Should data from case-control studies be included in systematic reviews alongside diagnostic cohort studies?

Tags: Poster
Burch J, Westwood M, Soares-Weiser K

Background: Over estimation of accuracy by case-control studies has been widely discussed.1 Evaluating data from cohort and case-control studies in one review is problematic. We undertook a systematic review comparing the accuracy of faecal occult blood tests (FOBTs), which included both designs. Few direct comparisons were available; indirect comparisons were attempted.

Objectives: To assess the impact of including case-control studies upon estimates of diagnostic accuracy, and validity of indirect comparisons.

Methods: We conducted a systematic review comparing diagnostic accuracy of FOBTs. With few direct comparisons, indirect

comparisons between diagnostic accuracy studies were attempted. Study design was a major consideration.

Results: Twenty-three of 33 studies evaluating guaiac FOBTs, and 17 of 35 evaluating immunochemical FOBTs were cohort studies. Heterogeneous data ruled out pooling, exacerbating the difficulty of comparing FOBTs. The impact of study design was primarily on sensitivity. For the detection of all neoplasms, the sensitivity of Haemoccult ranged from 4.3% to 45% in cohort studies, whereas casecontrol studies reported sensitivities between 50% and 71%. Two immunochemical FOBTs were evaluated in cohort and case-control studies. All other immunochemical FOBTS were only evaluated in case-control studies. The cohort study evaluating Immudia HemSp reported a sensitivity of 63%, sensitivities from case-control studies ranged from 43% to 98%. Two cohort studies evaluating OC Light reported sensitivities of 5.4% and 18%, sensitivities from case-control studies ranged from 39% to 69%.

Conclusions: The data presented consolidate the view that the case-control design over estimates accuracy. Diagnostic accuracy studies should preferentially use a cohort design. Those conducting systematic reviews should be wary of including both cohort and case-control studies.

References

1. Lijmer JG, Mol BW, Heisterkamp S, Bonsel GJ, Prins MH, van der Meulen JH, et al. Empirical evidence of design-related bias in studies of diagnostic tests. JAMA 1999; 282(11):1061-6.