Use of patient disease registries for regulatory purposes (new)
On 8 November 2018 the EMA cross-committee task force published a discussion paper on methodological and operational considerations in the use of patient disease registries for regulatory purposes:
- Discussion paper: Use of patient disease registries for regulatory purposes – methodological and operational considerations
Interested parties can send their comments and suggestions using the Form for submission of comments or an annotated version of the document mentioning the individual’s name, affiliation and contact details on the first page, to EMAregistries@ema.europa.eu by 30 June 2019.
EMA will consider the responses and finalise the document in consultation with the relevant EMA committees by the end of 2019.
The objective of this paper is to discuss methodological and operational aspects of patient disease registries. It focuses on important principles from a regulatory perspective and makes proposals on what might be considered good registry practice to support collection of high quality registry data. This paper also discusses important methodological aspects of studies performed with registry data to increase regulators’ confidence in their results.
A patient registry is defined as an organised system that uses observational methods to collect uniform data on a patient population defined by a particular disease, exposure or condition (e.g. age, pregnancy, specific patient characteristics), and which is followed over time. Patient disease registries may be established by public organisations such as academia or medical research associations of health care professionals or patients. They may have different overall objectives, such as to describe the natural history of a disorder, to monitor the efficacy or safety of treatments, to describe the impact of a disease on patients’ health and quality of life or to identify patients suitable for new treatments. The data collected may also be useful to support the regulatory evaluation of benefits and risks of medicines. Although product registries may have advantages in specific circumstances, e.g. to gather data in a subset of treated patients (such as geriatric patients) underrepresented in clinical trials, regulators generally prefer disease registries as they gather insights on clinical outcomes in patients receiving different treatments and may support a wider range of study designs, e.g. controlled designs without an external data source. The main focus of this document is therefore on patient disease registries even if many considerations also apply to product registries or registries of patients defined by a specific condition.
The data collected in a registry may be used for the purpose of a specific study, i.e. a detailed investigation of a research question or hypothesis. The differences between a registry and a registry-based study (a registry study) are highlighted as they must be well understood to avoid confusion in concepts and methods. This discussion paper therefore provides different sets of considerations for patient disease registries (“good registry practice”) and for registry studies.
Patient Disease Registries
For patient disease registries, the following considerations are emphasised:
- Patient population: great care should be exercised to ensure exhaustive enrolment of patients and avoid selection bias, e. the situation where enrolment is influenced by patient characteristics that may affect the validity of the analyses. Registries are prone to selection bias because numerous factors may influence the enrolment of patients in a registry and may be difficult to identify in advance and prevent. Four steps are proposed to be considered in the definition and enrolment of a study population.
- Time elements: accurate knowledge and recording of dates of important events are essential components of all registries. A list of core time elements that should normally be collected in all registries is proposed. Definitions should be standardised across different registries.
- Core data elements: a list of core data elements to be collected in all patients is proposed. They should be harmonised or mapped across registries for a same disease to support regulatory evaluations and facilitate implementation of a common data quality system, data exchange, common data analysis and interpretation of results from different registries.
- Terminologies: common terminologies should be used across registries for diseases, diagnostic tests, symptoms, medicinal products, active substances, adverse events and other relevant data. Where local or national terminologies are used, these should be mapped to international terminologies. A list of examples of recommended international terminologies for different data elements is provided.
- Quality management: quality planning, quality assurance, quality control and quality improvement are activities of quality management that should be continuously in place and considered as part of the management of registries. Data quality should address their consistency, completeness, accuracy and timeliness. Measures to improve data quality are described. They may be performed at management level and at operational level. Indicators of data quality are useful tools to measure and improve data quality.
- Safety analysis: disease registries conducted by organisations such as academia or medical research associations should follow the national requirements for the management of safety data. Reporting of suspected adverse reactions through the national or regional pharmacovigilance system should be encouraged. Any active data collection system put in place in a disease registry and initiated, managed or funded by a MAH to collect and record suspected adverse reactions to one of its medicinal products must follow the regulatory framework for PASS. Disease registries are generally not suitable for a rapid statistical analysis of new safety signals but they may be useful for the monitoring and characterisation of known or suspected adverse reactions. A list of adverse events of special interest (AESI) can be defined and integrated in the routine data collection system.
- Governance: Most registries have a governance model relying on principles and constraints based on their mandate, operating procedures, legal environment or funding sources. Effective collaboration between all parties is needed to ensure early identification of registries relevant for a regulatory procedure and evaluation of the adequacy and quality of the data collected. A number of activities that may be performed by registries, MAHs and regulators are proposed to strengthen the use of registry data. Amongst them is the agreement on principles of data sharing between registries, MAAs/MAHs and regulators. Principles of data ownership, informed consent and data security should be applied in accordance with the General Data Protection Regulation (GDPR).
For registry studies, the following aspects are particularly relevant:
- Regulatory context: post-authorisation safety (PASS) and efficacy (PAES) studies may be required by regulators to MAHs. For studies imposed as a legal obligation, the MAH holds the responsibility for their supervision, including the obligations to monitor the data generated, to consider its implications for the benefit-risk balance of the medicinal product and to communicate to authorities any new information which might influence this balance. The MAH must also ensure that the fulfilment of these obligations can be audited, inspected and verified, and register the study in a public database. These legal constraints should be taken into account by registry coordinators in their discussion with MAHs.
- Timelines: given the time needed to initiate a registry study, preparations and discussions about use of a patient registry post-authorisation should start before –during scientific advice- or at an early stage of the approval application process. It is recommended that such early discussions involve the MAH(s), regulators and registry coordinators. The study proposal should be sufficiently detailed to be used by registries and individual centres to assess whether they can participate in the study in terms of data availability and data quality requirements.
- Study protocol: an early decision to be made is the choice of the data collection method: secondary data collection, where the data for the study are already available and extracted from a dataset, or primary data collection, where the events of interest for the study are collected directly from patients as they come to the attention of the investigator. This choice has implications for safety reporting and should be clearly specified in the study protocol.
The study protocol should follow the recommendations of the Good pharmacovigilance practice (GVP) Module VIII and the technical guidance on the format and content of the protocol for non-interventional PASS. For studies addressing several products, all concerned MAHs should participate in a joint registry study based on a single protocol. If a registry study is to be conducted across multiple sites, a common protocol needs to be developed based on core common data elements and common design even if some aspects of the study may vary according to the characteristics of each registry.
The protocol should provide an estimation of the study size and the feasibility of attaining this sample size within the registry should be assessed using conservative assumptions (or a previous feasibility analysis), both in terms of number of patients and duration of follow-up. Milestones and timelines for completing the main study phases should be provided.
- Study population: where study data are extracted from data already collected routinely in the registry, methodological challenges to define the study population are comparable to those met in secondary data collection. When the safety or effectiveness of a new treatment need to be monitored in a disease registry, the study population will most often include new users who are either patients newly diagnosed with the disease and who received a first prescription (incident patients) or patients already included in the registry to whom the new treatment is prescribed (prevalent patients). The choice of including incident or prevalent patients in the study population has implications for the data analysis and the interpretation of the results. It is recommended to collect the data needed to distinguish incident and prevalent patients and identify possible differences in their characteristics. In case of prospective recruitment, it is critical that procedures are in place to ensure sequential inclusion of all eligible patients treated in the individual centres. Patients having been involved in clinical trials, often representing a subgroup with certain disease characteristics (e.g. genetic variants), could also be enrolled in a disease registry later on.
- Data collection: it is the investigators’ responsibility to collect for the study only the sets of data that are strictly needed to provide valid results. It is also their responsibility to collect all the data needed for this purpose including available data on potential confounders. Data to be collected include those needed for sensitivity analyses as outlined in the study protocol and statistical analysis plan (SAP). The legal status of a study with additional data collection should also be considered. Additional data collection may turn it into an interventional study and its relation to current clinical practice needs to be detailed in the study protocol.
- Data quality: the need for additional measures for data quality control in a registry study depends on the measures already applied routinely in the registry. In order to ensure acceptable data quality for individual registry studies conducted for regulatory purposes, source data verification and periodic auditing on a reasonable amount of data may need to be conducted on a risk analysis-based approach and following a strategy dependent on the scope of the study. Data source verification for a minimum of 10% of randomly selected patients registered in individual study centres would be considered adequate. Data quality measures also include quality checks at data entry and monitoring of patient follow-up.
- Data analysis: there is no single statistical method applicable to every study and the most appropriate statistical procedure would need to be selected on a case-by-case basis to specifically address the scientific question of interest. Analysis of registry data should take into account that they are selected observational cohorts and hence knowledge of techniques developed to analyse such data is essential. Commonly encountered methodological problems and how they could be approached are highlighted. The handling of missing data should be described in the SAP and justification should be provided for the assumption about their distribution.
- Safety reporting: a distinction should be made between studies with a design based on primary data collection directly from health care professionals (HCPs) or consumers and studies with a design based on secondary use of data; the recommendations from Module VI of the GVP should be followed. In case of primary data collection, the concerned MAH(s) should have a data collection or electronic system to collect, analyse and report information on adverse events notified by a HCP or a patient. Investigators of registry studies should also be informed of the mechanisms allowing them to report at any time to the national spontaneous reporting system any adverse event or suspected adverse reaction occurring during the course of the study.
- Reporting of study results: for imposed PASS and PAES, legal obligations apply to the final study report to be submitted by the MAH to regulatory authorities. MAHs should therefore be able to comment on the study results and their interpretation as well as on the format of the report. Following the submission of the final study report to the regulatory authority, additional information or clarifications may be requested to the MAH by the regulatory authority and the lead investigator should have the duty and scientific responsibility to address the scientific aspects of the request. The lead investigator should always have the right to independently prepare publications of the study results irrespective of the source of funding.
The EMA is willing to support interactions between MAAs/MAHs and registries and provide tools to facilitate recognition of disease registries as data sources to conduct studies for regulatory purposes. Early dialogue with regulators is encouraged. The Scientific Advice procedure on study protocols and the Qualification procedures of registries are available to pharmaceutical companies and registry coordinators to provide advice and opinions on the validity of registries and study protocols.