Based on the collected data, investigators then typically calculate a test statistic and the corresponding p value. The sample size of the trial is then calculated based on this pre-specified hypothesis and on the desired magnitude of the type I and type II errors. Most RCTs have as a pre-specified hypothesis that the intervention under evaluation improves health outcomes, which is statistically tested against the null hypothesis of no effect (Table 1). RCTs typically evaluate the benefits and harms of pharmaceuticals (and other interventions) by comparing health outcomes between one group of participants who receive the drug to be evaluated, and a second group of participants who receive a placebo or an alternative drug. The randomized controlled trial (RCT) has become the undisputed cornerstone of evidence-based health care. Minimally acceptable criteria for test performance should form the basis for hypothesis formulation and sample size calculations in diagnostic accuracy studies. Based on the role of the index test in the clinical pathway and the downstream consequences of test results, the consequences of test misclassifications can be weighed, to arrive at minimally acceptable criteria for pre-defined test performance: levels of sensitivity and specificity that would justify the test’s intended use.
In this paper, we propose a strategy for defining meaningful hypotheses in diagnostic accuracy studies.
In diagnostic accuracy studies, study hypotheses are rarely pre-defined and sample size calculations are usually not performed, which may jeopardize scientific rigor and can lead to over-interpretation or “spin” of study findings. Most randomized controlled trials evaluating medical interventions have a pre-specified hypothesis, which is statistically tested against the null hypothesis of no effect.
0 kommentar(er)
