How to Read Published Peptide Research: Study Types and Evidence Levels

Start with the question a study can answer#

Every study design is built to answer a particular kind of question. A cell-culture experiment can show that a model peptide binds a receptor in a dish. It cannot show what happens in a whole organism. A large randomized trial can test whether an intervention changes a defined outcome in a defined population. It cannot tell you why at the molecular level. Reading research well starts with matching the claim being made to the design that produced it.

Evidence-based medicine groups designs into levels of evidence. In the Oxford Centre for Evidence-Based Medicine (CEBM) scheme for questions about interventions, systematic reviews of randomized controlled trials sit at the top, and bench research, physiological reasoning and expert opinion without explicit critical appraisal sit at the bottom.^[1] The levels are a ranking of how well a design protects against bias for that question, not a verdict on any single paper.

Study types and what each can show#

The table below summarizes the main designs, from the laboratory bench to evidence syntheses. PubMed tags many articles with a publication type that matches these categories, which makes them searchable; the definitions quoted in the notes come from the National Library of Medicine's scope notes.^[2]

Regulatory approval is a separate thing#

Approval is not the next rung on the evidence ladder; it is a decision by a regulator about a specific product. In the United States, the FDA describes clinical research as a series of phases that grow in size, from early studies in tens of people to pivotal Phase 3 studies involving roughly 300 to 3,000 participants, followed by Phase 4 monitoring after approval.^[5] The FDA also notes that preclinical research is not a substitute for studies in people.^[5]

A marketing application to the FDA is meant to tell the full story of a drug, from preclinical data to Phase 3 data, along with proposed labeling and other required information, and a review team then makes a recommendation on approval.^[6] A single journal article, however well designed, is one input to that process. The phrase "research use only" on a material means it is not part of that process at all.

Preprints, peer review and publication status#

Peer review is the editorial step in which independent experts assess a manuscript before a journal publishes it. A preprint is, in the National Library of Medicine's definition, a scientific manuscript made available prior to peer review.^[2] Preprints let results circulate quickly, but they have not been through that check, and their conclusions sometimes change before formal publication. Peer review has limits as well: reviewers assess the manuscript they are sent and do not normally repeat the experiments.

Article type matters too. A narrative review summarizes literature at the authors' discretion, while a systematic review attempts to identify, appraise and synthesize all relevant studies using a defined method.^[2] Commentaries, letters and conference abstracts are shorter formats and often lack the detail needed to evaluate a claim.

Reading past the abstract#

An abstract is a summary written by the authors, and it emphasizes what they consider most important. The methods section is where you learn what was actually done. Before accepting a conclusion, check who or what was studied, what was compared, how outcomes were measured, and whether the analysis was planned in advance.

Sample size, controls and blinding#

Small studies produce unstable estimates. The Cochrane Handbook notes that larger effects are often observed in smaller trials, one possible explanation being that small studies are more likely to be published when they show large, statistically significant effects.^[3] A control group shows what happens without the intervention, and randomization keeps the groups comparable. Blinding keeps participants, investigators or outcome assessors from knowing who received what. The Handbook reports that inadequate or unclear allocation methods, and a lack of blinded outcome assessment, are associated on average with exaggerated effect estimates, especially for subjectively measured outcomes.^[3]

Effect sizes versus p-values#

A p-value describes how compatible the data are with a specified statistical model, usually one assuming no difference. The American Statistical Association's statement on p-values stresses that a p-value does not measure the probability that a hypothesis is true, and that it does not measure the size of an effect or the importance of a result.^[4] Look instead for the effect size (how large the difference is, in real units) and its confidence interval (the range of values compatible with the data). For example, a tiny difference can reach p < 0.05 in a very large sample, while a large difference in a small sample may not.

Funding and conflicts of interest#

Check the funding and disclosure statements. The Cochrane Handbook describes conflicts of interest as a possible source of bias through several mechanisms, and cites evidence that industry-funded trials were more likely to report positive conclusions and statistically significant results.^[3] A disclosed interest does not make a result wrong, but it is a reason to read the methods more closely.

Retractions and corrections#

Papers can be corrected or withdrawn after publication. PubMed labels a withdrawn article as a Retracted Publication and indexes the notice itself as a Retraction of Publication, so a retraction should be visible on the article record.^[2] Before relying on a paper, especially an older one found through a secondary source, check its PubMed record or the journal page for a retraction, correction or expression of concern.

Why cell and animal results often do not translate#

A result in cells or animals is a hypothesis about humans, not a demonstration. In one analysis of highly cited animal studies, about a third (37%) were replicated in human randomized trials, and few of the animal studies used random allocation.^[7] An earlier analysis in the BMJ argued that much animal research intended to inform human medicine is poorly conducted and rarely evaluated systematically against clinical results.^[8]

Common reasons include differences between species in metabolism and physiology, concentrations or exposures that do not resemble anything achievable in people, short study durations, small samples, and the same design weaknesses described above. Non-reporting bias can widen the gap: studies with statistically significant results are more likely to be published than those without.^[3]

A short reading checklist#

1. Identify the study type and the question it was designed to answer.
2. Read the methods: population or model, comparison, outcomes, and whether the analysis was pre-specified.
3. Check sample size, randomization and blinding.
4. Look for effect sizes and confidence intervals, not only p-values.
5. Read the funding and conflict-of-interest statements.
6. Confirm publication status: preprint or peer-reviewed, and whether it has been corrected or retracted.
7. Ask whether the result has been replicated, and whether a systematic review exists.

The same habits apply to analytical documents. For how to read a laboratory report rather than a journal article, see how to read a certificate of analysis and common purity report misreadings; unfamiliar terms are defined in the research peptide glossary.