Utility of registries for post-marketing evaluation of medicines. A survey of Swedish health care quality registries from a regulatory perspective

Aim The aim of this study was to describe content and procedures in some selected Swedish health care quality registries (QRs) of relevance to regulatory decision-making. Methods A workshop was organized with participation of seven Swedish QRs which subsequently answered a questionnaire regarding registry content on drug treatments and outcomes. Patient populations, coverage, data handling and quality control, as well as legal and ethical aspects are presented. Scientific publications from the QRs are used as a complementary measure of quality and scientific relevance. Results The registries under study collect clinical data of high relevance to regulatory and health technology agencies. Five out of seven registries provide information on the drug of interest. When applying external quality criteria, we found a high degree of fulfillment, although information on medication was not sufficient to answer all questions of regulatory interest. A notable strength is the option for linkage to the Prescribed Drug Registry and to information on education and socioeconomic status. Data on drugs used during hospitalization were also collected to some extent. Outcome measures collected resemble those used in relevant clinical trials. All registries collected patient-reported outcome measures. The number of publications from the registries was substantial, with studies of appropriate design, including randomized registry trials. Conclusions Quality registries may provide a valuable source of post-marketing data on drug effectiveness, safety, and cost-effectiveness. Closer collaboration between registries and regulators to improve quality and usefulness of registry data could benefit both regulatory utility and value for health care providers.


Introduction
In Europe, assessment of the safety and efficacy of a new drug before approval and during its entire life-cycle is performed within the European network of regulatory agencies, in Sweden represented by the Medical Products Agency (L€ akemedelsverket). This regulatory collaboration is coordinated by the European Medicines Agency (EMA). Decisions made within this regulatory network have important implications for the availability of safe drugs and vaccines to safeguard public health. In 2015 EMA set up a so-called Cross-Committee Registry Task Force to promote the use of data from disease registries for regulatory purposes. As a contribution to this effort the Medical Products Agency (MPA) in collaboration with national quality of care registries (QRs) performed a survey to describe the utility of registry-based clinical data generation in Sweden. Registries delivering high-quality data on drug exposure and/or relevant outcomes in clinical practice are valuable assets in the assessment of drug safety and effectiveness for all stakeholders. At present the knowledge of how far registries actually can meet this need is limited. This inventory of Swedish quality registries aims at filling this knowledge gap.
In Sweden different types of registries containing health care data have been organized to support clinical decisionmaking, quality improvement, as well as health technology assessment and policy-making. At the national level there are governmental National Health Care Registries (NHCR) held by the National Board of Health and Welfare (e.g. the Patient Registry, Cancer Registry, Cause of Death Registry, Prescribed Drug Registry [PDR], and Birth Registry) covering the entire Swedish population and with mandatory reporting. Regional health care databases cover county and regional populations, and there are also the QRs-the focus of this survey-that provide nationwide data, usually encompassing a specific disease, intervention, or patient group. QRs have been set up at the initiative of health care professionals primarily to support the improvement and sustainability of quality of care. In Sweden there are more than one hundred QRs, but the vast majority do not collect data on drug treatment (1). Data from QRs can be linked-by use of the personal identification number (PIN) given to all permanent residents in Sweden (2)-to other registries. In contrast to product registries which collect information on a single drug product, most QRs cover a disease which allows comparative studies. The utility of linkage to the Prescribed Drug Registry is reflected by the substantial output in the scientific literature as recently reviewed by Wallerstedt et al. (2).
International collaborations between registries, including Swedish QRs, have provided useful data (3,4). However, pooling individual-level data from registries located in different countries often raises problems of legal as well as logistic nature that may necessitate specific considerations (5,6). If such problems are overcome, these studies may enable comparison between country-specific settings or increase the size of patient populations for studies of orphan diseases or other situations with rare outcomes (7).
To describe the potential of QRs to meet a growing regulatory need for data from clinical practice we performed a survey including a selected subset of registries. All of these represent therapeutic areas of importance to public health and where new drugs have recently been introduced with requirements for post-marketing follow-up.

Definition of a national quality of care registry
The majority of Swedish QRs are organized and run by the medical profession. The aim is to improve the care of patients with a specific disease or the quality and outcome of a certain medical intervention. The registry is integrated in daily practices and sometimes also supports clinical decisionmaking. It is financed by public funding and governed by national laws and regulations (8).

Selection of registries
To be included, the registry should fulfill the national requirements for certification level 1 or 2 according to the Swedish Association of Local Authorities and Regions (SALAR). Requirements for certification level 1 include direct information to patients on registry results, active use of data for research and obtaining research funding in national or international competition, systematic validation of data quality, and control of coverage by cross-checking versus other data sources. The following five registries are certified at level 1: NDR (National Diabetes Registry), SRQ (Swedish Rheumatology Quality Registry), SWEDEHEART (The Swedish Web-system for Enhancement and Development of Evidencebased care in Heart disease Evaluated According to Recommended Therapies), Riksstroke, and NPCR (National Prostate Cancer Registry of Sweden); and the following two at level 2: SMSreg (Swedish Multiple Sclerosis Registry) and MACULAREG (Macula Registry). All registries are known to systematically collect data on drug treatment, and report at least five peer-reviewed publications in the field of drug efficacy or safety, facts also deciding selection for the study. These registries represented cardiology, neurology, multiple sclerosis (MS), stroke, rheumatoid arthritis, diabetes, prostate and breast cancer, and ophthalmology. All participated in a workshop organized by the MPA where the purpose of the project was presented and acceptance to participation confirmed. One invited registry, the breast cancer registry, chose not to participate.

Collection of information
The questionnaire used in this survey has been used previously (5,9). The questionnaire (MPA Quality Registry Questionnaire, All Rights Reserved; available online) was used to extract basic administrative data and information on data collection, handling, quality assurance, reporting, ethical and legal aspects, funding, and governance. Although the questionnaire was not formally validated, the reliability of data was controlled as registry holders verified on two occasions that the information from their respective registry was correctly transferred from the questionnaire to the tabular presentation in this report.

Assessment of regulatory usefulness
The information in the questionnaires was used to describe the usefulness of the QRs with focus on the following aspects: Completeness (number of participating units), Coverage (proportion of eligible patients included), Validity (clinically relevant and quality-assured data), Comparability (i.e. definitions and outcomes identical to those used in randomized controlled trials [RCTs], possibility to create control groups), and Organizational and financial robustness. The usefulness of QR data was assessed also from the regulatory relevance of their scientific publications dealing with drug-related issues. To illustrate this further, the publications were subdivided and presented in categories of safety, effectiveness, health economics, and issues on methodology, the last-mentioned category also including aspects on multinational collaboration.
To apply an external perspective to our description we used two sets of criteria elaborated by the National Institute for Health and Care Excellence (NICE) (10,11).
The first set is suggested to improve the quality of evidence generation for new treatments (i.e. when setting up a registry) and points out five areas of particular importance (10). These are: 1) Establishing a management structure; 2) Agreeing a mandatory data subset; 3) Preventing and monitoring incoherent entries; 4) Motivating those submitting data; and 5) Triangulation and data linkage of registry data to external data sources.
The second NICE set proposes the use of six main criteria when assessing the quality of a registry, which can be summarized as follows (11): 1) Data completeness in terms of patient population (as denominator); 2) Relevance of the data for answering the question; 3) Data granularity; 4) Independence of the registry; 5) Publications with data made from the registry; and 6) Aspects of data protection.

Organizational aspects
Some QRs have formed an umbrella organization containing several subregistries (Table 1). This has permitted new Certification level is a rating given to each registry and represents the level of development the registry has reached in terms of analyses, inclusion of relevant indicators, coordination with health services, use in research, data quality and reporting, coverage rate, technical solutions/tools, etc. There are four levels in total: 1 (highest), 2, 3, and the candidate level (lowest) (http://kvalitetsregister.se/englishpages/findaregistry/certificationlevels.2029.html). subregistries to develop from an ancestral QR (e.g. neuroregistry from MS only to Parkinson's disease, myasthenia, narcolepsy, etc.) which has provided IT platforms and practical experience facilitating the inclusion of additional diagnoses. In other cases QRs has evolved in the other direction, i.e. there has been a merger of initially independent registries into one larger body (e.g. SWEDEHEART). Some of the QRs have been active for 15 years or more, which has allowed the development of high coverage and robust systems for collaboration. All but one registry have websites in English facilitating contact with external parties like regulators, drug companies, academic groups, and other international stakeholders.

Patient selection and coverage
The number of included patients varied from 21,439 in the ophthalmological registry to more than 1.5 million in the cardiology registry ( Table 2). Coverage of the target population was high, for all estimated to be above 80%. It should be noted that registries classified as 'interventional' have coverage of 100%. The majority of the QRs can recruit patients and controls for clinical studies and also have the option to randomize to treatment within the registry (randomized registry controlled trials, RRCTs) ( Table 3).

Data recorded
Data recorded in the QRs are-for natural reasons-to a large extent disease-specific (Table 4). For registries focusing on interventional procedures the principal diagnosis may vary, as the inclusion is decided by the intervention and not the disease. Information on patients (sex, age, etc.) and on the disease in question (duration, scores for disease activity and organ damage), physical function, patient-reported outcome measures (PROMs), etc. is provided by all QRs or can be retrieved by linkage. Information on education and socioeconomic status can be obtained by linkage to other national registries, held by Statistics Sweden. The information on medication is of varying quality. Five out of seven registries provide information on the drug of interest, i.e. a targeted follow-up is included in the data collection. All prescribed medications can be found in the Prescribed Drug Registry and linked to other data by the PIN. Data on drugs used for inpatient care (i.e. non-prescribed) are collected by the QRs included in this survey. The outcome measures collected in the QRs are to a high extent the same as those used in the relevant clinical trials. Long-term safety can be adequately followed by means of data collected within each QR but importantly through the PIN and the possibility for linkage to other data sources. Some of the QRs are connected to the MPA for direct electronic reporting of suspected adverse drug reactions (ADRs).

Quality control procedures
The majority of the QRs have well-defined quality control procedures in place (Table 5). If specific research studies are performed using registry data, ethics approval and patient consent are obtained according to standard requirements and applicable legislation. As the collection of data for improvement of health care quality is seen as a part of routine care, specific permissions are not necessary. The basic regulation of this is laid down in the Swedish Personal Data Act (12) and the specific Patient Data Act (13), resulting in uniform processing of patient data by all registries.
Ethics committee approval is sought for all scientific projects, including all linkage studies (Table 6).

Governance
All registries are owned by public/governmental bodies ( Table  7). The funding for running the QRs is public, and yearly applications are needed (8). Decisions regarding funding are made by a committee nominated by the Government.
The financial and organizational robustness of these QRs seem reassuring as the funding comes from public sources and the governance is firmly integrated in the clinical professional organizations and the County Councils. Details on the proportions of public versus other funding were not asked for in this survey.

Reporting
Information on the results and specific studies is presented in scientific publications and in yearly reports to the Funding    The MS registry has the most elaborate real-time feed-back to reporting physicians (Table 8). For the within-registry communication, real-time feed-back of aggregated data at national, regional, and hospital level is becoming increasingly important. As the annual reports are key components in  Table 7. Legal, organizational, and financial aspects.
applications for continued public funding, they are comprehensive and give a good overview of the status of the registry. Registry data are also discussed at meetings with the respective national professional society. However, the scientific publications are the most important way to inform of results from the registries at the international level.
Applying external quality criteria to the registry content and procedures When applying the NICE criteria to the registries, we found that all QRs had taken such aspects into consideration when setting up their registry as well as when performing quality control over time. However, the item 'granularity', i.e. detailed information on medication, was not sufficient to answer all questions of regulatory interest.

Registry of scientific publications of regulatory relevance
A selection of publications from the registries is presented in a Supplementary Table (Publications of Regulatory Significance; available online) to illustrate their potential regulatory significance. They cover a broad range of scientific issues including drug safety, effectiveness, and utilization relating to multiple sclerosis, cardiovascular diseases, rheumatoid arthritis, diabetes, and prostate cancer. Health economic aspects including costs, sick leave, and work performance have been studied, as have quality of life and socioeconomic aspects in relation to drug treatment. Important information on changes in the target population characteristics over time can be captured, and long-term trends in prescribing patterns can be followed and reliably linked to data from other sources. The importance of accurate background incidence estimates has also been analyzed (14).
Some QRs have published reports of clinical trials using randomization of patients within the QR, so-called randomized registry controlled trials (RRCTs), occasionally described as 'a new disruptive scientific methodology' (15,16). A growing awareness of methodological and data quality aspects in registry research has generated several publications taking national as well as multinational aspects into account.

Discussion
The main finding of this survey is that Swedish National Quality of Care Registries may provide a feasible structure for managed introduction and long-term surveillance of new drugs and other medical interventions, including medical devices. The set-up, governance, and data management as well as in-registry clinical and scientific competence are of high quality as reflected by numerous publications in peerreviewed journals. They can also meet a need for real-time clinical decision support. The registries are willing to collaborate with regulatory and health technology assessment (HTA) bodies by providing relevant data from clinical practice. However, so far only a limited number of QRs-e.g. those participating in this survey-have the capacity to deliver high-quality data at short notice, which might be important when answering important safety issues. Thus, conclusions from this study cannot automatically be extrapolated to all Swedish QRs. Further support is therefore needed to continue the improvement of registry quality and to expand the concept to additional therapeutic areas, which also will be in the interest of public health. To fully explore the potential of QR data, linkage to other sources of information is often needed. This sometimes adds further ethical and legal requirements, complicating study performances. By revising some of these regulations to comply with current medical and regulatory needs, registry data could be used more effectively.
There are similar registry set-ups in other European countries, and bi-or multinational collaborations are established or underway in several therapeutic areas. Results of such collaborations can be found in publications on MS (17), myocardial infarction (3), cancer risk in biologics-treated patients (18), and diabetes (19). Collaborations have also been extended to include methodological and study design issues (20). Taking these efforts into account, regulatory, HTA agencies, and other public institutions should consider supporting or even initiating multinational registry collaborations to When is feed-back given to reporting physician/clinics? In real-time (Y/N) 7 0 All Annually (Y/N) 7 0 All Ad hoc, in case of need, e.g. safety problems (Y/N) 6 1 SMSreg, Riksstroke, MACULAREG, NDR, SRQ, SWEDEHEART NPCR answer specific questions, e.g. in orphan diseases or other situations with small study populations. A particular strength of the Swedish QRs is the possibility to link data on individual patient characteristics with treatments and outcomes, including PROMs for several drugs and not just a single product, as is the case with product registries. Several registries directly involve patients in the development of PROMs, internet-based patient reporting, educational efforts, etc. (21). These common patients/registries initiatives could support the ongoing efforts by regulatory agencies, IMI projects, and other activities to involve patients in regulatory procedures further. The most obvious weakness from a regulatory perspective is the insufficient granularity of information on medications, in particular regarding dosing, formulations, and duration of treatment. However, data retrieved by linkage to the Prescribed Drug Registry can often compensate for this lack.
Methods for quality control need to be further harmonized between registries. One way to facilitate this could be by offering inspections of registries in line with GCP standards, another to facilitate studies to validate registry content. This could ultimately result in a certification as a 'Good registry practice (GRP) registry'. A dialogue between regulators and registries may also facilitate the implementation of new EU regulations, for example the concept of 'low interventional studies' of obvious relevance to collection of clinical practice data in registries (22).

Conclusions
Swedish health care quality registries contain useful information on drugs in clinical practice. This can be used to improve assessments made by regulatory agencies but also to support health policy and public health decision-making regarding drug-related issues. We propose that regulators should interact directly with representatives from the registries to elaborate their role in a regulatory context and discuss common efforts to improve quality and usefulness of registry data. Such a dialogue could stimulate a fruitful development where registries could contribute substantially to the evaluation of drug safety and effectiveness. Reliable postmarketing data collection is imperative for a life-cycle benefit-risk assessment of drugs and also to support managed introduction of new drugs in routine clinical care.