Research methods
Trial literacy widgets
Compact widget-style explainers for the trial-reading skills that matter most when interpreting metabolic peptide evidence: hazard ratio, confidence interval, non-inferiority vs superiority, absolute vs relative risk reduction, mean vs median weight loss, placebo-adjusted vs absolute change, duration normalisation, and the limits of cross-trial comparison.
Compact widget-style explainers for the trial-reading concepts that come up most often when reading metabolic peptide evidence. Each widget gives a worked example, the signal it carries, and the most common caveat. For the longer-form primer see how to read metabolic peptide trials.
Hazard ratio
- Worked example
- SELECT reported a hazard ratio of approximately 0.80 for MACE-3 with semaglutide 2.4 mg vs placebo in adults with overweight or obesity and pre-existing CVD without diabetes.
- What that means
- A 20% lower rate of MACE-3 in the semaglutide arm than the placebo arm over the follow-up period.
- Common caveat
- The hazard ratio is a relative effect. The absolute event reduction depends on the underlying baseline event rate; the trial population (CVD + overweight/obesity) had a higher baseline rate than the general adult population.
Confidence interval
- Worked example
- Hazard ratio ~0.80 with a 95% CI of ~0.72-0.90.
- What that means
- The interval excludes 1.00 (no effect), so the result is statistically significant at conventional thresholds. The true hazard ratio is plausibly somewhere between ~0.72 and ~0.90 given the trial data.
- Common caveat
- Confidence intervals are not probability statements about the parameter itself — they are statements about the procedure that produced them across hypothetical repeated experiments.
Non-inferiority vs superiority
- Worked example
- SURPASS-CVOT compared tirzepatide with dulaglutide on MACE-3 in T2D + ASCVD and demonstrated non-inferiority.
- What that means
- Non-inferiority means tirzepatide is not unacceptably worse than dulaglutide on MACE-3. The trial was designed to rule out a meaningful worsening, not to demonstrate improvement.
- Common caveat
- Non-inferiority is not superiority. Headlines and editorial coverage sometimes blur this; the published trial protocol and statistical analysis plan are the authoritative reference.
Absolute vs relative risk reduction
- Worked example
- Hazard ratio 0.80 (relative effect) corresponds to an absolute event reduction that depends on the baseline rate. In a population with 6.5% three-year event rate in the placebo arm, a 20% relative reduction is ~1.3 absolute percentage points.
- What that means
- Relative reductions sound larger; absolute reductions are what matters for individual decisions and policy.
- Common caveat
- When comparing trials, ensure you're comparing absolute-to-absolute or relative-to-relative. Mixing the two misleads.
Mean vs median weight loss
- Worked example
- STEP 1 reported a mean percent body-weight reduction of ~14.9% with semaglutide 2.4 mg over 68 weeks; the median was lower because some participants achieved much larger reductions, pulling the mean up.
- What that means
- Mean values can be sensitive to outliers; medians describe the typical participant. Both belong in any weight-trial summary.
- Common caveat
- Trials commonly report mean values. The proportion achieving categorical thresholds (≥5%, ≥10%, ≥15%) is a complementary view.
Placebo-adjusted vs absolute change
- Worked example
- SURMOUNT-1 reported absolute mean body-weight reduction of ~20.9% with tirzepatide 15 mg and ~3.1% with placebo over 72 weeks. The placebo-adjusted reduction is ~17.8%.
- What that means
- Placebo-adjusted change isolates the pharmacological effect. Absolute change is what the patient experiences (which includes placebo and lifestyle effects).
- Common caveat
- Press releases sometimes report only absolute change without the placebo arm. Both should be quoted.
Trial duration normalisation
- Worked example
- SURMOUNT-1 ran 72 weeks; STEP 1 ran 68 weeks; SCALE ran 56 weeks. Weight trajectories continue to evolve over many months.
- What that means
- Comparing the percent body-weight reduction at week 56 vs week 72 vs week 104 across trials is not a like-for-like comparison.
- Common caveat
- Where possible, compare trials at matched time points and pre-specified analysis windows. Longer-duration trials (STEP 5 at 104 weeks; SURMOUNT-4 at 88 weeks total) provide durability evidence.
Cross-trial comparison
- Worked example
- Quoting tirzepatide's ~20.9% (SURMOUNT-1) alongside semaglutide's ~14.9% (STEP 1) as if it were a direct head-to-head comparison conflates two trials with different populations, comparators, and protocols.
- What that means
- Different populations, comparators, eras and protocols mean apparent differences may be artefact rather than real biological signal.
- Common caveat
- Head-to-head trials (SURMOUNT-5, SURPASS-2) and network meta-analyses with explicit methodology are the right places to read off comparative effect. See: how to compare trials without overclaiming.
What these widgets do not do
- They do not replace formal statistical training.
- They do not provide individualised clinical advice.
- They do not endorse use of any medicine or research compound.
- Numerical worked examples are illustrative; the cited trial publications are the authoritative source for exact values.
Related trials
Last reviewed 21 May 2026 · trial literacy widgets · Metabolic Peptides editorial team