Registries based on voluntary patients’ expression of interest have several advantages, as they are powerful tools to directly involve individuals with a specific condition in the process and grasp their daily life experience. On the other hand, the epidemiological validity of these registries may be limited, as they cannot reasonably map nationwide all existing individuals with that specific conditions [
15]; for this type of analysis one should refer to the National Healthcare resources. However, healthcare anonymised data are not easily accessible and may not be suitable to answer precise scientific questions relevant for a certain disease. In addition, individuals that provide their own data for the healthcare databases have a passive role and do not take part in any decision-making process.
Condition-specific registries may have different design and collect patient-reported or clinician-reported forms. Both registration modes have strengths and limitations that must be considered to make sure that the chosen model generates valid knowledge and fulfil all stakeholders’ expectations. Potential risks that may affect the quality of both type of registries concern population bias, data quality and validation, and loss to follow up.
Patient-driven registries
A bias in the population selection may derive from a different exposure of patients and families to communication regarding the registry existence and scopes. Those who are regularly followed by centre of expertise or are members of the POs of interest are in general better informed. Moreover, given that POs typically direct families and patients towards their referring experts where possible, it is reasonable to expect that these patients are also followed according to the most updated standards of care, thus affecting the information on adopted standards of care that can be inferred from the registry.
Collecting medical information directly from patients has a number of advantages. First, it captures the direct experience of individuals with a specific condition, it offers a fresh perspective on the quality of life and needs, and engages the family in a responsible way, favouring empowerment and participation. Importantly, it can help reducing the burden on clinicians regarding the workload of data entry. This type of collection proved its usefulness to design clinical trials and helped patient recruitment into trials [
6,
24,
25,
58] or, at least, promoted the identification of those that fulfilled the main criteria for a first screening (a clinical trial may have additional inclusion/exclusion criteria that are not identifiable through the registry). Nevertheless, the validity of patient-reported data is exposed to several risks that cannot be underestimated. First of all, the different participants may not uniformly record data and diagnoses. Although clinicians should not revise data inputted by patients [
15], at least their validation of the genetic and clinical diagnosis should be ensured [
60]. To this regard, if clinical data are required for purposes such as accurate natural history or post-marketing validation of a treatment, clinician-driven registries should be preferred. Another issue concerns patient retention. Filling in the registry forms and keeping data updated may not be a priority for affected individuals and their families or caregivers, especially when the health status of the patient worsens and the disease burden becomes high. Automatic reminders can favour the update of forms, but the support of the referring physician or the registry Curator can be very relevant to guarantee that the registry’s content is correctly updated.
Clinician-driven registries
Dual Registries have ambitious aims, which include defining a well-characterized and properly diagnosed patient population, not only for recruitment in clinical trials, but also for natural history studies, validation of outcome measures, creation of biorepositories, epidemiological analyses (if feasible) or surveys. They guarantee both direct input of patients in the registry and the collection of accurate and validated clinical data that are recorded in a systematic and scientifically sound manner and can be rapidly analysed. Of course, to fulfil the above scopes, data need to be validated and of high quality. In our experience, the organisation of training sessions on the functional scales used to record patients’ performance, the implementation of standards of care by the involved centers, as well as the regular monitoring and solicitation of data entry by the registry’s Curator, contribute to implement the databases with data of good quality. These registries may have limitations too, which depend on the nature of the registry itself and, for instance, on the rarity of the diseases and their related clinical and molecular heterogeneity. There may be enrolment biases due to several reasons. One, again, relates to the fact that registration is on a voluntary basis. Participation in the registry is often stimulated by POs or Patient Advocacy Groups linked to clinical study networks and clinicians; therefore, patients are more likely to register if they are already engaged in the patients’ community, and have instruments and contacts to become aware of the registry. In other cases, registration depends on clinical and genetic characteristics; the most severely affected patients are more likely to attend tertiary centers, which in turn are more often involved in registries. Clinicians also may be more prone to encourage registration of patients with genetically defined disorders as compared with those without a precise genetic diagnosis. Upcoming treatments specific for certain disease subgroups may also be a reason for asymmetrical recruitment. Therefore, how much the registry population corresponds to the real world, and how valid the epidemiological inferences are remain open questions. To address and overcome such selection biases is rather difficult. One way is to try to enlarge the registration pathways as much as possible, by clearly defining and disseminating the objectives through the involvement of clinicians in promoting the registry, dedicating time to carefully explain the Informed Consent document, obtaining the agreement and endorsement from the scientific societies, developing the Patient Advocacy Groups’ role and the potentialities of the web. When the registry concerns ultra-rare patients, such as in the case of the MGSD, registration supported by the clinicians themselves may be more effective than waiting for spontaneous registration by affected individuals or their families.
Mapping individuals living with an ultra-rare NMD condition may be even more effective when this occurs directly through an international registry [
6,
61,
62]. This could be the case also for the SBMA registry and discussion is ongoing to verify how this can be achieved [
12]. In this case, the Italian registry hosted on the NMD registry platform could become an important reference model. Opening the platform to international direct interaction may imply also that the registry structure is fully interoperable. Continuing efforts are ongoing to adopt common data elements and implement ontologies such as the Human Phenome Ontology [
59] in line with the international standards and guidelines [
58].
Theoretically, registries should define: how many patients they plan to recruit based on a sample size estimation in relation to the aims; how frequently patients should be re-assessed; for how long the registry should last; which statistical analyses will be applied. A registry’s steering committee should include statisticians to deal with such aspects. Plans should be in place for recruiting a reasonable number of patients in a predetermined timeframe for the purposes of the registry, and a statistical analysis plan should be defined right in the registry project design [
15].
The aims of the registries currently on the NMD platform are mainly of a descriptive nature and did not require preliminary sample size estimations per se. However, the Steering Committees that planned each database assessed their feasibility by careful evaluation of the expected number of individuals that the national registry could likely map, depending on the disease rarity and the patients’ cohorts of each participating center. The design of the registries was therefore based also on this knowledge, in addition to the international consensus on standardised items.
A very critical point that affects the timeframe and population size that allows the registry reaching statistical significance is its funding. This is particularly true when data collection puts a high burden on clinicians, because data items are numerous and/or patient population is relatively wide. In some cases, registries started as observational studies, with the declared intention of long-term maintenance, update and reuse of data. However, when the initial funding support ends it becomes more and more difficult for clinicians to keep the pace with data recording and updates. In order to overcome this problem, clinicians are encouraged to include fund request for structural sustainability and dedicated personnel in any grant application that entails using the registry as a tool.
Finally, a current limitation of the NMD registry platform is that, being owned by a private entity, it cannot merge data with public healthcare databases. This is clearly an obstacle when one considers the clinical burden in managing medical data and the missed opportunity to share individual biomedical information, which could already be available and complementary to the clinical data collected by the registry. All legally feasible efforts should be put in place to verify the possibilities of matching data with other systems, such as the regional or national healthcare registries [
15] or the new NMD European Reference Network initiative [
63].