Patient registries as a source of evidence
A patient registry is a standing, purpose-defined collection of uniform information about individual patients whose care follows clinical practice rather than a protocol. What a registry is, how one is designed and its data kept trustworthy, and what its data can and cannot support decide how far evidence from a registry can be relied on.
Definitions
Many definitions of a patient registry are in circulation. The World Health Organization's definition is relatively short and contains the three most important features: a patient registry is a file of documents containing uniform information about individual persons, collected in a systematic and comprehensive way, in order to serve a predetermined purpose. Uniform information is the first feature, systematic and comprehensive collection the second, and the predetermined purpose the third.
The US Agency for Healthcare Research and Quality defines a patient registry as an organised system that uses observational study methods to collect uniform data, clinical and other, to evaluate specified outcomes for a population defined by a particular disease, condition or exposure, and that serves one or more predetermined scientific, clinical or policy purposes. That definition captures the most important aspects of patient registries, such as the applicability of observational study methods, the need for standardised data collection, the evaluation of predefined outcomes and the identification of registry populations. It also underlines that the purpose of a registry should be determined beforehand, implying that data collection is purpose-driven and not the other way round, with the goal of the analysis driven by the data collected.
Patient registries are databases designed to systematically collect, store and analyse clinical data, and they can track patient demographics, diagnosis, treatments and outcomes, enabling longitudinal studies on a large scale.
Registries, registry-based studies and observational studies
The European Medicines Agency defines a patient registry as an organised system that collects uniform data over time on patients who are diagnosed with a particular disease or condition, or who receive particular medicines, and a registry-based study as a clinical trial or non-interventional study that investigates a research question using the data collection infrastructure or the patient population of one or several patient registries.
A patient registry uses observational study methodology, but a case-control study or a cohort study is a well-defined type of epidemiological study and is not a patient registry. A patient registry can be seen as a wider notion than a post-marketing observational study, because an observational study is usually shorter, and a post-marketing study is often a safety study and often thought of as a health authority requirement. Views on these terms differ, and the distinction made between them is arguably not very clear.
Purposes
The Agency for Healthcare Research and Quality distinguishes four purposes of patient registries:
- Describing the natural history of diseases, which is particularly important in rare disorders, where the natural progress of a disease and the symptoms of the people who have it may still not be known well enough.
- Evaluating the effectiveness of a treatment or any other medical intervention, both clinical effectiveness and cost-effectiveness; registry data are very often used for cost-utility analysis, cost-effectiveness analysis and other pharmacoeconomic evaluations.
- Collecting data on the safety of a medical intervention and any potential harm it can do, under the medical rule primum non nocere; health authorities often use this aim, for example for post-marketing commitments.
- Measuring or improving quality of care, including developing tools from the data that improve patient management and clinical care.
Each type of patient registry can be further characterised by the population enrolled, for example by a disease, phenotype, exposure to a drug or medical intervention, region of origin or other predefined features.
Design and uniform data collection
The purpose of a patient registry must be defined beforehand. Collecting data on anything and everything to get as much information as possible, and then turning to data mining, is the wrong way round: data collection must be purpose-driven, guided by a very clear view of the well-defined research questions the registry should answer. A frequent mistake is to think that the design of a registry needs no thought and that it is just a matter of collecting everything.
Data must be collected in a uniform manner to give meaningful results and for a statistician to be able to analyse them. In one centre, a few centres or one region the uniform manner may seem obvious, but clinical practice may differ between the regions of a registry run in many countries or across continents, so the uniform manner has to be agreed and clearly communicated, for example how a circumference is measured, so that the data are comparable.
Patient registries usually have no protocol but a manual or guidelines, which control how the data are collected. The guidelines can draw on international standards for laboratories, name recommended laboratories where the analyses are known to be performed properly, and describe how certain parameters should be measured. Where laboratory data are very important, a registry may offer a centralised laboratory, so that samples for certain variables are analysed centrally. Even so, discrepancies between countries and between centres can remain.
Data mining
There appear to be no truly regulatory guidelines on data mining of registry data comparable to the guidance for clinical trials. One important safeguard against misuse of the data is the registry's predefined purpose, which ensures that the data collected are the data that serve that purpose. Data mining itself is pretty widely used by pharmacovigilance institutions, such as the Uppsala Monitoring Centre, which collects adverse events from all over the world.
Duration, size and reach
Patient registries commonly run for a long time, collect information on a large number of patients and cover many countries and regions, whereas randomised trials are of limited duration and enrol a strictly limited number of subjects, based on sample size calculations. These properties are crucial particularly for registries of rare disorders, in which prevalence and incidence are by definition low, so that only registries are capable of providing enough data to draw meaningful conclusions.
Patient registries are often set up without limits, and usually without a time horizon. A registry run for many years allows analyses of how clinical practice changes over time, of how new medical interventions and progress in medicine affect clinical practice, of how the uptake of innovative medicines differs from region to region, and of secular trends in the patient population, which cannot be learned from randomised trials.
Patient numbers are often not limited either, and registries often hold thousands of patients. The large number matters because patient registries are not monitored as strictly as randomised trials: the many patients and the abundant information can compensate for the missing data and poor data quality that analysing registry data usually involves.
Patient registries are often run across borders, so regions and countries can be compared on how clinical practice is performed. They are often designed to use different sources of information: other registries contributing data, data from one registry cross-checked or matched with another, a specialist department and the patient's general practitioner both contributing, or data entered or pulled directly from a laboratory. Using different sources of information ensures a very comprehensive view of patient management and clinical practice.
Registries and randomised trials
The biggest difference from a randomised trial is that patients in a registry are treated according to standard clinical practice: their treatment and management are determined by physicians, caregivers and sometimes patients, at the treating physician's discretion, whereas in a trial the intervention is dictated by a protocol. This difference is extremely important to keep in mind, because it means viewing registry results somewhat differently: what a registry monitors and finally describes is real-life clinical practice.
A randomised trial is an attempt to construct a laboratory environment: a very homogeneous study population whose members are very much alike, from which background and baseline differences have been removed, so as to be as sure as possible that the difference seen at the end of the study is attributable only to the interventions applied in the study arms and is not confounded by other parameters. Once a drug is approved and used by ordinary physicians, its patients are of all sorts and very different from each other, obese and very thin patients, pregnant women and old people, and often none of them resembles the patients in whom the trial was run. A patient registry can replicate the diverse types of patients met in clinical practice, and so yields subgroups big enough to study treatment effects in them.
The main differences between the two can be set out in seven characteristics. Patient management follows the protocol in a trial and clinical practice in a registry; selection criteria are very stringent against much less stringent; duration is fixed against often open-ended; the comparator is often part of the study against various, such as part of the study or the general population; the risk of bias is low against high; generalisability is low against high; and applicability to the target population is various against high.
Evidence-based medicine strictly classifies sources of medical information by the strength and reliability of the evidence they provide, with meta-analyses and systematic reviews at the top of the pyramid, closely followed by randomised trials, particularly blinded and controlled ones, while patient registries fall under observational studies and are classified as low-level evidence. Randomised trials are considered the gold standard of medical evidence primarily because they are a reliable unbiased source for inferring causality from observed associations. The main concern about them as guidance in medical decision-making is their applicability to daily clinical practice and generalisability to the patient population at large.
As a general rule, studies performed with registry data do not prove causality and can only indicate associations between observations. Like observational studies, patient registries are a tool for identifying research questions. A strong causal relation should probably not be drawn from registry data, but an observation in a registry is a signal of something interesting happening in real life, and it helps define a scientific hypothesis and design prospective studies properly.
Whether a randomised trial or a patient registry is needed follows from the basic question being asked. If the question is whether a medicine works at all, a laboratory environment is needed and the answer comes from a randomised trial, because a clean and convincing answer is wanted. If the question is whether and how it works in real life, the answer comes from patient registries, which use observational study methodology. Randomised trials answer the question "Can it work?", and studies based on patient registry data answer "Does it work? How does it work in real life?" Both answers are important, and the two sources of information are not competing but complementary. This difference between trials and registries should be kept in mind when writing papers or any other medical communication about patient registries.
Ideally a randomised controlled trial answers whether a drug works, but for very rare diseases there are sometimes not enough patients to run one, and a registry can then be very important in providing supplemental data to answer that question.
Registries and their findings
A paediatric registry of children with short stature treated with growth hormone ran for 25 years, from 1987 until active data collection stopped in 2012, and collected data on more than 80,000 children from 51 countries, some 300,000 patient-years of follow-up. KIMS, started in 1994, collected data on adults with growth hormone deficiency, that is with hypopituitarism, also a very rare disorder, from 16,000 patients in approximately 31 countries, with 90,000 patient-years of follow-up. InfCare HIV, a mainly Swedish database with some Danish centres, had almost 6,000 patients with HIV infection in 2012 and covered almost all, if not all, Swedish patients with HIV, so it covers one country comprehensively rather than reaching across borders.
A growth hormone registry run between 1987 and 2012 was initiated at the request of regulatory authorities as a post-marketing surveillance study to follow 500 patients treated with Genotropin for five years, and it evolved into one of the largest and longest-running pharmacoepidemiological studies. Its four primary objectives were to evaluate the long-term safety of growth hormone and the outcomes of treatment with Genotropin, to determine relationships between clinical status, dosage schedule and response to treatment, to develop clinical tools for individualised growth hormone treatment of children, and to contribute to knowledge of growth and growth disorders. It was shared in the public domain and yielded 129 publications cited in PubMed.
Other published registry analyses include one of fracture risk in adults with growth hormone deficiency, untreated and treated, from the Hypopituitary Control and Complications Study (HypoCCS); a comparison in KIMS of adult growth hormone deficiency by underlying aetiology; and an analysis of cardiovascular risk factors in 2,589 adults with growth hormone deficiency in KIMS. Another used the UK General Practice Research Database, a longitudinal general practice database collecting data from a representative sample of general practices in the UK, to study the incidence and prevalence of overactive bladder symptoms and the use of anticholinergic or antispasmodic drugs and healthcare resources. An analysis of InfCare HIV data looked for characteristics of suboptimally treated patients, starting from a clear definition of suboptimal treatment.
PETHEMA, the Programa Español de Tratamientos en Hematología, is a cooperative research group of virtually all Spanish haematologists and many from Portugal and Latin America. Its AML epidemiological registry holds all the clinical data and much of the correlated biological data of more than 24,600 patients diagnosed with acute myeloid leukaemia over several decades, about 5,300 of them with the acute promyelocytic subtype. Drawing on cases of predefined clinical profiles numbering mostly in the thousands, its published research has described special patient populations; validated certain treatments and shown advantages of some over others; shown the suitability of allogeneic transplantation and analysed the incidence of secondary malignancies; allowed comparisons with historical external controls; demonstrated the prognostic value of minimal residual disease and its key role in risk stratification; and characterised clinical and genetic patterns linked to treatment responses and prognoses.
Sponsorship
Patient registries are very often sponsored by a company, and the sponsor can define which products or patients data are collected on. Some company-sponsored registries collect data only on the sponsor's own product, as the Pfizer Pharmacia registries did, and others collect data on all patients regardless of the product they take. Collecting data on a comparator product that is also on the market is acceptable.
A registry need not be company-driven: InfCare HIV was not, and it collected data on all kinds of treatment.
Consent, ethics review and privacy
Patients in a registry generally need to sign informed consent, and ethics committee approval, or at least its opinion, is needed as well. The requirements vary between countries, but in the majority of countries both are needed, and obtaining at least the ethics committee's opinion should be standard. The exception is very large databases such as Medicare, probably more popular in the United States, from which completely anonymous parts of patient files are retrieved.
Names, addresses and the like are not collected, but under the European Union's data privacy law personal data are any data by which a patient can be identified, so the line is very difficult to draw. In a registry of a rare disorder with perhaps 100 patients in a country, a combination of data can identify a patient. One registry collected only the year, or the month and year, of birth rather than the full date. Because of privacy law it is better to collect less personal information, while making sure the data can still be analysed and understood.
Data quality
The quality of a registry can be checked against independent national registries: the Swedish population of the KIMS database was cross-checked against the national cancer registry and the national death registry to see whether it had collected all the cases those registries held.
Quality can also be increased by designing monitoring, such as stop values and warning signs, into the data collection tool, and a web-based tool with enough checkpoints can yield very clean data. Such monitoring is always a trade-off against what the investigators or physicians entering the data will stand, since getting stop values every couple of minutes is irritating, but the technical means to improve data quality exist.
In October 2021 the EMA published guidance on methods and good regulatory practices for pharmaceutical organisations planning and conducting registry-based studies. Requirements for the types, structures and processing of data can differ significantly between existing registries, which often makes it hard to assess whether an existing registry is suitable for use in clinical studies. The guideline aims to help those involved in registry-based studies define study populations and design study protocols, gives guidance on data collection, data quality management and data analysis, and includes an annex with good practices for establishing and managing patient registries and for their use for other regulatory purposes. It is based on a discussion paper on the methodological and operational aspects of using patient registries for regulatory purposes, whose public consultation generated almost 1,000 comments from 68 stakeholder organisations, and on experience from the qualification opinions of the agency's human medicines committee on two networks of registries and input from five workshops on specific patient registries. The guideline is to facilitate a more data-driven, robust regulation of medicines, as foreseen in the Big Data Steering Group Workplan.
Statistical analysis
The lack of randomisation is the main problem with registry data. Statistical techniques allow models to be built that correct for all the background differences and yield reliable results, compensating for that lack, so collaboration with statisticians is extremely important.
Many years ago there was resistance to and suspicion of results based on registry data. Evidence-based medicine has arguably changed its principles somewhat over the years, and with appropriate statistical methodology new, reliable knowledge can be produced from registry data, which forms a basis for publication and presentation.
Simple descriptive studies need only simple descriptive statistics, but comparisons between treatment groups, prediction of treatment outcomes and mortality analysis need more advanced techniques, because treatments in a registry are not randomised and patients in one group, such as a treated group, may be systematically different from those in another, such as an untreated group. The analysis must consider such selection bias, and it should account for all known or potential confounders. In an analysis of change in body mass index during long-term growth hormone treatment in the KIGS registry of children with short stature, the primary aetiology groups differed before treatment started, so advanced statistical methods had to be used.
Uses of registry data
One working categorisation sorts the uses of registry data into five fields: science, clinical practice, safety, regulatory and health economics.
Science. Registry data provide new knowledge, and they help in understanding diseases and their natural history and in seeing differences between healthcare systems and geographical regions and changes over time. They are also instrumental in answering critical questions that differ between fields of medicine, such as treatment patterns and responses to treatment in different patient groups. In pituitary disorders such questions included mortality in adults with growth hormone deficiency receiving growth hormone, the almost everlasting concern about cancer risk in growth-hormone-treated children, and prediction models.
A finding in a registry can lead to more precise information. An analysis of an entire registry database showed a surprising increase in mortality from infectious diseases that could not be investigated precisely on the whole database, and later work in the Swedish population found what lay behind it.
Clinical practice. Registry data can monitor patients' adherence to treatment, and they can be made into patient management tools: individual patient feedback reports and clinic performance reports.
Safety. The role of patient registries in safety can hardly be overestimated: they detect rare adverse events and rare or atypical treatment reactions, and teach much more about safety than a typical clinical trial. A randomised trial usually has limited follow-up, standardised treatment conditions that leave no room for variation, and a rather small patient cohort, so rare adverse events are unlikely to be captured in it. A registry or surveillance study has long follow-up, includes patients who are not eligible for trials and so a population much more like daily clinical practice, is much bigger, allows subgroup analyses, and has a much higher chance of detecting very rare adverse effects. In almost 60,000 children treated with growth hormone, the standardised incidence ratio for cancer did not differ from the general population among those who had no other risk factor for developing a malignancy. Registries from compassionate use arguably add value from a pharmacovigilance point of view, since for safety more data are better and information from compassionate use is much closer to normal clinical practice than that from randomised trials. A patient registry may also provide clinical context for adverse events reported spontaneously. Patient registries are often used to fulfil health authority requirements, for example by running post-marketing authorisation surveillance studies.
Regulatory. Data from patient registries and large databases are the basis for many regulatory reports, both ad hoc safety reports and the regular safety reports such as PSURs and DSURs. Patient registries, observational studies and databases are part of post-marketing work and drug development.
A registry can also be a condition of marketing authorisation. In June 2018, when the EMA's human medicines committee recommended granting a marketing authorisation for Kymriah and Yescarta, the first two chimeric antigen receptor T-cell medicines in the European Union, the use of a patient registry to monitor their long-term safety and efficacy was a risk management measure and a condition for the marketing authorisation. The EMA has qualified a registry for collecting post-authorisation safety and efficacy data.
The EMA's revised guidelines on the tests and studies needed to support marketing authorisation applications for certain haemophilia medicines were published in July 2018 for factor VIII deficiency and on 30 November 2018 for factor IX deficiency. They changed the investigation of recombinant and human plasma-derived factor VIII and factor IX medicines in previously untreated patients: for this very small subset, data should be collected from patient registries rather than from small clinical trials that may not be fully representative of how the medicine is used day to day once it is on the market. The approach relies on registries as a source of high-quality real-world data to support regulatory decision-making, and reflects that clinical data are difficult to obtain in this subset because it is highly selected and the numbers available for trials are very small. The updated guidelines aim to optimise and facilitate the use of these registries and provide parameters for the core data sets that should be collected. Waiving the requirement for a clinical trial in previously untreated patients was discussed at an EMA workshop on haemophilia registries in July 2015. After a public consultation in 2017, a second workshop, on 8 June 2018, aimed to define the requirements for practical implementation using existing registries to support post-authorisation observational studies of haemophilia medicines, and discussed the governance of registries, patient consent, data collection, data quality and data sharing, and interoperability between registries. A questions-and-answers document covers the practical implementation. The EMA's initiative on patient registries is supported by a task force of representatives from its scientific committees and working parties, representatives from the European Commission and experts from national competent authorities.
Medicine regulators may sometimes suggest that pharmaceutical companies use the data collection infrastructure or population of a patient registry to exploit information from clinical use and to monitor the safety and effectiveness of authorised medicines in the real-world setting.
Health economics. Registry data can provide information on the burden of disease and healthcare utilisation, and they can be used for pharmacoeconomic modelling.
Rare-disease registries
The low prevalence of rare diseases limits the data available to build evidence on the safety and effectiveness of drugs, on diagnosis and for research in general, and patient registries are one possible solution. Patient registries have been set up for many rare diseases to gather the information required for treatment development in one place. They can be based on a single disease or a group of related diseases, assembled to gather data to test a new product in a clinical trial, or drawn from a particular population. They are set up either by a company developing a product, which restricts the data to internal use, or by patient organisations or clinical consortia, in which case the data may be available for others to use. They provide information on, among other things, the natural history of a rare disease, its incidence, the expected number of patients eligible for a clinical trial, the choice of trial endpoints, the tracking of treatment outcomes, and quality-of-life assessments that lead to healthcare resource utilisation. Patient heterogeneity arises from the underlying genetic cause in allelic diseases, and it may also reflect other genomic influences beyond the disease gene and environmental factors such as diet and living conditions; patient registries that include genetic variation data can be invaluable here.
Families with rare diseases are often very willing to allow the collection of their data, because they understand its importance in research and development, but the holding of personal data, including health data, is regulated differently worldwide, which makes sharing the data challenging. To provide real-world data, registries need quality assurance processes for data organisation and data quality, consideration of potential biases, and a database fit for purpose.
A recent survey of rare-disease registries found most of them based in Europe, predominantly led from Germany, the UK or the USA. Some held more than 30,000 cases, two-thirds covered a range of diseases and a third a single disease. Most were national, and many others continental or global, which requires interoperable data elements, ontologies and common terminologies so that data collected in different places can be combined. They aimed to provide participants for clinical studies, evaluate or improve clinical care, describe epidemiology or improve understanding of natural history, and collected sociodemography, diagnosis, medical history, care pathway and treatment history. Approximately one-fifth used common or core data or ontological coding language, and nearly half had no clear governance. Many included patient-reported outcomes, but not all involved all potential users, such as patient organisations, in their design. Funding came mainly from federal or European Union bodies, with many registries funded by private pharmaceutical or technical companies.
By 2025, Sanofi's Rare Disease Registries, set up 30 years earlier and expanded to four lysosomal storage disorders, Fabry disease, Gaucher disease, mucopolysaccharidosis type I and Pompe disease, held data on more than 18,000 patients at over 800 sites in 64 countries, had a patient council, and had led to more than 100 peer-reviewed articles. DEM-CHILD, a registry for the neuronal ceroid lipofuscinoses, or Batten disease, was initiated by collaborating European clinicians and led by Angela Schulz to improve early diagnosis and optimise standards of care. It measures the prevalence of each form in the participating countries, links retrospective and prospective clinical data with genetic data to describe each form's clinical course and correlate genotype with phenotype, and provides a tool for evaluating experimental and palliative therapies. It held more than 250 patients. It has ethical approval and follows European data protection guidelines, with an approved audit trail to ensure data safety and data stored on different servers with emergency power and daily backup. It planned to let parents contribute data, and it shares data non-exclusively with third parties worldwide, such as scientists and pharmaceutical companies. It applies established and novel clinical rating scales to assess disease progression, and quality-of-life questionnaires, and has a biobank of serum and cerebrospinal fluid samples. The EMA and the FDA accepted its natural-history data for late-infantile CLN2 disease as valid natural-history controls for the efficacy evaluation of experimental therapies, which led to the expedited approval of intracerebroventricular enzyme replacement therapy with cerliponase alfa in May 2017.
The NCL Mutation Database, freely accessible alongside DEM-CHILD, collects published genetic and phenotype data on more than 700 variants in the NCL genes and more than 1,700 patients, to inspire study design, predict disease severity and consider the implications for therapeutic development. Consistent checks by an expert curator increase the quality and accuracy of its data. Potentially duplicated patient records are highlighted and investigated with the clinicians or researchers concerned, and the age of onset is kept distinguishable from the age at diagnosis, since the time between them may be considerable. Each variant is checked for accurate Human Genome Variation Society nomenclature, because errors in variant nomenclature are relatively common and have in some cases led to the publication of supposedly new variants that were misdescribed known ones.
In rare diseases registries have proliferated, which has implications for their long-term sustainability. The critical factor influencing that sustainability will be a registry's quality, which can be broadly divided into two categories: the registry's operation systems and the data it collects. Data quality matters even more in rare diseases, where populations are very small and poor data quality may skew the results or make them inconclusive, limiting the acceptability of the findings, and registries with a higher level of data quality are more likely to be accepted by healthcare providers. Data quality may be defined in several ways, including completeness, interoperability, accuracy, validity, consistency, timeliness, uniqueness and traceability; the definition of registry quality may vary considerably between registries, and the level of consensus on data quality is unclear.
A systematic review of English-language peer-reviewed literature from 2010 to 2025 with a clear focus on data quality and rare-disease registries included 78 studies and identified nine themes in them: completeness, selection bias, validity, accuracy, consistency, interoperability, duplication, standardisation and common data set elements. Term occurrences grew from 50 in 2010–2013 to 876 in 2022–2025, a 17.5-fold increase, and totalled 1,618. Completeness terms were among the most frequent, with completeness appearing 224 times, complete 166 and completed 90. Interoperability, with no mentions before 2014 and 145 in 2022–2025, was the most cited single term overall, with 229 mentions, while common data set elements never appeared as a phrase and minimum data set only 13 times. The review's search terms may not have captured the frequency of terms such as common data elements or minimum data set accurately. Completeness was the dominant theme in 2010–2013 and still the most frequent term in 2014–2017, closely followed by validity; interoperability and duplicate were the most frequent terms in 2018–2021, and interoperability and completeness in 2022–2025. Every theme co-occurred with at least one other, and common data set elements, though infrequent, was well interconnected, because the articles that had it also had several other themes.
The trend shows a progression from basic aspects of quality in 2010–2013, particularly completeness and whether data had been fully recorded, towards system-level issues such as interoperability and duplication from 2018 onwards, which matter when data are shared across settings or used for secondary purposes. The overlap between themes suggests that the definition of data quality in rare-disease registries is inherently multidimensional. Many healthcare professionals are not aware of rare-disease registries, those who are take limited part, and clinicians and their administrative and care staff often have heavy workloads that leave little time for data entry or patient follow-up. If data were sufficiently interoperable they could flow between sources, minimising manual entry and the transcription errors it may cause, although without local approval free flow of highly sensitive data is likely to be challenging at institutional level. Rare-disease registries rarely need real-time data, so systems that bulk-download source data and upload them at a convenient time are one solution to the time constraints. Such systems still require agreement on standardised data sets that can be collected universally, referred to as common data elements, core outcome sets or minimum data sets. Projects such as GloBE-Reg, a global registry for novel therapies in rare bone and endocrine conditions, aim to improve data quality by minimising the amount of data collected.
In a rare-disease registry, the informatics platform matters for the researchers' long-term compliance in entering the right data, and so does the relationship with the statisticians who elaborate the data.
Every rare disease should be linked with a registry, each standardised as necessary to capture global real-world data, and access to those data should not be restricted unnecessarily. Real-world data from digital health technology could be improved by home-based regular longitudinal assessments appropriate to the disease, and every registry should be fit for use by regulatory bodies.