Studies may show results that are statistically significant but clinically irrelevant. If a study uses a large enough sample, small effects of questionable relevance may be identified as significant factors in an outcome. The reason is that various independent factors with little relevance, autonomous trends as well as random fluctuations can lead to small effects, which can show up as statistically significant if the sample is large and the confidence interval small.
Patients often present with a complex mix of psychological, physiological, and social problems as well as other comorbidities, which can be difficult to separate statistically. There may be correlations of these symptoms with a factor which has not been identified yet, leading to an apparent direct correlation among them.
The cause and effect may not be direct if there are unknown variables that play a role or if either the cause or the effect are not reliably measurable. Especially in mental health, measuring symptoms can be difficult, and may vary from one research team to the next, even if they use the same questionnaires, for example.
As people often engage in multiple therapies, pharmacological and non-pharmacological, accounting for them and making individual cause and effect correlations and predictions may be difficult.
Commercial interests may also play a role by putting the focus on something that is less relevant, while ignoring a relevant factor.