Informing diagnostic recommendations by test accuracy - the case of POS

Article type
Authors
Rummer A1, Kasper J2, Sturtz S1, Scheibler F1
1Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG) , Germany
2Department of Health and Caring Sciences, Faculty of Health Sciences, University of Tromsø, Norway
Abstract
Background: While the GRADE working group translates information given in 2x2 tables of diagnostic accuracy studies into assumptions of possible benefits and harms, the Cochrane Diagnostic Test Accuracy Working Group more rigorously states in its Handbook: that in principle, recommendations regarding use of diagnostic tests based on accuracy data only should be avoided (Ch 11.9.1). However, exceptions are described for contexts within which certain assumptions regarding the relation of test accuracy and benefit exist. We report on our reasoning in the process of drawing recommendations from an IQWiG benefit assessment of neonatal pulse oximetry screening (POS) for the detection of critical congenital heart disease (cCHD).
Methods: Natural history of cCHD is detrimental in most cases within a relatively short period of time. Per definition, earlier detection (and intervention) in the maternity clinic would gain better outcomes than detection by symptoms after hospital discharge, which most likely leads to emergency admissions and interventions. Therefore, we based our recommendations on 3 assumptions: 1. Earlier detection and intervention are better 2. Additional detection is always a benefit 3. False positive results have only minor adverse effects, as any intervention would require additional echocardiography.
Results: In our systematic review, we included 1 CCT and 5 DTA. The CCT supports the hypothesis, that POS as add-on to routine clinical screening reduces morbidity in newborns with cCHD (severe acidosis: OR: 0.268 (0.110; 0.654), P = 0.003). The DTA studies reported additional identification of cCHD. The NNS ranged from 421 to 7100, PPV from 25.9 to 75% (corresponding to the same amount of fp results).
Discussion: Due to high potential of bias and lacking mortality data, we considered the CCT not sufficiently indicating a benefit of POS. Consistent results of 5 DTAs, however, seem to give enough reason for a positive conclusion. We are in line with recommendations of other HTA reports, none of them reporting on underlying assumptions.
Conclusion: In explicitly justified cases practical recommendations derived from DTA studies might be reasonable.