Piroxicam and Levonorgestrel Co-treatment for Emergency Contraception: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Piroxicam and Levonorgestrel Co-treatment for Emergency Contraception (NCT03614494), a randomized phase 2, phase 3 study. The primary outcome is Percentage of Pregnancies Prevented.

2. Clinical Question

This is a question about the contribution of piroxicam when levonorgestrel is given in both groups. It is not a comparison of levonorgestrel with no emergency contraception, and the registry record does not report a group that received neither levonorgestrel nor piroxicam. The primary outcome is a model-based percentage: expected pregnancies are calculated using a model attributed in the record to Trussell and colleagues, while observed pregnancies are counted in the trial. This makes the estimate dependent on both the observed outcome and the expected-pregnancy calculation. The registry record does not provide further details about how expected pregnancy was calculated beyond its description of cycle-day input and the cited model.

3. Trial Design

The registry record describes a randomized, parallel-group trial with triple masking and phases listed as PHASE2 and PHASE3. Its status is COMPLETED. The recorded enrollment is 860. The study start date is 2018-08-20, and the primary completion date is 2022-10-05. The sponsor is Dr. Hang Wun Raymond Li.

Participants were assigned to receive either piroxicam 40 mg plus levonorgestrel 1.5 mg as a single oral dose, or placebo plus levonorgestrel 1.5 mg as a single oral dose. The primary outcome lists 418 participants in each arm. Those arm figures are the reported primary-outcome denominators and should not be treated as an account of every participant enrolled, randomized, or included in every analysis. The registry record does not report a participant-flow account explaining the relationship between enrollment and the primary-outcome denominators.

Randomization supports a comparison between the co-treatment groups by aiming to balance baseline factors, including factors not measured or reported. Masking can reduce the influence of expectations on participant behavior, care, outcome ascertainment, or analysis, depending on who was masked; the record gives the masking designation as triple but does not identify the masked roles. The parallel model means the groups are compared as separate assigned arms, rather than as repeated treatment periods within the same participant.

The registry record does not report the randomization sequence method, allocation concealment procedure, details of masking, protocol deviations, or the sample-size justification used when designing the trial. It records planned enrollment and completed status, but those details alone do not establish the rationale for the enrollment target or describe the conduct of the trial.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: No per-assignment drop-out.

Overall Study

Analysis population for the primary outcome, as defined in the registry: efficacy-evaluable participants

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as Percentage of Pregnancies Prevented, measured over 1 month.

The primary endpoint is titled “Percentage of Pregnancies Prevented” and has a timeframe of 1 month. Its definition is the number of expected pregnancies minus the number of observed pregnancies, divided by the number of expected pregnancies. Expected pregnancies are estimated from the cycle day on which unprotected sexual intercourse occurred, using the model attributed in the registry record to Trussell and colleagues. The stated population is efficacy-evaluable participants.

This endpoint is used to express observed pregnancy counts relative to an estimate of how many pregnancies would have been expected. It therefore differs from a simple percentage of participants who did not become pregnant. Its value depends on the expected-pregnancy model as well as the observed pregnancy count, so interpretation requires attention to both components. The registry record does not give the expected and observed pregnancy counts underlying each arm’s percentage, nor does it report uncertainty intervals for the arm-specific percentages.

As a statistical concept, a model-based expected count can provide a reference when pregnancy risk varies with circumstances such as cycle day. However, it also makes the endpoint sensitive to the suitability and implementation of that reference model. The registry record describes the model’s source and the cycle-day basis, but further details about implementation and validation in this trial are not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does the primary risk difference express?

The primary analysis reports a risk difference of 31.4. A risk difference expresses the contrast between two group values on their original scale; it is not a relative change. Here the endpoint is a percentage of expected pregnancies prevented, so the difference describes the separation between the two arm-level percentages on that scale. The registry labels the analysis as a chi-squared comparison. A chi-squared test assesses whether the observed pattern of categorical outcome data is compatible with a null hypothesis of no group difference, under the test’s assumptions.

The posted 95% confidence interval, from 26.2 to 36.4, conveys uncertainty around the reported difference under the analysis method. In general, a confidence interval is a range generated by a procedure that, over repeated use under its assumptions, captures the underlying quantity at the stated rate. It is not a guarantee that the true value lies inside this particular interval. For this endpoint, interpretation also depends on how expected pregnancies were modeled and on which participants entered the efficacy-evaluable analysis.

How should the p-value be read?

The primary p-value is reported as p < 0.001. A p-value describes the compatibility of the observed data with a specified null model, assuming the statistical procedure and its assumptions apply. A small value indicates that the observed separation would be unusual under that null model. It does not measure the size or clinical importance of the effect, and it is not the probability that the null hypothesis is true. The primary result is also described as a superiority analysis, but the registry record does not report the full testing plan or any multiplicity strategy.

Why do methods differ across outcomes?

The registry’s primary comparison uses a chi-squared method and reports a risk difference. The secondary pregnancy-rate analysis reports a logistic regression method with Firth’s bias reduction and an odds ratio. These measures answer related but different questions. A risk difference describes absolute separation on a probability scale, while an odds ratio compares odds between groups. Odds are not the same as probabilities, particularly when outcomes are not rare. Firth’s bias-reduction approach is a general method that can stabilize logistic regression estimates in settings where ordinary maximum-likelihood estimation may be biased or unstable, including sparse outcome patterns. The registry does not provide further details about the model specification or covariates.

A separate recalculation of the secondary pregnancy-rate summary uses the posted event counts and denominators to estimate an unadjusted risk difference and associated uncertainty. Such a recalculation is a simple comparison of the posted summaries, not a substitute for the sponsor’s own analysis. It does not adjust for participant characteristics, and it need not reproduce a regression estimate because the estimands and methods differ. The registry record does not indicate a discrepancy between the posted secondary analysis and the recalculation. The sponsor’s analysis remains the primary reported inferential analysis for that endpoint.

What does over-dispersion mean for count outcomes?

As a general statistical concept, over-dispersion occurs when the variability in count data exceeds what a chosen probability model expects. If a model assumes less variability than the data contain, standard errors may be too small and uncertainty may be understated. This trial’s primary measure is a count-derived percentage, but the registry does not report a dispersion estimate or a model for count variability. It would therefore be inappropriate to claim that over-dispersion was present, absent, or addressed in the primary analysis.

How should the planning illustrations be interpreted?

The planning illustrations are based on the observed secondary effect and specify two-sided significance level 0.05, equal allocation, and a normal approximation. They show participants per arm under different fractions of the observed effect and for the stated power targets. These are prospective planning illustrations, not observed power and not evidence that the trial was designed using those assumptions. The registry record does not report the trial’s own sample-size justification, assumed effect, target power, or planning method. A calculation based on an observed effect is especially sensitive to the effect estimate itself; considering smaller fractions illustrates that designs intended to detect more modest effects require more participants, all else equal.

Power is a design property defined relative to a specified effect, variability, analysis method, and significance threshold. It is not a measure of the probability that a reported result is correct. Retrospective “observed power” based on the estimated effect does not add meaningful evidence beyond the estimate and its confidence interval. The planning figures are useful only as conditional illustrations of design requirements under their stated assumptions, not as a description of what the completed trial could reliably detect or of the sponsor’s actual planning.

What cannot be assessed from the reported methods?

The registry record does not report a prespecified statistical analysis plan, detailed model covariates, interim analyses, missing-data methods, or sensitivity analyses. It also does not give sufficient detail to evaluate the assumptions behind the expected-pregnancy model. These gaps matter because outcome definitions, exclusions, model choices, and analysis populations can affect estimates and uncertainty. The posted numerical results can be described as reported, but a complete audit of the analysis cannot be reconstructed from the registry record alone.

10. Analyses Reported by the Sponsor

The sponsor’s reported primary analysis uses a chi-squared method and estimates a risk difference of 31.4, with a 95% confidence interval from 26.2 to 36.4 and p < 0.001. The posted primary result is therefore the sponsor’s own reported comparison, rather than a separate recalculation. The registry record does not identify a discrepancy between the sponsor’s analysis and a recalculation for this outcome.

For the secondary pregnancy-rate outcome, the sponsor reports a Firth bias-reduced logistic regression analysis, expressed as an odds ratio, with a 95% confidence interval and p-value. A separate unadjusted recalculation from the posted counts expresses the contrast as a risk difference. Because the estimands and methods differ, those results need not be numerically identical. The recalculation does not replace or override the sponsor’s regression analysis, and the registry record does not flag a discrepancy. In general, when a discrepancy is identified, the sponsor’s own analysis takes precedence over a simple recalculation from summary data.

The registry does not provide enough detail to reconstruct the sponsor’s model specification, covariates, or full analysis plan. The reported analyses should therefore be interpreted according to the methods and estimates stated in the record, without assuming unreported adjustments or procedures.

11. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when several outcomes, comparisons, subgroups, or analytic choices are tested. Each additional test can create further opportunities for a result to appear statistically notable by chance, so a prespecified testing hierarchy or an adjustment may be important when drawing confirmatory conclusions. The registry record reports a primary outcome and a secondary pregnancy-rate outcome, but a multiplicity plan, adjustment, or hierarchy is not reported in the registry record. It is not possible to determine from the posted information whether other outcomes or analyses were planned, conducted, or omitted from the record.

The record identifies the primary analysis population as efficacy-evaluable participants. It does not report a definition of efficacy-evaluable status, an intention-to-treat analysis, a per-protocol analysis, participant flow by reason for exclusion, or how participants were handled if outcome information was unavailable. Missing-data methods and sensitivity analyses are not reported in the registry record. These omissions matter because excluding participants or handling incomplete pregnancy follow-up in different ways can change the composition of an analysis and the uncertainty of its estimates.

Randomization supports causal comparison when assigned groups are retained and analyzed appropriately, but an analysis restricted to a subset can weaken that protection if inclusion depends on post-randomization events or outcome availability. The record does not establish whether such selection occurred; it simply does not provide the details needed to assess it. Nor does the posted enrollment figure alone explain the primary analysis denominators.

Interim analyses, stopping rules, and any adjustments for repeated looks at the data are not reported in the registry record. The registry also does not report a detailed statistical analysis plan, so the prespecification of endpoints, transformations, models, and subgroup analyses cannot be evaluated. These are limits on what can be concluded about the analysis process, not evidence that a particular problem occurred.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 2% reporting threshold.

Most frequent non-serious adverse event terms

The registry reports serious adverse events for the piroxicam and placebo groups. In each group, the serious-event count is 0 among 430 participants listed at risk, and deaths are reported as 0. Other adverse events are reported for 117 participants in the piroxicam group and 133 in the placebo group, with 418 listed at risk in each group for that category. The record does not provide serious adverse-event terms, and no top serious-event terms are listed.

These figures describe the registry’s recorded counts and denominators; they do not establish that no other events occurred or that event ascertainment was complete. The serious-event denominator differs from the denominator listed for other adverse events, and the record does not explain the reason. It is therefore important not to infer a participant-flow explanation or treat the figures as directly interchangeable. The registry does not provide event definitions, severity detail for other events, attribution to study treatment, or a fuller safety narrative. Zero reported serious events and deaths in the posted data should be read as recorded observations, not as proof of absence of risk.

13. What the Result Does, and Does Not, Mean

The primary result is a statistically clear separation between the randomized co-treatment groups on the registry’s percentage-of-pregnancies-prevented measure. The reported confidence interval does not include no difference, and the reported p-value is small under the stated analysis. In the secondary pregnancy-rate outcome, the direction of the posted results is consistent with the primary comparison. Together, these findings describe what the registry reports for this trial and its analyzed population.

They do not show that the result applies to every person seeking emergency contraception. The registry record does not report detailed eligibility criteria, participant characteristics, or subgroup estimates sufficient to assess variation in the result. It also does not report the trial’s full analysis plan, missing-data approach, or participant-flow details. Those limits constrain judgments about generalizability and robustness.

The primary endpoint is not simply the proportion of participants who avoided pregnancy. It depends on expected pregnancies calculated using a cycle-day model and on observed pregnancy counts. Consequently, the result should be understood as a contrast on that defined measure, not as a direct estimate of an individual’s chance of pregnancy or of a treatment effect under every setting. The record does not give enough detail to independently evaluate the expected-pregnancy calculation.

The findings also do not establish a complete balance of benefit and harm. The registry reports adverse-event counts, but it does not provide detailed event descriptions, attribution, or enough safety information to characterize all risks. Statistical significance is not the same as clinical importance, and a trial result does not itself provide medical advice. Interpretation should stay within the outcomes, population, and methods actually reported.

14. Why This Trial Matters Statistically

This trial illustrates why the estimand—the exact quantity an analysis aims to compare—matters. Its primary endpoint is a percentage of expected pregnancies prevented, not a simple pregnancy proportion. That distinction changes what the numerical difference means and makes the expected-pregnancy model part of the interpretation. A reader must understand both the observed outcome and the reference calculation rather than treating the endpoint as a direct count of pregnancies avoided.

The record also illustrates the value of distinguishing primary and secondary analyses. The primary analysis reports a risk difference, while the secondary pregnancy-rate analysis includes a logistic-regression odds ratio and an unadjusted recalculation on a risk-difference scale. Those measures are related but answer different statistical questions. Reporting them side by side can be informative, but their estimates should not be treated as directly interchangeable.

Finally, a numerical result is only one part of an evidence assessment. The trial is described as randomized and triple masked, yet the record does not explain the masking roles, the efficacy-evaluable definition, missing-data handling, or multiplicity plan. Such information helps readers assess whether the analysis preserves the design’s strengths and how much confidence to place in the precision of the estimate. Planning illustrations based on an observed effect can demonstrate how assumed effect size influences required sample size, but they do not reveal the trial’s original design rationale or observed power.

The statistical lesson is to interpret estimates in the context of endpoint construction, analysis population, uncertainty, and reporting completeness. A small p-value does not replace those considerations. Nor does incomplete registry detail establish that the analysis was flawed; it defines what cannot be assessed from this record.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The primary endpoint depends on an expected-pregnancy calculation based on cycle day and a cited model. The registry record does not provide the detailed implementation, validation, or expected and observed counts underlying the arm-level percentages. This limits independent assessment of the endpoint’s construction and its sensitivity to model assumptions.

The primary analysis is described as efficacy-evaluable, but the record does not define that population or explain how participants moved from enrollment to the reported arm denominators. It does not report missing-outcome methods, exclusions, or sensitivity analyses. The posted denominators therefore cannot be fully evaluated as an account of the randomized population. The serious and other adverse-event denominators also differ, without a reported explanation.

Further trial-specific details are absent: eligibility criteria, baseline characteristics, allocation-concealment procedures, the roles included in triple masking, and a detailed statistical analysis plan are not reported in the registry record. The record also does not report a multiplicity strategy, interim analyses, or the sponsor’s original sample-size justification. These gaps limit assessment of potential selection, analysis flexibility, and the intended precision of the trial.

The adverse-event information is sparse. Serious-event terms are not listed, and detail about the nature, severity, timing, and attribution of other events is not reported in the registry record. The posted counts alone cannot characterize the full safety profile. Finally, the registry’s outcome and analysis entries do not provide enough information to determine whether additional planned outcomes or analyses were unposted. These limitations affect how comprehensively the evidence can be appraised; they do not by themselves show that any specific bias or reporting problem occurred.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Piroxicam and Levonorgestrel Co-treatment for Emergency Contraception — NCT03614494.
• Registry record last updated 2024-05-06; status last verified 2023-11.
• Results first posted 2024-05-06.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

NCT03614494 is a completed randomized, triple-masked, parallel-group trial of piroxicam plus levonorgestrel versus placebo plus levonorgestrel for emergency contraception. The registry lists PHASE2 and PHASE3 and records enrollment of 860. The primary outcome is the percentage of expected pregnancies prevented over 1 month among efficacy-evaluable participants. The posted primary comparison favors the piroxicam arm on this endpoint, with a reported risk difference of 31.4 and a 95% confidence interval from 26.2 to 36.4.

A secondary pregnancy-rate outcome is also reported, along with adverse-event counts. The registry record does not report several details needed for a full appraisal, including the efficacy-evaluable definition, missing-data methods, multiplicity plan, and detailed safety information. The primary endpoint’s model-based expected-pregnancy component is central to interpreting the result.