Research on a compound is usually built in stages: cell work (in vitro), then animal models, then, for compounds that reach that stage, regulated clinical trials. Each stage answers a narrower question. A result describes what was measured in that system under those conditions. It does not, by itself, describe any other system.
- Cell, animal and human studies answer different questions and are not interchangeable.
- A surrogate marker stands in for an outcome and has to be validated against it.
- Trial design features such as concealed allocation and blinding change how far a result can be trusted.
- An effect size says how big a difference is, not how reliable the study is.
- Read the model first: cells, animals or people.
- Then read what was measured, and whether it was the outcome itself or a stand-in.
- Then read how the study was run before reading the number.
Five terms, defined
In vitro
Work done on cells or tissue outside a living organism, usually in a dish or a tube. It isolates one mechanism cleanly, which also means it leaves out everything else a living system does.
Animal model
A study in a living animal chosen to stand in for a question about another species. How well a finding transfers between species is itself a research question (Hackam and Redelmeier, 2006), and reporting standards exist so that animal studies state their design clearly (Kilkenny et al., 2010).
Surrogate marker
A measurement used in place of the outcome that actually matters, because it is easier or faster to measure. Prentice (1989) proposed a strict test: a valid surrogate has to capture the relationship between the intervention and the true outcome. Fleming and DeMets (1996) reviewed cases where surrogates misled.
Placebo-controlled trial
A trial in which one group receives the thing under test and a comparison group receives an inactive substitute, so that the two can be compared. How the groups are assigned and whether those running the trial know who is in which group both matter. In an analysis of 250 controlled trials, trials with inadequately concealed allocation reported effect estimates exaggerated by 41%, and trials that were not double-blind by 17% (Schulz et al., 1995).
Effect size
A measure of how large a difference between groups is, as opposed to whether a difference was detected at all. Sullivan and Feinn (2012) set out why a p-value alone is not enough: a small difference can be statistically detectable and still be small.
What a single result does and does not show
- It shows what was measured, in the model used, under the stated conditions.
- It does not show that the same measurement would hold in a different species, a different preparation or a different setting.
- It does not show that a stand-in measurement tracks the outcome unless that link has been tested.
- It does not show that the study was run well. Allocation, blinding and reporting have to be read separately.
Reading a body of work with these terms
The same questions apply to any compound. For a worked example of sorting published work by model, see RU-58841: what types of studies have been published. For a different kind of evidence, one measured sample at a time, see what a Janoshik sample report shows.
Methodology and review
Written by the Norwood Labs Team. Each source below was looked up on PubMed on 17 November 2026 and its title, authors, journal, year and PMID were checked against the PubMed record. Figures quoted from a paper come from its PubMed abstract. Last reviewed 17 November 2026.
Research use only
Norwood Labs sells RU-58841 and pyrilutamide (KX-826) as unlicensed research compounds, strictly for laboratory research use. They are not medicines and are not for human consumption. This page explains research methods and makes no claim about any compound.