Start with the population
Find out whether the paper studied cells, animals, healthy volunteers, or patients with a specific diagnosis. Results in one setting cannot simply be carried into another. Small early studies are useful for finding signals, but they often cannot establish uncommon risks or durable benefit.
Check the comparator and the endpoint
A trial without a meaningful control group may confuse treatment effects with time, expectations, or other care. A biomarker is not automatically a clinical outcome. More receptor activity, a higher hormone concentration, or a change in a laboratory test does not by itself mean someone feels better or lives longer.
Read the size of the effect
Absolute differences, uncertainty intervals, duration, and withdrawals provide context that a percentage alone leaves out. Group averages can hide different individual responses. Look for adverse events and missing data as carefully as you look for the main efficacy result.
Trace the claim back to the source
Use the paper, registered trial, or current official product label. A review can provide context but should not replace the original source for a precise quantitative claim. Notice whether a commercial seller is citing research on a different molecule, formulation, route, or patient population.