When a group submits a study to a journal, the editor sends it to two or three experts in the field, usually anonymous. They review the design, methods, analysis and whether the conclusions hold up, and recommend accepting it, asking for changes or rejecting it.
It is a valuable filter, but it has limits. Reviewers do not repeat the experiments and do not usually see the raw data, so they do not catch every error or fraud. Their rigour varies widely between journals, and some low-quality outlets accept almost any work.
It is also worth distinguishing a peer-reviewed paper from a preprint, which is published before any review. Databases such as PubMed index papers from biomedical journals, but being indexed is not the same as quality.
A result becomes reliable when it is reproduced in independent studies, not just when it is published. Peer review is the first step, not the last.
Related terms
Definition for educational and scientific purposes. It is not medical advice or a recommendation for use. NeoPeptidos products are sold labelled for research use only (RUO).