What are clinical diagnostic studies?
Planning of diagnostic studies
Diagnostic studies serve to investigate, evaluate, and assess diagnostic procedures in medicine, with diagnostic accuracy, usually expressed as sensitivity and specificity, being the primary focus. Currently, the effects and consequences of the diagnosis are also coming into focus: What is the effectiveness of a therapy chosen based on the diagnostic test? Health insurance providers, in particular, are keeping a close eye on this aspect.
Diagnostic studies are a particularly interesting and challenging type of clinical trial for two reasons. This may seem surprising at first, since they ultimately only involve evaluating a contingency table comparing two true patient states (reference standard: ill: D , not ill: D-) with two diagnostic test results (test result: positive: T , negative: T-). However, the complexity of planning and analyzing clinical diagnostic studies becomes clear when one considers the checklist of the STARD statements
(Standards for the Reporting of Diagnostic Accuracy Studies) illustrates this. Although primarily intended for publication, it is helpful to consider the checklist during the planning phase. This helps to define the correct study objectives and population, avoid biases, appropriately address measurement quality, and draw the right conclusions.
The interesting aspect of this type of study is initially the planning phase. It is crucial to accurately represent (i) the intended use, (ii) the clinical application setting, and (iii) the point in the diagnostic pathway. This is more difficult than one might think; particularly in the development of new diagnostics, it has become apparent that these aspects are sometimes neglected during development.
Diagnostic studies are divided into different phases. The phase classification according to Köbberling et al. has proven very useful here; see this. link.
Biases in diagnostic studies
A second important point concerns the potential biases that can arise during the execution and analysis of diagnostic studies. Spectrum bias is of primary concern here. In the case of a case-control design, i.e., the recruitment of patients based on their disease status, there is an unintentional exclusion of "unclear" cases, meaning the inclusion of only "clear" cases, and thus an underrepresentation of mild cases, low-stage diseases, and "gray area" cases. Among the healthy individuals, there will be a bias towards healthier, younger individuals, as well as patients/participants without age-related comorbidities. Consequently, the application scenario is not accurately reflected, and the diagnostic accuracy of the procedure is overestimated. In extreme cases, the non-disease group is chosen incorrectly: If the intention is to use a diagnostic tool to differentiate between two disease states (e.g., to differentiate between tumor and inflammation when symptoms are present), then the study should not include an examination of "sick" vs. "healthy".
The inadequate consideration of selecting suitable study populations and their composition is the reason for the failure of many initially promising biomarkers. The solution is to conduct the study as a cohort study, with the inclusion criterion being... required diagnosisto be used in cases of unknown disease status.
Biases are still frequently associated with the reference standard, i.e., the determination of a patient's group assignment (D , D-). The reference standard itself may, for example, have limited diagnostic accuracy. Or it may be determined differently for groups D and D-, potentially with varying degrees of accuracy.t.
ACOMED statistik specializes in the planning, execution, analysis, and evaluation of diagnostic studies. Contact us if you require biometric support for planning and analysis.
