¹Institute of Medical Biometry and Epidemiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Text

Introduction: Diagnostic accuracy studies are conducted to evaluate a diagnostic test procedure based on its ability to detect or rule out a medical condition. Various issues can arise during the test that can influence the test procedure or lead to a biased or missing test decision. It is common practice to exclude cases with missing or inconclusive test results from the analysis [1]. This approach may lead to biased results. For example, excluding patients with inconclusive results may result in the selection of patients who are clearly healthy or at a more advanced stage of the disease, leading to an overestimation of sensitivity and specificity [2]. We refer to these issues, which may bias the accuracy of the test being evaluated, as interfering events.

We adapted the estimand framework from the ICH E9 (R1) addendum [3] to diagnostic studies. During the planning phase, the clinician and the statistician should jointly define the attributes of the estimand (index test, target condition, target population, accuracy measure and handling of interfering events) in accordance with the overarching study objective. The estimand defines WHAT is to be estimated. It is, therefore, important to identify possible interfering events and how to deal with them before the study starts. We suggest six different strategies: The diagnostic policy strategy, the indicative event strategy, the setback strategy, the hypothetical strategy, the while under monitoring strategy and the principal stratum strategy. It is possible to choose different strategies for different interfering events.

Methods: We investigated the characteristics and the corresponding clinical implications of these strategies. Using an analytical approach, we calculated the estimand for the various strategies at different scenarios. These included varying the proportion of independent interfering events and of interfering events dependent on either the index test or the disease state.

Results: The indicative event and setback strategies show the greatest difference to an estimation without an interfering event. Furthermore, depending on the strategy, the difference becomes relevant for a proportion of interfering events between 15% and 60%.

Conclusion: In summary, the different strategies change the accuracy of the index test. It is crucial to decide beforehand how to account for the interfering events to match the clinical question.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.

Literatur

[1] Shinkins B, Thompson M, Mallett S, Perera R. Diagnostic accuracy studies: how to report and analyse inconclusive test results. BMJ. 2013;346:f2778. DOI: 10.1136/bmj.f2778
[2] Kohn MA, Carpenter CR, Newman TB. Understanding the direction of bias in studies of diagnostic test accuracy. Acad Emerg Med. 2013;20(11):1194–1206. DOI: 10.1111/acem.12255
[3] International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH). ICH E9 (R1) addendum on estimands and sensitivity analysis in clinical trials to the guideline on statistical principles for clinical trials. European Medicines Agency; 2020 [updated 2020 Feb 17; cited 2025 Mar 21]. Available from: https://www.ema.europa.eu/en/documents/scientific-guideline/ich-e9-r1-addendum-estimands-sensitivity-analysis-clinical-trials-guideline-statistical-principles_en.pdf