Statistical inference is essential for evaluating empirical support for medical research findings. Still, much medical research is methodologically poor. The problem is not that medical research fails to employ statistical methods; on the contrary, research reports usually include large numbers of p-values. The poor quality is more related to the misconception that statistical methodology is only helpful for producing p-values.
Assuming that physicians conduct most medical research, the problem can perhaps be explained by a conflict between fundamental differences in treating patients and doing scientific research. Statistical inferences about an “average patient” may not help when treating a specific patient, as patients respond differently to treatment due to sex, age, ethnicity, environment, etc. The best treatment for most patients is not necessarily the best for a specific one. A physician may have to rely on an accepted treatment strategy that is assumed to work even if it is not based on evidence. Such treatment strategies can often be described as dogma, opinions that must not be questioned. Failing to comply with dogma can perhaps even lead to litigation.
Statistical significance can probably play a role in reports that aim to support or extend dogma. However, unquestionable opinions have no place in scientific research based on questioning opinions and making critical evaluations; the search for truth is paramount, and evidence is central. For this purpose, statistical methodology has more to offer.
