Research
How to read a study run by the seller
Industry funds most clinical research, and sponsored trials are not run worse than other trials. The measured difference sits between a study's results and the sentence its authors wrote about them.
Who paid is a fact, not a verdict
Most clinical trials are funded by the company that makes the product. Refusing to read those studies would mean refusing to read most of the evidence that exists for most medicines.
The useful question is not whether to trust a sponsored study. It is where sponsorship actually shows up in one, so a reader knows which part of the paper to slow down on.
The answer has been measured, and it is more specific than most people expect.
What the research on sponsorship found
A Cochrane methodology review pooled 75 papers comparing industry-sponsored drug and device studies with studies funded by other sources.
Industry-sponsored studies more often had favorable efficacy results, with a pooled risk ratio of 1.27. They more often had favorable conclusions, with a risk ratio of 1.34.
Then comes the part that changes how you read a paper. On the standard measures of study quality the reviewers found no difference: sequence generation, allocation concealment, follow-up and selective outcome reporting were comparable between the two groups. On one measure the industry studies did better, with a risk ratio of 1.25 for having low risk of bias from blinding.
The gap they did find was elsewhere. In industry-sponsored studies there was less agreement between the results and the conclusions, with a risk ratio of 0.83.
The authors concluded that sponsorship by the manufacturing company leads to more favorable efficacy results and conclusions than sponsorship by other sources. They described this as an industry bias that standard risk-of-bias assessment does not explain.
The practical instruction falls out of that. The methods are usually sound. Read the results table, then read the conclusion, and check that the second describes the first.
The gap between a result and a conclusion has been counted
A study published in JAMA looked specifically at trials that missed their primary endpoint, and counted how their reports were written.
The researchers screened 616 published reports and found 72 randomized trials with a clearly identified primary outcome that did not reach statistical significance. Then two readers appraised each one for what the authors called spin: reporting strategies that highlight a benefit despite a nonsignificant primary result, or that distract from it.
The abstract was where it concentrated. Spin appeared in the results section of the abstract in 37.5 percent of reports and in the conclusions section of the abstract in 58.3 percent. In 23.6 percent, the abstract conclusion focused only on treatment effectiveness. The title itself carried spin in 18 percent.
It ran through the main text too: 29.2 percent in the results, 43.1 percent in the discussion, and 50 percent in the conclusions. More than 40 percent of reports had it in at least two of those sections.
These are trials that did not show what they set out to show. In more than half of them the abstract's closing sentence was written to suggest otherwise.
This is not a peptide problem or an industry problem. It is a publishing habit, and it is the reason an abstract's last line is the least reliable sentence in a paper.
The comparator decides more than the drug does
A trial result is a comparison, so the thing being compared against is half the answer. Federal regulation on adequate and well-controlled studies lists the recognized types: placebo, dose comparison, no treatment, active treatment and historical control.
The regulation is unusually blunt about the trap in an active comparator. If the intent is to show that two treatments are similar, the report should assess whether the study could have detected a difference at all. It then says the quiet part: similarity of test drug and active control can mean either that both drugs were effective or that neither was.
A well-run trial discloses its comparator's weaknesses in its own labeling. The semaglutide prescribing information describes a study against titrated insulin, and states that only 26 percent of patients had been titrated to goal by the primary endpoint at week 30.
That sentence is in the approved labeling, published by the sponsor. It is exactly the kind of detail that vanishes when a result is repeated elsewhere, and it is the model for what disclosure looks like when it is done properly.
The levers that settle an answer before any data arrives
Several design choices shape a result, and all of them are made before the first patient is enrolled.
Who was enrolled. Federal regulation requires that subject selection give adequate assurance the participants actually have the condition being studied. A population selected to respond is a different study from a representative one.
How long it ran. A short trial can show a marker moving and cannot show whether the effect persists.
What counted as a response. The regulation requires that the methods of assessing response be well defined and reliable, and that the report explain the variables measured and the criteria used.
Whether anyone was blinded. Adequate measures to minimize bias among subjects, observers and analysts are a listed characteristic, and the report is expected to describe them.
What the product was. One requirement is easy to skip and matters enormously here: for a study to count, the test drug has to be standardized as to identity, strength, quality, purity and dosage form. A trial of material that was not characterized cannot give significance to its own results.
What a single uncontrolled study is worth
Much of what gets cited in this market is not a controlled trial at all. It is a case series, an open-label observation, or a report with no comparison group.
Federal regulation states the position on those directly. Uncontrolled or partially controlled studies are not acceptable as the sole basis for a claim of effectiveness. Carefully conducted and documented, they may provide corroborative support for controlled studies and may yield valuable safety data.
The same passage draws a line under the weakest tier. Isolated case reports, random experience, and reports lacking the details that permit scientific evaluation will not be considered.
That is a workable hierarchy for a reader. Corroborating evidence is real and it is not the same as the evidence it corroborates.
Six checks on a study a seller is citing
Read the results table before the abstract. The measured gap in sponsored research is between results and conclusions, and the table is the part that does not get rewritten.
Find the primary endpoint and check whether it was met. If the quoted claim comes from something else, the trial's own answer was different.
Read the comparator. Against placebo, against an active treatment, or against nothing, and whether the study was capable of detecting a difference.
Check who was enrolled and for how long, against the claim being made.
Check whether the product studied was characterized. A result about uncharacterized material is a result about that batch and nothing else.
Note the funder, and then keep reading. Sponsorship predicts a favorable conclusion. It does not predict a badly run study, and treating it as disqualifying would throw away most of the evidence that exists.
Key takeaways
- Industry-sponsored studies were no worse on standard quality measures and were better on blinding, in a Cochrane review of 75 papers.
- The measured difference is between results and conclusions: risk ratios of 1.27 for favorable results, 1.34 for favorable conclusions, and 0.83 for the two agreeing.
- In 72 trials that missed their primary endpoint, spin appeared in the abstract conclusion 58.3 percent of the time.
- Federal regulation warns that a similarity result can mean both treatments worked or neither did, and asks whether the study could have detected a difference.
- A study only counts if the product tested was standardized as to identity, strength, quality, purity and dosage form.
- Uncontrolled studies may corroborate controlled ones; isolated case reports and reports lacking evaluable detail are not considered at all.
Frequently asked questions
Should I ignore a study funded by the company selling the product?
No. Industry funds most clinical research, so ignoring sponsored studies would mean ignoring most of the evidence for most medicines. A Cochrane review of 75 papers found industry-sponsored studies were no worse on standard quality measures such as allocation concealment and follow-up, and were more likely to have low risk of bias from blinding. What it did find was more favorable results, more favorable conclusions, and less agreement between the results and the conclusions.
Where does sponsorship actually show up in a paper?
In the conclusion rather than the methods. The Cochrane review found a risk ratio of 1.27 for favorable efficacy results and 1.34 for favorable conclusions in industry-sponsored studies, alongside a risk ratio of 0.83 for agreement between results and conclusions. The authors described it as an industry bias that standard risk-of-bias assessment does not explain. In practice that means reading the results table first and then checking that the conclusion describes it.
What is spin in a trial report?
It is a reporting strategy that highlights a benefit despite a nonsignificant primary result, or distracts from it. A JAMA study appraised 72 randomized trials that missed their primary endpoint. Spin appeared in the abstract's conclusions in 58.3 percent of them, in the abstract's results in 37.5 percent, and in the title in 18 percent. In nearly a quarter, the abstract conclusion discussed only treatment effectiveness. It is a general publishing habit rather than an industry one.
Why does the comparator matter so much?
Because a trial result is a comparison, so what the product was compared against is half of it. Federal regulation warns that when a study aims to show two treatments are similar, the report should assess whether it could have detected a difference at all. It adds that similarity can mean either that both drugs worked or that neither did. Approved labeling sometimes discloses comparator weaknesses directly, as the semaglutide labeling does when it reports that only 26 percent of patients on the insulin comparator had been titrated to goal at the primary endpoint.
Is a case series useful evidence?
It has a defined and limited role. Federal regulation states that uncontrolled or partially controlled studies are not acceptable as the sole basis for a claim of effectiveness. Carefully conducted, they may corroborate controlled studies and may yield valuable safety data. The same passage excludes the weakest tier outright: isolated case reports, random experience, and reports lacking the details that permit scientific evaluation.
Sources
Each document below is named as it names itself, with the date printed on that document rather than the day it was read.
- Industry sponsorship and research outcome (Cochrane methodology review) — Cochrane Database of Systematic Reviews (Lundh, Lexchin, Mintzes, Schroll and Bero), February 2017
- Reporting and interpretation of randomized controlled trials with statistically nonsignificant results for primary outcomes — JAMA (Boutron, Dutton, Ravaud and Altman), May 2010
- 21 CFR 314.126 — Adequate and well-controlled studies — Office of the Federal Register, Electronic Code of Federal Regulations, January 2026
- OZEMPIC (semaglutide) injection — FDA-approved prescribing information — National Library of Medicine, DailyMed, June 2026