Systematic underestimation of treatment effects as a result of diagnostic test inaccuracy: implications for the interpretation and design of thromboprophylaxis trials.

Many clinical trials that have assessed strategies for the prevention of deep vein thrombosis have employed diagnostic tests that are less accurate than venography. The correct interpretation of these trials has been the subject of considerable debate. This paper attempts to quantify the likely effects of the use of inaccurate diagnostic tests (in particular fibrinogen scanning) on the validity and precision of estimates of treatment effects. It is shown that, if there is no difference in the rate of misclassification of deep vein thrombosis between study and control groups, then treatment effects will tend to be systematically underestimated in trials that only use inaccurate diagnostic tests. However, for trials in which there are no false positive diagnoses, such as those employing venography, the estimate of relative treatment effects will be unbiased. This phenomenon is demonstrated empirically by data from randomised trials of antiplatelet therapy. The implications of these findings are two-fold. First, so long as the accuracy of the test is the same in study and control groups, positive trial results cannot be ascribed to diagnostic test inaccuracy. Second, unbiased estimates of relative treatment effects can be obtained by using cheap, practicable and non-invasive screening tests that are supplemented by venographic assessment of positive test results. This would facilitate the conduct of much larger randomised trials than has generally been possible when mandatory venography was used; if future trials randomised some thousands of patients and employed confirmatory venography then they would provide both unbiased and precise estimates of relative treatment effects.