This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
ETAPPH was a completed phase 3, randomized, double-blind, parallel-group trial evaluating tranexamic acid injection for prevention of postpartum haemorrhage after caesarean section. The registry reports 1226 participants, with 611 assigned to the tranexamic acid study group and 613 assigned to the placebo control group.
| Feature | ETAPPH |
|---|---|
| Phase | Phase 3 |
| Condition | Postpartum Hemorrhage |
| Population | Participants undergoing caesarean section for whom postpartum haemorrhage prevention was evaluated |
| Design | Randomized, double-blind, parallel-group |
| Primary purpose | Prevention |
| Enrollment | 1226 |
| Primary endpoint | Number of Participants With Postpartum Hemorrhage (PPH) |
| Primary endpoint type | Binary |
| Primary endpoint time frame | Up to day 2 postpartum |
| Primary statistical method | Chi-squared test |
| Effect measure | Risk Ratio (RR) |
| ClinicalTrials.gov | NCT04733157 |
| Lead sponsor | University of Zimbabwe |
2. Clinical Question
The central statistical question was whether tranexamic acid injection changed the risk of postpartum haemorrhage compared with normal saline placebo after caesarean section.
Population
Participants undergoing caesarean section in a phase 3 prevention trial evaluating postpartum haemorrhage.
Intervention
Tranexamic acid injection.
Comparator
Normal saline placebo.
Primary question
Does tranexamic acid change the occurrence of postpartum haemorrhage through up to day 2 postpartum?
3. Trial Design
Tranexamic acid
- Tranexamic acid injection
- Randomized study group
- Double-blind comparison
- PPH assessed up to day 2 postpartum
Normal saline placebo
- Normal saline placebo
- Randomized control group
- Double-blind comparison
- PPH assessed up to day 2 postpartum
The allocation structure is important for interpreting the risk ratio. Randomization creates the basis for comparing the incidence of the binary outcome between the two groups, while double-blinding reduces the opportunity for knowledge of treatment assignment to influence trial conduct or outcome assessment.
4. Trial Timeline
Study start
The registry records March 23, 2021 as the trial start date.
Primary completion
The registry records December 14, 2021 as the primary completion date.
Registry status
The ClinicalTrials.gov record lists the study status as completed.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame | Type |
|---|---|---|---|
| Number of Participants With Postpartum Hemorrhage (PPH) | Calculated estimated blood loss exceeding 1000ml. Estimated Blood Loss (EBL) was calculated using laboratory values of hemoglobin levels before and after procedure. Postpartum hemorrhage is defined as blood loss exceeding 1000mL after cesarean section. | Up to day 2 postpartum | Binary |
Statistically, this endpoint converts a continuous clinical quantity—estimated blood loss—into a binary classification: whether the participant met the registry's threshold for postpartum haemorrhage. The primary analysis therefore compares the proportion classified as having PPH between randomized groups.
Why the binary definition matters
Once estimated blood loss is classified using the stated threshold, the primary analysis no longer directly estimates an average difference in blood loss. Instead, it asks whether the probability of crossing the PPH threshold differs between the two groups.
This distinction is important because two treatments can have similar rates of crossing a threshold while still producing different distributions of continuous blood loss. Conversely, a difference in the proportion crossing the threshold can occur even when the underlying continuous measurements overlap substantially.
6. Statistical Methodology
Chi-squared test
The posted statistical analysis used a chi-squared test for the binary PPH endpoint. This method evaluates whether the observed distribution of participants across the two outcome categories differs between the randomized treatment groups beyond what would be expected from sampling variation under the null hypothesis of no association between treatment group and PPH status.
The chi-squared statistic summarizes how far the observed cell counts depart from the counts expected under independence between treatment assignment and the binary outcome.
Risk ratio
The effect measure reported for the primary endpoint was the risk ratio (RR). For a binary outcome, the risk ratio compares the probability of the event in the treatment group with the probability of the event in the control group.
An RR of 1 corresponds to equal observed risks. An RR below 1 corresponds to a lower observed risk in the tranexamic-acid group, while an RR above 1 corresponds to a higher observed risk.
Confidence interval
The reported 95% confidence interval gives an interval estimate around the observed risk ratio. It is useful because the point estimate alone does not communicate how precisely the randomized comparison estimates the underlying relative risk.
Missing data
The registry specifically reports missing hemoglobin assessments for the PPH endpoint: 93 participants in Group A and 87 participants in Group B had missing assessments. The statistical analysis record identifies multiple imputation / missing data as an additional concept associated with the analysis.
Missing laboratory measurements matter because the binary PPH classification depends on hemoglobin measurements before and after the procedure. If outcome status cannot be determined for some participants, the effective information available for the primary endpoint differs from the total randomized enrollment.
7. Primary Result: Postpartum Haemorrhage
The primary endpoint analysis compared the occurrence of postpartum haemorrhage through up to day 2 postpartum between the tranexamic acid group and the placebo group.
Risk ratio for postpartum haemorrhage
95% CI: 0.69–1.09 · P = 0.33
Chi-squared analysis; tranexamic acid versus normal saline placebo
| Primary endpoint | Time frame | Analysis | Effect estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Number of Participants With Postpartum Hemorrhage (PPH) | Up to day 2 postpartum | Chi-squared | RR 0.87 | 0.69–1.09 | 0.33 |
The estimated risk ratio of 0.87 indicates that the estimated risk of postpartum haemorrhage in the tranexamic acid group was 0.87 times the corresponding risk in the placebo group. Expressed as a relative comparison, an RR of 0.87 corresponds to a 13% lower estimated risk in the tranexamic acid group relative to placebo.
The estimate does not mean that 13% of participants were protected from PPH, nor does it establish a 13% absolute reduction in risk. A risk ratio is relative: the absolute difference depends on the underlying event risks in the two groups.
The 95% confidence interval of 0.69–1.09 describes uncertainty around the estimated relative risk. Because the interval spans 1, the data are compatible with a lower relative risk as well as a relative risk slightly above 1 under the stated analysis framework.
The P-value of 0.33 is evidence about the compatibility of the observed data with the null hypothesis used by the statistical test. It is not a measure of the magnitude or clinical importance of the effect. A P-value does not tell us that the treatment effect is 33%, nor does it quantify the probability that the treatment has no effect.
The interpretation also depends on the handling of missing hemoglobin measurements. The registry reports 93 missing PPH assessments in Group A and 87 in Group B and identifies multiple imputation / missing data as part of the analysis concepts.
8. Understanding the Primary Estimate
Point estimate
RR 0.87 is the single estimated relative comparison reported for PPH between tranexamic acid and placebo.
Direction
Because the estimate is below 1, its direction corresponds to lower estimated PPH risk in the tranexamic acid group.
Precision
The 95% CI of 0.69–1.09 shows that the estimate has uncertainty broad enough to include both a reduction and a value above the null value of 1.
Statistical evidence
The reported P-value is 0.33. The P-value should be interpreted as evidence from the specified statistical test, not as an effect-size measure.
Relative risk is not absolute risk
Suppose two groups had event risks of 10% and 8.7%. Their ratio would be 0.87, but the absolute difference would be 1.3 percentage points. The same risk ratio could arise from very different underlying risks. This is why relative measures should not be interpreted without considering the event probabilities from which they are calculated.
The registry's posted primary analysis reports the relative measure rather than a separately reported absolute risk difference. The statistical interpretation should therefore remain centered on the reported RR, confidence interval, and P-value rather than deriving additional numerical quantities.
9. Missing Data and Multiple Imputation
Missing outcome information is one of the most important statistical features of the posted primary analysis. The registry states that some participants did not have laboratory measurements for hemoglobin and therefore had missing assessments for PPH.
| Group | Participants with missing PPH assessments |
|---|---|
| Group A — Tranexamic Acid | 93 |
| Group B — Placebo | 87 |
The endpoint depends on change in hemoglobin values, so missing laboratory measurements can translate directly into missing binary outcome classifications. The registry identifies multiple imputation / missing data as a concept in the analysis.
Multiple imputation is designed to incorporate uncertainty about missing values rather than treating an imputed value as if it were known with certainty.
The important statistical issue is not simply how many observations are missing. It is also whether the probability of missingness is related to observed characteristics, treatment assignment, outcome status, or unobserved outcome information. Different assumptions about the missing-data mechanism can produce different estimates.
10. Statistical Methods Explained
Why was a chi-squared test used?
The primary endpoint is binary: each participant is classified according to whether postpartum haemorrhage occurred. A chi-squared test is designed to compare categorical outcome distributions between groups. It therefore matches the structure of the primary endpoint.
What does a risk ratio of 0.87 mean?
An RR of 0.87 means that the estimated probability of PPH in the tranexamic acid group was 0.87 times the estimated probability in the placebo group. The relative difference corresponding to that ratio is 13% lower estimated risk. This is not the same as saying that the absolute probability fell by 13 percentage points.
Why is 1 the key value for the confidence interval?
For a risk ratio, the null value is 1 because a ratio of 1 represents equal risk in the two groups. The reported 95% CI of 0.69–1.09 crosses 1, so the interval includes the no-relative-difference value.
Why does the P-value not measure effect size?
The P-value summarizes how compatible the observed data are with the null hypothesis under the specified statistical model or test. It depends on both the magnitude of the observed difference and the amount of information available. It is therefore not a substitute for the RR or its confidence interval.
Why does missing hemoglobin information matter?
PPH is defined using estimated blood loss derived from hemoglobin values before and after the procedure. If those laboratory measurements are unavailable, the participant's PPH status may not be directly classifiable. Missing-data methods are therefore part of the inferential pathway for the primary endpoint.
What does double-blinding contribute statistically?
Blinding does not change the mathematical definition of the risk ratio or chi-squared statistic. Its role is upstream: reducing opportunities for knowledge of treatment assignment to influence trial behavior, outcome assessment, or other processes that could introduce systematic differences between groups.
11. Risk Ratio, Confidence Interval, and P-value Together
The three reported statistical quantities answer different questions and should be read together.
| Quantity | Reported value | Primary question it addresses |
|---|---|---|
| Risk ratio | 0.87 | What is the estimated relative risk in tranexamic acid versus placebo? |
| 95% confidence interval | 0.69–1.09 | How uncertain is the estimated relative risk? |
| P-value | 0.33 | How compatible are the data with the null hypothesis under the chi-squared analysis? |
A useful statistical reading starts with the effect estimate, then examines its uncertainty, and finally considers the hypothesis-test result. Treating the P-value as the main result can obscure the magnitude and precision of the estimated treatment effect.
The reported RR of 0.87 points toward a lower estimated PPH risk with tranexamic acid. The 95% CI of 0.69–1.09 indicates uncertainty that includes the null value of 1. The reported P = 0.33 indicates that the chi-squared analysis did not provide strong statistical evidence against the null hypothesis under its stated framework.
These statements are complementary. The P-value does not replace the effect estimate, and the effect estimate alone does not communicate its statistical precision.
12. Analysis Population and Missing Assessments
The posted analysis defines the endpoint analysis population in terms of PPH assessment using hemoglobin values. Participants with missing laboratory measurements were therefore relevant to the analysis population because their PPH status could not be directly determined from the required measurements.
| Analysis feature | Registry description |
|---|---|
| Endpoint population | PPH defined as blood loss greater than or equal to 1000 mL using change in hemoglobin values |
| Group A missing assessments | 93 participants |
| Group B missing assessments | 87 participants |
| Missing-data concept | Multiple imputation / missing data |
| Primary test | Chi-squared |
| Effect measure | Risk Ratio |
This structure illustrates an important principle in clinical-trial statistics: the randomized enrollment count and the number of participants with evaluable measurements for a particular endpoint are not necessarily identical. The statistical analysis must account for the distinction without confusing missing outcome information with an absence of the event.
13. Safety Results
The registry reports serious adverse events by randomized group. These figures describe participants affected by serious adverse events relative to the participants at risk in each group.
| Group | Participants with serious adverse events | Participants at risk |
|---|---|---|
| Group A — Tranexamic Acid | 0 | 611 |
| Group B — Placebo | 1 | 613 |
The registry reports one affected participant among 613 in the placebo group and none among 611 in the tranexamic acid group. These counts are descriptive safety information. They do not constitute a formal comparative safety analysis, and the ClinicalTrials.gov record does not provide a confidence interval or P-value for this safety comparison.
The observed serious-adverse-event counts were 0/611 in the tranexamic acid group and 1/613 in the placebo group. With such small event counts, estimates of relative or absolute differences would be highly sensitive to the statistical method used. The reported data are therefore best read as the observed counts rather than as evidence of a quantified between-group safety effect.
14. What the Primary Analysis Does — and Does Not — Establish
What it estimates
The analysis estimates the relative risk of the binary PPH endpoint for tranexamic acid compared with placebo.
What it tests
The chi-squared analysis tests the relationship between randomized treatment group and the categorical PPH outcome.
What the CI adds
The confidence interval communicates uncertainty around the reported RR rather than merely identifying a single point estimate.
What it does not show
The primary RR does not directly quantify an absolute risk difference, mean blood loss difference, or individual participant benefit.
The primary analysis is therefore most appropriately interpreted as a randomized comparison of a prespecified binary clinical endpoint. The continuous laboratory measurements underlying the endpoint are important for defining PPH, but the posted effect measure is the risk ratio for the binary classification.
15. Statistical Interpretation of the Confidence Interval
The interval from 0.69 to 1.09 contains a range of relative-risk values. Its lower boundary corresponds to a substantially lower relative risk than the point estimate, while its upper boundary is slightly above the null value of 1.
For an RR, values below 1 favor a lower event risk in the treatment group and values above 1 favor a higher event risk. The fact that the confidence interval crosses 1 means that the uncertainty interval includes both directions relative to the null.
The point estimate is below 1, but the confidence interval extends across 1. The interval therefore conveys more information than the point estimate alone.
The confidence interval should not be interpreted as saying that there is a 95% probability that the true risk ratio lies between 0.69 and 1.09. In conventional frequentist inference, the 95% confidence level refers to the long-run behavior of the interval-generating procedure under repeated sampling.
16. Why This Trial Matters Statistically
ETAPPH is a useful teaching example because the primary endpoint and statistical analysis illustrate several core principles of clinical-trial biostatistics without requiring a complex time-to-event model.
| Concept | How it appears in ETAPPH |
|---|---|
| Randomization | Participants were randomized to tranexamic acid or normal saline placebo. |
| Blinding | The trial was double-blind. |
| Parallel-group design | Two randomized groups were compared in parallel. |
| Binary endpoint | PPH was classified as a binary outcome. |
| Chi-squared test | The posted primary analysis used a chi-squared test. |
| Risk ratio | The treatment effect was reported as a risk ratio. |
| Confidence interval | The RR was accompanied by a two-sided 95% CI. |
| P-value | The primary analysis reported P = 0.33. |
| Missing data | Missing hemoglobin assessments affected PPH assessment in both groups. |
| Multiple imputation | Multiple imputation / missing data was identified as an analysis concept. |
The combination of these features makes ETAPPH particularly useful for understanding the distinction between estimand, endpoint definition, effect measure, and statistical test. The endpoint defines what is being measured; the risk ratio describes the comparative effect; and the chi-squared test addresses the statistical evidence under the specified categorical framework.
17. Limitations
- Missing hemoglobin assessments: 93 participants in Group A and 87 in Group B had missing PPH assessments. Because PPH classification depends on hemoglobin measurements, missingness is directly relevant to the primary endpoint.
- Missing-data assumptions: The registry identifies multiple imputation / missing data as an analysis concept, but the information reported here does not provide the detailed imputation model or its assumptions.
- Binary endpoint: The primary result summarizes whether participants crossed the PPH threshold rather than describing the complete distribution of estimated blood loss.
- Confidence-interval interpretation: The reported 95% CI of 0.69–1.09 spans the null value of 1, so the point estimate should not be interpreted independently of its uncertainty.
- Safety-event counts: Serious adverse events were uncommon in the reported groups, with 0 affected participants in Group A and 1 in Group B. Sparse events limit the information available for comparative safety inference.
- Scope of posted analysis: The registry identifies one posted statistical analysis for the primary endpoint. Additional analyses should not be inferred from the existence of other posted outcome measures.
18. Planned and Posted Statistical Framework
The registry describes a randomized, double-blind, parallel-group prevention trial with a binary primary endpoint. The posted analysis used a chi-squared test and reported a risk ratio with a two-sided 95% confidence interval and P-value.
| Component | ETAPPH framework |
|---|---|
| Design | Randomized, double-blind, parallel |
| Primary endpoint | Number of Participants With Postpartum Hemorrhage (PPH) |
| Endpoint scale | Binary |
| Time frame | Up to day 2 postpartum |
| Primary comparison | Tranexamic acid versus placebo |
| Primary test | Chi-squared |
| Effect measure | Risk Ratio |
| Confidence interval | 95%, two-sided |
| Reported P-value | 0.33 |
| Missing-data concept | Multiple imputation / missing data |
This is a coherent analysis structure for a randomized trial whose principal outcome is binary. The main interpretive complication is the missing laboratory information used to determine PPH status, because the statistical conclusion depends not only on the observed outcome classifications but also on how missing assessments are handled.
19. Related Tutorials
Learn more about the methods used in this trial:
20. Related Calculators
21. Sources
- ClinicalTrials.gov: NCT04733157 — ETAPPH.
- PubMed: PMID 37106343.
Continue through Clinical Biostats
Connect the statistical concepts used in ETAPPH with broader biostatistics tutorials, calculators, and clinical-trial analyses.
22. Record Summary
ETAPPH provides a clear example of statistical analysis for a randomized binary clinical endpoint. The phase 3 trial randomized 1226 participants in a double-blind, parallel-group design comparing tranexamic acid injection with normal saline placebo. The primary endpoint was postpartum haemorrhage through up to day 2 postpartum, defined using estimated blood loss derived from hemoglobin measurements before and after the procedure.
The posted primary analysis used a chi-squared test and reported a risk ratio of 0.87, with a two-sided 95% confidence interval of 0.69–1.09 and a P-value of 0.33. The point estimate corresponds to a 13% lower estimated relative risk in the tranexamic acid group, while the confidence interval includes the null value of 1. The interpretation is therefore appropriately centered on the estimated relative effect together with its uncertainty rather than on the P-value alone.
Missing hemoglobin assessments are an important part of the statistical story: the registry reports 93 missing PPH assessments in Group A and 87 in Group B and identifies multiple imputation / missing data as an analysis concept. The serious-adverse-event data were also sparse, with 0 affected participants among 611 in the tranexamic acid group and 1 among 613 in the placebo group. Together, these features illustrate why a clinical-trial result should be interpreted through its endpoint definition, analysis population, missing-data structure, effect measure, uncertainty, and statistical test rather than through a single numerical result.