, Brookhart MA, Carpenter WR, Meyer AM, Sandler RS, Stürmer T. Calendar time-specific propensity scores and comparative effectiveness research for stage III colon cancer chemotherapy. Pharmacoepidemiol Drug Saf. 2013;22(8):810-818. I4V-MC-B011 Non-interventional PASS Protocol Page 64
Lund JL, Richardson DB, Sturmer T. The active comparator, new user study design in pharmacoepidemiology: historical foundations and contemporary application. Curr Epidemiol Rep. 2015;2(4):221-8. Madsen KM, Schønheyder HC, Kristensen B, Nielsen GL, Sørensen HT. Can hospital discharge diagnosis be used for surveillance of bacteremia? A data quality study of a Danish hospital discharge registry. Infect Control Hosp Epidemiol. 1998;19(3):175-180. Marks JL, Edwards CJ. Protective effect of methotrexate in patients with rheumatoid arthritis and cardiovascular comorbidity. Ther Adv Musculoskelet Dis. 2012;4(3):149-157. Neovius M, Simard JF, Askling J, ARTIS study group. Nationwide prevalence of rheumatoid arthritis and penetration of disease-modifying drugs in Sweden. Ann Rheum Dis. 2011;70(4):624-629. Nilssen Y, Strand T-E, Wiik R, Bakken IJ, Yu XQ, O’Connell DL, Møller B1. Utilizing national patient-register data to control for comorbidity in prognostic studies. Clin Epidemiol. 2014;6:395-404. Nørgaard M, Skriver MV, Gregersen H, Pedersen G, Schønheyder HC, Sørensen HT. The data quality of haematological malignancy ICD-10 diagnoses in a population-based hospital discharge registry. Eur J Cancer Prev. 2005;14(3):201-206. Olumiant [Summary of Product Characteristics]. Available at: http://www.ema.europa.eu/docs/en_GB/document_library/EPAR_- _Product_Information/human/004085/WC500223723.pdf. Accessed August 14, 2017. O’Shea JJ, Kontzias A, Yamaoka K, Tanaka Y, Laurence A. Janus kinase inhibitors in autoimmune diseases. Ann Rheum Dis. 2013;72(suppl 2):ii111-ii115. Pedersen AG, Ellingsen CL. Data quality in the causes of death registry. Tidsskr Den Nor Laegeforening Tidsskr Prakt Med Ny Raekke. 2015;135(8):768-770. Pedersen CB. The Danish Civil Registration System. Scand J Public Health. 2011;39(7 suppl):22-25. doi:10.1177/1403494810387965. Puolakka K, Kautiainen H, Pohjolainen T, Virta L. Rheumatoid arthritis (RA) remains a threat to work productivity: a nationwide register-based incidence study from Finland. Scand J Rheumatol. 2010;39(5):436-438. Ramiro S, Sepriano A, Chatzidionysiou K, Nam JL, Smolen JS, van der Heijde D, Dougados M, van Vollenhoven R, Bijlsma JW, Burmester GR, Scholte-Voshaar M, Falzon L, Landewé RBM. Safety of synthetic and biological DMARDs: a systematic literature review informing the 2016 update of the EULAR recommendations for management of rheumatoid arthritis. Ann Rheum Dis. 2017;76(6):1101-1136. Ringbäck Weitoft G, Ericsson O, Löfroth E, Rosén M. Equal access to treatment? Population-based follow-up of drugs dispensed to patients after acute myocardial infarction in Sweden. Eur J Clin Pharmacol. 2008;64(4):417-424. Rikala M, Hartikainen S, Sulkava R, Korhonen MJ. Validity of the Finnish Prescription Register for measuring psychotropic drug exposures among elderly Finns. Drugs Aging. 2010;27(4):337-349. I4V-MC-B011 Non-interventional PASS Protocol Page 65
Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41. Royston P. Multiple imputation of missing values. The Stata Journal. 2004;4(3):227–241. Royston P, White I. Multiple Imputation by Chained Equations (MICE): Implementation in Stata. J Stat Softw. 2011;45(4). Available at: http://dx.doi.org/10.18637/jss.v045.i04. Rubin DB. The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials. Stat Med. 2006;26(1):20-36. Schmidt M, Pedersen L, Sørensen HT. The Danish Civil Registration System as a tool in epidemiology. Eur J Epidemiol. 2014;29(8):541-549. Schneeweiss S. Developments in post-marketing comparative effectiveness research. Clin Pharmacol Ther. 2007;82(2):143-156. Schneeweiss S, Gagne JJ, Glynn RJ, Ruhl M, Rassen JA. Assessing the comparative effectiveness of newly marketed medications: methodological challenges and implications for drug development. Clin Pharmacol Ther. 2011;90(6):777-790. Seeger JD, Walker AM, Williams PL, Saperia GM, Sacks FM. A propensity score-matched cohort study of the effect of statins, mainly fluvastatin, on the occurrence of acute myocardial infarction. Am J Cardiol. 2003;92(12):1447-1451. Selmer R, Sakshaug S, Skurtveit S, Furu K, Tverdal A. Statin treatment in a cohort of 20 212 men and women in Norway according to cardiovascular risk factors and level of education. Br J Clin Pharmacol. 2009;67(3):355-362. Singh JA, Saag KG, Bridges SL Jr, Akl EA, Bannuru RR, Sullivan MC, Vaysbrot E, McNaughton C, Osani M, Shmerling RH, Curtis JR, Furst DE, Parks D, Kavanaugh A, O’Dell J, King C, Leong A, Matteson EL, Schousboe JT, Drevlow B, Ginsberg S, Grober J, St Clair EW, Tindall E, Miller AS, McAlindon T; American College of Rheumatology. 2015 American College of Rheumatology Guideline for the Treatment of Rheumatoid Arthritis. Arthritis Care Res. 2015;68(1):1-25. Skatteetaten. The National Registry [Skatteetaten – Det sentrale folkeregisteret] NORWEGIAN. 2012 [cited 2017 Jul 25]. Available at: http://www.skatteetaten.no/en/About- Skatteetaten/Statistikk-og-analyse/En-oversikt-over-datainnhold/Det-sentrale- folkeregisteret/Det-sentrale-folkeregisteret/. Smolen JS, Aletaha D, Koeller M, Weisman MH, Emery P. New therapies for treatment of rheumatoid arthritis. The Lancet. 2007;370(9602):1861-1874. Smolen JS, Steiner G. Therapeutic strategies for rheumatoid arthritis. Nat Rev Drug Discov. 2003;2(6):473-488. Solomon DH, Reed GW, Kremer JM, Curtis JR, Farkouh ME, Harrold LR, Hochberg MC, Tsao P, Greenberg JD. Disease activity in rheumatoid arthritis and the risk of cardiovascular events. Arthritis Rheumatol. 2015;67(6):1449-1455. Spreeuwenberg MD, Bartak A, Croon MA, Hagenaars JA, Busschbach JJV, Andrea H, Twisk J, Stijnen T. The multiple propensity score as control for bias in the comparison of more than two treatment arms. Med Care. 2010;48(2):166-174. I4V-MC-B011 Non-interventional PASS Protocol Page 66
Statistics Denmark. Containing detailed statistical information on the Danish Society. 2012 [cited 2012 Dec 10]. Available at: http://www.dst.dk/en. Statistics Finland – Education. [cited 2017 Jul 27]. Available at: http://www.stat.fi/til/kou_en.html. Statistics Finland – Health – Causes of death. [cited 2017 Jul 25]. Available at: http://tilastokeskus.fi/til/ksyyt/index_en.html. Statistics Finland – Income and Consumption. [cited 2017 Jul 27]. Available at: http://www.stat.fi/til/tul_en.html. Statistics Sweden. Statistiska Centralbyrån. [cited 2017 Jul 27]. Available at: http://www.scb.se/en/. [Swedish Cause of Death Registry] Dödsorsaksregistret – SWEDISH. [cited 2017 Jul 25]. Available at: http://www.socialstyrelsen.se/register/dodsorsaksregistret. Swedish Register of Education. [cited 2017 Jul 27]. Available at: http://www.jpi- dataproject.eu/Home/Database/348?topicId=4. Tamariz L, Harkins T, Nair V. A systematic review of validated methods for identifying venous thromboembolism using administrative and claims data. Pharmacoepidemiol Drug Saf. 2012;21(suppl 1):154-162. The Finish Register of Completed Education and Degrees and student flow statistics. Available at: http://www.stat.fi/til/kou_en.html. The Swedish Register of Education. Statistics Sweden; Available at: http://www.scb.se/statistik/UF/UF0506/Produktbeskrivning_short_English_UF0506_2004010 1r.doc. THL – Care Register for Health Care. The National Institute for Health and Welfare (THL), Finland. [cited 2017 Jul 25]. Available at: http://www.thl.fi/en/web/thlfi- en/statistics/information-on-statistics/register-descriptions/care-register-for-health-care. THL – Medical Birth Register. The National Institute for Health and Welfare (THL), Finland. [cited 2017 Jul 25]. Available at: http://www.thl.fi/en/web/thlfi-en/statistics/information-on- statistics/register-descriptions/newborns. Thode HC. Power and sample size requirements for tests of differences between two Poisson rates. J R Stat Soc Ser Stat. 1997;46(2):227-230. Thygesen SK, Christiansen CF, Christensen S, Lash TL, Sørensen HT. The predictive value of ICD-10 diagnostic coding used to assess Charlson comorbidity index conditions in the population-based Danish National Registry of Patients. BMC Med Res Methodol. 2011;11:83. van Buuren S, Boshuizen HC, Knook DL. Multiple imputation of missing blood pressure covariates in survival analysis. Stat Med. 1999;18(6):681-694. van Dartel SA, Fransen J, Kievit W, Flendrie M, den Broeder AA, Visser H, Hartkamp A, van de Laar MA, van Riel PL. Difference in the risk of serious infections in patients with rheumatoid arthritis treated with adalimumab, infliximab and etanercept: results from the Dutch Rheumatoid Arthritis Monitoring (DREAM) registry. Ann Rheum Dis. 2013;72(6):895-90 I4V-MC-B011 Non-interventional PASS Protocol Page 67
Wasserman AM. Diagnosis and management of rheumatoid arthritis. Am Fam Physician. 2011;84(11):1245-1252. Webster PC. Sweden’s health data goldmine. CMAJ Can Med Assoc J. 2014;186(9):E310. Wettermark B, Hammar N, Fored CM, Leimanis A, Otterblad Olausson P, Bergman U, Persson I, Sundström A, Westerholm B, Rosén M. The new Swedish Prescribed Drug Register – opportunities for pharmacoepidemiological research and experience from the first six months. Pharmacoepidemiol Drug Saf. 2007;16(7):726-735. I4V-MC-B011 Non-interventional PASS Protocol Page 68
14. Annex I4V-MC-B011 Non-interventional PASS Protocol Page 69
Annex 1. List of Standalone Documents None. I4V-MC-B011 Non-interventional PASS Protocol Page 70
Annex 2. ENCePP Checklist for Study Protocols Study title:
Study I4V-MC-B011: A Retrospective Cohort Study to Assess the Safety of Baricitinib Compared with Other Therapies Used in the Treatment of Rheumatoid Arthritis in Nordic countries
Study reference number:
Section 1: Milestones Yes No N/A Section Number
1.1 Does the protocol specify timelines for
1 1.1.1 Start of data collection 6 2 1.1.2 End of data collection 6
1.1.3 Study progress report(s) 6
1.1.4 Interim progress report(s)
1.1.5 Registration in the EU PAS register 6
1.1.6 Final report of study results. 6
Comments:
Section 2: Research question Yes No N/A Section Number
2.1 Does the formulation of the research question and objectives clearly explain:
2.1.1 Why the study is conducted? (e.g. to address an
1 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. 2 Date from which the analytical dataset is completely available. I4V-MC-B011 Non-interventional PASS Protocol Page 71
Section 2: Research question Yes No N/A Section Number
important public health concern, a risk identified in the risk 7 management plan, an emerging safety issue)
2.1.2 The objective(s) of the study? 8
2.1.3 The target population? (i.e. population or subgroup to whom the study results are intended to be generalised) 8
2.1.4 Which formal hypothesis(-es) is (are) to be tested?
2.1.5 If applicable, that there is no a priori hypothesis?
Comments:
Section 3: Study design Yes No N/A Section Number
3.1 Is the study design described? (e.g. cohort, case-control, 9.1 cross-sectional, new or alternative design)
3.2 Does the protocol specify whether the study is based on primary, secondary or combined data 9.1 collection?
3.3 Does the protocol specify measures of occurrence? 9.7 (e.g. incidence rate, absolute risk)
3.4 Does the protocol specify measure(s) of association? (e.g. relative risk, odds ratio, excess risk, incidence rate ratio, 9.7 hazard ratio, number needed to harm (NNH) per year)
3.5 Does the protocol describe the approach for the collection and reporting of adverse events/adverse 8 reactions? (e.g. adverse events that will not be collected in case of primary data collection) I4V-MC-B011 Non-interventional PASS Protocol Page 72
Comments:
Section 4: Source and study populations Yes No N/A Section Number
4.1 Is the source population described? 9.4
4.2 Is the planned study population defined in terms of:
4.2.1 Study time period? 6
4.2.2 Age and sex? 9.1
4.2.3 Country of origin? 9.1
4.2.4 Disease/indication? 9.1
4.2.5 Duration of follow-up? 9.5
4.3 Does the protocol define how the study population will be sampled from the source population? (e.g. 9.2.1.1 event or inclusion/exclusion criteria)
Comments:
Section 5: Exposure definition and measurement Yes No N/A Section Number
5.1 Does the protocol describe how the study exposure is defined and measured? (e.g. operational details for 9.3.1 defining and categorising exposure, measurement of dose and duration of drug exposure)
5.2 Does the protocol address the validity of the exposure measurement? (e.g. precision, accuracy, use of 9.3.1 validation sub-study)
5.3 Is exposure classified according to time windows? (e.g. current user, former user, non-use) I4V-MC-B011 Non-interventional PASS Protocol Page 73
Section 5: Exposure definition and measurement Yes No N/A Section Number
9.3.1
5.4 Is exposure classified based on biological mechanism of action and taking into account the 9.3.1.3.1 pharmacokinetics and pharmacodynamics of the drug?
Comments:
Section 6: Outcome definition and measurement Yes No N/A Section Number
6.1 Does the protocol specify the primary and secondary 9.3.2. (if applicable) outcome(s) to be investigated?
6.2 Does the protocol describe how the outcomes are defined and measured? 9.3.2
6.3 Does the protocol address the validity of outcome measurement? (e.g. precision, accuracy, sensitivity, 9.3.2 specificity, positive predictive value, prospective or retrospective ascertainment, use of validation sub-study)
6.4 Does the protocol describe specific endpoints relevant for Health Technology Assessment? (e.g. HRQoL, QALYs, DALYS, health care services utilisation, burden of disease, disease management)
Comments:
Section 7: Bias Yes No N/A Section Number
7.1 Does the protocol describe how confounding will be 9.7 and 9.9 addressed in the study? I4V-MC-B011 Non-interventional PASS Protocol Page 74
Section 7: Bias Yes No N/A Section Number
7.1.1 Does the protocol address confounding by 9.7 and 9.9 indication if applicable?
7.2 Does the protocol address:
7.2.1 Selection biases (e.g. healthy user bias) 9.9
7.2.2 Information biases (e.g. misclassification of 9.9 exposure and endpoints, time-related bias)
7.3 Does the protocol address the validity of the study 9.3.2 covariates?
Comments:
Section 8: Effect modification Yes No N/A Section Number
8.1 Does the protocol address effect modifiers? (e.g. 9.7.8.4.2 collection of data on known effect modifiers, sub-group analyses, and anticipated direction of effect) 9.7.8.5.2
Comments:
Section 9: Data sources Yes No N/A Section Number
9.1 Does the protocol describe the data source(s) used in the study for the ascertainment of:
9.1.1 Exposure? (e.g. pharmacy dispensing, general practice 9.2.2 and prescribing, claims data, self-report, face-to-face interview) 9.4
9.1.2 Outcomes? (e.g. clinical records, laboratory markers or 9.4 values, claims data, self-report, patient interview including scales I4V-MC-B011 Non-interventional PASS Protocol Page 75
Section 9: Data sources Yes No N/A Section Number
and questionnaires, vital statistics)
9.1.3 Covariates? 9.3.3
9.2 Does the protocol describe the information available from the data source(s) on:
9.2.1 Exposure? (e.g. date of dispensing, drug quantity, 9.2.2 and dose, number of days of supply prescription, daily dosage, 9.4 prescriber)
9.2.2 Outcomes? (e.g. date of occurrence, multiple event, 9.4 severity measures related to event)
9.2.3 Covariates? (e.g. age, sex, clinical and drug use 9.3.3 history, co-morbidity, co-medications, lifestyle)
9.3 Is a coding system described for:
9.3.1 Exposure? (e.g. WHO Drug Dictionary, Anatomical 9.3.1 Therapeutic Chemical (ATC) Classification System)
9.3.2 Outcomes? (e.g. International Classification of Diseases 9.3.2 (ICD)-10, Medical Dictionary for Regulatory Activities (MedDRA))
9.3.3 Covariates? 9.3.3
9.4 Is the linkage method between data sources 9.4.1 described? (e.g. based on a unique identifier or other)
Comments:
Section 10: Analysis plan Yes No N/A Section Number
10.1 Is the choice of statistical techniques described? 9.7 I4V-MC-B011 Non-interventional PASS Protocol Page 76
Section 10: Analysis plan Yes No N/A Section Number
10.2 Are descriptive analyses included? 9.7
10.3 Are stratified analyses included? 9.7.5
10.4 Does the plan describe methods for adjusting for confounding? 9.7
10.5 Does the plan describe methods for handling missing data?
10.6 Is sample size and/or statistical power estimated? 9.5
Comments:
Section 11: Data management and quality control Yes No N/A Section Number
11.1 Does the protocol provide information on data 9.6 storage? (e.g. software and IT environment, database maintenance and anti-fraud protection, archiving)
11.2 Are methods of quality assurance described? 9.8
11.3 Is there a system in place for independent review 9.8 of study results?
Comments:
Section 12: Limitations Yes No N/A Section Number
12.1 Does the protocol discuss the impact on the study results of:
12.1.1 Selection biases? 9.9.3
12.1.2 Information biases? I4V-MC-B011 Non-interventional PASS Protocol Page 77
Section 12: Limitations Yes No N/A Section Number
12.1.3 Residual/unmeasured confounding? 9.7
(e.g. anticipated direction and magnitude of such biases, validation sub-study, use of validation and external data, analytical methods)
12.2 Does the protocol discuss study feasibility? (e.g. 9.5 study size, anticipated exposure, duration of follow-up in a cohort study, patient recruitment)
Comments:
Section 13: Ethical issues Yes No N/A Section Number
13.1 Have requirements of Ethics 10 Committee/Institutional Review Board been described?
13.2 Has any outcome of an ethical review procedure been addressed?
13.3 Have data protection requirements been 10 described?
Comments:
Section 14: Amendments and deviations Yes No N/A Section Number
14.1 Does the protocol include a section to document future amendments and deviations?
Comments: I4V-MC-B011 Non-interventional PASS Protocol Page 78
Section 15: Plans for communication of study Yes No N/A Section results Number
15.1 Are plans described for communicating study 12 results (e.g. to regulatory authorities)?
15.2 Are plans described for disseminating study results 12 externally, including publication?
Comments:
Name of the main author of the protocol: ______
Date: //
Signature: ______I4V-MC-B011 Non-interventional PASS Protocol Page 79
Annex 3. Additional Information Data on medication use, disease state, clinical information including comorbidities and lifestyle factors, pregnancy records, and socioeconomic status, among others, that are collected in national registers in Denmark, Finland, Norway, and Sweden will be included in the analytic dataset. The list of registers available in each country is provided in Table 14.1.
Table 14.1. Description of Data Sources and Their Respective Variables of Interest Used in the Present Study Sample Variables of Interest Register Brief Description Available for This Study Denmark Research register and a data source DMARD treatment, RA disease The Danish Rheumatology Database for rheumatologic diseases for activity (DAS28), functional status (DANBIO) (Ibfelt et al. 2016) monitoring clinical quality. (HAQ), RA disease duration The Danish National Patient Information on all patients in Discharge diagnoses and their date Registry (Andersen et al. 1999) contact with a Danish hospital. Contains information on the total Date, type, strength and quantity of The Danish National Prescription redemption of prescriptions in drug dispensed Registry (Gaist et al. 1997; Denmark at community pharmacies Kildemoes et al. 2011) since 1994. Registration is to monitor the health Mothers’ age, parity, BMI, and Medical Birth Registry (Knudsen of the newborns and of the quality smoking. Offspring’s time of and Olsen 1998) of the antenatal and delivery care gestation and conception. services. The Danish Civil Registration Information on all Danish citizens, Date of death or emigration System (Pedersen 2011; Schmidt et including date of death, al. 2014) immigration, or emigration. Contains information on all causes Date and cause of death The Danish Registry of Causes of of death in Denmark: date and place Death (Helweg-Larsen 2011) of death, sex, age, and municipality of residence. Information on education, level, and Education level and duration The Danish Education Register duration on all people educated in (Statistics Denmark [WWW]) Denmark or immigrated to Denmark. The Danish register on personal Information on income and tax Household income before tax income and transfer payments payment, earned income, pensions, (Statistics Denmark [WWW]) and benefits. Sweden Research register that collects clinical data on patients with RA, as well as other rheumatic diseases, The Swedish Rheumatology Quality DMARD treatment, RA disease and may be enriched with data on Register (SRQ) (Eriksson et al. activity (DAS28), functional status comorbid conditions, prescription 2014) (HAQ), RA disease duration drug dispensings, and mortality from national data sources. I4V-MC-B011 Non-interventional PASS Protocol Page 80
Sample Variables of Interest Register Brief Description Available for This Study Information on all completed in- and National Inpatient Registry (IPR) Hospital admission and discharge, out-patient admissions at public (Ludvigsson 2011 [WWW]) diagnoses, surgery, including dates hospitals. Contains information on the total The Swedish Prescribed Drug Date, type, strength, and quantity of redemption of prescriptions in Registry (Wettermark et al. 2007) drug dispensed Sweden since 2005. Contains data on practically all Swedish Medical Birth Registry deliveries in Sweden. The register’s For example, infant diagnoses, (Centre for Epidemiology 2003 key data contains information about maternal smoking [WWW]) prenatal care, delivery care, and neonatal care. Information on all Swedish citizens, The Swedish Population Registry including date of death, Date of death or emigration (Ludvigsson et al. 2016) immigration, or emigration. Contains data on deceased persons Swedish Cause of Death Registry and death causes that were (Swedish Cause of Death Registry Date and cause of death registered in Sweden at the time of [WWW]) death. The Swedish Register of Education Information on education, level, and (Swedish Register of Education length on all people educated in Education level and length [WWW]) Sweden or immigrated to Sweden. Information on income and tax The Swedish Income Register payment, earned income, pensions Household income before tax (Statistics Sweden [WWW]) and benefits. Norway Register-based longitudinal observational study of which the main objectives are to study the The Norwegian Antirheumatic Drug DMARD treatment, RA disease effectiveness of treatment of Register (NOR-DMARD) (Kvien et activity (DAS28), functional status inflammatory joint diseases with al. 2005) (HAQ), RA disease duration biological disease modifying anti-rheumatic drugs (DMARDs) in clinical practice. Contains a combination of data recorded at treatment sites, when Norwegian Patient Registry Hospital admission and discharge, patients have received referral or (NPR) (Nilssen et al. 2014) diagnoses, surgery, including dates treatment in a hospital, an outpatient clinic or a contract specialist. Contains a complete listing of all The Norwegian Prescription Date, type, strength, and quantity of prescription drugs dispensed by Database (NorPD) (Furu 2008 drug dispensed pharmacies since 2004. Contains information about all births in Norway from 1967. The Mothers’ age, parity, BMI, and Medical Birth Registry of Norway register is based on mandatory smoking. Offspring’s time of (MBRN) (Irgens 2000) reporting. gestation and conception. I4V-MC-B011 Non-interventional PASS Protocol Page 81
Sample Variables of Interest Register Brief Description Available for This Study Information on all Norwegian The National Registry (DSF) citizens, including date of death, Date of death or emigration (Skatteetaten 2012 [WWW]) immigration, or emigration. Includes all people nationally Cause of Death Registry (Pedersen registered in Norway who died in Date and cause of death and Ellingsen 2015) another country and all people who died in Norway. Information on education, level, and Education Register – Statistics length on all people educated in Education level and length Norway (THL [WWW]) Norway or immigrated to Norway. Information on income and tax The Norwegian Income Register payment, earned income, pensions Household income before tax (THL [WWW]) and benefits. Finland ROB-FIN was designed to monitor the safety, effectiveness, and cost-effectiveness of biological The National Register for Biologic treatments in rheumatic diseases. In DMARD treatment, RA disease Treatment in Finland(ROB-FIN) addition to the data collected by activity (DAS28), functional status (Konttinen et al. 2006) rheumatologists at routine care (HAQ), RA disease duration visits, supplementary information is retrieved from national healthcare registers. The purpose of the register is to collect data on the activities of Care Register for Health Care (THL health centres, hospitals, and other Discharge diagnoses and their date [WWW]) institutions providing inpatient care and on the clients treated in them. Contains information on all The Finnish Prescription Register medications purchased in Date, type, strength, and quantity of (Furu et al. 2010; Rikala et al. 2010) accordance with a doctor’s drug dispensed prescription. Includes data on live births and on stillbirths of foetuses with a birth Mothers’ age, parity, BMI, and Medical Birth Register (THL – weight of at least 500 g or with a smoking. Offspring’s time of Medical Birth Register [WWW]) gestational age of at least 22 weeks, gestation and conception as well as data on the mothers. National register that contains basic The Finnish Population Register information about Finnish citizens Date of death or emigration (Frontpage [WWW]) and foreign citizens residing permanently in Finland. Statistics Finland produces statistics on causes of death and on the Statistics Finland (Cause of Death) development of mortality. The (Statistics Finland – Health – Causes Date and cause of death statistics on causes of death cover of death [WWW]) the persons who have died in Finland. I4V-MC-B011 Non-interventional PASS Protocol Page 82
Sample Variables of Interest Register Brief Description Available for This Study The Finish Register of Completed Information on education, level, and Education and Degrees and student length on all people educated in Education level and length flow statistics (Statistics Finland – Finland or immigrated to Finland. Education [WWW]) Household income (Consider to Finish income data – Statistics adjust by broad income groups to Finland (Statistics Finland – Income Wages, Salaries, and Labour Costs address potential confounding by and Consumption [WWW]) differences in socioeconomic status) Abbreviations: DAS28 = Disease Activity Score modified to include the 28 diarthrodial joint count; DMARD = disease-modifying anti-rheumatic drug; HAQ = Health Assessment Questionnaire; RA = rheumatoid arthritis.
Access to the data of the healthcare and other registries requires regulatory approvals, which is associated with various timelines for the processing of the data access applications. Details about these requirements are provided in Table 14.2. I4V-MC-B011 Non-interventional PASS Protocol Page 83
Table 14.2. Regulatory Requirement for Access to Registries and Databases Estimated Time Average Data Lag Registry Regulatory Requirements for Data Access (if Applicable) Approval Denmark
The Danish Rheumatology Access requires approval from the Danish 6 months Database (DANBIO) Clinical Registries (RKKP) The Danish National Patient Access requires regional approval and 2–3 months Registry approval from Statistics Denmark (SD) The Danish National Access requires regional approval and 2–3 months Prescription Register approval from SD The Medical Birth Registry Access requires regional approval and 2–3 months approval from SD The Danish Civil Access requires regional approval and 2–3 months Registration System approval from SD The Danish Registry of Access requires regional approval and 2–3 months Causes of Death approval from SD The Danish Education Access requires regional approval and 2–3 months Register approval from DST The Danish register on Access requires regional approval and 2–3 months personal income and transfer approval from DST payments Sweden
The Swedish Rheumatology Approval is needed from Regional ethics 2–4 months Quality Register (SRQ) review board (EPN) and from SRQ
National Inpatient Registry Approval is needed from EPN and from 2–4 months National Board of Health and Welfare (SOC) The Swedish Prescription Approval is needed from EPN and from 2–4 months Registry SOC The Swedish Medical Birth Approval is needed from EPN and from 2–4 months Registry SOC The Swedish Population Approval is needed from EPN and from 2–4 months Registry Statistics Sweden (SCB) Swedish Cause of Death Approval is needed from EPN and SOC 2–4 months Registry The Swedish Register of Approval is needed from EPN and from 2–4 months Education SOC The Swedish Income Approval is needed from EPN and from 2–4 months Register SOC Norway
The Norwegian Approval is needed from Regional 2.5–5 months Antirheumatic Drug Committees for Medical and Health Register (NOR-DMARD) Research Ethics (REK) I4V-MC-B011 Non-interventional PASS Protocol Page 84
Estimated Time Average Data Lag Registry Regulatory Requirements for Data Access (if Applicable) Approval The Norwegian Patient Approval is needed from REK 2.5–5 months Registry The Norwegian Prescription Approval is needed from REK and 2.5–5 months Database Norwegian Data Protection Agency The Medical Birth Registry Approval is needed from REK 2.5–5 months of Norway The National Registry Approval is needed from REK 2.5–5 months Cause of Death Registry Approval is needed from REK 2.5–5 months Education Register–Statistics Approval is needed from REK 2.5–5 months Norway The Norwegian Income Approval is needed from REK 2.5–5 months Register Finland
The National Register for ROB-FIN scientific leadership 2–3 months Biologic Treatment in Finland (ROB-FIN) Care Registry for Health Access requires approval from National 2.5–5 months Care Institute for Health and Welfare (THL) The Finnish Prescription Access requires approval from the Social 6 months Register Insurance Institution (KELA) Medical Birth Registry Access requires approval from THL 2.5–5 months
The Finnish Population Access requires approval from THL 2.5–5 months Register Statistics Finland (Causes of Access requires approval from THL 2.5–5 months Death) The Finish Register of Access requires approval from THL 2.5–5 months Completed Education and Degrees and Student Flow Statistics Finish Income Data – Access requires approval from THL 2.5–5 months Statistics Finland