Retatrutide Once Weekly in Adults with Knee Osteoarthritis Pain and Obesity
New England Journal of Medicine, 2026 · paper · Clinical trials
Two phase 3 trials of weekly retatrutide, a triple-hormone-receptor drug, in adults with painful knee osteoarthritis and obesity. Retatrutide reduced knee pain and body weight more than placebo; the most common adverse events were gastrointestinal.
11 findings
Substance TRIUMPH-4 baseline blood pressure appears in different arms in two tables. Table S1 gives SBP/DBP for 9 mg, 12 mg and placebo as "128.9 (12.5) / 81.6 (8.3)", "130.2 (13.0) / 81.5 (8.3)" and "129.1 (11.8) / 82.0 (7.6)". Tables S8 and S9, with columns in the same order, give SBP "130.2 (1.0) 129.1 (1.0) 128.9 (1.0)" and DBP "81.5 (0.7) 82.0 (0.7) 81.6 (0.7)": the same values rotated across the arms.
- p. 53 Table S1, SBP/DBP row, TRIUMPH-4
128.9
- p. 67 Table S8, Systolic blood pressure baseline, TRIUMPH-4
130.2 (1.0)
- p. 71 Table S9, Systolic blood pressure baseline, TRIUMPH-4
130.2 (1.0)
- p. 53 Table S1, SBP/DBP row, TRIUMPH-4 (RETA 9-mg, RETA 12-mg, Placebo)
128.9 (12.5) / 81.6 (8.3) 130.2 (13.0) / 81.5 (8.3) 129.1 (11.8) / 82.0 (7.6)
Only the authors can say what is right here.
Judge's note: Checked the column headers in Table S1 (p52–53: RETA 9-mg, RETA 12-mg, Placebo) and Table S8 (p67–68: RETA 9-mg, RETA 12-mg, Placebo). The values match to one decimal but are cyclically shifted, and SBP and DBP shift together. A difference between observed and model-based estimates would not produce the same numbers in permuted order. The TRIUMPH-1-OA columns agree between the tables. Only the authors can say which assignment is correct.
Second check: another model agreed
Report a problem with this finding
Reference or label Figure 1 legend: "P<0.001 for all comparisons with placebo", but Panel B plots the TRIUMPH-1-OA ITT 4-mg estimate (−3.41), whose comparison with placebo has P=0.005.
- p. 9 Figure 1 legend, last sentence
P<0.001 for all comparisons with placebo.
- p. 9 Results, Primary End Points, TRIUMPH-1-OA ITT
for 4 mg (P = 0.005)
- p. 8 Figure 1 Panel B, ITT 4-mg point
−3.41
Judge's note: Figure 1B (p8, upright image) plots ITT points including 4 mg at −3.41 against placebo at −2.46. The p9 text gives that comparison as −0.96 (95% CI, −1.61 to −0.30), P=0.005. Table S7 (supp-01 p63) shows the same, p=0.005. The abstract itself says 'P≤0.005 for all comparisons'. The blanket P<0.001 in the legend therefore does not cover every comparison drawn in the figure.
Second check: another model agreed
Report a problem with this finding
Reference or label Table S6 (TRIUMPH-4) headers give "RETA 9-mg N=148" and "Total N=445", with percentages computed on those Ns (e.g. 36 (24.3), 125 (28.1)), but its footnote defines "N=all randomly assigned participants who took ≥1 dose of study drug". Figure S5 shows only "444 (99.8%) exposed", and Table 4 and Table S12 give 9 mg N=147.
- p. 62 Table S6 header and footnote, TRIUMPH-4
N=all randomly assigned participants who took ≥1 dose of study drug
- p. 39 Figure S5
444 (99.8%) exposed
- p. 12 Table 4 header, 9 mg
(N = 147)
- p. 76 Table S12 TRIUMPH-4 column header
N=147
Only the authors can say what is right here.
Judge's note: Checked Table S6 (p61/62), Figure S5 (p38/39), and the reported N=147 in the main Table 4 header and Table S12. The percentages use the randomized denominators (36/148 = 24.3% vs 36/147 = 24.5%; 125/445 = 28.1% vs 125/444 = 28.2%). The population label in the footnote does not match the N used.
Second check: another model agreed
Report a problem with this finding
Reference or label In the Multiplicity Control text of supplement pp.32–33, each figure pointer is one too low: "0.025 in TRIUMPH-1-OA (Figure S1)" should be Figure S2, "0.05 (Figure S2)" for TRIUMPH-4 should be Figure S3, and "testing scheme outlined in Figure S3" for the pooled SWER scheme should be Figure S4.
- p. 33 Statistical Analysis Methods, Multiplicity Control
the initial two-sided significance level was 0.025 in TRIUMPH-1-OA (Figure S1)
- p. 33 Statistical Analysis Methods, Multiplicity Control
controlled for FWER at a two-sided significance level of 0.05 (Figure S2)
- p. 35 Figure S1 title
Figure S1. Study Designs
- p. 33 Statistical Analysis Methods, Multiplicity Control
initial two-sided significance level was 0.025 in TRIUMPH-1- OA (Figure S1)
Judge's note: Opened the figure pages. Figure S1 (PDF p35) is 'Study Designs'. Figure S2 (p36) is the TRIUMPH-1-OA FWER scheme. Figure S3 (p37) is the TRIUMPH-4 FWER scheme. Figure S4 (p38) is the pooled-analyses SWER scheme with α1p = 0.02375. All three text pointers are off by one.
Second check: another model agreed
Report a problem with this finding
Reference or label Supplement p.33: "Similar approach as Case Study 2 of Vandemeulebroecke et al. (2024) was applied.3" carries superscript 3, which is Bretz & Xi 2019; Vandemeulebroecke et al. 2024 is reference 4.
- p. 34 Statistical Analysis Methods
Similar approach as Case Study 2 of Vandemeulebroecke et al. (2024) was applied.3
- p. 81 Supplement References 3 and 4
Bretz F, Xi D. Commentary on “Statistics at FDA
- p. 81 Reference list
3. Bretz F, Xi D. Commentary on “Statistics at FDA
- p. 81 Reference list
4. Vandemeulebroecke M, Häring DA, Hua E, Wei X, Xi D.
Judge's note: Checked the supplement reference list on PDF p81. Ref 3 is Bretz F, Xi D, Stat Biopharm Res 2019. Ref 4 is Vandemeulebroecke M et al., Clin Trials 2024. The sentence names the 2024 paper but cites ref 3.
Second check: another model agreed
Report a problem with this finding
Reference or label Table S5, Age row (supplement p. 60): "Younger patients have a stronger association between BMI and knee osteoarthritis than older patients.8" Reference 8 (DeCleene et al., JAMA 2026;335:975-985, doi:10.1001/jama.2025.26817) estimates US obesity prevalence by state, race and ethnicity, sex and age; its full text never mentions osteoarthritis, the knee or joints. It does support the sentence before ("Obesity prevalence is generally the highest between ages 45 and 64 years"), so the superscript may belong there, leaving the knee osteoarthritis claim without a source.
- p. 61 Table S5, Age row
Younger patients have a stronger association between BMI and knee osteoarthritis than older patients.8
- p. 81 Reference 8
US State-Level Prevalence of Adult Obesity by Race and Ethnicity From 1990 to 2022
Judge's note: Rests on outside knowledge: the content of DeCleene et al. (JAMA 2026) beyond its title. Checked Table S5 on p60/61 and the reference list on p80/81. Judged by its title, ref 8 is about obesity prevalence by race and ethnicity. Ref 6 (knee OA burden attributable to high BMI) would seem a closer fit for this claim. The authors should check the citation.
Second check: another model agreed
Report a problem with this finding
Reference or label supp-02 p1164: "The categories of treatment period discontinuation based on the disposition reason is outlined in Table GZBN.4.2." In SAP v3, Table GZBN.4.2 is "Imputation Procedure – Pain-Related Endpoints". The categorization is in Table GZBN.4.3, "Categorization of Treatment Period Discontinuation Reasons".
- p. 1164
The categories of treatment period discontinuation based on the disposition reason is outlined in Table GZBN.4.2.
Judge's note: The crop shows the pointer to Table GZBN.4.2. In supp-02 p1167, Table GZBN.4.2 is the pain-related imputation table and Table GZBN.4.3 is the categorization of discontinuation reasons. Table GZBN.4.1 on p1166 itself refers to Category 5A/5B 'in Table GZBN.4.3'. The change log on p~1130 refers to 5A/5B categories 'in Table GZBN.4.2', which suggests an earlier version used that numbering and this reference was not updated. As printed, the pointer is wrong.
Second check: another model agreed
Report a problem with this finding
Wording Abstract: "In both trials, serious adverse events were more common with retatrutide than with placebo", but in TRIUMPH-1-OA Table 4 shows 6 (4.2) with 4 mg against 10 (6.9) with placebo.
- p. 1 Abstract, Results
In both trials, serious adverse events were more common with retatrutide than with placebo
- p. 12 Table 4, Serious adverse event row, TRIUMPH-1-OA
6 (4.2)
- p. 78 Table S13, TRIUMPH-1-OA, Participants with ≥1 SAE
6 (4.2) 14 (9.8) 21 (14.6) 10 (6.9)
- p. 1 Abstract, Results
serious adverse events were more common with retatrutide than with placebo
Only the authors can say what is right here.
Judge's note: On main p12 (Table 4, image), the serious adverse event row for TRIUMPH-1-OA reads 4 mg 6 (4.2), 9 mg 14 (9.8), 12 mg 21 (14.6), placebo 10 (6.9). Table S13 and the Safety text on p13 give the same figures. The 4-mg group had fewer serious adverse events than placebo. The abstract sentence makes no exception for dose and is not restricted to 9 and 12 mg or to pooled arms. The authors should say whether the sentence means the 9-mg and 12-mg doses or pooled retatrutide.
Second check: another model agreed
Report a problem with this finding
Wording Results, Trial Participants (p. 7): participants "were screened between August 1, 2023, and November 14, 2025" (TRIUMPH-4) and "between July 10, 2023, and April 30, 2026" (TRIUMPH-1-OA). These are the trials' start and completion dates on ClinicalTrials.gov (NCT05931367: start 2023-08-01, completion 2025-11-14; NCT05929066: start 2023-07-10, completion 2026-04-30), not screening windows: screening cannot have run to the last visit of a 68- or 80-week trial. "Screened between" should describe the trial period, or give the screening dates.
- p. 7 Results, Trial Participants
screened between August 1, 2023, and November 14, 2025
- p. 1 Front matter
This article was published on November 8, 2026
Judge's note: Main p7 (image) gives the screening window. Main p6 gives the treatment period as 68 weeks, and 380 of 445 participants (85.4%) completed the trial. The front matter (p1) gives publication as November 8, 2026. A participant screened on November 14, 2025 would reach week 68 no earlier than about March 2027 (68 weeks is about 476 days), after publication. The screening end date, most likely its year, appears to be wrong. Only the authors can give the correct date.
Second check: another model agreed
Report a problem with this finding
Wording Discussion: participants with BMI "of 35 or higher had significant reductions in the WOMAC pain score, improvements in physical function, and reductions in body weight", but no physical-function result is reported for the BMI ≥35 subgroup.
- p. 13 Discussion, paragraph on BMI ≥35
of 35 or higher had significant reductions in the WOMAC pain score, improvements in physical function, and reductions in body weight
- p. 5 Table 2, 'Results in participants with baseline BMI ≥35' block
Results in participants with baseline BMI
- p. 6 Methods, End Points and Assessments
among participants with a BMI of at least 35 were additional key secondary end points
Only the authors can say what is right here.
Judge's note: Methods (p6) name only WOMAC pain and percent body-weight change as key secondary end points in the BMI ≥35 subgroup. The Table 2 'Results in participants with baseline BMI ≥35' block (p5 image) contains only those two. A search of supp-01 for BMI ≥35 finds pain, weight and weight-threshold results only (Tables S2–S3, S8, S9), with no physical-function result for the subgroup. The Discussion sentence has no reported data behind its physical-function claim.
Second check: another model agreed
Report a problem with this finding
Wording Methods (p.33) and the Figure S4 description say the pooled key secondary end points "were tested initially at a two-sided significance level of 0.025, respectively, with the overall SWER at a two-sided significance level of 0.05". Table S4 footnote b says the "Global Type-I error rate (alpha) used for the graphical multiple testing procedure of SWER endpoints is 0.0475", and its first local level is 0.0238, not 0.025.
- p. 34 Statistical Analysis Methods
were tested initially at a two-sided significance level of 0.025, respectively, with the overall SWER at a two-sided significance level of 0.05
- p. 60 Table S4 footnote b
Global Type-I error rate (alpha) used for the graphical multiple testing procedure of SWER endpoints is 0.0475
- p. 38 Figure S4 description
tested under α1p = α2p =0.025, with overall SWER at two-sided 0.05
Only the authors can say what is right here.
Judge's note: Read Table S4 on p60 and the text on pp33/38. The local levels in Table S4 (0.0238, 0.0428, 0.0322, 0.0475) match a starting alpha of 0.02375 per dose (total 0.0475), not the 0.025 per dose (total 0.05) stated in the methods. All P values are <0.001, so no rejection changes. The authors should say which alpha was actually used.
Second check: another model agreed
Report a problem with this finding
18 minor (typos and the like; most checked by the judge only)
Reference or label "hypothetical strategy (Table S8 and Fig. S6)" is cited for physical-function and week-36 pain results, but Fig. S6 is "WOMAC Pain Reduction Targets — Efficacy Estimand Results".
- p. 10 Results, Key Secondary End Points
hypothetical strategy (Table S8 and Fig. S6)
- p. 40 Figure S6 title
Figure S6. WOMAC Pain Reduction Targets — Efficacy Estimand Results
Judge's note: Main p10 cites Fig. S6 in a sentence about physical function at week 68/80 and pain at week 36. Fig. S6 (supp-01 p40) shows only the proportions reaching ≥30%, ≥50% and ≥70% WOMAC pain reduction under the efficacy estimand. It has no physical-function or week-36 data. This is the only place the article cites Fig. S6. The figure is related material but does not show the end points named.
Second check: another model agreed
Report a problem with this finding
Reference or label "Similar results were observed with the intention-to-treat estimand (Table S7)" follows the week-36 pain result, but Table S7 has no week-36 data.
- p. 10 Results, Key Secondary End Points
Similar results were observed with the intention-to-treat estimand (Table S7)
- p. 63 Table S7
Table S7. Primary and Key Secondary Pain-related Endpoints – ITT Estimand
Judge's note: Table S7 (supp-01 p63–64) has ITT results for the primary pain end point, the 4-mg pain end point, physical function at week 68/80, and the ≥50% responder end points. It has no WOMAC pain at week 36. The pointer supports the physical-function statement but not the week-36 statement in the preceding sentence. This is minor ambiguity about scope.
Second check: another model agreed
Report a problem with this finding
Reference or label Cardiometabolic results for the treatment-regimen "and the efficacy estimand ... are shown in Table S9 and Figure S11", but the efficacy-estimand cardiometabolic results are in Table S8.
- p. 11 Results, Additional Secondary End Points
estimand and the efficacy estimand that followed the hypothetical strategy, are shown in Table S9 and Figure S11
- p. 70 Table S9 title
Treatment Regimen Estimand for Non-pain-related, Cardiometabolic parameters
Judge's note: The Table S9 title (supp-01 p70) restricts it to the treatment-regimen estimand for cardiometabolic parameters. The efficacy-estimand values for blood pressure, lipids, hsCRP and fasting insulin are in Table S8 (supp-01 p67–69); the p69 footnote names the efficacy estimand data-points set. Figure S11 covers systolic blood pressure only. The pointer should include Table S8.
Second check: another model agreed
Report a problem with this finding
Reference or label supp-03: the same rationale sentence cites "(CDC 1995; Hunter et al. 2014)" in the Synopsis (p895) but "(CDC 1995; Hunter et al. 2019)" in Section 2.1 Study Rationale (p925).
- p. 925
(CDC 1995; Hunter et al. 2019)
Only the authors can say what is right here.
Judge's note: Both page images confirm the different years for the same sentence and claim. The protocol's reference list (protocol page 166) is not within the in-scope pages, and 'Hunter' appears nowhere else in the extracted supp-03 text. So I cannot tell which year is intended. The authors should say which one is correct.
Second check: another model agreed
Report a problem with this finding
Wording "Enrollment of female participants was capped at approximately 70%" sits among criteria for both trials, but Table 1 shows 435 (75.8) women in TRIUMPH-1-OA.
- p. 6 Methods, Participants
Enrollment of female participants was capped at approximately 70%
- p. 3 Table 1, Female sex, TRIUMPH-1-OA total
435 (75.8)
Only the authors can say what is right here.
Judge's note: On main p6 the sentence comes straight after 'Key exclusion criteria for both trials...' and is not restricted to one trial. Table 1 (p3 image) shows female sex at 75.5–76.4% in every TRIUMPH-1-OA arm and 435/574 overall. The '70% upper limit' wording appears only in the TRIUMPH-4 protocol and SAP (supp-03 and supp-02, both GZBN). The TRIUMPH-1 protocol is not among the files. The authors should say whether the cap applied to TRIUMPH-1-OA and, if so, why enrolment exceeded it.
Second check: another model agreed
Report a problem with this finding
Wording The TRIUMPH-4 Study Investigator List includes a "United Kingdom" section (3 sites), whereas TRIUMPH-4 is described as run "in 5 countries" and Table S1 shows no UK participants for TRIUMPH-4.
- p. 6 TRIUMPH-4 Study Investigator List
United Kingdom
- p. 52 Table S1, Country, United Kingdom row, TRIUMPH-4
United Kingdom - - - -
- p. 61 Table S5
in 5 countries across North America, Europe, and Asia Pacific
Only the authors can say what is right here.
Judge's note: Supp-01 p6 (image) lists three UK investigators under the TRIUMPH-4 list. The full TRIUMPH-4 list (p4–7) covers six countries with 48 named sites. Main p2 says '49 centers in 5 countries', and Table S5 says 5 countries. Table S1 shows '-' for United Kingdom in every TRIUMPH-4 column. The UK sites may have been activated without enrolling anyone, but the supplement does not say so.
Second check: another model agreed
Report a problem with this finding
Wording Figure 2A gives "Retatrutide, 12 mg 147" at baseline and for EE, HYB and ITT, whereas 148 were randomized (Table 1, Table 2, Figure 1A, and Table S7 ITT N=148).
- p. 10 Figure 2A, No. of participants, Retatrutide 12 mg
Retatrutide, 12 mg 147
- p. 8 Figure 1A, No. of participants, Retatrutide 12 mg
Retatrutide, 12 mg 148
Only the authors can say what is right here.
Judge's note: Figure 2A (main p10 image) shows 147 for the 12-mg group in every column, including baseline and ITT. Figure 1A shows 148 at baseline, EE, HYB and ITT. Table S7, the ITT physical-function analysis, lists 12-mg N=148. Main p6 says efficacy analyses included all randomized participants. A missing physical-function baseline value could explain the 147, but the paper does not say so.
Second check: another model agreed
Report a problem with this finding
Typo Supplement p.32: "stratification factors (defined by joint levels of sex and prediabetes status [Yes or No], baseline WOMAC pain severity (severe [≥7] and moderate [<7]), an indicator of concomitant pain medication use at baseline, and their interactions with treatment." The parenthesis opened at "(defined" is never closed.
- p. 32
an indicator of concomitant pain medication use at baseline, and their interactions with treatment.
Judge's note: Checked the page image. The only ')' closes '(severe ...'. The '(defined by' bracket stays open to the end of the sentence.
Second check: none (judge only)
Report a problem with this finding
Typo Table S5, race row: "The prevalence of obesity is the highest among Black adults (49.9%, followed by Hispanic adults (45.6%)..." The parenthesis before 49.9% is never closed.
- p. 61 Table S5, Race or ethnic group
Black adults (49.9%, followed by Hispanic adults (45.6%)
Judge's note: The page image shows the same missing closing parenthesis, so it is not an extraction artefact.
Second check: none (judge only)
Report a problem with this finding
Typo Table S6, TRIUMPH-1-OA, Discontinued study treatment, RETA 4-mg: "20 (14)" is printed without the decimal used throughout the table (should be 14.0).
- p. 62 Table S6, TRIUMPH-1-OA, Discontinued study treatment, RETA 4-mg
20 (14)
Judge's note: The page image shows '20 (14)'. Figure S5 gives the same count as '20 (14.0%)'.
Second check: none (judge only)
Report a problem with this finding
Typo Table footnote, supplement p.62/63: "Data presented as model-based estimate (standard error) unless specified for week 68 (TRIUMPH-4) and week 80 (TRIUMPH-1 OA) unless stated otherwise." The 'unless' clause is redundant.
- p. 63
unless specified for week 68 (TRIUMPH-4) and week 80 (TRIUMPH-1 OA) unless
Judge's note: Checked the page image. The two 'unless' clauses are not adjacent duplicated words, but the sentence is redundant and garbled. This is cosmetic only.
Second check: none (judge only)
Report a problem with this finding
Typo Table S8, hsCRP treatment comparison, TRIUMPH-4 9 mg: "--33.3 (-50.4), -10.3)" has a doubled minus sign and a misplaced parenthesis; intended "-33.3 (-50.4, -10.3)".
- p. 69 Table S8, hsCRP treatment comparison, TRIUMPH-4 9 mg
--33.3 (-50.4), -10.3)
Judge's note: The page image shows the malformed cell exactly as quoted. The other cells use the form 'estimate (lower, upper)'.
Second check: none (judge only)
Report a problem with this finding
Typo Supplement reference 5 prints an unrendered HTML entity: "1990–2020" instead of "1990–2020".
- p. 81 Supplement reference 5
1990–2020
Judge's note: The entity is visible on the page image of the reference list (PDF p81).
Second check: none (judge only)
Report a problem with this finding
Typo supp-03 p940, exclusion criterion 16: "there is an any condition that could confound the participant's assessment". The stray word "an" should be removed.
- p. 940
there is an any condition that could confound the
Judge's note: The page image shows 'there is an any condition'. Cosmetic.
Second check: none (judge only)
Report a problem with this finding
Typo SAP amendment table (supp-02 p.1137): "described more fully in the compound compound-level safety standards" has a duplicated word.
- p. 1137
e described more fully in the compound compound-level safety standa
Judge's note: The page image shows 'compound' at the end of one line and 'compound-level' at the start of the next.
Second check: code check
Report a problem with this finding
Typo SAP endpoint table (supp-02 p.1147): "percent change in hsCR" should read hsCRP.
- p. 1147
percent change in hsCR
Judge's note: The page image shows 'hsCR' with the P missing.
Second check: none (judge only)
Report a problem with this finding
Typo supp-02 p1156: study identifier printed as "Study J1-I-MC-GZBJ (GZBJ)". Everywhere else the document uses the form J1I-MC-xxxx (e.g. "J1I-MC-GOA1" in the next sentence, "J1I-MC-GZBN" in the running header).
- p. 1156
Of note, Study J1-I-MC-GZBJ (GZBJ) is a randomized
Judge's note: The page image shows 'J1-I-MC-GZBJ' with an extra hyphen. This is the only occurrence of that form in the extracted text. The sibling identifiers on the same page and in the header use 'J1I-MC-'. Cosmetic formatting slip.
Second check: none (judge only)
Report a problem with this finding
Typo supp-02 p1214: "Approximately 1400 participants will be screened to achieved approximately 435 participants". The phrase "to achieved" should read "to achieve".
- p. 1214
Approximately 1400 participants will be screened to achieved approximately 435 participants
Judge's note: The page image (SAP section 5, Sample Size Determination) shows 'to achieved'. Grammatical slip.
Second check: none (judge only)