Research Literacy
How Small Is Too Small? Sample Size, Explained
The short answer
Sample size determines what a study can detect, not just how confident it can be. Early dose-finding work on the growth-hormone secretagogue MK-677 tested groups of just 10-12 healthy elderly adults and found a clear, statistically significant rise in bone-turnover markers — a large, easy-to-detect biochemical signal. A later 161-patient trial of the same peptide in hip-fracture patients found IGF-I levels jumped 84% versus 17% on placebo (clearly significant, non-overlapping confidence intervals) — but functional recovery, the outcome patients actually cared about, showed no significant difference, despite favorable trends. The study's own authors said functional measures may not be responsive enough for a trial that size. Contrast that with the SELECT trial, which enrolled 17,604 people specifically because the real effect it was chasing was small.
Quick answers
- What is it?
- How small a study's sample size can be before its results stop being trustworthy.
- Why are people interested in it?
- People reading about small trials of peptides or drugs want to know when to trust the results and when not to.
- What has actually been studied?
- Early dose-finding trials of the growth-hormone secretagogue MK-677 in groups of 10 to 50 people, a later 161-patient hip-fracture trial of the same peptide, and the 17,604-person SELECT semaglutide trial.
- How strong is the evidence?
- Sample size determines what a study can detect, not just how confident it can be — a small trial caught a large biochemical signal reliably, but the same peptide's 161-patient trial found a clear biochemical effect alongside no significant difference on the noisier functional outcome patients cared about.
- What don't we know yet?
- Whether a larger MK-677 hip-fracture trial would find a genuine functional benefit or confirm none exists, whether the bone-turnover changes translate into fewer fractures over time, and how large a trial needs to be in general, since there's no fixed rule.
- Where can I read the original research?
- Sources at the bottom of this page: Oral administration of the growth hormone secretagogue MK-677 increases markers of bone turnover in healthy and functionally impaired elderly adults and The effects of MK-0677, an oral growth hormone secretagogue, in patients with hip fracture.
The short answer: Sample size determines what a study can detect, not just how confident it can be. Early dose-finding work on the growth-hormone secretagogue MK-677 tested groups of just 10-12 healthy elderly adults and found a clear, statistically significant rise in bone-turnover markers — a large, easy-to-detect biochemical signal. A later 161-patient trial of the same peptide in hip-fracture patients found IGF-I levels jumped 84% versus 17% on placebo (clearly significant, non-overlapping confidence intervals) — but functional recovery, the outcome patients actually cared about, showed no significant difference, despite favorable trends. The study's own authors said functional measures may not be responsive enough for a trial that size. Contrast that with the SELECT trial, which enrolled 17,604 people specifically because the real effect it was chasing was small.
What the research shows
A small sample can still catch a big signal
Sample size and statistical power aren't about a fixed threshold of "enough people" — they're about matching the sample to the size of the effect being chased and how much natural variability surrounds it. According to PubMed, a 1999 study in the Journal of Bone and Mineral Research tested the growth-hormone secretagogue MK-677 (ibutamoren) in healthy elderly adults using groups of just 10 to 12 people per dose, given 10 mg or 25 mg daily for two weeks. Even with a sample that small, urinary markers of bone turnover rose significantly compared to placebo (10 mg: +10%; 25 mg: +17%; both P<0.05), because the biochemical effect was large and consistent enough to separate clearly from the noise even with few participants. (DOI)
Watching one research program scale its sample size up
The 1999 MK-677 paper is really three studies stacked together, and tracking how the sample size grew across them shows the logic in action. The first was that 2-week, 10-to-12-per-group dose-response pilot. The second scaled up modestly: 50 healthy elderly adults, 20 on placebo for 4 weeks and 30 on MK-677 (25 mg for 2 weeks, then 50 mg for 2 weeks), which was enough to detect a smaller, subtler signal — a significant 8% rise in serum osteocalcin (P<0.05 vs placebo) — that the tiny pilot groups likely weren't built to catch reliably. The third study scaled up again to 105 elderly adults with functional impairment, of whom 63 on MK-677 and 28 on placebo completed the full 9 weeks; at that size, osteocalcin rose 29.4% and bone-specific alkaline phosphatase rose 10.4% (both P<0.001 vs placebo), with urinary bone-resorption markers also significant (P<0.05). (DOI) Each step up in sample size wasn't arbitrary — it tracked a shift toward smaller, more clinically relevant effects and a more variable, harder-to-study population.
The same peptide, a bigger trial, and a much noisier outcome
A 2004 trial in the Journal of the American Geriatrics Society tested MK-677 in 161 older adults recovering from hip fracture, across 13 medical centers, over six months of treatment. On the biochemical measure, the result was unambiguous: serum IGF-I rose 84% (95% CI 63-107%) on MK-677 versus 17% (95% CI 8-28%) on placebo — confidence intervals that don't overlap at all. But the trial's actual focus, functional recovery measured with a panel of physical performance tests, showed no statistically significant difference between groups, even though patients on MK-677 trended toward better scores on three of four lower-extremity measures. (DOI) The researchers wrote plainly that "present functional performance measures may not be sufficiently responsive for use as the primary endpoint of small intervention studies" — a direct acknowledgment that a trial sized adequately for one outcome, a large, low-noise biochemical marker, can be undersized for another, a smaller, noisier functional one, even using the exact same participants.
What it takes to catch a smaller, real effect
At the opposite extreme, the SELECT cardiovascular outcomes trial of semaglutide enrolled 17,604 people specifically because the effect it needed to detect was real but modest: an absolute difference of about 1.5 percentage points in cardiovascular events between semaglutide and placebo. A trial that size produced a tight, statistically significant confidence interval (hazard ratio 0.80, 95% CI 0.72-0.90) around that modest effect. (DOI) Detecting a small true effect reliably requires a large sample; detecting a large true effect can be done with far fewer people. The MK-677 bone-marker result and the SELECT cardiovascular result were both real, but they needed very different amounts of data to prove it.
Why underpowered studies create false confidence in both directions
A study that's too small for its outcome doesn't just risk missing a real effect — a false negative. It can also produce a result that looks dramatic purely because small samples generate noisier, more extreme estimates by chance, inflating an effect that would shrink with more data. Neither failure mode announces itself in a headline number; both require looking at the sample size and the width of the confidence interval, not just whether a result cleared the conventional significance threshold.
What we don't know yet
Whether a larger MK-677 hip-fracture trial would find a genuine functional benefit or confirm that none exists. The 161-patient trial wasn't sized to distinguish between those two possibilities, and no larger follow-up trial was located in this search.
Whether the bone-turnover changes seen in the small dose-response studies translate into fewer fractures over time. That would require a much larger, longer trial designed around fracture outcomes specifically, which doesn't appear to exist for MK-677.
How large a trial "needs" to be, in general. There's no fixed number from any of these studies — the right sample size depends entirely on the size of the effect being chased and how much variability surrounds the outcome being measured, and that has to be estimated fresh for every new question.
Sources
- Murphy MG, Bach MA, Plotkin D, et al. Oral administration of the growth hormone secretagogue MK-677 increases markers of bone turnover in healthy and functionally impaired elderly adults. Journal of Bone and Mineral Research, 1999;14(7):1182-1188. DOI
- Bach MA, Rockwood K, Zetterberg C, et al. The effects of MK-0677, an oral growth hormone secretagogue, in patients with hip fracture. Journal of the American Geriatrics Society, 2004;52(4):516-523. DOI
- Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes. New England Journal of Medicine, 2023;389(24):2221-2232. DOI
Article records maintained via PubMed. Peptide Proof reports what studies found. It does not provide medical advice, dosing guidance, or treatment recommendations.
| Evidence Index: mk-677 · tesamorelin · semaglutide
Sources
- Oral administration of the growth hormone secretagogue MK-677 increases markers of bone turnover in healthy and functionally impaired elderly adults
- Journal of Bone and Mineral Research, 1999 · n=187 · DOI
- The effects of MK-0677, an oral growth hormone secretagogue, in patients with hip fracture
- Journal of the American Geriatrics Society, 2004 · n=161 · DOI
- Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes
- New England Journal of Medicine, 2023 · n=17604 · DOI
Get the evidence, not the hype.
New research briefs, plainly written and fully sourced. No dosing advice, ever.