|
UPDATE: INTERNATIONAL REPORTUsing Pharmacoepidemiology To Inform Drug Coverage Policy: Initial Lessons From A Two-Province Collaborative
Prescription drug plan managers seek "real-world" evidence regarding the safety and effectiveness of drugs and drug coverage policies. University-based pharmacoepidemiologists with access to administrative health data can help meet these information needs and, by collaborating with other health plans, exploit variations in plan policy. Canada, with its universal health insurance, public drug benefit plans, and population-based linked health care databases, is an ideal setting for such research. Here we describe our initial experience in collaborating with researchers and drug plan managers in British Columbia and Ontario.
DRUG PLAN MANAGERS in Canada and the United States face the impossible task of controlling costs while providing access to a growing list of costly new drugs. Pharmacoepidemiological research using routinely gathered administrative health data, such as those generated by the U.S. Medicare prescription drug benefit, offers a tool to help plan managers understand the impacts of drugs and drug coverage policies.1 In this paper we briefly review drug licensing and drug coverage decision making in Canada, and we describe a collaborative effort by university researchers and provincial drug plan managers to conduct and use pharmacoepidemiological research.
Evidence for drug licensing and drug coverage in Canada. In Canada, as in most countries, drug licensing decisions are made by a federal program after review of research evidence presented by the drug’s manufacturer. Although there is room for interpretation, the process for identifying products as safe and efficacious is relatively clear and unambiguous. Likewise, market authorization, if granted, is for specific drug dosages and clinical indications. In many countries, including Canada, this evidence-based framework is extended to the review of drugs by public drug benefit plans.2 Drug payment decisions move beyond issues of individual drugs’ safety and efficacy to include issues of their relative effectiveness, cost-effectiveness, and budget impact. In Canada, drug coverage decisions are made by individual provincial drug plans. To reduce duplication of effort, in 2003 the provinces agreed to a centralized Common Drug Review (CDR) for all new submissions to the provincial formularies.3 As in the past, this process draws on existing prelicensure evidence of how the drug compares with relevant alternatives in terms of safety and efficacy, therapeutic advantages and disadvantages, and cost-effectiveness. The outcome is one of three recommendations: (1) cover (with or without restrictions); (2) do not cover; or (3) further evidence required. Each province then weighs the recommendation along with other factors, such as competing priorities and budget impact, to arrive at a funding decision.4 Other evidence of interest to drug plan managers. Although evidence from prelicensure clinical trials is important for formulary decision making, a logical extension of the role of evidence in drug licensing and listing decisions is a move from reliance on premarket evidence to the use of evidence drawn from research conducted in real-world, observational settings.5 Questions of particular interest to drug plan managers include the following: What is the impact of reimbursement of one drug upon the use of other drugs? Are patients meeting prespecified criteria for coverage? Are they achieving the benefits predicted from premarket trials? What are the impacts of drug use on health outcomes and costs? Prescription drug utilization studies began in Canada in the late 1970s, with development of population-based drug insurance claims databases. In-house provincial drug claims analyses traditionally focused on tracking costs and usage of individual drugs or drug classes. More recently, university-based researchers have used more sophisticated methods and linked drug claims to other health service and outcome data to study a broad range of important health policy questions. Although some provinces promote collaboration between researchers and drug plan managers, formal cross-provincial links remain underdeveloped. In British Columbia, the provincial government has a history of working with university-based researchers to evaluate the impact of drug coverage policies.6 It recently created the Therapeutics Initiative (TI) to facilitate collaboration between researchers and the provincial drug plan, BC PharmaCare.7 By comparison, in Ontario, links between managers of the Ontario Drug Benefit (ODB) program and scientists at Toronto’s Institute for Clinical Evaluative Sciences (ICES) are less formal. In British Columbia, members of our research team (Ken Bassett and Jim Wright) are university based, lead the TI, and advise the province’s Drug Benefits Committee—the committee that decides which drugs the province will cover.8 Other team members (Leanne Warren and Greg Carney) spend at least part of their time as program managers or analysts within the BC Ministry of Health. This contrasts with the situation in Ontario, where team members are affiliated with ICES and have no formal role in provincial drug funding decisions. (Gary Naglie and Andreas Laupacis were members of the province’s expert Drug Quality and Therapeutics Committee prior to our study.)9 Although some of us have written about establishing a national network of pharmacoepidemiologists that could serve as a resource for governments, such as the new Developing Evidence to Inform Decisions about Effectiveness (DEcIDE) Network of the U.S. Agency for Healthcare Research and Quality (AHRQ), there has been little formal analysis or discussion of how such links could be structured or how they might influence formulary decision making.10 We therefore undertook a two-year demonstration project to establish and evaluate links between our university-based research teams and drug plan managers in British Columbia and Ontario. The project involved defining information needs of provincial drug plan managers, working with plan managers in the two provinces to develop pharmacoepidemiological evidence, and then assessing the value of that evidence.
Baseline interviews with drug plan managers and advisers. During late 2002, as part of a larger project looking at drug plan managers’ information needs, we conducted thirty semistructured interviews with drug plan managers (19) and advisers (11) in nine of thirteen Canadian provinces, including British Columbia and Ontario. Four themes surfaced from these discussions. (1) All drug plans used clinical and economic evidence (based largely on phase III randomized controlled trials) before listing. (2) Virtually all did at least some postmarket monitoring, but this was limited mainly to budget impact and only rarely considered safety issues or health outcomes. (3) Very few studies evaluated health outcomes associated with specific formulary policies. (4) "Lack of resources" and "lack of expertise" were common reasons for not doing more. Respondents consistently identified five drug categories as priorities for evidence on use, costs, and health impacts: coxibs [celecoxib (Celebrex), rofecoxib (Vioxx), valdecoxib (Bextra), and meloxicam (Mobicox)]; atypical neuroleptics [olanzapine (Zyprexa), risperidone (Risperdal), and quetiapine (Seroquel)]; tumor necrosis factor (TNF)–alpha inhibitors [infliximab (Remicade) and etanercept (Enbrel)]; statins; and proton pump inhibitors (PPIs). These drugs were cited because their use was perceived as being "out of control"; because they were being "used for everything"; and because of concern about the growing use of biologics, whose more complex nature will require increase follow-up because of "cost and safety concerns." These findings are largely consistent with those of Roy West and colleagues, who surveyed the staffs of five provincial drug plans (British Columbia, Alberta, Ontario, Quebec, and New Brunswick) in 2000.11 They reported that plans sometimes engaged in systematic follow-up on the economic impact of drug listing decisions but were less likely to evaluate the quality or appropriateness of drug use and rarely looked at real-world cost-effectiveness. Methods for producing pharmacoepidemiological evidence. To provide decisionmakers with a better understanding of the types of postmarketing studies that were possible, we developed the following hierarchy: Level I studies examine temporal trends in drug use and costs, and they permit comparison of patterns of use across plans with different coverage policies. Level II studies use data on prior medical history and concurrent drug use for people exposed to specific drugs, to provide information about appropriateness of use and to test the validity of assumptions made in cost-effectiveness and budget impact analyses. Level III studies examine outcomes associated with drug exposure to explore safety and effectiveness questions. Computer programs for each type of study were developed at ICES for use with Ontario’s data platforms and variables and were then adapted for use in British Columbia. By the end of our two-year project, programs for levels I and II studies were fully functional in both provinces; programs for level III studies were implemented shortly thereafter. To date, we have completed eight cross-provincial analyses: five level I (one for each drug group of interest to the drug plans); two level II; and one level III. Below is a summary of our level I study on nonsteroidal anti-inflammatory drugs (NSAIDs). The example demonstrates the importance of long-term relationships between researchers and decision-making processes, which often extend over many months and even years. It also shows that the role of primary research can vary depending on the initial coverage decision. That is, unlike decisions to fund drugs, decisions to limit funding often lead to debate and research in defense of the decision. Examples of the other types of studies have been published elsewhere.12 Case study: level I study of use and costs of NSAIDs in British Columbia and Ontario. Background. Health Canada’s Therapeutics Products Directorate (TPD) licensed the coxibs celecoxib, rofecoxib, and meloxicam in April 1999, October 1999, and October 2000, respectively. In Ontario, celecoxib and rofecoxib were added to the public drug benefit formulary in April 2000 as Limited Use (LU) products, which meant that they were reimbursed if the physician used a special numeric code on the prescription to indicate that the patient had a previously documented peptic ulcer or gastrointestinal bleed or had failed with three conventional NSAIDs. Meloxicam was granted General Benefit status in March 2001, meaning that it was reimbursed without restrictions. In British Columbia, public coverage of coxibs was made available in April 2001, a year later than in Ontario, and then only through a more tedious and time-consuming prior authorization process called a Special Authority (SA) request. These SA requests were reviewed case by case. To be eligible for coverage, patients must have tried and failed four lower-cost NSAIDs from two price categories. Data sources and methods. British Columbia and Ontario have electronic drug claims databases that accurately record information about prescriptions dispensed in retail pharmacies, including the dispensation date, name and quantity of the drug dispensed, and a unique patient identifier.13 The BC PharmaNet Database includes all prescriptions dispensed in the province, regardless of payer. The ODB Database includes only prescriptions paid for by the provincial drug plans for people age sixty-five and older and those on social assistance. Our study was of all residents age sixty-five and older, and it covered the period 1997–2002. Use was measured as the number of prescriptions dispensed and total costs (including dispensing fee). Use rates were calculated using census data from Statistics Canada.
Findings and follow-up.
From 1997 to 2002 there was a small but steady decrease in the number of NSAID prescriptions reimbursed by the BC government (Exhibit 1
We presented these results to drug plan managers in both provinces. Plan managers in British Columbia found the level I studies reassuring. They showed that use of the new, more expensive NSAIDs is limited. However, because plan managers faced intense criticism for their more restrictive coverage policy, they sought level II (appropriateness) and level III (adverse event) studies, including a cross-provincial, population-based analysis of hospitalization for gastrointestinal hemorrhage before and after the introduction of coxibs.14 Ontario drug plan managers, by comparison, had less interest in the NSAID studies, because further restricting coxib access was unlikely. Closing interviews with drug plan managers in BC and Ontario. Two senior staff members from each drug plan were interviewed in spring 2004. They were presented with a list of the reports and presentations that team members had prepared for the drug plans during the preceding three years (nineteen in Ontario and ten in BC) and were asked to rate the extent to which this evidence influenced plan decision making or policy. They were then asked to identify the three most and least influential studies and why. Our analysis of their responses led us to three main conclusions. (1) Although the simple, cross-provincial level I studies could be useful when defending or considering changes to policy, levels II and III study designs provided more valuable evidence. (2) Evidence was of greatest perceived value when it informed a specific listing decision or policy; was available at the right time in the decision-making process; and was presented in a way that made application easy. (3) The perceived value of individual research projects varied considerably according to the roles and responsibilities of the assessor.
We demonstrated that it is possible for teams of researchers in two provinces to collaborate with their respective drug plan managers, agree upon issues of mutual importance, and undertake research using common methods. Our initial level I studies have led to a level III study of NSAID use and gastrointestinal hemorrhage and to level II studies of the characteristics of new statin users and of antipsychotic drug use in long-term care. The experience has taught us some lessons about doing cross-provincial research, what evidence matters most to drug plan managers, and the value and limitations of administrative data. Feasibility. Although it is feasible to conduct collaborative research with drug plan managers, getting their input can be challenging. During our project, Ontario researchers and plan managers met only three times. The close proximity of the British Columbia researchers to plan staff was clearly preferable in terms of keeping abreast of burning issues and getting questions answered. But it also contributed to some loss of independence. In both settings, regular, dedicated, brief team meetings and explicit processes for planning, priority setting, and triggering studies could improve the usefulness of the research. We did not transfer data between provinces. Rather, we developed computer programs and analytic plans jointly, performed the analyses separately using our own provincial databases, and then combined results using summary statistics and analytical intermediates. This avoided logistical and privacy issues related to data transfer, and it also meant that researchers were not required to learn the intimate details of each other’s databases.
Importance of research for drug plan decision making.
Timeliness of information is critical for effective policy making. Our research was most valuable to provincial drug plan managers when it was designed to inform or evaluate specific coverage decisions and when results were available at the right time in the decision-making process. At the outset, we were fortunate to have some dedicated resources and mechanisms to guarantee access to key databases in each province. However, routine updates to the databases occur at various intervals (some more frequently than others), access to some data is available only on a project-by-project basis by application, and research protocols involving primary data collection or data transfer require review by external research ethics boards. For example, obtaining an extract from the BC PharmaCare Database for our level I NSAID analysis (Exhibit 1
Limitations of administrative health data.
Although there are advantages to using administrative data, there are also limitations. As shown by our NSAID analysis, restrictive public drug coverage policies have the potential to shift costs to private insurers and consumers (Exhibit 1 Drug claims alone can provide data for only level I studies that look at drug utilization and costs. The levels II and III studies that are more valued by decisionmakers require linkage to data on clinical history and outcomes. In some cases, that information can be obtained from other administrative data sources. However, even linked administrative data cannot answer many important clinical questions related to drug use. For example, detailed information on the clinical characteristics of patients receiving TNF-alpha inhibitors or the quality-of-life benefits associated with NSAIDs or neuroleptics are not available from administrative data. These questions call for judicious use of primary data collected through costly and time-consuming reviews of patients or medical records. We need new models for funding and conducting such research. One approach to primary data collection being tested in Australia and in the province of Alberta with high-cost biologic therapies makes providing data and evidence of adequate response to treatment conditions for ongoing reimbursement by the public drug plans.15 These programs generate data for ongoing safety and effectiveness assessment, while ensuring that patients who do not benefit are withdrawn from therapy. Some benefits of "inequity." Finally, although it is hoped that Canada’s CDR will eventually reduce interprovincial variation in drug coverage, each province will fund medications as it sees fit. Variation in coverage might persist because of legitimate differences in priorities or values. Ironically, it was just such variation that initially motivated our collaboration and made it possible to conduct research that would not have been done independently. Provinces will invest in administrative data research programs that serve their own information needs; however, federal funding and foresight are needed to build and sustain the kinds of cross-program relationships that are necessary to make the most of these resources.
J. Michael Paterson (paterson{at}ices.on.ca) is a senior research coordinator at the Institute for Clinical Evaluative Sciences (ICES) in Toronto, Ontario. Andreas Laupacis is president and chief executive officer of ICES. Ken Bassett is acting director of the Therapeutics Initiative in Vancouver, British Columbia. Geoff Anderson is an adjunct scientist at ICES. The members of the collaborative are named in an acknowledgment. The BC-Ontario Pharmacosurveillance for Decision-Making Collaborative, in addition to the authors, consists of James M. Wright, Leanne Warren, Greg Carney, Alan Cassels, Muhammad Mamdani, and Gary Naglie. This work was supported in part by a financial contribution from Health Canada’s Health Policy Research Program. Geoff Anderson holds the Liberty Health Chair in Health Management Strategies at the University of Toronto. Naglie holds the Mary Trimmer Chair in Geriatric Medicine Research at the University of Toronto. Ken Bassett and Andreas Laupacis are members of the Canadian Expert Drug Advisory Committee and receive honoraria for this work. At the time this research was done, Mamdani was a senior scientist at ICES; he is now employed by Pfizer Inc. ICES is a nonprofit research corporation sponsored by the Ontario Ministry of Health and Long-Term Care (OMOHLTC). The Centre for Health Services and Policy Research (CHSPR) and the Therapeutics Initiative (TI) are funded by the British Columbia Ministry of Health (BCMOH). The opinions, results, and conclusions expressed in this paper are those of the authors; no endorsement by the OMOHLTC, BCMOH, ICES, TI, or CHSPR is intended or should be inferred.
| ||||||||||||||||||||||||||||||||||||
Top
Background
Study Methods And Results
