# Reviewer A

## Comments

There are a number of issues relating to the methodology, the statistical analysis, and the interpretation of results that need to be addressed.

### Statistical points

1. The use of the composite primary outcome needs to be justified. Surely the decision for "red cell transfusion within 2 days" is very subjective? What is the overlap between the two components of the outcome? For those having red cell transfusion, what was their blood loss?

2. Statistical testing of baseline characteristics between randomized groups is incorrect and should be omitted.

3. The statistical model used to assess the primary outcome (log-binomial mixed-effect regression model with study centre as a random effect) is appropriate. However, there are no covariates in the model. It is convention to include all variables which have been stratified for in the randomization (i.e. age and placenta praevia type). From Figure 2 it is clear that the PPH incidence is very different for the placenta praevia types.

4. PPH incidence was found to be 29.7% and 35.1% in the tranexamic group and placebo group respectively with a RR = 0.83. Thus, the observed (adjusted) relative reduction was 17%. This is less than the "minimal important relative difference" of 20% on which the sample size calculation was based. Hence, the conclusion should be moderated.

### General points

1. The composite primary outcome blends a formula-based blood loss estimate with a clinical decision (transfusion). Given acknowledged discrepancies between calculated vs gravimetric EBL and the influence of antepartum bleeding, clinical thresholds, and center practice, more discussion of the interpretability and measurement variability would be helpful.

2. Following above, is the primary outcome patient-centered enough? Have the authors considered using certain clinical outcome(s) that matter more directly to patients?

3. "Participants were interviewed 6 weeks after delivery via telephone. They were asked about … and the development of adverse events." Are the adverse events gathered through interviews with participants? How was this data corroborated by medical records, or verified in another way? How would the authors comment on potential recall bias, or the accuracy of information from patients on, for example, acute liver injury or pelvic abscess diagnosis?

4. The trial is underpowered to detect SAEs, therefore "no increase in serious adverse events" should be interpreted as "no signal detected," not proof of safety equivalence. This statement should be modified.

5. Please revise the data and code sharing statement to ensure it fully complies with the requirement that the data underlying the results be available at the time of publication, in a repository with open licences and a stable persistent identifier (DOI).

6. To facilitate the review process, the statistical reviewer may require access to the data when assessing the revised manuscript. Please make the data available upon submission of the revision.

7. Reporting of adverse events is sparse in the main manuscript. A clear definition of AEs and SAEs, with more detailed results, should be included in the main text, rather than referring readers elsewhere. Similarly, please also summarise results for maternal satisfaction and Edinburgh scores in the main text.

8. The acronym of the trial is confusing in the context of the French trials. Can you omit or amend it?