gms | German Medical Science

Diagnostic accuracy studies serve to investigate, evaluate and assess diagnostic procedures. They differ from therapeutic studies in their endpoints. For example, to verify the accuracy of the diagnosis, the sensitivity and specificity are calculated as co-primary endpoints. Further, diagnostic accuracy studies distinguish in their design. A paired design is set up in which each patient is examined with both the gold standard and the experimental test, and possibly also with a comparator test [1], [2].

In diagnostic accuracy studies problems can arise, when incidents after the recruitment of the patients occur. This can lead to missing values or to a biased result of the diagnostic accuracy. These protocol violations are for example unblinding before the reference test is performed or starting treatment between both tests. These incidents are called intercurrent events.

In the ICH E9 (R1) addendum [3] estimands were introduced for randomized controlled trials to consider their intercurrent events. Their aim is to define the effect size of the clinical trial objective [4]. They consist of the attributes population, treatment, endpoint, population-level summary and intercurrent events [3]. The addendum outlines five different strategies to handle intercurrent events. These strategies are the treatment policy strategy, the hypothetical strategy, the composite variable strategy, the while on treatment strategy and the principal stratum strategy [3]. Thus, the intercurrent events could be treated differently in order to control their influence on the treatment effect.

Based on the attributes, the intercurrent events and types of bias an estimand framework for diagnostic accuracy studies was developed. It translates the concept of the estimand for therapeutic studies to diagnostic studies and conducts their different characteristic and events. For example, the treatment item is substituted by the diagnostic tests. This includes the specification of the index test and the description of the gold standard or reference test.

A literature review was conducted to search for different examples of diagnostic accuracy studies. Diverse types of tests and their attributes were identified. Furthermore, various types of bias were investigated to detect intercurrent events and different intercurrent events from the examples of diagnostic accuracy studies are described. New strategies as well as modified strategies were gathered in addition to the existing five strategies to handle intercurrent events.

The authors declare that they have no competing interests.

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