THE INFLUENCE OF SOCIOECONOMIC STATUS ON SELECTION OF ANTICOAGULATION FOR ATRIAL FIBRILLATION: A POPULATION BASED STUDY

Transcription

1 THE INFLUENCE OF SOCIOECONOMIC STATUS ON SELECTION OF ANTICOAGULATION FOR ATRIAL FIBRILLATION: A POPULATION BASED STUDY by Michelle Sholzberg A thesis submitted in conformity with the requirements for the degree of Master of Science (Clinical Epidemiology and Health Care Research) Graduate Department of Health Policy, Management and Evaluation University of Toronto Copyright by Michelle Sholzberg 2016

2 Abstract THE INFLUENCE OF SOCIOECONOMIC STATUS ON SELECTION OF ANTICOAGULATION FOR ATRIAL FIBRILLATION: A POPULATION BASED STUDY Michelle Sholzberg Master of Science 2016 Graduate Department of Health Policy, Management and Evaluation University of Toronto Introduction: Without third-party insurance, access to marketed drugs is limited to those who can afford to pay. We examined this phenomenon in the context of anticoagulation for nonvalvular atrial fibrillation (NVAF). Methods: To determine whether, wealthier patients with NVAF were more likely to switch from warfarin to dabigatran prior to its addition to the provincial formulary, we conducted a population-based retrospective cohort study. Results: We found that higher SES was associated with switching to dabigatran prior to its coverage on the provincial formulary (p<0.0001). Subjects in the highest quintile were 50% more likely to switch to dabigatran than those in the lowest income quintile. Conclusions and Relevance: We have documented socioeconomic inequality in access to dabigatran among patients receiving warfarin for NVAF. This highlights the importance of timely reimbursement decisions for equitable drug access. Copyright by Michelle Sholzberg 2016

3 Acknowledgments Sincerest thanks to my thesis committee members Drs. Muhammad Mamdani, David Juurlink, Tara Gomes and Andreas Laupacis. Your collective guidance has been invaluable and career altering. A special thank you to my supervisor Andreas for his capable, patient and inspiring mentorship throughout the thesis process and beyond. I also wanted to specifically thank Tara Gomes for patiently teaching me how to access and analyze administrative data and big thanks to Zhan Yao for her incredible work in distilling the data for this project. Thank you to Dr. Thomas Parker, Dr. Victor Tron, Dr. Christine Brezden-Masley, Dr. Paula James, Dr. Rita Selby, Dr. Jerry Teitel, Ms. Hina Chaudhry, Mr. Syed Mahamad, Mr. Aziz Jiwajee, Ms. Natalya O Neill, Ms. Jessica Petrucci and to the Hematology-Oncology Clinical Research Group of St. Michael s Hospital for the ongoing support of my research. Finally, thank you to my husband, Eddie, to my son, River and to my entire family, for your endless love and encouragement. iii

4 Table of Contents Title Page... i Abstract... ii Acknowledgments... iii Table of Contents... iv List of Tables... vi List of Figures... viii List of Appendices... ix Chapter 1: Introduction and Background Description of drug approval and funding processes in Canada and Ontario Description of the Ontario Ministry of Health and Long-Term Care (MOHLTC) and the Ontario Drug Benefit Program (ODBP) Description of the Pan-Canadian Pharmaceutical Alliance (PCPA) Limited Use Criteria for Dabigatran in Ontario Description of socioeconomic strata in Ontario and Canada Description of the association between patient socio-economics/-demographics and health services/medication use Description of evolution of treatment of non-valvular atrial fibrillation Chapter 2: Rationale, Specific Objectives and Hypotheses Study rationale Specific research questions Primary research question Secondary research question Study objectives Primary study objective Secondary study objective Research hypothesis Hypothesis for primary research question Hypothesis for secondary research question iv

5 Chapter 3: Methods Study design and overview Data sources Study participants Inclusion criteria Exclusion criteria Outcome variables Study timeline diagrams Examination of causation and confounding bias Examination of selection and information bias Selection bias Information bias Sample size calculation Statistical analysis Study cohort and baseline characteristics Primary objective: Cochran-Armitage trend test and logistic regression Secondary objective: Cochran-Armitage trend test and logistic regression Chapter 4: Results Study cohort and baseline characteristics Primary analysis Secondary analysis Chapter 5: Discussion Main study findings and interpretation Study strengths Study limitations Conclusions and future directions References Appendices v

6 List of Tables Table 1. Dates that Dabigatran Became Available on the Provincial Formularies Table 2. Dabigatran 110 mg Dose ODBP Limited Use Criteria Table 3. Dabigatran 150 mg Dose ODBP Limited Use Criteria Table 4. Diagnosis Codes Used to Define Hemorrhage Outcomes Table 5. Estimation of Proportions for Sample Size Calculation Table 6. Sample Size Calculation with Varying Incremental Proportion of Switch Across Income Quintiles Table 7. Covariates Adjusted for within Logistic Regression Model Table 8. Baseline Characteristics of Elderly Ontarians with Non-Valvular Atrial Fibrillation Table 9. Primary Analysis: Assessment for Trend of Switchers across Income Quintiles Prior to ODBP listing of Dabigatran Table 10. Primary Analysis: Odds of Switching to Dabigatran Prior to ODBP listing of Dabigatran Unadjusted Logistic Regression Model Table 11. Primary Analysis: Odds of Switching to Dabigatran Prior to ODBP listing of Dabigatran Adjusted Model Table 12. Secondary Analysis: Assessment for Trend of Switchers across Income Quintiles After ODBP listing of Dabigatran vi

7 Table 13. Secondary Analysis: Odds of Switching to Dabigatran after ODBP listing of Dabigatran Unadjusted Model Table 14. Secondary Analysis: Odds of Switching to Dabigatran after ODBP listing of Dabigatran Adjusted Model vii

8 List of Figures Figure 1. Timeline Showing Approval of Dabigatran Health Canada to Provincial Listing Figure 2. Process of Funding Medications in Canada And Ontario Figure 3. Study Outcomes in Aggregate Figure 4. Causation Diagram Figure 5. Primary Analysis: Cochran-Armitage Test for Trend Figure 6. Cohort and Sub-Cohort Patient Numbers Figure 7. Forest Plot of Secondary Analysis: Odds of Switching to Dabigatran after ODBP listing of Dabigatran Adjusted Model Figure 8. Forest Plot of Secondary Analysis: Odds of Switching to Dabigatran after ODBP listing of Dabigatran Adjusted Model viii

9 List of Appendices Appendix 1. Diagnosis codes used to define hemorrhage outcomes Appendix 2. Covariates adjusted for within logistic regression model Appendix 3. ICD-10 codes for valvular disease types Appendix 4. ICD-9 codes for valvular disease types Appendix 5. ICD-10 codes for vascular disease types Appendix 6. ICD-10 codes for renal disease types Appendix 7. ICD-10 codes for hepatic disease types Appendix 8. Relevant OHIP diagnostic fee codes Appendix 9. List of abbreviations ix

10 Chapter 1: Introduction and Background 1.1 Introduction: Description of drug approval and funding processes in Canada and Ontario Approximately 30 billion Canadian dollars were spent on publicly and privately funded prescription drugs in 2013, accounting for 14% of total health expenditures. 1 Outpatient prescription funding in Canada is a patchwork of public and private insurance, and private payment by patients. On average, 36% of funding comes from public sources, 36% from private insurers and 22% out of patient pocket in Canada. 1,2 Federal drug plans in Canada, for example plans providing coverage for the First Nations, account for only 2% of prescription drug costs in the country. 3 Therefore, all remaining public drug coverage for prescriptions is organized at the provincial level. As a result, there is considerable variation from province to province in public access to prescription medications, and the proportion of publicly funded drug spending relative to total drug spending. 1,2,4,5 In order for a drug to be listed on the Ontario formulary, it must go through various regulatory stages, starting with Health Canada approval and ending with provincial drug plan approval for public coverage and eligibility (see Figure 1). The process is lengthy, not particularly transparent and involves committees advising other committees who then advise those making governmental decisions. Drugs are approved for sale by Health Canada on the basis of safety and efficacy. Thereafter, the Patent Medicine Price Review Board (PMPRB) sets a maximum drug price that can be charged by manufacturers for patented medicines. Individual drug or therapeutic (drug class) reviews occur through the Canadian Drug Expert Committee 1

11 2 (CDEC) of the Common Drug Review (CDR) which is run through the Canadian Agency for Drugs and Technologies in Health (CADTH). 6 The CDEC makes a funding recommendation to participating drug plans based on clinical and economic evidence, and includes patient/public input when developing their recommendation. CDEC s recommendation options for individual drugs are: 1) list, 2) list with clinical criteria and/or conditions (limited use criteria), 3) do not list at submitted price or 4) do not list. A drug is listed for a specific indication when it has demonstrated comparable or added clinical benefit and acceptable cost-effectiveness relative to its comparators. 6 A list with limited use recommendation is used, for example, when there is comparable or added clinical benefit and acceptable cost-effectiveness in a subgroup of patients, but concern that the drug is not cost-effective in other patients. 6 Examples of limited use clinical criteria include the following: 1) specific requirements regarding comorbidity status (e.g. the requirement for adequate renal function when a drug is cleared by the kidney), 2) inadequate clinical response, intolerance or inability to use a comparator drug(s) (e.g. the requirement for a trial of anticoagulation with warfarin with inadequate response in the case of the direct oral anticoagulant reimbursement), 3) specific requirements regarding expertise of prescriber (e.g. the limitation of filgastrim reimbursement to prescriptions written by an oncologist or hematologist), and 4) specific rules regarding initiation and cessation of drug based on response to treatment (e.g. the requirement for evidence of improvement of visual acuity on monthly monitoring in the setting of diabetic macular edema for the continued reimbursement of ranibizumab). 6 Participating province s or territory s Drug Advisory Committees make final recommendations regarding drug coverage and eligibility. Decisions at the

12 3 provincial/territorial level are made on the basis of clinical and economic evidence taking CDR recommendations into consideration. The Ontario provincial committee is called the Committee to Evaluate Drugs (CED). 6 Canadian provinces/territories make their own decisions regarding drug funding and consider local budgetary impact when making these decisions. However, participating provincial/territorial drug plans agree with the CDEC recommendation in the majority of cases. 6,7 In Ontario, a new drug product is considered for funding if the manufacturer makes a complete submission to the Ontario Ministry of Health and Long-Term Care (MOHLTC). Once the MOHLTC receives a submission and CDEC s recommendation, the CED considers the Ontario context for funding and conducts its own assessment of clinical evidence, patient perspective and the potential impact of the drug on health services compared to existing treatments in Ontario. 6 The CED then makes a recommendation regarding whether the drug should be listed in the ODBP Formulary (as a general benefit (no restrictions to their use), limited use(only funded if the patients meet specified criteria, such as having a certain severity of disease) or considered for funding through the Exceptional Access Program (EAP) on a case-by-case basis. 6 Prior to the listing decision, the Ministry of Health may negotiate price reductions, price-volume agreements and other arrangements, which are not made public. The final decision is made by the Executive Officer of the Drug Programs Branch who is accountable to the Deputy Minister of Health. 6

13 4 1.2 Description of the Ontario Ministry of Health and Long-Term Care (MOHLTC) and the Ontario Drug Benefit Program (ODBP) The Ontario MOHLTC, which funds public drugs and other health services (e.g hospitals, physicians) in Ontario has a budget of more than $45 billion, representing close to 40% of provincial spending. 6 The ODBP funds (as a general benefit or LU drug) close to 4000 drugs listed in the formulary and an additional 1000 through the EAP. 6 The ODBP provides drug benefits for Ontarians who are 65 years of age or older, residents of long-term care homes and homes for special care, recipients of professional home care services, recipients of social assistance or recipients of the Trillium Drug Program (TDP). 6 The TDP is intended for Ontarians who have high prescription drug costs relative to their household income. Many drugs paid for by the ODBP formulary are reimbursed only if the physician specifies the LU code, implying that the patient meets specific LU criteria Description of the Pan-Canadian Pharmaceutical Alliance (PCPA) In 2010, the Pan-Canadian Pharmaceutical Alliance (PCPA) for New Drugs was established. This alliance allows for participating provinces/territories to determine whether they should negotiate price with a drug manufacturer collectively, individually or not at all. 8 The goals of the PCPA are to: increase access to drug options; to enhance consistency of drug listing decisions across the country; to capitalize on combined buying power, to achieve consistent pricing; to lower drug costs; to reduce duplication of negotiations; to minimize delays in drug access and to optimize resource utilization. 8 There is substantial potential for negotiating power through the PCPA and it claims to

14 5 have already resulted in significant governmental savings since its inception. 8 The North Western Territories, Nunavut and Quebec are the only jurisdictions not included in the PCPA. Many brand name drugs have gone through the PCPA pricing process including the direct oral anticoagulant (DOACs). The PCPA committee estimates what degree of price reduction is required to meet a threshold of cost-effectiveness. However, the process of negotiation taken for the first drug in a family (e.g. dabigatran) is particularly challenging, because the starting negotiating price is unclear. 9 An important concern for the funders during the PCPA dabigatran negotiation was the potential cost impact if there was a major shift away from warfarin use, because the cost differential between the DOACs and warfarin was so large (verbal communication Anne Holbrook). This meant that the negotiations took a considerable length of time. At the end of PCPA negotiations, a Letter of Intent (LOI) is drafted and signed to document the ultimate funding decision. 9 The entire process from Health Canada approval to the date that the PCPA LOI was signed, was 17 months for dabigatran. It took an additional 1 month for dabigatran to be listed on the ODBP public formulary (see Figure 2). This was at least 1 month faster than other Canadian jurisdictions (for those where public formulary listing dates were available) (See Table 1).

15 6 Figure 1. Process of funding medications in Canada and Ontario Health Canada (federal) Reviews drugs for safety and efficacy Approves drugs for sale in Canada Patented Medicine Prices Review Board Limits the prices set by manufacturers for patented medicines Common Drug Review and pan-canadian Oncology Drug Review for cancer products (national, all F/P/T plans except Quebec) Reviews clinical data, cost effectiveness data & patient input Makes funding recommendations Pan-Canadian Pharmaceutical Alliance Participating provinces/territories negotiate the price of new brand name drugs Provinces, Territories, Federal Plans Make final decisions on drug coverage & eligibility Pay for the drugs Ontario General benefit product Limited use (LU) product (with LU criteria) Exceptional access product Decision not to fund Abbreviations: F/P/T Federal/Provincial/Territorial

16 7 TABLE 1. Dates that dabigatran became available on the provincial formularies Province Date Prince Edward Island February 13, 2013 Manitoba September 24, 2012 New Brunswick June 8, 2012 Nova Scotia May 2, 2012 Saskatchewan May 2012* Health Benefits for First Nations and Inuit May 27, 2013 British Columbia May 23, 2012 Alberta May 1, 2012 Ontario April 12, 2012 Newfoundland May 2013* *exact date unknown Modified from a communication with the ODPRN

17 8 Figure 2. Timeline from Health Canada approval of dabigatran to provincial formulary listing LOI Signed Mar 22 Provincial Formularies HC Approval October 26 CDR Completion June 22 File Engagement July* ON April 12 AB May 1 NS May 2 BC May 23 NIHB May 27 Provincial Formularies PEI February 13 NFLD May *exact dates unknown PCPA Time July 2011-March 2012 Abbreviations: HC Health Canada; CDR Common Drug Review; LOI Letter of Intent; ON Ontario; AB Alberta; NS Nova Scotia; BC British Columbia; NIHB Non-Insured Health Benefits for First Nations and Inuit; SK Saskatchewan; MB Manitoba; PEI Prince Edward Island; NFLD Newfoundland; PCPA Pan-Canadian Pharmaceutical Alliance Adapted from Pan Canadian Drugs Negotiations Report, March 22, 2014

18 9 1.4 Limited Use Criteria for Dabigatran in Ontario Dabigatran, or dabigatran etexilate, its prodrug, is available in two different dosages in Canada. The standard dose is 150 mg by mouth twice daily and 110 mg by mouth twice daily is recommended for patients at increased risk of bleeding, including those over 75 years with more than one risk factor for bleeding. However, efficacy in stroke prevention may be diminished at the lower dose On April 24 th, 2012, dabigatran, at both dosages, was listed on the ODBP public formulary with specific limited use criteria (see tables 2 and 3).. 13 The limited use criteria for dabigatran define specific requirements regarding comorbidity status including renal function, age specifications, non-valvular nature of atrial fibrillation, atrial fibrillation with a high thromboembolic risk stratification according to the CHADS2 score (Congestive heart failure, Hypertension, Age over 75 years, and Diabetes mellitus, and 2 points for prior Stroke/transient ischemic attack) as well as previously documented inadequate clinical response, intolerance or inability to use warfarin (Table 2).

19 10 TABLE 2. Dabigatran 110 mg Dose ODBP Limited Use Criteria For the prevention of stroke and systemic embolism in at risk patients with non-valvular atrial fibrillation (AF), AND in whom: 1) Anticoagulation is inadequate following a reasonable trial on warfarin; OR 2) Anticoagulation with warfarin is contraindicated or not possible due to inability to regularly monitor via International Normalized Ratio (INR) testing (i.e., no access to INR testing services at a laboratory, clinic, pharmacy, and at home). Exclusion Criteria: Impaired renal function (creatinine clearance or estimated glomerular filtration rate less than 30mL/min); OR greater than or equal to 75 years of age without documented stable renal function; OR hemodynamically significant rheumatic valvular heart disease, especially mitral stenosis; OR prosthetic heart valves. Note: (a) Documented stable renal function is defined as creatinine clearance or estimated glomerular filtration rate maintained for at least 3 months (i.e., 30-49mL/min for 110mg twice daily dosing and, greater than or equal to 50mL/min for 150mg twice daily dosing for at least 3 months). (b) At risk patients with atrial fibrillation are defined as those with a CHADS2 score of greater than or equal to 1. (c) Inadequate anticoagulation is defined as INR testing results that are outside of the desired INR range for at least 35% of the tests during the monitoring period (i.e., adequate anticoagulation is defined as INR test results that are within the desired INR range for at least 65% of the tests during the monitoring period). (d) A reasonable trial on warfarin is defined as at least 2 months of therapy. (e) Since renal impairment can increase bleeding risk, renal function should be regularly monitored. Other factors that increase bleeding risk should also be assessed and monitored (see product monograph). (f) Patients starting dabigatran should have ready access to appropriate medical services to manage a major bleeding event. (g) There are currently no data to support that dabigatran provides adequate anticoagulation in patients with rheumatic valvular disease or those with prosthetic heart valves; dabigatran is not recommended in these populations. 13

20 11 TABLE 3. Dabigatran 150 mg Dose ODBP Limited Use Criteria For the prevention of stroke and systemic embolism in at risk patients with non-valvular atrial fibrillation (AF), AND in whom: 1) Anticoagulation is inadequate following a reasonable trial on warfarin; OR 2) Anticoagulation with warfarin is contraindicated or not possible due to inability to regularly monitor via International Normalized Ratio (INR) testing (i.e., no access to INR testing services at a laboratory, clinic, pharmacy, and at home). Exclusion Criteria: Impaired renal function (creatinine clearance or estimated glomerular filtration rate less than 30mL/min); OR greater than or equal to 75 years of age without documented stable renal function; OR hemodynamically significant rheumatic valvular heart disease, especially mitral stenosis; OR prosthetic heart valves. Note: (a) Documented stable renal function is defined as creatinine clearance or estimated glomerular filtration rate maintained for at least 3 months (i.e., 30-49mL/min for 110mg twice daily dosing and, greater than or equal to 50mL/min for 150mg twice daily dosing for at least 3 months). (b) At risk patients with atrial fibrillation are defined as those with a CHADS2 score of greater than or equal to 1. (c) Inadequate anticoagulation is defined as INR testing results that are outside of the desired INR range for at least 35% of the tests during the monitoring period (i.e., adequate anticoagulation is defined as INR test results that are within the desired INR range for at least 65% of the tests during the monitoring period). (d) A reasonable trial on warfarin is defined as at least 2 months of therapy. (e) Since renal impairment can increase bleeding risk, renal function should be regularly monitored. Other factors that increase bleeding risk should also be assessed and monitored (see product monograph). (f) Patients starting dabigatran should have ready access to appropriate medical services to manage a major bleeding event. (g) There are currently no data to support that dabigatran provides adequate anticoagulation in patients with rheumatic valvular disease or those with prosthetic heart valves; dabigatran is not recommended in these populations. 14

21 Description of socioeconomic strata in Ontario and Canada Socioeconomic status (SES) is a conceptual amalgamation of a person s economic and sociologic position relative to others, and can be measured in a number of ways including income, level of education, type of occupation and accumulated wealth. Statistics Canada estimates SES by combining the International Socio-Economic Index of Occupational Status (ISEI) with information on the highest level of education among a student s parents, family assets, educational resources at home and family cultural assets. 15 Working poor is a term used to describe individuals who maintain regular employment but remain in relative poverty; whose household income is at or below the Low Income Measure (LIM). 16,17 The majority of Canadians living in poverty are the working poor and are not welfare recipients. The working poor in Ontario have different characteristics compared to those receiving social assistance. They are more likely to be immigrants, women, from racial minorities, have completed secondary school and live in households Administrative data only have a limited number of variables that can be used to estimate SES. Most frequently, SES is categorized into five quintiles according to postal code-associated median household income at the level of the dissemination area (from Statistics Canada census information using Statistics Canada Postal Code Conversion File Version 5F). The dissemination area is defined as the smallest geographical unit that is used for census data reporting. 21 The small area-base is important, as it has been shown to more accurately reflect individual SES than larger areas. 21 Each quintile is assigned to approximately 20% of the population. This method of estimating SES has been validated

22 13 as a useful approach to overcome the absence of socioeconomic data in most Canadian medical records and administrative databases Validation of this area-based method of SES estimation has included matching large volume socioeconomic data from census tracts and block groups to individual records of those belonging to a large prepaid health plan in the United States and their residential address. 24 Furthermore, this method has been used in many studies studying the relationship between health outcomes and SES in Ontario. 22,23, Description of the association between patient socio-economic status, demographics and health services/medication use The health of the poor, and of the working poor in particular, is quite complicated. Both groups have higher rates of a multitude of physical and mental health problems compared to the wealthier. The working poor are particularly vulnerable because despite working, they cannot afford to pay out-of-pocket for expensive medicines and do not have private insurance nor social assistance to cover additional health related costs. 8,9,25 These inequities regarding health service and medication availability amongst the well off and the poor are worrisome, and tactics that focus on the promotion of health equity are of paramount importance. A universal Canadian pharmacare program is an example of one such tactic that might minimize these inequities. 4,5,26-28 For all of these reasons we became interested in investigating access to dabigatran for NVAF before and after its inclusion on Ontario s provincial formulary.

23 Evolution of anti-thrombotic treatment of non-valvular atrial fibrillation For many decades, most patients with atrial fibrillation have been anticoagulated with vitamin K antagonists (VKAs), such as warfarin, to diminish their risk of arterial thromboemboli, particularly stroke. The introduction of the direct oral anticoagulants (DOACs) as alternatives to VKAs has been met with cautious enthusiasm among clinicians. 29 These agents are currently available for prevention and treatment of venous thromboembolism and for prevention of arterial thromboembolism. DOACs differ significantly from traditional oral VKAs mechanistically and pharmacokinetically. Dabigatran directly inhibits the final effector of coagulation, thrombin (factor IIa), while rivaroxaban and apixaban directly inhibit the rate-limiting enzyme of coagulation, factor Xa. Thrombin and factor Xa are targeted by the DOACs, given their pivotal roles in clot formation. The major advantages of the DOACs include their rapid onset of action, shorter half-lives, lack of requirement for regular laboratory monitoring and the absence of food interactions when compared to VKAs. 30,31 DOACs are now recommended for the prevention and treatment of thromboembolism. 5,6 The RE-LY multicenter, non-inferiority trial, comparing the use of dabigatran versus warfarin for AF, found essentially similar rates of stroke or systemic embolism but lower rates of life-threatening and major bleeding. 7 The annual drug cost of dabigatran is $1,168 CAD per patient, compared with approximately $50 CAD for Warfarin. Thus, purely from the perspective of drug costs, without consideration of pharmacy dispensing fees, dabigatran is approximately 20 fold more expensive than warfarin. However, despite dabigatran s high drug specific cost, it has been shown to be a cost-effective thromboprophylactic strategy for NVAF in multiple

24 15 studies because the cost of anticoagulation with warfarin includes laboratory monitoring and patient travel, resulting in a total annual cost of $750 CAD for warfarin management. 6,32-37 The high percentage of time in therapeutic range on warfarin (the accepted metric of effective and safe anticoagulation) in Canada and risk of bleeding on dabigatran in the elderly with renal dysfunction were considered in the national and provincial cost-effectiveness analyses of dabigatran. 12,38,39 When compared with warfarin, the reported cost per quality-adjusted life-year (QALY) was $9,041 (dabigatran 150 mg twice daily) and $29,994 (dabigatran 110 mg twice daily). The relatively favourable cost per QALY results despite the marked difference in drug costs were influenced predominantly by the decrease in risk of intracranial hemorrhage with dabigatran and decrease in expense related to drug monitoring. Therefore, dabigatran was approved by Health Canada on October 26, 2010 for thromboprophylaxis in non-valvular AF. On April 24, 2012 the ODB program began to reimburse dabigatran for patients with non-valvular AF who met specific LU criteria (see tables 2 and 3). Socioeconomic and prescriber characteristics have previously been shown to affect anticoagulant choice for NVAF in the United States. A retrospective claims analysis conducted by AbuDagga et al. in 2014 identified patient health plan type and a prescribing physician who was a cardiologist as amongst the strongest factors associated with dabigatran use. 40 Also, the American Outcomes Registry for Better Informed Treatment of Atrial Fibrillation (ORBIT-AF) showed that private insurance and patient education was associated with switching from warfarin to dabigatran 41. Multivariable logistic regression analysis of dabigatran versus warfarin use in this cohort of 20,320

25 16 people with NVAF calculated an odds ratio of 3.12 (95% CI ) for cardiology specialty and an odds ratio of 1.68 (95% CI ) for preferred provider organization (PPO) insurer (with health maintenance organization 42 as the reference). 40 Both PPOs and HMO are networks of medical providers who have agreed with a thirdparty insurer to provide health care at reduced rates, but the PPO provides more flexibility when picking a physician or hospital and is generally more expensive than an HMO. 3 Therefore, there is previous evidence of inequity in anticoagulation selection for NVAF from one study conducted in the United States, which has a health care system very different from Canada s.

26 Chapter 2: Rationale, Specific Objectives and Hypotheses 2.1 Study rationale Canadian provincial drug plans generally decide which drugs they will reimburse based upon an assessment of value for money. This means that drugs that have been approved by Health Canada may be not reimbursed at all, are only funded if patients meet specific clinical criteria, 26 or are fully funded for all indications. 43 Patients can pay for drugs that are approved by Health Canada but are not covered by the public plans through private insurance or out-of-pocket. Supporters of this approach to drug reimbursement point out that all patients have equal access to drugs that are felt to be good value for money and that public resources are not spent on drugs that are not cost-effective. 2,43-45 However, it can take time for public plans to incorporate new evidence about a drug s benefits in their decisionmaking. Thus, in these instances poorer patients are less likely to be able to access effective new drugs because they do not have private insurance plans and cannot afford to pay for them out-of-pocket. 46,47 Data from publicly funded drug programs may provide insight into this equity issue. In Canada, dabigatran etexilate was approved by Health Canada on October 26, 2010 for thromboprophylaxis in patients with NVAF. It was added to the Ontario Drug Benefit Program (ODBP) formulary about 1.5 years later, on April 24, This provided an opportunity to empirically assess whether wealthier patients in Ontario were more likely to receive dabigatran than poorer patients during the time between regulatory approval and provincial funding. 17

27 18 Studies have previously suggested income-related differences in treatment selection and we hypothesize that the same is true for the selection of anticoagulant therapy in the setting of NVAF. 4,8 If the results of this study show that people living in wealthier neighborhoods with AF were more likely to be receiving dabigatran prior to ODB approval, this would support the argument for more attention to rapid decision making about cost-effective drugs as part of a more equitable pharmacare strategy in Canada. 4,5, Specific research questions Primary research question Are elderly Ontarians, treated with dabigatran for non-valvular atrial fibrillation (NVAF) prior to its reimbursement by the Ontario Drug Benefit Program (ODBP), from October 24, 2011 to April 24, 2012, more likely to live in wealthier neighborhoods compared to those treated with warfarin during the same time period? Secondary research question Are elderly Ontarians treated with dabigatran for NVAF after its reimbursement by the ODBP, from April 24, 2012 to October 24, 2012, more likely to live in wealthier neighborhoods compared to those treated with warfarin during the same time period?

28 Study objectives Primary study objective In Canada, dabigatran etexilate was approved by Health Canada on October 26 th, 2010 for thromboprophylaxis in patients with nonvalvular atrial fibrillation (NVAF). It was added to the provincial formulary almost 18 months later, on April 24 th, We sought to determine whether older Ontarians who switched from warfarin to dabigatran during this period were more likely to live in wealthier neighborhoods, as compared with those who remained on warfarin Secondary study objective To test whether any identified socioeconomic gradient persisted once dabigatran became available through the public drug program. 2.4 Research hypothesis Hypothesis for primary research question Our hypothesis was that wealthier patients were more likely to be on dabigatran prior to its reimbursement by the ODBP Hypothesis for secondary research question Our hypothesis was that after dabigatran was reimbursed by the public formulary, the economic gradient would disappear (i.e. wealthier patients were not more likely to be on dabigatran after its reimbursement by the ODBP)

29 Chapter 3: Methods 3.1 Study design and overview We conducted a population-based retrospective cohort study of Ontarians aged 66 and older with NVAF who were treated with warfarin between October 28, 2008 and October 26, This study was approved by the Research Ethics Board of Sunnybrook Health Sciences Centre, Toronto, Ontario. 3.2 Data sources We used the Registered Persons Database (RPDB), which contains basic demographic data and information on vital status, to characterize SES, based on neighbourhood income quintile. 22,23,48 The RPDB was also used to identify the geographic location of patient residence (urban vs. rural) We used the Ontario Drug Benefit Program (ODBP) database to identify prescriptions for medications. We used the Canadian Institute for Health Information Discharge Abstract Database (CIHI DAD), the CIHI National Ambulatory Care Reporting System (NACRS) and the Ontario Health Insurance Plan (OHIP) database to identify patients with NVAF and other comorbidities (including major hemorrhage). We approximated SES based on each patient s place of residence on October 26, 2010 (the date of Health Canada approval for dabigatran), and patients were divided into quintiles according to median neighborhood income, as done previously

30 Study participants Inclusion criteria We identified Ontarians with NVAF aged 66 years and older treated with warfarin prior to Health Canada s approval of dabigatran, which occurred on October 26, 2010 (Figure 1). The diagnosis of atrial fibrillation was made if there was one or more diagnostic code for atrial fibrillation (see below) within the five years preceding October 26, This was done in effort to maximize the likelihood that atrial fibrillation was the reason for the warfarin prescription. To focus on patients with evidence of chronic warfarin use at the time of Health Canada approval of dabigatran, we limited our analysis to patients with two or more warfarin prescriptions in the 6 months before Health Canada s approval of dabigatran. Patients were identified on the basis of inpatient hospitalization or emergency department visit for atrial fibrillation using the International Classification of Diseases and Related Health Problems, 10th revision codes (ICD-10) I48.0 and I48.1, or a physician visit with the diagnosis of atrial fibrillation (OHIP diagnosis code 427 in context of previous warfarin use). Validation studies suggest ICD codes for the diagnosis of atrial fibrillation have a positive predictive value between 70 and 96% Exclusion criteria We did not include patients with only one warfarin prescription, those with inpatient diagnoses of mitral stenosis, prosthetic heart valves, or mitral or aortic valve surgery, as well as residents of chronic care facilities in the three years prior to formulary listing of dabigatran. 53 We also excluded patients who died prior to the end of the study

31 22 period (i.e. ineligible for early or late switch) and those with a hemorrhage leading to a hospital visit in the three years prior to cohort entry or during follow-up, to avoid the possibility that anticoagulant choice was influenced by a previous hemorrhage. The ICD- 10 codes used for the definition of major hemorrhage are detailed in the Table 4, and have a positive predictive value of 87% and a negative predictive value of 92% for the identification of major bleeding events. 54 Some patients who were excluded from the RE- LY study were included in our study because some data forming the basis of their eligibility criteria were not available in the provincial databases (e.g. creatinine clearance, active liver disease and severe stroke) Study Timelines as they Pertain to the Definition of Study Participants Study participants with NVAF were accrued from April 26, 2010 to October 26, 2010 (i.e. from the beginning of Look-back Window 1 to the date of Health Canada approval of dabigatran) to document continuous anticoagulation with warfarin ( 2 prescriptions). Look-back Window 1 was the six months prior to the Health Canada approval date. Look-back Window 2 was the six month time period prior to dabigatran LU formulary listing date (i.e. October 24, 2011 to April 24, 2012). In this time period warfarin prescriptions were reviewed to define the outcome (see below). The follow-up period for the study ended six months after the dabigatran formulary listing date (i.e. October 24, 2012) (see Figure 3). For all of the above prescription counts, only one prescription per person per day was counted.

32 23 TABLE 4. Diagnosis codes used to define hemorrhage outcomes Hemorrhage Type ICD-10 Codes Intracerebral I60, I61, I62.0, I62.1, I62.9 K92.0, K92.1, I85.0, I98.20, I98.3, K22.10, K22.12, K22.14, K22.16, K25.0, K25.2, K25.4, K25.6, K26.0, Upper Gastrointestinal K26.2, K26.4, K26.6, K27.0, K27.2, K27.4, K27.6, K28.0, K28.2, K28.4, K28.6, K29.0, K63.80, K31.80 Lower Gastrointestinal K55.20, K62.5, K92.2 N02.0, N02.1, N02.2, N02.3, N02.4, N02.5, N02.6, N02.7, N02.8, N02.9, K66.1, N93.8, Other Hemorrhage N93.9, N95.0, R04.1, R04.2, R04.8, R04.9, R31.0, R31.1, R31.8, R58, D68.3, H35.6, H43.1, H45.0, M25.0

33 Outcome variables We stratified patients into two mutually exclusive groups. We defined switchers as patients whose warfarin use ceased in the interval between dabigatran s approval by Health Canada and its inclusion on the provincial formulary, but who had evidence of dabigatran use ( 1 prescription) within six months thereafter (April 24, 2012 to October 24, 2012). These patients were assumed to have switched to dabigatran between October 26, 2010 and April 24, We defined non-switchers as patients who continued warfarin therapy during the same interval, and had ongoing use of warfarin ( 1 prescription) even after dabigatran s inclusion on the public formulary. We excluded all individuals (6952 out of a total of 41,749) who did not fit into one of these mutually exclusive groups (31,614 non-switchers and 3183 switchers ). We did not have data about drugs paid for privately, either out-of-pocket or by private insurance companies. Because switchers were patients receiving warfarin for NVAF who later transitioned to dabigatran, we inferred that discontinuation of warfarin in the interval between dabigatran s approval and formulary listing, reflected initiation of dabigatran during this same period. In a secondary analysis, we stratified non-switchers into those who initially remained on warfarin, but transitioned to dabigatran in the six months following its addition to the ODBP formulary ( late switchers ) and those who continued to use warfarin exclusively ( continuous warfarin users ), to explore whether addition of dabigatran to the provincial formulary attenuated any SES gradient identified in our primary analysis. (Figure 3). The initial starting dose of dabigatran (110 mg versus 150 mg) was examined for switchers and late switchers.

34 Study timeline diagrams Figure 3. Study outcomes in aggregate Abbreviations: NVAF Non-Valvular Atrial Fibrillation

35 Examination of causation and confounding bias The exposure, income quintile (surrogate for SES), was identified prior to the outcome (using dabigatran), in keeping with Bradford-Hill s temporality criterion suggesting causation Previous studies have suggested strong, consistent, progressive income-related differences in drug treatment selection and we hypothesize that the same is true for the selection of anticoagulant therapy for NVAF in Ontario, Canada. 40,55-57 Also, the association between the exposure and outcome seems plausible given that the cost of dabigatran in Ontario is approximately 20-fold higher than warfarin and that this association has previously been described in the United States. 40,41 Socioeconomics however, is not specifically (i.e. solely) associated with drug treatment selection and does have multiple other effects. As per the application of concepts from the counterfactual causation model, the true association measure will likely differ from the measured effect size in this population-based study due to numerous confounders and one effect modifier. 55 The confounders include age, gender, relevant comorbid conditions (renal disease, hepatic disease, myocardial infarction, cerebrovascular disease, peripheral vascular disease, diabetes mellitus, Charlson Comorbidity Index), number of drugs prescribed in the last year prior to cohort entry, drug with antiplatelet effect prescription, prescription of a drug thought to potentially interact with dabiagatran and patient urban or rural residence. These variables are confounders because they: 1) are associated with selection of anticoagulation apart from its association with income quintile, 2) are plausibly associated with income quintile, and 3) are not intermediates in the causal path from income quintile to anticoagulant selection. 55 Specialist visit between Health Canada

36 27 approval and dabigatran listing is, however, an effect modifier given that it could vary the strength of the association (magnitude of the effect) between income quintile (the exposure) and anticoagulant selection (the outcome). 55

37 28 Figure 4. Causation diagram CONFOUNDERS: (e.g. age, relevant comorbid conditions, gender, number of drugs prescribed in last year prior to cohort entry, patient urban or rural residence, prescription of a drug thought to interact with dabigatran, drugs with antiplatelet effect) EXPOSURE: Income Quintiles EFFECT MODIFIER: Specialist visit between HC approval and formulary listing OUTCOME: A/C Selection Abbreviations: HC Health Canada; A/C anticoagulant

38 Examination of selection and information bias Selection bias Exclusion of those who died and those with hemorrhagic complications prior to October 24, 2012 could have resulted in selection bias if patients in the lower income quintiles were more likely to die or bleed. To assess for the potential for selection bias due to this exclusion, we reviewed how many people were excluded per income quintile. We found evidence that individuals in lower income quintiles were more likely to be excluded (6.86% were excluded from income quintile 1 versus 4.86% in income quintile 5). This was not a surprising finding since individuals of lower SES are more likely to have comorbid medical illnesses and social conditions separately associated with an increased risk of death. However, this exclusion enriched the lower income quintiles with people who were more likely to convert to dabigatran and therefore attenuated, if anything, the association between high SES and dabigatran use. Also, some patients who were excluded from the RE-LY study were included in our study because some data forming the basis of their eligibility criteria were not available in the provincial databases (e.g. creatinine clearance, active liver disease and severe stroke). Therefore, selection bias could have influenced associations found between anticoagulant selection and the absence of renal disease since those who may have had an absolute contraindication to dabigatran (i.e. creatinine clearance < 30 ml/min) were included in the study. However, this would not have impacted the assessment of the association between SES and the frequency of switching to dabigatran.

39 Information bias Differential outcome misclassification is present given the lack of confirmation that dabigatran is being received during look-back window 2 (6 month time period prior to ODBP funding of dabigatran) in the switchers group. Further misclassification is possible given that the income quintile variable in the Registered Persons Database (RPDB) is based on neighborhood income, not patient income. OHIP diagnostic fee codes do not distinguish between atrial fibrillation and atrial flutter, and may result in some misclassification. However, ICD-10 codes do distinguish between these two atrial arrhythmias. Atrial fibrillation is far more common than atrial flutter, the two often coexist and are managed in a similar fashion. Finally, we excluded individuals who did not have atrial fibrillation in the five years prior to the first warfarin prescription in effort to maximize the chances that atrial fibrillation was the reason for anticoagulation. It is possible however, that another diagnosis, such as venous thromboembolism, was the cause for warfarin use - resulting in misclassification. 3.8 Sample size calculation To design a study with sufficient power to assess for a trend in proportions, we had to anticipate the proportions of patients who switched to dabigatran early ( switchers ) in each of the ordered SES quintiles. From Gomes work describing rates of hemorrhage during warfarin therapy for NVAF, we anticipated that income quintiles will be evenly split among patients with NVAF treated with Warfarin. 23 We hypothesized that

40 31 the proportion (P2 in table 1) of those switching from warfarin to dabigatran would show a linear trend across income quintiles. Based on prescription data on warfarin and dabigatran (from the time between Health Canada approval and ODBP formulary listing of dabigatran) in the IMS Brogan Canadian CompuScript database, we hypothesized that approximately 3% (range 2% to 4%) of continuous warfarin users would discontinue warfarin and pay privately for dabigatran in the time period between Health Canada approval and formulary listing. The database contains prescription records from a sample of pharmacies that are included in a panel of over 5600 pharmacies in Canada. The panel is stratified by province, size and type of pharmacy 58. This database is used frequently for research studies. 59,60 We further hypothesized that the proportion of people switching would vary linearly by income at 1% (range 0.25% to 1.5%) increments per income quintile. Alpha, type 1 error, represents the probability of concluding that there is a trend when there is no such trend, i.e. to incorrectly reject the null hypothesis. Beta, type 2 error, represents the probability of concluding that there is no trend in proportions when there is a trend, i.e. failure to reject the false null hypothesis. By convention, we used an alpha at 0.05 and beta of 0.10 (for power of 90%) in this one-sided trend test. Using the hypothesized proportions in Table 5, we calculated that 253 (range 113 to 5280) patients per income quintile would be required to assess for the presence of a linear trend using the Cochran-Armitage trend test. We determined that we would be sufficiently powered at this level given the large number of patients with atrial fibrillation in Ontario s administrative databases (Table 6). In our study, we had a total of 34,797