PIONEER 4 is the trial that put a semaglutide tablet against a GLP-1 injection in the same double-blind study, and it is quoted almost everywhere as having shown the pill to be at least as good. What it actually reported is more specific and more interesting: on the preregistered primary endpoint, under the trial’s own primary analysis, oral semaglutide was not superior to liraglutide, and on the questionnaire measuring what participants thought of their treatment the injection matched it. What changes when semaglutide is swallowed rather than injected is worked through in the route article, and liraglutide’s own case is in the liraglutide article. What follows is PIONEER 4 read as an experiment.
What was run
A randomized, double-blind, double-dummy phase 3a trial recruited from 100 sites in 12 countries. Every participant took a daily tablet and gave a daily subcutaneous injection, one of which was active in the two treatment arms and neither of which was in the third — the only way to blind a tablet against an injection. Of 950 screened, 711 were eligible and were randomized 2:2:1 by an interactive web-response system, stratified by background glucose-lowering medication and country, to once-daily oral semaglutide escalated to 14 mg (n = 285), once-daily subcutaneous liraglutide escalated to 1.8 mg (284), or placebo (142), for 52 weeks.[1]
Entry required type 2 diabetes for at least 90 days, a glycated hemoglobin of 7.0% to 9.5%, and a stable dose of metformin of at least 1,500 mg or the maximum tolerated, alone or with an SGLT2 inhibitor, for at least 90 days.[2] Mean age was 56 (SD 10), 341 of 711 (48%) were women, and mean baseline glycated hemoglobin was 8.0% (SD 0.7).[1][2] Trial completion ran 277 (97%), 274 (96%) and 134 (94%).[1]
The trial prespecified two estimands and named which one governed. Treatment policy — every participant analyzed as randomized, regardless of stopping the drug or taking rescue medication — was the primary estimand. Trial product, which assumes everyone stayed on drug and took no rescue, was secondary.[1] That distinction decides the headline.
The primary endpoint had three parts, and one of them failed
The primary endpoint was the change in glycated hemoglobin from baseline to week 26, tested as oral semaglutide’s superiority to placebo, and its noninferiority to liraglutide at a 0.4-point margin, and its superiority to liraglutide.[1]
Under the primary estimand, glycated hemoglobin fell 1.2 points (SE 0.1) on oral semaglutide, 1.1 (0.1) on liraglutide and 0.2 (0.1) on placebo. Superiority to placebo was comfortable, at an estimated treatment difference of −1.1 points (95% CI, −1.2 to −0.9; p < 0.0001). Noninferiority to liraglutide was met, at −0.1 (−0.3 to 0.0).[1] The superiority test on that same difference is posted in the registry as p = 0.0645.[2] It did not pass.
Superiority arrives only under the secondary trial-product estimand, where the difference widens to −0.2 points (95% CI, −0.3 to −0.1; p = 0.0056).[1][2] That is a real result and it is not the preregistered one. The honest summary of this trial’s glucose finding is that a 14 mg tablet matched a 1.8 mg daily injection and did not beat it, and the abstract’s own interpretation says as much: noninferior to liraglutide, superior to placebo, and superior to both only on bodyweight.
The weight comparison, which the tablet did win
Body weight fell 4.4 kg (SE 0.2) on oral semaglutide against 3.1 kg (0.2) on liraglutide and 0.5 kg (0.3) on placebo at week 26, an estimated treatment difference of −1.2 kg (95% CI, −1.9 to −0.6; p = 0.0003) against liraglutide and −3.8 kg (−4.7 to −3.0) against placebo. Under the trial-product estimand those become −1.5 kg (−2.2 to −0.9) and −4.0 (−4.8 to −3.2).[1]
One-point-two kilograms is the entire weight advantage of the tablet over the older daily injection, and it is the only endpoint on which this trial found oral semaglutide superior to liraglutide under the primary analysis. In relative terms the posted results give weight change of −4.89% (SD 4.95) at week 26 and −4.94% (6.37) at week 52, against −3.33% and −3.25% on liraglutide and −0.60% and −0.99% on placebo.[2]
Under five percent, in a year. Neither the abstract nor the registry posts a baseline body weight, so the trial cannot be converted between units without assuming one — dividing 4.4 kg by 4.89% implies a baseline near 90 kg, which is a derivation and not a reported figure. What the injectable doses studied for weight loss produce, in people selected for weight rather than for glucose, is in the semaglutide weight-loss article. This is a diabetes trial of a diabetes dose, and the gap between the two pictures is large.
Both active arms lost ground in the second half
The primary endpoint sits at week 26 and the trial ran to week 52, which makes the second half a free look at durability. Reaching a glycated hemoglobin below 7.0% went from 188 of 278 to 167 of 275 on oral semaglutide, and from 168 of 272 to 148 of 269 on liraglutide, while placebo went from 19 of 134 to 20 of 133. Reaching below 6.5% fell from 133 to 119 and from 116 to 88.[2]
Mean glycated hemoglobin holds at −1.2 on the tablet and drifts from −1.1 to −0.9 on liraglutide, so the attainment loss is partly people slipping back across a threshold rather than a collapse.[2] But the direction is the same in both active arms and the opposite of the control arm’s, and it is invisible in any summary that stops at week 26. Weight, by contrast, held: −4.4 kg at both time points on the tablet.[2]
The rescue columns tell a related story. By week 52, 20 participants on oral semaglutide, 18 on liraglutide and 43 on placebo had taken rescue medication, a hazard ratio of 0.15 (95% CI, 0.09 to 0.26) for the tablet against placebo and 1.17 (0.62 to 2.22; p = 0.6252) against liraglutide. Starting any additional antidiabetic medication ran 39, 29 and 46, a hazard ratio of 1.17 (0.75 to 1.80; p = 0.4915) for the tablet against the injection.[2] Neither comparison against liraglutide is significant, and neither points the direction the tablet is marketed in.
What participants said about the pill
This is the reversal that matters commercially, because avoiding an injection is the tablet’s entire proposition. The trial administered the Diabetes Treatment Satisfaction Questionnaire, and the posted results break it out item by item.
Satisfaction with treatment improved by 0.72 points (SD 1.49) on oral semaglutide and 0.73 (1.54) on liraglutide at week 26, and by 0.67 against 0.70 at week 52. Willingness to continue the present treatment improved by 0.64 on the tablet and 0.74 on the injection at week 26. Recommending the treatment to others: 0.57 against 0.63. Total treatment satisfaction: 3.59 (SD 6.12) against 3.44 (6.51) at week 26, and 3.41 against 3.11 at week 52.[2]
One item moved the way the premise predicts. Convenience of treatment improved by 0.55 on the tablet against 0.39 on the injection at week 26, and 0.50 against 0.42 at week 52.[2] That is the tablet’s advantage, and on a six-point item it is about a sixth of a point. No significance test is posted against any of these columns and the standard deviations are large relative to the means, so the correct reading is not that the injection won — it is that a year of swallowing a pill instead of injecting produced no measurable advantage in satisfaction or in willingness to keep going. Where this leaves the format decision generally is in the oral-against-injectable article.
Side effects, and three columns that do not favor the tablet
Adverse events were reported by 229 (80%) on oral semaglutide, 211 (74%) on liraglutide and 95 (67%) on placebo.[1] The posted table gives 973, 927 and 300 treatment-emergent events, and the gastrointestinal rows are the expected ones: nausea 55 of 285, 51 of 284 and 5 of 142; diarrhea 43, 31 and 11; vomiting 24, 13 and 3; constipation 22, 11 and 4; abdominal pain 16, 6 and 3.[2] Lipase rose to 1.33 and 1.40 times baseline at week 26 on the two active arms against 0.99 on placebo, and pulse rose 2 and 3 beats per minute against 0.[2]
Three columns run against the tablet. Serious adverse events were 31 of 285 (10.9%) on oral semaglutide, 22 of 284 (7.7%) on liraglutide and 15 of 142 (10.6%) on placebo. Deaths over the 52 weeks were 3, 4 and 1. And severe or blood-glucose-confirmed symptomatic hypoglycemia affected 2 participants on oral semaglutide, 7 on liraglutide and 3 on placebo, across 2, 9 and 3 episodes.[2]
The hypoglycemia column is the only one of the three that favors the tablet, and none of them is a powered comparison: a trial sized to detect a 0.4-point difference in glycated hemoglobin cannot separate seven hypoglycemic participants from two, or four deaths from three. They are printed here because leaving them out would be a choice as well.
What PIONEER 4 did not establish
It adjudicated no cardiovascular outcome and was not designed to; the oral semaglutide safety trial is a separate study, covered in the PIONEER 6 article. It ran 52 weeks, so nothing in it speaks to a second year. It never compared the tablet with injected semaglutide — the comparator was a different molecule at its own maximum dose — so it settles nothing about which route of the same drug is better. It ran no weight endpoint at a weight-management dose, and no participant was enrolled for obesity.
The population is a metformin-treated diabetes population with a mean glycated hemoglobin of 8.0% and no upper weight criterion. Of 711 participants, 519 were White, 94 Asian and 29 Black or African American.[2] Somebody without diabetes, not on metformin, buying a compounded oral preparation to lose weight, sits outside every entry criterion the trial had.
A limit on the record rather than the design: the paper carries a published correction, filed as a Department of Error notice, with no abstract and no open access. What it amended could not be established, and the figures above are quoted as the abstract and the registry state them.
What was in the tablet
Every figure here belongs to a manufactured, FDA-approved oral semaglutide tablet supplied free, co-formulated with an absorption enhancer, taken each morning under protocol conditions with a fixed water volume and a fixed fasting window afterward, escalated to 14 mg over eight weeks, in adults with type 2 diabetes on metformin. Sellers listed on the oral semaglutide board dispense compounded preparations, which are not FDA-approved, are not reviewed by the FDA for safety, efficacy or quality before they reach a patient, and were not the product in this trial or any other.
The sentence to retire about PIONEER 4 is that the tablet beat the injection. On the trial’s own primary analysis of its own primary endpoint, the superiority test returned p = 0.0645; what the tablet won was 1.2 kilograms and a sixth of a point on a convenience item. That is a defensible reason to prefer it. It is not the reason it is usually sold.