Effect of Early Tranexamic Acid Administration on Mortality, Hysterectomy, and Other Morbidities in Women with Post-Partum Haemorrhage (WOMAN)
2017 · paper · Clinical trials
An international placebo-controlled trial of tranexamic acid in about 20,000 women with post-partum haemorrhage. Death from bleeding was modestly reduced, especially with treatment within three hours of birth; hysterectomy and the combined death-or-hysterectomy outcome were not reduced.
Checked: PDF from europepmc.org, fetched 4 Oct 2026. Page numbers are those of that PDF.
9 findings
Substance Hysterectomy result printed as "358 [3·6%] ... vs 351 [3·5%] ... RR 1·02, 95% CI 0·88–1·07; p=0·84" (Abstract p1 and Results p6). The upper limit 1·07 does not fit the counts, which give about 0·88–1·17.
- p. 1 Abstract, Findings
RR 1·02, 95% CI 0·88–1·07; p=0·84
- p. 6 Results, hysterectomy paragraph
RR 1·02, 95% CI 0·88–1·07; p=0·84
Only the authors can say what is right here.
Judge's note: I checked both page images (p1 Abstract Findings, p6 Results right column), and the interval is printed identically in both places. From 358/10036 vs 351/9985 I get RR=1·015, SE(log RR)=sqrt(1/358-1/10036+1/351-1/9985)=0·0738, and a 95% CI of exp(0·0147±0·1446)=0·878–1·173. The printed lower bound and p=0·84 (z≈0·20) fit this calculation; the printed upper bound does not. On the log scale the printed interval runs about −0·15 and +0·05 around the estimate, so it cannot be a Wald-type interval around 1·02. No adjusted hysterectomy analysis is reported that could account for it, and there is no erratum in the files. 1·17 is the likely intended value, but only the authors can confirm it.
Second check: another model agreed
Report a problem with this finding
Substance Results p6 prints "546 [5·6%] placebo group" for the composite primary endpoint; the Abstract and Table 3 print "546 [5·5%]" / "546/9985 (5·5%)".
- p. 6 Results, primary endpoint paragraph
546 [5·6%] placebo group
- p. 1 Abstract
546 [5·5%]
- p. 6 Table 3, All patients
546/9985 (5·5%)
Judge's note: Seen on the p6 image (both the text and Table 3) and in the p1 abstract. The count, population and endpoint are the same in all three places. 546/9985 = 5·47%, so 5·5% is correct and the 5·6% in the text is not.
Second check: another model agreed
Report a problem with this finding
Substance Results p5 say "20 002 (99·7%) received the first dose ... (10 037 received tranexamic acid and 9975 received placebo; figure 1)", but Figure 1 shows "10 027 received allocated dose 1", and 10 027+9975=20 002.
- p. 5 Results, first paragraph
10 037 received tranexamic acid and 9975 received placebo
- p. 4 Figure 1 trial profile
10 027 received allocated dose 1
Judge's note: Checked on the images of p5 and p4. The text's own total, 20 002, requires 10 027. The 10 037 in the text conflicts with both that total and Figure 1, which the sentence cites.
Second check: another model agreed
Report a problem with this finding
Substance Results p5: "20 002 (99·7%) received the first dose ... (10 037 received tranexamic acid and 9975 received placebo; figure 1)". 10 037 + 9975 = 20 012, and Figure 1 gives "10 027 received allocated dose 1".
- p. 5
of whom 20 002 (99·7%) received the first dose of the
Judge's note: On p5 the text prints 10 037. Figure 1 on p4 prints 10 027 for the tranexamic acid arm and 9975 for placebo. 10 027 + 9975 = 20 002, which matches the printed total, so the 10 037 in the text looks like a transcription slip for 10 027.
Second check: another model agreed
Report a problem with this finding
Reference or label "We published a statistical analysis plan before the allocation was unblinded.18" (p3), and "...time to treatment interaction.18" (p5), cite ref 18 (Prieto-Merino et al, Dangers of non-specific composite outcome measures, BMJ 2013). The WOMAN statistical analysis plan is ref 17 (Shakur et al, Trials 2016).
- p. 3 Statistical analysis, first sentence
We published a statistical analysis plan before the allocation was unblinded.18
- p. 5 Methods, subgroup paragraph end
data on the time to treatment interaction.18
- p. 12 Reference 17 and 18
statistical analysis plan for the WOMAN trial. Trials 2016; 17: 249.
- p. 3 Statistical analysis, first sentence
allocation was unblinded.18
Judge's note: Reference list on p12 checked: 17 is the WOMAN SAP and 18 is the composite-outcomes BMJ paper. Ref 17 is first cited only in the Discussion ("post-partum haemorrhage death.17"), after ref 18 has already been cited twice in the Methods, so the citation order is broken. Ref 18 is cited appropriately in the Discussion on all-cause mortality. The Methods sentences describe the published SAP and its prespecification, which is ref 17.
Second check: another model agreed
Report a problem with this finding
Wording The primary outcome is defined as "hysterectomy within 42 days of randomisation" in Methods (p3), but as "within 42 days of giving birth" in the Abstract, Results (p6) and Discussion.
- p. 3 Outcomes
hysterectomy within 42 days of randomisation
- p. 6 Results
hysterectomy within 42 days of giving birth occurred in 1080 women
- p. 1 Abstract Methods
hysterectomy within 42 days of giving birth
Only the authors can say what is right here.
Judge's note: Methods Outcomes (text lines 263–264) says randomisation, and follow-up is also stated as 42 days after randomisation. The Abstract, the Results p6 and the Discussion say giving birth. Some women were randomised more than 3 h after birth, so the time origin is stated two ways.
Second check: another model agreed
Report a problem with this finding
Wording In Table 1, "Placenta fully delivered" for placebo is Yes 9016 (90%) and No 990 (10%), which sums to 10 006 against n=10 009, and no Unknown row is shown.
- p. 4 Table 1, Placenta fully delivered, placebo
9016 (90%)
Only the authors can say what is right here.
Judge's note: Checked on the image of p4. Other placebo variables sum to 10 009 and list Unknown explicitly, even when it is 0 (blood loss Unknown 0). The tranexamic acid column for this row sums to 10 051. The 3 placebo women are not accounted for.
Second check: another model agreed
Report a problem with this finding
Wording In Table 1, "Clinical signs of haemodynamic instability" for placebo is Yes 5898 (59%) and No 4110 (41%), which sums to 10 008 against n=10 009, and no Unknown row is shown.
- p. 4 Table 1, Clinical signs of haemodynamic instability, placebo
4110 (41%)
Only the authors can say what is right here.
Judge's note: Checked on the image of p4. The tranexamic acid column sums to 10 051. Other variables show Unknown explicitly. One placebo woman is not accounted for.
Second check: another model agreed
Report a problem with this finding
Wording Figure 5 prints 'p<0·0000*' (footnote: '*Heterogeneity p value'), which is not a valid bound; it is also unclear which comparison it refers to, since the >3 h stratum already carries p=0·17*.
- p. 9 Figure 5, below >3 hours Overall row
p<0·0000*
Only the authors can say what is right here.
Judge's note: The Figure 5 image on p9 shows 'p<0·0000*' beside the >3 hours Overall row. A p value cannot be below 0, so the intended threshold and test (presumably the heterogeneity between time strata) need the authors to clarify.
Second check: another model agreed
Report a problem with this finding
25 minor (typos and the like; most checked by the judge only)
Reference or label Reference 15 cites 'Good Clinical Practice E6(R1). 1996 June 10.' but its URL points to '.../E6/E6_R2__Step_4.pdf'.
- p. 12 Reference 15
Good Clinical Practice E6(R1). 1996 June 10.
- p. 12 Reference 15 URL
E6_R2__Step_4.pdf
Only the authors can say what is right here.
Judge's note: The p12 reference list image confirms that the R1 version and 1996 date sit with a URL to the R2 document, so the entry is internally inconsistent about which version is cited. The 'International Conference' (p2 text) versus 'International Council' (reference) difference is not counted as a defect: the text names the body that issued the GCP guideline, and the reference uses its current name. The claim that the organisation was renamed rests on outside knowledge.
Second check: another model agreed
Report a problem with this finding
Wording Placebo composite primary endpoint is "546 [5·5%]" in the Abstract and Table 3 (546/9985 (5·5%)) but "546 [5·6%] placebo group" in the Results text on p6. 546/9985 = 5·47%, so the Results-text figure is the one that is off.
- p. 1
in the placebo group, RR 0·97, 95% CI 0·87-1·09; p=0·65
Judge's note: Abstract p1, Table 3 p6 and Results text p6 all checked on the images. The abstract value the cluster quotes is correct. The inconsistent value is in the Results paragraph on p6 (text line 707), where 5·6% is printed for the same quantity.
Second check: another model agreed
Report a problem with this finding
Wording Table 3 uses "534/10 032 (5·3%)" for the tranexamic acid composite primary endpoint. However, Figure 1 and the Abstract say "10 036 included in analysis", and the Figure 1 footnote defines the excluded women as those without primary-endpoint information.
- p. 6 Table 3, All patients
534/10 032 (5·3%)
- p. 4 Figure 1
10 036 included in analysis
- p. 1 Abstract
of whom 10 036 and 9985, respectively, were included in the analysis
Only the authors can say what is right here.
Judge's note: Seen on the p4 Figure 1 and on Table 3 (p6). Table 4 and Figure 4 also use 10 032, while the death analyses (Table 2, Figure 3) use 10 036. The methods (p4) mention a complete-case analysis with no imputation, which could account for 4 women missing hysterectomy or surgery data. But Figure 1 says all 10 036 have primary-endpoint information, and the 4 women are not explained. Only the authors can reconcile this.
Second check: another model agreed
Report a problem with this finding
Wording The death-due-to-bleeding result within 3 h is printed as "RR 0·69, 95% CI 0·52–0·91" in the Abstract and Results but as "0·69 (0·53–0·90)" in the Figure 5 WOMAN ≤3 h row, with the same counts (89 vs 127).
- p. 6 Results
RR 0·69, 95% CI 0·52–0·91; p=0·008
- p. 9 Figure 5, ≤3 hours WOMAN row
0·69 (0·53–0·90)
- p. 1 Abstract, Findings
RR 0·69, 95% CI 0·52–0·91; p=0·008
Only the authors can say what is right here.
Judge's note: Seen on the p1, p6 and p9 images. A crude log-RR interval from 89/7520 vs 127/7408 gives SE about 0·137 and an interval of about 0·527–0·903, which matches Figure 5. The text's 0·52–0·91 may come from a different method (for example a regression model), but neither place says so. Query: which method or model gives each interval?
Second check: another model agreed
Report a problem with this finding
Wording In Table 1, "Baby delivered in the randomising hospital" for placebo shows "8756 (88%)" and "1251 (13%)". With n=10 009 these round to 87% and 12%.
- p. 4 Table 1, Baby delivered in the randomising hospital, placebo
8756 (88%)
- p. 4 Table 1, same row, No
1251 (13%)
Judge's note: Checked on the image of p4. The row sums to 10 009 (8756+1251+2), so the denominator is the group total. 87·48% and 12·50% (12·499%) were both rounded up. The tranexamic acid percentages in the same row are correct.
Second check: another model agreed
Report a problem with this finding
Wording In Table 2, organ failure for tranexamic acid is "25 (0·3%)". 25/10 036 = 0·249%, which rounds to 0·2%.
- p. 5 Table 2, Organ failure, tranexamic acid
25 (0·3%)
Judge's note: Checked on the image of p5. The denominator n=10 036 is given in the table header. The other cells in the table round correctly.
Second check: another model agreed
Report a problem with this finding
Wording In Table 2, sepsis for tranexamic acid is "15 (0·2%)". 15/10 036 = 0·149%, which rounds to 0·1%.
- p. 5 Table 2, Sepsis, tranexamic acid
15 (0·2%)
Judge's note: Checked on the image of p5. The value is below 0·15, so it is not a rounding tie.
Second check: another model agreed
Report a problem with this finding
Wording Results p6: "191 (27%) were within 1 h of randomisation (figure 2)". In Figure 2B, the total hour-0 bar reaches about 250; only its 'for bleeding' segment is about 190.
- p. 6 Results, hysterectomy paragraph
191 (27%) were within 1 h of randomisation (figure 2)
- p. 5 Figure 2B, hour 0 bar
Number of hysterectomies*
Only the authors can say what is right here.
Judge's note: Seen on the zoomed Figure 2B crop. The hour-0 bar totals about 248 and its bleeding segment tops out near 191. For deaths, the text's '43 within 1 h' matches the full hour-0 bar in Figure 2A (about 40 + 3). Under the same convention, 191 matches only the bleeding hysterectomies. The figure is also hard to reconcile internally: its footnote excludes 311 pre-randomisation hysterectomies, yet the bars sum to about 600. Values read from a bar chart are approximate, so only the authors can say which number is meant.
Second check: another model agreed
Report a problem with this finding
Typo Table 5 row header reads 'ED-5Q+' where the instrument is the EQ-5D ('quality of life measured using the EQ5D' in Methods); the '+' mark has no footnote, since the footnote defines only '*'.
- p. 8 Table 5 row label
ED-5Q+
- p. 3 Outcomes
quality of life measured using the EQ5D
Judge's note: The Table 5 image on p8 shows 'ED-5Q+'. The footnote reads 'Data are n (%), unless otherwise indicated. *Fatal or non-fatal. RR=relative risk.' and has no '+' entry. Methods (p3) say EQ5D. The letters are transposed and the marker is undefined.
Second check: none (judge only)
Report a problem with this finding
Typo Figure 1 has "10 051 baseline data availble" and "10 009 baseline data availble", a misspelling of "available".
- p. 4 Figure 1
10 051 baseline data availble
- p. 4 Figure 1 trial profile, both arms
baseline data availble
Judge's note: Visible in both arms of the trial profile on the image of p4.
Second check: none (judge only)
Report a problem with this finding
Typo Methods say "independent data safety and monitoring committee" (p4), while the collaborators list says "Data Monitoring and Ethics Committee" (p10).
- p. 4 Statistical analysis
independent data safety and monitoring committee
- p. 10 Collaborators
Data Monitoring and Ethics Committee
Judge's note: Both passages were read in the text. Only one oversight committee is listed, so the two names appear to refer to the same body. The Methods wording may simply be descriptive, so this is a naming consistency point only.
Second check: none (judge only)
Report a problem with this finding
Typo In Table 2, pulmonary embolism for placebo is printed "11 (0·1)" without a % sign, and bleeding for placebo is "191 (1·9 %)" with a stray space.
- p. 5 Table 2, Pulmonary embolism, placebo
11 (0·1)
Judge's note: Both are visible on the image of p5. All other cells use the n (x%) format.
Second check: none (judge only)
Report a problem with this finding
Typo RR is defined as "RR=risk ratio" in the Table 2 footnote and the text, but as "RR=relative risk" in the Table 4 and Table 5 footnotes.
- p. 7 Table 4 footnote
RR=relative risk.
- p. 5 Table 2 footnote
RR=risk ratio.
- p. 7 Table 4 footnote
RR=relative risk
- p. 5 Table 2 footnote
RR=risk ratio
Judge's note: Seen on the p5 Table 2 footnote and the p7 and p8 footnotes. The two terms are near-synonyms, so this is only a terminology inconsistency.
Second check: none (judge only)
Report a problem with this finding
Typo Table 3, ≤1 h, placebo: "229/4726 (4·9%)". 229/4726 = 4·85%, which rounds to 4·8%.
- p. 6 Table 3, ≤1 h, placebo
229/4726 (4·9%)
Judge's note: Seen on the p6 Table 3 image. The value is not a rounding tie. The other percentages in the table round correctly.
Second check: code check
Report a problem with this finding
Typo Table 3, Caesarean section, placebo: "257/2873 (9·0%)". 257/2873 = 8·95%, which rounds to 8·9%.
- p. 6 Table 3, Caesarean section, placebo
257/2873 (9·0%)
Judge's note: Seen on the p6 Table 3 image. 8·945% is not a tie, so it should round to 8·9%.
Second check: code check
Report a problem with this finding
Typo Table 3, Other or unknown, placebo: "272/3652 (7·5%)". 272/3652 = 7·45%, which rounds to 7·4%.
- p. 6 Table 3, Other or unknown, placebo
272/3652 (7·5%)
Judge's note: Seen on the p6 Table 3 image. 7·448% should round to 7·4%.
Second check: code check
Report a problem with this finding
Typo Results p6: "578 (81%) of 709 hysterectomies". 578/709 = 81·5%, which rounds to 82%.
- p. 6 Results, hysterectomy paragraph
578 (81%) of 709 hysterectomies
- p. 6
578 (81%) of 709
Judge's note: Seen on the p6 image. The same paragraph rounds 608/709 to 86% and 191/709 to 27%, so it rounds elsewhere. 81·52% should round to 82%.
Second check: code check
Report a problem with this finding
Typo Results p6: "Brace sutures were used more often in the tranexamic group (300". The word 'acid' is missing; every other mention reads 'tranexamic acid group'.
- p. 6 Results, surgical interventions paragraph
more often in the tranexamic group (300
Judge's note: Seen on the p6 image. This is a cosmetic, missing-word inconsistency.
Second check: none (judge only)
Report a problem with this finding
Typo Results p6: "p=0·70).There was no heterogeneity". The space after the full stop is missing.
- p. 6 Results, time to treatment
p=0·70).There
Judge's note: The p6 image shows the missing space in the typeset text as well, so this is not only a text-layer artefact.
Second check: none (judge only)
Report a problem with this finding
Typo Figure 4, Other/unknown, tranexamic acid: "45/3606 (1·3%)". 45/3606 = 1·25%, which rounds to 1·2%.
- p. 7 Figure 4, Other/unknown, tranexamic acid
45/3606 (1·3%)
Judge's note: Seen on the p7 Figure 4 image. 1·248% is not a tie.
Second check: code check
Report a problem with this finding
Typo Results p7: "18 (48·7%) were in the tranexamic acid group and 19 (51·4%)". 18/37 = 48·6%, and the printed shares sum to 100·1%.
- p. 7 Results, transfusion paragraph
18 (48·7%) were in the tranexamic acid group and 19 (51·4%)
- p. 7 Results, transfusion
18 (48·7%) were in the tranexamic acid group
- p. 7
Of these, 18 (48·7%) were in the tranexamic acid group
Judge's note: Seen on the p7 image. 18/37 = 48·649%, which is not a tie and rounds to 48·6%. 19/37 = 51·35%, which rounds to 51·4%.
Second check: none (judge only)
Report a problem with this finding
Typo Discussion p8: 'the decision to enrol a women into the trial' should read 'a woman'.
- p. 8 Discussion, second paragraph
the decision to enrol a women into the trial
- p. 8 Discussion, para 2
enrol a women into the trial
Judge's note: The page image confirms the text as quoted.
Second check: none (judge only)
Report a problem with this finding
Typo The National Coordinators list (p10) reads 'Chrstine Biryabarema (Uganda)'; the Uganda site list (p11) gives 'Christine Biryabarema'.
- p. 10 National Coordinators
Chrstine Biryabarema (Uganda)
- p. 11 Uganda site list
Christine Biryabarema
Judge's note: Both images confirm the two spellings.
Second check: none (judge only)
Report a problem with this finding
Typo The trial coordinating team list (p10) contains the misspellings 'statisticial support' and 'project administator 2010–2015'.
- p. 10 Trial coordinating team
statisticial support
- p. 10 Trial coordinating team
administator 2010–2015
- p. 10 Trial coordinating team
project administator
Judge's note: Both misspellings are visible in the page image.
Second check: none (judge only)
Report a problem with this finding
Typo 'Pakistan(5282)' has no space before the count, whereas other entries read e.g. 'Nigeria (5711)'.
- p. 10 Trial sites list
Pakistan(5282)
Judge's note: In the p10 image there is no visible gap between the italic 'Pakistan' and '(5282)', while 'Nigeria (5711)' has a clear space. This is purely cosmetic.
Second check: none (judge only)