When the entire population is the sample: strengths and limitations in register-based epidemiology

B ias and confounding are the twin banes of epidemiology, but of the 2, bias is worse.Confounding can often be detected and at least partly corrected by appropriate statistical analyses, but if bias is built into the design of a study, it can be difficult to detect, and even more difficult, if not impossible, to correct.Different types of epidemiological studies are prone to different types of biases.One type of particular concern in case-control studies is survival bias, which can arise when risk factors for a disease that is often fatal are measured only in the patients who survive it.An exposure that is more frequent in the survivors may appear to be related to the disease when it is actually associated with survival.Alternatively, if an exposure is related both to disease and its lethality, its association with the disease may be diminished by survival bias.

<h3>Background</h3> Predictors of poor outcome in systemic lupus erythematosus (SLE) may lead to the identification of high-risk patients at the onset of disease (incident cases) and/or when we first assess them in our clinics (prevalent cases). We tested whether the Lupus Severity Index (LSI) can help characterize high versus low risk lupus patients. <h3>Methods</h3> Population: Patients from six lupus centers were recruited according to a standard data collection protocol. We characterized incident cases and prevalent cases as those with a diagnosis made within or after the previous 15 months. Data collected: Demographic, socioeconomic, disease specific and medication data were collected at baseline and annually. We collected: the American College of Rheumatology (ACR) and the Systemic Lupus International Collaborating Clinics (SLICC) classification criteria, the SLE Disease Activity Index (SLEDAI), the Systemic Lupus Activity Questionnaire (SLAQ), and the SLICC Damage Index (SDI). The LSI was derived from the ACR classification criteria and used as a predictor variable. Statistical analyses: Kruskal-Wallis test and Spearman correlations were used to see the association of LSI with categorical and continuous variables respectively. The baseline LSI was used to predict outcomes at follow-ups using logistic regressions and Spearman correlations for dichotomous and continuous variables respectively. <h3>Results</h3> We enrolled 639 lupus patients and 440, 324 and 168 were re-evaluated at 1, 2 and 3 years. Baseline characteristics (table 1) [median (IQR)] were: age=49.0 (36.8–58.5) years, female=92%, Caucasian=74%, disease duration=10.1 (2.7–20.6) years. We had 129 (20%) incident cases and 471 (74%) prevalent cases with missing information in 39 (6%). Twelve patients died during follow-up. Table 1 summarizes baseline associations between LSI and several characteristics for the incident and prevalent cases. We found that age, sex, ethnicity (Asian worse LSI), SLICC classification criteria, SLEDAI, prednisone use and daily dose were associated with LSI in both incident and prevalent groups while the SDI and the use of immunosuppressors drugs was associated with LSI only in the prevalent cases. In follow-up, baseline LSI predicted SDI in prevalent cases (p=0.02) with a trend in incident cases (p=0.07). LSI predicted death in the prevalent group. <h3>Conclusions</h3> The LSI is easy to derive from the ACR classification criteria and a useful measure of severity in lupus. The LSI is associated with baseline characteristics, some of them - like disease activity and prednisone dose - modifiable. LSI can predict adverse outcome such as damage or death over time. <h3>Funding Source(s):</h3> Lupus Canada.

Understanding the long-term safety of disease-modifying therapies for multiple sclerosis (MS) in routine clinical practice can be undertaken through registry-based studies. However, variability of data quality across such sources poses the challenge of data fit for regulatory decision-making. CLARION, a non-interventional cohort safety study of cladribine tablets, combines aggregated data from MS registries/data sources, except in Germany (which utilizes primary data collection). We describe the application of key data quality indicators (DQIs) within CLARION to evaluate data quality over time, as recommended by the European Medicines Agency (EMA) guideline on registry-based studies. DQIs were defined with participating registries/sources; they were used to assess data quality according to the EMA Data Quality Framework, addressing consistency, accuracy, completeness, and study representativeness. DQIs were associated with potential remedial measures if data quality was not met. DQIs were summarized overall and for individual MS registries/data sources to November 1, 2022. A total of 28 DQIs were analyzed using data from 5069 patients arising from eight MSregistries/data sources and 14countries. The Representativeness DQIs showed that 72.0% of patients were female, median age at MS diagnosis was 29.0 to 43.3 years, and 93.5% had relapsing-remitting MS. Consistency DQIs showed a total of 2899 patients had achieved at least two years of follow-up; 6.9% did not have any recorded visits during this timeframe. Discrepant values were assessed as part of Accuracy DQIs, and improvements over time were noted for recorded dates of MS onset and diagnosis. Regarding Completeness DQIs, 191/5069 (3.8%) patients were lost to follow-up. The application of 28 DQIs within the CLARION study has helped with understanding, not only intrinsic and question-specific determinants of data quality, but also tracking the quality of post-authorization safety data obtained from MS registries/data sources, thereby providing a foundation for the regulatory decision-making process.

Background: Data on type 1 diabetes incidence and prevalence are limited, particularly for adults. This study aims to estimate global numbers of incident and prevalent cases of type 1 diabetes in 2017 for all age groups by country and areas defined by income and region. Methods: Incidence rates of type 1 diabetes in children from existing literature were used and extrapolated to adults using available data on adult to child incidence ratios. Age-specific incidence rates were applied to population estimates to obtain incident case numbers. Duration of diabetes was estimated from available data and adjusted using differences in childhood mortality between countries. Prevalent case numbers were derived by modelling the relationship between prevalence, incidence, and disease duration. Findings: Global numbers of incident and prevalent cases were estimated to be 234,710 and 9,004,610, respectively. High income countries, with 17% of the global population, account for 49% of global incident cases and 52% of prevalent cases. Asia has the largest number of incident (32%) and prevalent (31%) cases and the largest proportion of the population (60%). Globally, the prevalent cases were 6%, 35%, 43%, and 6% in the age groups 0-14, 15-39, 40-64, and 65+ years, respectively. Interpretation: Globally type 1 diabetes represents about 2% of the estimated total cases of diabetes, ranging from less than 1% in certain Pacific countries to more than 15% in Northern European populations. This study provides information for development of health care and policy approaches to manage type 1 diabetes. Funding Statement: The study was funded by the US Centers for Disease Control and Prevention, through grant 5NU2GGH001235-05-00 of the WHO-CDC Cooperative Agreement, and by The Leona M. and Harry B. Helmsley Charitable Trust as part of the ACCISS Study. Declaration of Interests: AG and SMH report grants from the Institute of Applied Economics and Health Research, Copenhagen, Denmark outside the submitted work. SW reports non-financial support from Novo Nordisk, outside the submitted work. All other authors declare no competing interests.

Aims/hypothesisData on type 1 diabetes incidence and prevalence are limited, particularly for adults. This study aims to estimate global numbers of incident and prevalent cases of type 1 diabetes in 2017 for all age groups, by country and areas defined by income and region.MethodsIncidence rates of type 1 diabetes in children (available from 94 countries) from the IDF Atlas were used and extrapolated to countries without data. Age-specific incidence rates in adults (only known across full age range for fewer than ten countries) were obtained by applying scaling ratios for each adult age group relative to the incidence rate in children. Age-specific incidence rates were applied to population estimates to obtain incident case numbers. Duration of diabetes was estimated from available data and adjusted using differences in childhood mortality rate between countries from United Nations demographic data. Prevalent case numbers were derived by modelling the relationship between prevalence, incidence and disease duration. Sensitivity analyses were performed to quantify the impact of alternative assumptions and model inputs.ResultsGlobal numbers of incident and prevalent cases of type 1 diabetes were estimated to be 234,710 and 9,004,610, respectively, in 2017. High-income countries, with 17% of the global population, accounted for 49% of global incident cases and 52% of prevalent cases. Asia, which has the largest proportion of the world’s population (60%), had the largest number of incident (32%) and prevalent (31%) cases of type 1 diabetes. Globally, 6%, 35%, 43% and 16% of prevalent cases were in the age groups 0–14, 15–39, 40–64 and 65+ years, respectively. Based on sensitivity analyses, the estimates could deviate by ±15%.Conclusions/interpretationGlobally, type 1 diabetes represents about 2% of the estimated total cases of diabetes, ranging from less than 1% in certain Pacific countries to more than 15% in Northern European populations in 2017. This study provides information for the development of healthcare and policy approaches to manage type 1 diabetes. The estimates need further validation due to limitations and assumptions related to data availability and estimation methods.Graphical abstract

Response

A nationwide registry study on heart failure in Norway from 2008–2018: variations in lookback period affect incidence estimates

Most studies that have examined the clinical features of gonorrhea and chlamydial infection have been based on prevalent cases (cases of undetermined onset). In our investigation, we compared signs and symptoms of incident (new) cases of these infections with those observed in prevalent cases (involving the same women) that were diagnosed at enrollment in a prospective study of female prostitutes in Kinshasa, Zaire. Neisseria gonorrhoeae or Chlamydia trachomatis was present at enrollment in 29.2% (225 of 771) of the women in the study. As they were followed during the study, 509 (66.0%) had at least one episode of gonorrhea or chlamydial infection. No symptom was significantly associated with these infections at enrollment or during follow-up. Clinical signs, such as endocervical mucopus (P < .001) and vaginal discharge (P = .001), were associated with both the prevalent and incident cases. However, none of these signs was simultaneously sensitive and specific for detection of these infections. The frequency of clinical signs was significantly reduced after successful treatment (all P values, < .05). This study shows that a syndromic approach to screening for gonococcal and chlamydial infections in female prostitutes is as problematic for acute incident cases as for prevalent cases. Therefore, there is still an urgent need for simple, cheap, reliable tests that could be used in sexually transmitted disease intervention programs in developing countries.

Characteristics of patients with cytomegalovirus retinitis in the era of highly active antiretroviral therapy

Monitoring the Quality of Clinical Inquiry Projects

BackgroundOur knowledge regarding dementia epidemiology for American Indians and Alaska Natives (AI/ANs) is very limited. Longitudinal electronic health record (EHR) data available in the Indian Health Service (IHS) provides an opportunity to estimate dementia prevalence and incidence among AI/ANs at the national level. However, as with other EHR‐based studies, identifying dementia patients via clinical diagnostic codes likely underestimates the prevalence of dementia. Furthermore, longitudinal studies of dementia using EHR could be challenging if some prevalent cases of dementia cannot be identified at baseline, which might be mistakenly considered as incident cases in subsequent years.MethodsWe extracted data from the IHS National Data Warehouse and related EHR databases between fiscal year (FY) 2007‐2013. Adults were identified as having dementia if they had at least one qualifying ICD‐9 diagnostic code for all‐cause dementia. A total of 1,117 AI/AN adults who were 45+ years old were identified as dementia patients in FY2007. Among these patients, we evaluated the number of years needed to correctly classify them as prevalent cases using data after FY2007. We also examined the impact of different dementia definitions on our ability to identify these prevalent cases using FY2008‐2013 data.ResultsAmong the FY2007 dementia patients who used IHS services in FY2008, only 63.7% of them were identified as having dementia in FY2008. The remaining 36.3% prevalent cases might be classified as incident cases if they have a qualifying diagnostic code after FY2008. Even when we used a 5‐year window (FY2008‐FY2012), only 78.9% of the FY2007 dementia patients were identified as prevalent cases. Among dementia patients who were 65+ years old in FY2007, a 5‐year window correctly identified 87.7% of the FY2007 dementia patients as prevalent cases. Altering the definition of dementia by adding dementia medications or diagnostic codes for other types of cognitive disorders did not substantially change our ability in distinguishing prevalent vs. incident cases of dementia.ConclusionsIt is challenging to distinguish dementia prevalent and incident cases using EHR data from the IHS. The accuracy of estimating prevalence and incidence of dementia using this data source might be higher among older patients.

This paper describes the use of Research Electronic Data Capture (REDCap) to conduct one of the follow-up waves of the 2004 Pelotas birth cohort. The aim is to point out the advantages and limitations of using this electronic data capture environment to collect data and control every step of a longitudinal epidemiological research, specially in terms of time savings and data quality. We used REDCap as the main tool to support the conduction of a birth cohort follow-up. By exploiting several REDCap features, we managed to schedule assessments, collect data, and control the study workflow. To enhance data quality, we developed specific reports and field validations to depict inconsistencies in real time. Using REDCap it was possible to investigate more variables without significant increases on the data collection time, when comparing to a previous birth cohort follow-up. In addition, better data quality was achieved since negligible out of range errors and no validation or missing inconsistencies were identified after applying over 7,000 interviews. Adopting electronic data capture solutions, such as REDCap, in epidemiological research can bring several advantages over traditional paper-based data collection methods. In favor of improving their features, more research groups should migrate from paper to electronic-based epidemiological research.

What is next for occupational cancer epidemiology?

There are various initiatives ongoing to collect meaningful data on haemophilia A and B (HA and HB). At present, haemophilia medicinal products obtain marketing authorization (MA) after appropriate performance of clinical trials (CTs). However, the number of patients recruited for CTs, as required in the corresponding guidelines 1, 2, rarely reflects the number of patients needed to obtain statistical relevance. Additional data collection after MA is needed to complement the information on Safety and Efficacy (S&E). Registries might be a source to contribute to this information. For new and upcoming FVIII and FIX products, the Pharmacovigilance Risk Assessment Committee (PRAC) recommendations include enrolment in registries to obtain longitudinal real-life data. The collection of data in registries has the additional value of allowing comparison of products. It is of inherent importance to maximize the quality of registry data to improve the benefit to public health. Moreover, strengths and weaknesses of data collection systems such as regulatory CTs and registries need to be identified. Providing robust and sufficient clinical data on efficacy and safety to support MA for products used in rare diseases can be challenging. In general, the basic regulatory needs on the clinical trial concept for FVIII and FIX products are laid down in the corresponding European clinical guidelines 1, 2. The potential delay in market access for novel factor concentrates for the treatment of haemophilia patients in Europe has been criticized 3-8. Thus, in accordance with European Paediatric Regulation, timely access to novel products has to be balanced against the need to assess S&E of a product in previously treated patients (PTPs) below the age of 12 years. In addition, for novel factor concentrates, it is a requirement to investigate PUPs under controlled conditions in CTs. With the fast-growing number of novel factor concentrates that are applying for MA, it might be challenging to perform clinical trials in PUPs with HA and HB in countries where the products will be marketed. Data from registries might be able to complement information derived from CTs (Table 1). Standardized study conditions, including frequency and method of inhibitor testing and predefined case definitions Consistent with regulatory requirements Close monitoring of inhibitors High data quality Same parameters and data collected for all products All PUPs at participating centres included (no selection bias) Real-life data Large number of patients Long follow-up Comparison among products Exploration of treatment variations/differences in standard care that may influence outcome Selection of PUPs: only approximately 30–50% are able to enter CTs (inclusion and exclusion criteria, Good Clinical Practice criteria) Small sample size Open, uncontrolled study design No enrolment of patients with moderate or mild haemophilia Long duration of recruitment (no suitable PUPs if too many products entering market at same time) Limited value of single PUP study No direct comparison of products feasible Limited study period Unknown confounding factors Unknown confounding factors No standardized dataset yet No standardized time schedule for, e.g. inhibitor testing or blood sampling Missing data Changing treatment practice over time Parameters are defined now but must be able to answer questions in 10 years, including about unknown new side effects Switch to other FVIII products (challenging for statistical analysis) Frequently driven by independent investigators, and no direct access to data for regulators or industry For novel products, e.g. those developed for their long-acting properties, it is crucial to identify and mitigate new safety issues that might emerge once a product is extensively used postmarketing, and data from registries could be predestined to fill this gap. According to the European Commission expert group on rare diseases, a patient registry is an organized 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 serves one or more predetermined scientific, clinical or policy purposes 9. Independent haemophilia registries might be a potential source of S&E data for haemophilia products 10-14. Once a significant number of patients have been entered, registries are able to minimize selection bias and to collect long-term outcome data. A lack of transparency complicates the discussion on whether existing registries already provide the data that regulators are seeking 15. Even if the collected parameters are known, the quality of the data is often unknown. Educating and training patients and centres to improve compliance and adequate data monitoring, which should be determined in a harmonized international or at least European protocol, are needed to enhance data quality. Examples for information transparency of existing registries are the PedNet registry (pednet.nl), the FranceCoag registry (francecoag.org) and the UKHCDO registry (ukhcdo.org). Cohort studies are preferred above registries, optimization of patient numbers and preventing selection bias improves the quality of data. Harmonization of study protocols and CRF forms enables pooling of data from different registries. Therefore, it is necessary to obtain informed consent from patients and caregivers to share their data for specific requests on S&E. Patients might be anonymized using centre codes and unique identifiers that might be stored by an independent entity. Whether a legal European solution could be found in that respect should be investigated. National registries might be the starting points for a European collaboration, and in the future merged into a European umbrella registry. Collaboration of all stakeholders is needed and must be improved, and basic rules must be established for handling and analysing data. The definition of an adequate core dataset (an ‘interoperable model’) is seen as the way forward. The number of parameters must be balanced between the workload for the physician and the need for monitoring of confounders. Additionally, different datasets are required for PUPs and PTPs, and a method, such as use of unique identifiers, must be found to link the data of a PUP to the corresponding data when the PUP becomes a PTP. Collection of longitudinal data must be ensured so that each patient's status is adequately mirrored contributing to the knowledge on the demographic change in the haemophilia population. Furthermore, it should be possible to distinguish extended-half-life products from other products in registries, and therefore, a solution must be found to enter the type of product, even before a product receives MA. However, an adequate core dataset, adjustment for confounding factors and methods of analysis need to be established, harmonized and, finally, approved by regulators. Registries can contribute significantly to clinical knowledge and providing data on epidemiology, public health profiles, quality of life, consumption and pharmacoeconomic aspects. Specific, well-defined clinical, laboratory or genetic questions may be addressed providing guidance on future drug development. The data coming from registries and cohort studies should be robust and as complete as possible to allow for sound evaluation. Therefore, the objectives should be clear and established upfront. A strong argument for the additional effort needed for the establishment of a well-designed and well-managed registry is the possibility of increasing sample size and surveying an entire population to complement the information coming from regulatory CTs. In principle, registries can be an important source of data for the evaluation of inhibitor development in PUPs, allowing for comparison of different products. All patients who participate in CTs should be included in registries. How exactly this objective can be achieved and which legal avenues exist to do so must be discussed. Regarding the current paediatric investigation plans (PIPs) published on the European Medicines Agency (EMA) website, one might calculate that approximately 500 PUPs will be needed for CTs in the coming years. Data from those PUPs are invisible for the community as long as the CTs last. In addition, treatment information of PUPs which do not fit to CT inclusion criteria might also not become available. In this context, more collaboration and cooperation are urgently needed. The EMA organized a workshop on 01–02 July 2015, to investigate whether current European haemophilia registries can provide valuable data for regulatory purposes and to evaluate how to best capture data, building on existing expertise in this area. This workshop revealed that patient organizations support the inclusion of all haemophilia patients in a registry, despite the hurdles provided by data protection laws. Furthermore, the implementation of an international personal identifier system is encouraged, although establishing such a system would be challenging. During the EMA workshop, it was noted that financial and technical support is crucial, e.g. providing appropriate facilitation options, such as technical tools. However, this support needs to be balanced against the independence of registries. The accuracy and robustness of data entry requires a proactive review and control programme. A predefined adequate dataset might serve as a common denominator for existing or emerging registries and a funding system needs to be established to cover most important questions and financial hurdles, to comply with monitoring requirements and to assist in setting up centralized data entry processes. The following conclusions were, with regard to content, agreed upon by all stakeholders at the EMA workshop: The current regulatory and scientific situation regarding strategy and outcome of data collection is complex due to the legal background and different needs and interests of involved parties (Fig. 1). To obtain MA, the performance of CTs following a stepwise approach dedicated to pre- and postauthorization investigation is mandatory. The concept and conduct of CTs as well as data availability are tightly controlled via established procedures between applicants and regulatory authorities. The majority of haemophilia registries are run by physicians, consequently, outcome reporting and access to data vary depending on individual conditions. Based on European law, regulators do have access to the protocol, conduct and results of CTs but are not similarly involved in the structure, management, evaluation and data access of registries. To optimize and complement the information coming from regulatory CTs with longitudinal data coming from registries, a suitable strategy needs to be developed. Identification of, agreement on and approval of an adequate dataset serving as a common denominator for registries could be a starting point. Additionally, the possibility of anonymized data sharing with regulatory authorities upon request has to be established. The details of data sharing, access, timelines and reporting structures need to be developed in consultation with the involved parties and legal advisors. The way forward will be to develop a collaboration involving all stakeholders, in which regulators have a key role. The availability of data coming from CTs for a single product balanced with the accessibility of data derived from registries will have an impact on reconsideration of the EMA PUP approach. C.K., H.M.v d B., B. K-S. and A.H. wrote the paper. The authors stated that they had no interests which might be perceived as posing a conflict or bias.

We propose and study a fully efficient method to estimate associations of an exposure with disease incidence when both, incident cases and prevalent cases, i.e., individuals who were diagnosed with the disease at some prior time point and are alive at the time of sampling, are included in a case-control study. We extend the exponential tilting model for the relationship between exposure and case status to accommodate two case groups, and correct for the survival bias in the prevalent cases through a tilting term that depends on the parametric distribution of the backward time, i.e., the time from disease diagnosis to study enrollment. We construct an empirical likelihood that also incorporates the observed backward times for prevalent cases, obtain efficient estimates of odds ratio parameters that relate exposure to disease incidence and propose a likelihood ratio test for model parameters that has a standard chi-squared distribution. We quantify the changes in efficiency of association parameters when incident cases are supplemented with, or replaced by, prevalent cases in simulations. We illustrate our methods by estimating associations of single nucleotide polymorphisms (SNPs) with breast cancer incidence in a sample of controls, incident and prevalent cases from the U.S. Radiologic Technologists Health Study.