Reading the evidence
Case Report Evidence: What Case Reports and Safety Signals Can Show
AR
· 9 min read
A summary of published research, with citations. It states no amount, schedule or route, and makes no claim about what this compound does in a person. Supplied for laboratory research use only.
On this page · 8 sections
- What is a case report, and what is it good for?
- What does a case series add?
- What are pharmacovigilance databases?
- What is disproportionality analysis?
- What do real GLP-1 disproportionality studies show?
- What is a signal, and what is it not?
- How should you read a case report or a FAERS study?
- Frequently asked questions
A case report describes what happened to one patient; a case series describes several. Pharmacovigilance databases such as the US FDA Adverse Event Reporting System (FAERS) and Canada Vigilance collect many such reports. Disproportionality analysis looks for drug-event pairs reported more often than expected. All of these can raise a signal: a possible link worth investigating. None can say how often harm occurs or prove that a drug caused it, because nobody knows how many people took the drug without incident, or how many events went unreported.

What is a case report, and what is it good for?
A case report is a detailed clinical account of a single patient: what they took, what happened, what else might explain it and how it resolved. It sits at the bottom of the evidence hierarchy for estimating effects, and at the top for noticing something new. Rare, unexpected or severe events are often first described this way, long before any trial or cohort study could detect them.
Two recent reports concern products sold as retatrutide outside clinical trials. Retatrutide is investigational and approved nowhere.
Buzzi and Huber. A woman in her mid-20s was admitted with five days of persistent nausea, repeated vomiting and loose stools. Blood tests, urinalysis, abdominal ultrasound and brain MRI found no alternative cause, and standard anti-nausea drugs did not help. Only on repeated questioning did she report receiving an unverified product said to contain retatrutide, from a self-described life coach without medical qualification; she had not considered it medication. Symptoms resolved within a week. The authors scored causality with the Naranjo scale, a standard questionnaire covering timing, alternative causes and what happened when exposure stopped. Her score was 7, which they report as "probable".
Branine. A man in his mid-30s with long-standing type 1 diabetes developed severe vomiting, diarrhoea, high blood glucose, ketonaemia peaking at 4.3 mmol/L and acute kidney injury after taking an online-purchased product marketed as retatrutide. His partner, who took the same product, also had gastrointestinal symptoms. But both had eaten a possible source of food poisoning, and his stool culture grew Shigella flexneri. The author states that causation cannot be established.
The second report shows a case report's main weakness more plainly than the first: a competing explanation sitting in the same patient. Both share a further limit. Nobody analysed what was in the vials, so neither can say anything about retatrutide as a characterised drug. Gray-market retatrutide and the evidence explains why that gap cannot be closed after the fact, and retatrutide side effects in the trials gives the controlled record.
What does a case series add?
A case series gathers several patients with a shared exposure or outcome. It can show a pattern: similar timing, similar features, similar resolution. It still has no comparison group and no denominator, so it cannot show the pattern is more common in exposed people than in anyone else.
Case series and clinical anecdote often start the chain of evidence. Hathaway and colleagues' study of semaglutide and NAION, a form of sudden optic nerve damage, opens by noting that anecdotal experience raised the possibility. They then tested it with a matched cohort, and others followed with larger designs; correlation vs causation follows that story.
What are pharmacovigilance databases?
Regulators collect spontaneous reports of suspected adverse reactions from health professionals, manufacturers and the public. FAERS is the US database; Canada Vigilance is Health Canada's; EudraVigilance is the European one. A report records a patient, a suspected drug and an event. It is not a confirmed diagnosis and not a confirmed cause.
| Strength | Weakness |
|---|---|
| Covers every marketed drug, in real-world use, indefinitely | No denominator: the number of people exposed is unknown |
| Captures rare and unexpected events trials are too small to see | Massive under-reporting |
| Collects reports from patients as well as clinicians | Reporting rises with publicity and with newly marketed drugs |
| Can be searched quickly when a question arises | Duplicates, missing details and unverified exposures |
Under-reporting is not a minor footnote. Hazell and Shakir's systematic review of 37 studies from 12 countries estimated a median under-reporting rate of 94%. Even for specific serious reactions the median was 85%. The reports that do arrive are not a random sample of the events that occurred.
What is disproportionality analysis?
Because there is no denominator, databases cannot give rates. Disproportionality analysis compares proportions instead. Of all reports mentioning drug A, what share mention event X, compared with the share mentioning X among reports for all other drugs? If X is unusually prominent among drug A's reports, that is a disproportionality signal.
Bate and Evans described the standard measures: the proportional reporting ratio (PRR), the reporting odds ratio (ROR), the information component (IC) and the empirical Bayes geometric mean (EBGM). The Bayesian versions shrink estimates based on few reports, to avoid chasing noise. The READUS-PV guideline, developed by 34 international experts, now sets out what such analyses should report. Its authors note that these are the most commonly used quantitative methods for detecting signals, and that their methods are generally poorly reported.
What do real GLP-1 disproportionality studies show?
Pancreatic and biliary events. Wu and colleagues analysed FAERS reports for liraglutide, semaglutide, dulaglutide and tirzepatide. All four showed a pancreatic signal, but so did the comparison drugs, metformin and SGLT2 inhibitors. A signal shared with unrelated diabetes drugs points at least partly to who is being treated and reported, not only to the drug. Biliary signals appeared for the four study drugs and not for the comparators, which is more specific. The authors state that such signals cannot estimate incidence or prove causality. Tirzepatide is authorised in Canada as Mounjaro; its trial-based record is in tirzepatide side effects.
Gastric motility. Qian and Jiang screened FAERS for drugs associated with delayed gastric emptying and checked results against Canada Vigilance. Semaglutide's ROR for impaired gastric emptying was 80.27 in FAERS and 54.17 in the Canadian data. A huge ROR for an effect expected from a drug's known pharmacology is a check that the method works. It is not a measure of risk: an ROR of 80 does not mean 80 times as many people are affected.
Compounded and counterfeit products. McCall and colleagues compared FAERS reports for compounded and branded GLP-1 drugs. Of 81,078 reports, 707 involved compounded products, which had higher RORs for several gastrointestinal events and much higher RORs for preparation errors and contamination. Health Canada's warnings about unauthorised products are covered in Health Canada GLP-1 warnings.
What is a signal, and what is it not?
| A signal is | A signal is not |
|---|---|
| A hypothesis that a drug and an event may be linked | Proof that the drug caused the event |
| A reason to look at better data | An incidence or a risk |
| Prioritised by strength, seriousness and plausibility | Comparable across drugs by the size of its ROR |
| Something regulators assess alongside trials and cohorts | A finding to quote as "X times the risk" |
The path from signal to answer runs through designs that have denominators: cohort studies with an active comparator, pooled trial data and, where feasible, randomised trials. Converting the result into absolute terms is covered in relative vs absolute risk. Case reports that never reach a database, and signals that were checked and found nothing, are also subject to publication bias.
How should you read a case report or a FAERS study?
- Was the exposure confirmed, and was the product characterised?
- What alternative causes were present, and were they excluded?
- Did symptoms stop when exposure stopped, and return if it restarted?
- For a database study: what comparator was used, and did comparator drugs show the same signal?
- Does the paper say, as it should, that it cannot estimate incidence?
- Has the signal been tested in a design with a denominator?
For where each design sits in the wider hierarchy, see how to read a peptide study. Nothing here is medical, dosing, or therapeutic guidance.
Frequently asked questions
Can a case report prove a drug caused a side effect? Rarely on its own. A single case can be compelling when the timing is tight, other causes are excluded and the event recurs on re-exposure. Even then it cannot show how often the event happens. Causality scales like Naranjo's grade the link in one patient as possible, probable or definite; they do not establish a general effect.
Why do FAERS studies on the same drug reach different conclusions? They differ in time window, comparator drugs, how duplicates are removed, which event terms are grouped, and which statistic is used. Reporting itself changes over time with publicity and prescribing. Without a denominator, small methodological choices shift results noticeably, which is why the READUS-PV reporting guideline was developed.
What is Canada Vigilance? It is Health Canada's programme for collecting reports of suspected adverse reactions to health products marketed in Canada, with a publicly searchable online database. Like FAERS, it holds unverified spontaneous reports. It is useful for checking whether a signal seen in US data also appears in Canadian reporting, as Qian and Jiang did.
Does a high reporting odds ratio mean a high risk? No. It means the event makes up an unusually large share of that drug's reports compared with other drugs. That can reflect a real effect, an expected pharmacological action, stimulated reporting after publicity, or who takes the drug. The absolute risk could still be very small or very large; the ROR cannot tell you which.
Selected Research References
- Buzzi RM, Huber LC. Emerging risks of health optimisation: refractory nausea associated with peptide injections reported to contain retatrutide. BMJ Case Reports. 2026;19(10):e276106. pubmed.ncbi.nlm.nih.gov/42843917
- Branine N. Online-Sourced Retatrutide Complicating Impending Diabetic Ketoacidosis in a Patient With Type 1 Diabetes and Concurrent Shigella Gastroenteritis. Cureus. 2026;18(8):e115260. pubmed.ncbi.nlm.nih.gov/42669023
- Naranjo CA, Busto U, Sellers EM, et al. A method for estimating the probability of adverse drug reactions. Clinical Pharmacology and Therapeutics. 1981;30(2):239-245. pubmed.ncbi.nlm.nih.gov/7249508
- Hazell L, Shakir SA. Under-reporting of adverse drug reactions: a systematic review. Drug Safety. 2006;29(5):385-396. pubmed.ncbi.nlm.nih.gov/16689555
- Bate A, Evans SJ. Quantitative signal detection using spontaneous ADR reporting. Pharmacoepidemiology and Drug Safety. 2009;18(6):427-436. pubmed.ncbi.nlm.nih.gov/19358225
- Fusaroli M, Salvo F, Begaud B, et al. The Reporting of a Disproportionality Analysis for Drug Safety Signal Detection Using Individual Case Safety Reports in PharmacoVigilance (READUS-PV): Development and Statement. Drug Safety. 2024;47(6):575-584. pubmed.ncbi.nlm.nih.gov/38713346
- Wu Y, Li K, Wang Y, et al. Pancreatic and biliary safety signals of GLP-1 receptor agonists and tirzepatide: a disproportionality analysis of FAERS reports. Frontiers in Endocrinology. 2026;17:1941948. pubmed.ncbi.nlm.nih.gov/42865361
- Qian Z, Jiang N. Drug-induced gastric motility disorders: A disproportionality analysis from the FAERS and CVARD databases. PLoS One. 2026;21(6):e0351731. pubmed.ncbi.nlm.nih.gov/42284324
- McCall KL, Mastro Dwyer KA, Casey RT, et al. Safety analysis of compounded GLP-1 receptor agonists: a pharmacovigilance study using the FDA adverse event reporting system. Expert Opinion on Drug Safety. 2026;25(3):581-588. pubmed.ncbi.nlm.nih.gov/40285721
- Hathaway JT, Shah MP, Hathaway DB, et al. Risk of Nonarteritic Anterior Ischemic Optic Neuropathy in Patients Prescribed Semaglutide. JAMA Ophthalmology. 2024;142(8):732-739. pubmed.ncbi.nlm.nih.gov/38958939
All products sold by NovaEvum are for research and laboratory use only. They are not intended for human or animal consumption, diagnosis, treatment, cure, or prevention of any disease.
