How to read psychiatric study headlines

Was it animal or human? Mouse data doesn't transfer directly to clinical practice. Was it randomized or observational? Observational studies show association, not causation. How big was the sample? Single small studies are easy to overinterpret. Was it preregistered? Preregistered protocols reduce p-hacking risk.

Was the outcome measure clinically meaningful, such as symptom-scale change or remission rate, or a surrogate, such as a biomarker or brain imaging signal? Was the comparator a placebo, an active control, or treatment as usual? Each gives a different read.

What does the headline absolute risk reduction look like versus the relative risk reduction? Relative numbers always look bigger. Has the finding been replicated?