Some problems with non-inferiority tests in psychotherapy research: psychodynamic therapies as an example.

In virtually every field of medicine, non-inferiority trials and meta-analyses with non-inferiority conclusions are increasingly common. This non-inferiority approach has been frequently used by a group of authors favoring psychodynamic therapies (PDTs), concluding that PDTs are just as effective as cognitive-behavioral therapies (CBT). We focus on these examples to exemplify some problems associated with non-inferiority tests of psychological treatments, although the problems also apply to psychopharmacotherapy research, CBT research, and others. We conclude that non-inferiority trials have specific risks of different types of validity problems, usually favoring an (erroneous) non-inferiority conclusion. Non-inferiority trials require the definition of non-inferiority margins, and currently used thresholds have a tendency to be inflationary, not protecting sufficiently against degradation. The use of non-inferiority approaches can lead to the astonishing result that one single analysis can suggest both, superiority of the comparator (here: CBT) and non-inferiority of the other treatment (here PDT) at the same time. We provide recommendations how to improve the quality of non-inferiority trials, and we recommend to consider them among other criteria when evaluating manuscripts examining non-inferiority trials. If psychotherapeutic families (such as PDT and CBT) differ on the number of investigating trials, and in the fields of clinical applications, and in other validity aspects mentioned above, conclusions about their general non-inferiority are no more than a best guess, typically expressing the favored approach of the lead author.