Reading Early-Stage Research: Animal Studies, Small Trials and Preprints

Why early results need care

New health ideas usually start in a laboratory, and most never reach the clinic. This page covers the four kinds of early evidence you will see most often: cell and animal studies, small trials, preprints and stand-in measures. It is education, not personal medical advice. The evidence labels it uses are explained in how we label evidence.

Cell and animal studies

These are preclinical studies: in vitro, meaning in the laboratory outside the body, and in vivo, meaning in the body. [1][2] They help decide whether something is worth testing in people and what might be harmful. [3] The FDA says preclinical research is not a substitute for studies of how a drug interacts with the human body. [4]

A systematic review looked at interventions where human trials had already given clear answers. Corticosteroids helped in animal models of head injury but not in clinical trials, and tirilazad improved animal stroke outcomes but was associated with a worse outcome in patients. [5] So the right wording for a mouse result is "in mice", never "in people". Our label for this is preclinical (cells/animals only). The full comparison is in cell studies vs animal studies vs human trials.

Small trials

Small early trials are useful for safety and for generating ideas, but they are easy to overread. The FDA notes that Phase 2 studies, typically a few hundred patients, are not large enough to show whether a drug will be beneficial. [4] Across 93 meta-analyses, treatment effects averaged 48% larger in trials with fewer than 50 patients than in trials of 1,000 or more. [6] In a review of 49 highly cited clinical studies, 16% were later contradicted and another 16% had found stronger effects than later studies, and among randomized trials those with contradicted or stronger effects were smaller. [7]

Questions to ask: How many people took part? Was there a comparison group, such as a placebo (an inactive look-alike) or a standard treatment? [8] Were people assigned at random? Did others repeat it? See replication and reproducibility.

Preprints

A preprint is a report shared before journal peer review. [9] In a study of bioRxiv and medRxiv preprints that were later published, the conclusions of 7.2% of non-COVID-19 abstracts and 17.2% of COVID-19 abstracts changed in a discrete way by publication, though most changes did not qualitatively change the paper's conclusions. [9] Treat a preprint as a draft: read it for ideas, but do not base health decisions on it. See preprints vs peer-reviewed papers.

Stand-in measures

A study may report a blood level or scan result rather than how people feel or live. NCI notes that such surrogate endpoints are not always true indicators of how well a treatment works. [10] In oncology, more than half of 65 reported links between such stand-ins and survival were of low strength. [11] Prefer studies that measure outcomes you would notice.

A quick checklist

• Who was studied: cells, animals or people like you?
• How many, for how long, compared with what?
• Was the outcome something people feel or experience, or only a lab marker?
• Was it published after review, and has anyone repeated it?
• Who paid for it, and who is selling something?
• Does a regulator or a systematic review say anything on it?

A regulator's page or a systematic review carries more weight than a single study. See research and safety sources and evaluating safety claims.