Facilitating Prospective Registration of Diagnostic Accuracy Studies: A STARD Initiative

Although the introduction of prospective trial registration policies has been successful in reducing waste in research, diagnostic accuracy studies are rarely registered. We describe why diagnostic accuracy studies should be registered, and where and how this can be done. Advantages of registration include the identification of unpublished studies, prevention of selective outcome reporting, prevention of unnecessary duplication of research, collaboration between researchers, and linkage of study materials. In a survey among representatives of 16 major trial registries, such as ClinicalTrials.gov, ANZCTR (Australian New Zealand Clinical Trials Registry), and the UK-based ISRCTN registry (International Standard Randomised Controlled Trial Number), 13 responded, of which 8 (62%) indicated they always accept registration of diagnostic accuracy studies and 5 (38%) do so in some cases. However, all but one of them (92%) indicated that their registry currently does not provide specific guidance for registering diagnostic accuracy studies. A second survey among the 85 members of the STARD Group (Standards for Reporting Diagnostic Accuracy) resulted in the identification of 14 essential protocol items and was used for developing a guide on how these items can be registered in existing major trial registries. We propose that investigators responsible for diagnostic accuracy studies should register their study, before recruiting patients, in 1 of the existing major trial registries that are willing to host such studies. We also propose that governmental, research, and academic institutions that provide funding for and journals that publish diagnostic accuracy studies require such registration. Over the past decade, there have been growing concerns about a wide range of sources of avoidable waste in biomedical research (1). Examples include studies addressing low-priority clinical questions, study results that are unreported, and published study reports that are not as informative as they could be (1, 2). One approach to reduce research waste has been the introduction of prospective trial registration policies (3). In 2005, the International Committee of Medical Journal Editors (ICMJE) declared that its member journals would only consider reports of trials for publication if they had been registered in a publicly accessible trial registry prospectively, before enrollment of the first study participant (4). Subsequently, similar requirements have been implemented by several governmental organizations, funders, and academic institutions (2). These policies led to a large increase in registered studies, (5), and about half of recently published trials now report a registration number (6). Unfortunately, registration rates remain low for diagnostic accuracy studies, (7, 8) despite the compelling rationale for prospective registration that applies to this type of study design (9–12). One reason may be that registration policies have thus far mainly addressed randomized trials of interventions, whereas diagnostic accuracy studies, which compare medical tests against a clinical reference standard, are often considered to be observational. Another reason may be the unavailability of specific guidance for registration of these studies. In this article, we explained why we believe that there is a need to register diagnostic accuracy studies and where and how this can be done. We provide guidance for authors and readers based on 2 online surveys (Fig. 1). Diagnostic tests are indispensable in clinical practice. Physical examination, imaging techniques, laboratory tests, and other forms of medical testing inform clinicians about the likelihood that a tested patient has the suspected target disease or condition and guide subsequent decisions about further testing or treatment. Most diagnostic tests are not perfectly accurate in classifying patients as having the target condition or not, and some of those tested will have a false-positive or a false-negative test result (13). Diagnostic accuracy is typically evaluated by comparing test results with those of a reference standard in a series of patients suspected of having a target condition. The diagnostic accuracy of the test under investigation is then expressed in measures such as sensitivity and specificity, positive and negative predictive values, likelihood ratios, and area under the receiver operating characteristic curve (13). Every year, thousands of these diagnostic accuracy studies are being undertaken. Some are based on a retrospective review of chart data; many others prospectively collect data of patients who have specifically consented to data collection. Sometimes, evaluations of diagnostic accuracy are incorporated into larger overarching projects that have other non-diagnostic study objectives (7). There are numerous reasons why prospective registration of diagnostic accuracy studies is beneficial for those involved in conducting research as well as those who rely on the research findings. Full disclosure of all study materials, including the protocol, is increasingly considered to be an ethical obligation, not only toward individuals who participated in the study on the understanding that their data would contribute to scientific knowledge, but also toward future patients and public funders of research, including tax payers (14). The Declaration of Helsinki states: “Every research study involving human subjects must be registered in a publicly accessible database before recruitment of the first subject” (15). As described for trials of interventions, many initiated diagnostic accuracy studies are never published, whereas others take years to get published (8, 16–19). This may lead to reporting bias when summarizing the available evidence in systematic reviews, meta-analyses, and clinical practice guidelines (19). Such bias could result in invalid implementation of medical tests in clinical practice and suboptimal clinical decision-making, potentially threatening patient safety and efficient use of healthcare funds. Registration facilitates the identification of unpublished studies, thereby providing an opportunity to prevent this major source of research waste (20). If a registered record is available for the study, principal investigators can be contacted to obtain study results and identify reasons for not publishing them. By helping the inclusion of unpublished studies in systematic reviews, registration could also increase precision of summary estimates in meta-analyses and power for assessing sources of heterogeneity. An analysis of 117 systematic reviews assessing medical interventions published in 2012–2013 identified that 35% contained a screen of trial registries for unpublished trials (21). Selective reporting of pre-defined primary outcomes appears to be common among published diagnostic accuracy studies (16). Such selective publication may provide a distorted view of diagnostic accuracy, leading to erroneous impressions of the usefulness of the test in clinical practice. If studies are prospectively registered with pre-specified outcomes, peer reviewers and journal editors will have the opportunity to identify and prevent selective reporting in study reports submitted for publication. In a survey among peer reviewers, it was shown that one-third examined corresponding trial registries and reported discrepancies with the peer-reviewed manuscript to journal editors (22). Unreported outcomes can be identified, and corresponding results can be obtained by contacting the principal investigators. The registered record provides a more comprehensive knowledge of the methods, allows for a comparison between what was originally proposed and what was actually done, facilitates the identification of potential reporting biases as registered, and allows for reported outcomes to be easily compared. Unpublished or ongoing diagnostic accuracy studies may be unnecessarily duplicated if other researchers are not aware of their existence (10). Before initiating a new study, researchers and funders could search trial registries to identify ongoing or completed studies with similar objectives. This may help them appreciate the research gaps that further studies could address. Such practices may lead to a more efficient allocation of research funds and efforts. In addition, unnecessary exposure of study participants to potentially harmful tests or other interventions can be prevented. Reports of diagnostic accuracy studies are often incomplete in their description of essential study design features, making it difficult for readers to assess the risk of bias and the applicability of the study findings (23, 24). In such cases, corresponding completed registered records may be useful as an additional source of information on study methods, as has been shown for trials of therapeutic interventions (25). By facilitating the identification of studies that are about to start or are ongoing, registration has the potential to improve collaborations between researchers that are operational in similar fields of research. Patients are among the main users of trial registries (26). Those willing to participate in scientific research, and clinicians looking for relevant research projects for their patients to participate in, will have the opportunity to identify and participate in eligible studies if these are registered. Diagnostic accuracy studies are particularly sensitive to bias, which may jeopardize the trustworthiness of study results (27). Registration can improve protocol development, as it may encourage researchers to carefully consider a standardized set of critical protocol elements that are likely to affect the results of their studies, thereby improving the overall methodological quality of the study. Increasingly, individual studies lead to multiple reports (28). Researchers can publish the trial protocol and report results in conference abstracts, in applications with regulatory bodies, in patient information leaflets, on websites, and in one or more full publications in biomedical journals. These reports often focus on different research questions and outcomes, sometimes based on a subset of the total study group. Linkage of reports that originate from the same study can be difficult (28). This is challenging for researchers who want to identify all the relevant evidence for a systematic review and may come across publications with potentially overlapping participants. Linkage will be highly facilitated if all diagnostic accuracy studies have a unique registration number that is included in any report generated from the study. In contrast to randomized trials, as performed for pharmaceuticals, diagnostic accuracy studies can also be retrospective in nature, relying on chart records or previously collected data. Some of the benefits of registration may not fully apply to such retrospective studies, such as the potential for collaboration, but many other advantages still apply, if the authors register their study before they extract the data and start performing their analyses. No central trial register exists that focuses specifically on diagnostic accuracy studies, and we do not see the need to start a separate register. We believe that registration of diagnostic accuracy studies within existing major trial registries is a better option than developing a registry specifically designed for such studies, because (a) these registries are widely known and used in the scientific and clinical community; (b) some already contain numerous diagnostic accuracy studies; (c) a limited number of large, well-designed registries has practical advantages over many small ones that focus on specific types of research; and (d) diagnostic accuracy studies are often part of broader studies that also contain nondiagnostic outcomes. The major existing trial registries are ClinicalTrials.gov, a service of the U.S. National Institutes of Health, and 15 primary registries in the World Health Organization (WHO)15 Registry Network (see http://www.who.int/ictrp/network/primary/en/). These are considered as “Primary Registries” because they “meet specific criteria for content, quality and validity, accessibility, unique identification, technical capacity and administration”. At present, ClinicalTrials.gov and the 15 primary registries in the WHO Registry Network primarily focus on the registration of trials of therapeutic interventions, although some welcome studies with other designs as well. We performed an online survey among representatives of these 16 registries to assess their policy toward registration of diagnostic accuracy studies (Fig. 1) [The survey details and results are provided in Data Supplement 1 and 2 that accompanies the online version of this report at http://www.clinchem.org/content/vol63/issue8]. In total, 13 registry representatives (81%) responded; of these, 8 (62%) indicated they always accept registration of diagnostic accuracy studies and 5 (38%) do so in some cases. A list of major trial registries where diagnostic accuracy studies can be registered is provided in Table 1. Major trial registries where diagnostic accuracy studies can be registered. ICMJE = International Committee of Medical Journal Editors; RCT = randomized clinical trial; CCT = controlled clinical trial. Major trial registries where diagnostic accuracy studies can be registered. ICMJE = International Committee of Medical Journal Editors; RCT = randomized clinical trial; CCT = controlled clinical trial. In our survey of trial registries, 12 of 13 (92%) respondents indicated that their registry currently does not provide specific guidance for registering diagnostic accuracy studies, with the UK-based ISRCTN registry (International Standard Randomised Controlled Trial Number) being the exception (29). Lack of explicit instructions for registering diagnostic accuracy studies may be an important reason why researchers believe that it is difficult or impossible to register these studies in existing trial registries. Yet, registering diagnostic accuracy studies can be as straightforward as registering a randomized trial of a new drug. The WHO, which has encouraged trial registration for years, has proposed a “Trial Registration Data Set”, containing 20 items that cover essential study protocol information that must appear in a registered record for a trial to be considered fully registered (see http://www.who.int/ictrp/network/trds/en/). Each Primary Registry in the WHO Registry Network mentioned in Table 1 requires registered records to at least contain these 20 items. The format of ClinicalTrials.gov is slightly different, although there is a major overlap between items that need to be registered here as well. Because many items on WHO's Trial Registration Data Set apply to any type of research involving humans, there should be no barriers to registration of diagnostic accuracy studies. Items unique to diagnostic accuracy studies can be incorporated in the Trial Registration Data Set. To identify such items, we performed a second online survey, among the members of the STARD (Standards for Reporting Diagnostic Accuracy) group (Fig. 1). This group includes 85 experts in the field of diagnostic research.‡ (23, 24, 30) [see online Data Supplements 3 and 4 for survey and results]. In brief, we built a list of 20 items potentially relevant when registering a diagnostic accuracy study that are not currently covered by WHO's Trial Registration Data Set. This list was established on the basis of STARD 2015 reporting guideline (23, 24, 30). The items were proposed to survey respondents, and they could indicate whether they felt that the item should or should not be included in the registered record of a diagnostic accuracy study. Items that were tagged as relevant by two-thirds or more of the respondents were included in our final list. Respondents also had the opportunity to suggest items that were not included in the list. The inclusion of these items was decided based on discussion by the authors of this manuscript, consisting of 5 trial registry representatives (LMA, HF, KEH, HL, LH), 3 current or former editors-in-chief of journals that publish diagnostic accuracy studies (VB, DM, NR), and 5 persons with broad experience in performing diagnostic accuracy studies and systematic reviews thereof (DAK, JFC, CAG, MDM, PMB). In total, 71 STARD group members (84%) responded. On the basis of the survey results, we recommend that 14 items that are currently not (explicitly) covered by WHO's Trial Registration Data Set be incorporated for diagnostic accuracy studies. Table 2 provides WHO's Trial Registration Data Set, a proposal of where and how the 14 newly identified items can be mentioned when registering diagnostic accuracy studies, and examples from existing registered records. This applies to 7 of the 20 items within WHO's Trial Registration Data Set; the current definition of the 13 other items equally applies to diagnostic accuracy studies and does not need any modification. WHO's Trial Registration Data Set (TRDS), modified for diagnostic accuracy studies. WHO's Trial Registration Data Set is available at http://www.who.int/ictrp/network/trds/en/. WHO's Trial Registration Data Set also contains several required or optional data fields related to the “study type” (item 15), where only fixed options can be selected (e.g., whether the study is “interventional” or “observational”). Suggestions on how to address these for diagnostic accuracy studies are provided in Supplemental Data 5. WHO's Trial Registration Data Set (TRDS), modified for diagnostic accuracy studies. WHO's Trial Registration Data Set is available at http://www.who.int/ictrp/network/trds/en/. WHO's Trial Registration Data Set also contains several required or optional data fields related to the “study type” (item 15), where only fixed options can be selected (e.g., whether the study is “interventional” or “observational”). Suggestions on how to address these for diagnostic accuracy studies are provided in Supplemental Data 5. The Trial Registration Data Set also contains several required or optional data fields related to the “study type” (item 15), where only fixed options can be selected: for example, whether the study is “interventional” or “observational,” or whether the “method of allocation” is “randomized” or “nonrandomized.” To improve uniformity across registered records, we provide suggestions on how to address these required data fields for diagnostic accuracy studies in online Data Supplement 5. With the list of 14 recommended items identified in this project, we evaluated the content of recently registered diagnostic accuracy studies. Using the search term “diagnostic accuracy,” we searched WHO's International Clinical Trials Registry Platform Search Portal on November 1, 2016. This engine searches through trial registration data provided by the Primary Registries and ClinicalTrials.gov. The 30 most recently registered records of diagnostic accuracy studies were selected. For each record, 1 author (DAK) assessed if the newly identified recommended items listed in Table 2 were reported and in which data field; difficult cases were discussed with a second author (PMB). A summary of results is provided in Table 3, with results as per the registered record in online Data Supplement 6. Some recommended items were almost always registered: for example, identification of the record as a diagnostic accuracy study (item 1; 90%), the target condition (item 2; 97%), and the index test whose diagnostic accuracy is under evaluation (item 3; 100%). Other recommended items were much less frequently registered: for example, information available to the performers or readers of the index test (item 5; 47%), information available to the assessors of the reference standard (item 6; 40%), the setting in which patients will be recruited (item 8; 50%), how data will be collected (item 10; 47%), and how participants will be sampled (item 11; 37%). We found that 6 records (20%) contained at least 11 of the 14 recommended items, illustrating that it should be possible to register all the information proposed in Table 2 in the existing format of trial registers. of registered records of diagnostic accuracy studies = 30). of registered records of diagnostic accuracy studies = 30). With this we to several mentioned in the We on researchers to prospectively register their diagnostic accuracy studies in one of the major trial registries. We suggest that the guidance proposed will help to that registered records are fully informative and across registries. By registered records of diagnostic accuracy studies, trial registries a We on existing major trial registries to that registered records of diagnostic accuracy studies are Registries could the proposed guidance into their data field by such diagnostic accuracy studies, report A increase in the number of clinical trials was the International Committee of Medical Journal Editors implemented its clinical trial registration policy we on journals and to encourage authors in registering their diagnostic accuracy studies. This could be by addressing registration in instructions to a such recommend researchers to prospectively register their diagnostic accuracy studies in a publicly accessible trial registry and to report the corresponding registration number in the final study or by making the reporting of a trial registration field in the and in the reporting of the funding or researchers could recommend or require registration of diagnostic accuracy studies as a condition for full funding or ethical The an to improving through research and with for example, requires all clinical trials to be registered. At some clinical trials must have a registration number before they can get (2). of can also policies could apply to diagnostic accuracy studies. registration provides major to prevent research The and guidance provided here can be used to prospective registration of diagnostic accuracy studies and improve registration We propose that investigators responsible for diagnostic accuracy studies register their study, before recruiting patients, in 1 of the 13 existing major trial registries that are willing to host such studies. We also propose that existing major trial registries consider the proposed items, as these are unique to diagnostic accuracy studies, and that funding and journals publishing diagnostic accuracy studies require and prospective registration of such studies. World Health Organization for Reporting Diagnostic We the for in the first survey, which to identify the policy of major trial registries toward registration of diagnostic accuracy Clinical Trials Registry), Clinical Trial Registry), Registry of Clinical Clinical Trials Clinical Trials Registry of Clinical Clinical Trial Registry), Clinical Trials Registry), Clinical Trials Registry), and We the members of the STARD 2015 group for in the second survey, which to identify important protocol items that need to be registered for diagnostic accuracy studies.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.