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
PACT was a randomized, parallel, single-masked phase 3 trial evaluating an individualized opioid prescription protocol after cesarean delivery. The individualized protocol allowed a prescription of 0 to 20 tablets of oxycodone 5 mg, while the comparator was a fixed opioid prescription.
| Feature | PACT |
|---|---|
| Trial name | Opioid Prescription After Cesarean Trial |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Psychiatry & Mental Health |
| Conditions | Pregnancy Related; Opioid Use; Pain |
| Design | Randomized, parallel, single-masked |
| Primary purpose | Treatment |
| Enrollment | 5,521 |
| Interventions | 0 to 20 tablets of oxycodone 5 mg; fixed opioid prescription |
| Primary endpoint | Moderate to severe worst pain score at 1 week post hospital discharge |
| ClinicalTrials.gov | NCT04296396 |
2. Clinical Question
The primary statistical question was whether an individualized opioid prescription could provide pain outcomes that were not inferior to those associated with a fixed opioid prescription after cesarean delivery. The registry defines the primary endpoint as the number of participants with moderate to severe pain at 1 week post-discharge.
Population
Participants in a phase 3 randomized trial addressing opioid prescribing, pain, and pregnancy-related care after cesarean delivery.
Intervention
Individualized Opioid Prescription (IOPP), consisting of 0 to 20 tablets of oxycodone 5 mg.
Comparator
Fixed Opioid Prescription.
Primary question
Is the rate of moderate to severe worst pain at 1 week post-discharge sufficiently similar that IOPP can be considered non-inferior to the fixed prescription?
3. Trial Design
Individualized Opioid Prescription
- 0 to 20 tablets of oxycodone 5 mg
- Individualized Opioid Prescription protocol
- Primary comparison against the fixed opioid prescription
Fixed Opioid Prescription
- Fixed opioid prescription
- Comparator for the individualized prescription protocol
- Primary comparison based on pain at 1 week post-discharge
The trial was randomized and parallel rather than crossover or factorial. The registry describes the masking as single. No stratification factors, crossover procedure, Bayesian analysis, or interim-analysis procedure are reported in the available registry information.
Trial start
The registry records the study start date as September 21, 2020.
Primary completion
The registry records primary completion on April 7, 2022.
Registry status
The trial is listed as completed.
4. Endpoints
| Endpoint | Time frame | Type | Definition / analysis |
|---|---|---|---|
| Number of Participants With Worst Pain Score of Moderate to Severe on the Brief Pain Inventory | 1 week post hospital discharge | Binary | Moderate to severe pain is defined as a value of 4 or higher on the “Worst Pain” question of the Brief Pain Inventory in the last 24 hours. |
| Number of Participants Who Filled One or More Opioid Prescriptions Beyond the Amount Prescribed at Discharge | After hospital discharge and up to 90 days postpartum | Binary | Risk ratio reported. |
| Number of Opioid Tablets Unused Since Discharge | 90 days postpartum | Continuous | Median difference reported using quantile regression. |
| Number of Morphine Milligram Equivalents Used at 2 Weeks Post Discharge | 2 weeks post discharge | Continuous | Median difference reported using quartile regression. |
| Worst Pain Severity Score at 2 Weeks Post Discharge | 2 weeks post discharge | Continuous | Median difference reported using quantile regression. |
| Pain Interference Score ≥ 4 at 2 Weeks Post Discharge | 2 weeks post discharge | Binary | Risk ratio reported using a chi-squared analysis. |
| Number of Participants Who Indicated an Improved Global Impression of Change | 2 weeks post discharge | Binary | Risk ratio reported; registry method listed as linear regression. |
| Number of Opioid Prescriptions Filled | 90 days postpartum | Binary | Chi-squared analysis reported; no effect estimate is listed in the registry analysis entry. |
Primary endpoint definition
The Brief Pain Inventory measures pain severity on a numeric scale from 0 to 10, with 0 representing no pain and 10 representing the highest level of pain. For the primary endpoint, a participant counted as having moderate to severe pain if the “Worst Pain” value for the last 24 hours was 4 or higher at 1 week post-discharge.
5. Statistical Methodology
Non-inferiority framework
The primary endpoint was analyzed within a non-inferiority framework. The registry states that the non-inferiority margin was 5 percentage points. The design assumed 90% power and a 0.025 one-sided significance level, with a required sample size of 4,300 participants before accounting for the anticipated missing data.
Because the registry states that a 20% missing rate was incorporated to account for the short window for assessing pain at 1 week post-discharge, the final planned sample size was 5,500 participants, or 2,750 per group. Actual enrollment was 5,521.
The registry defines the risk difference as the rate in the fixed group minus the rate in the IOPP group. Thus, a negative value favors the fixed group on the observed primary pain rate, while a positive value favors IOPP under the registry's stated direction. Non-inferiority is established when the lower confidence limit is no more than 5 percentage points below zero.
Risk difference
The primary effect measure was a risk difference, rather than a hazard ratio or odds ratio. This is directly interpretable on the absolute probability scale. If the risk difference is 0, the two groups have the same observed event rate. A positive value in the registry's direction means the event rate was higher in the fixed group than in the IOPP group.
Categorical secondary endpoints
The registry reports chi-squared analyses for several binary secondary outcomes. These analyses compare categorical outcome distributions between the randomized prescription groups. Risk ratios were reported for several of these endpoints, providing a relative rather than absolute measure of event frequency.
Continuous and distributional outcomes
The registry reports quantile regression for the number of opioid tablets unused and for worst pain severity, with median differences as the effect measure. Quartile regression was used for morphine milligram equivalents used at 2 weeks post-discharge, again with a median difference as the reported effect measure.
Linear regression entry
The analysis of participants indicating an improved Global Impression of Change is listed in the registry as “Regression, Linear,” while the reported effect measure is a risk ratio. Because the registry does not provide additional modeling specifications in the available analysis entry, the precise model parameterization cannot be determined from that entry alone.
Multiplicity
Several secondary endpoints have confidence intervals wider than the conventional 95% level. The registry specifically states that the 96.25%, 97.5%, and 98.75% confidence intervals were used because of false discovery rate adjustment. This is an important distinction: the confidence level is part of the multiplicity-adjusted inferential framework and should not be silently treated as an ordinary 95% interval.
| Endpoint analysis | Method reported | Effect measure | Confidence level |
|---|---|---|---|
| Primary moderate-to-severe pain | Not reported | Risk difference | 95% |
| Additional opioid prescription beyond discharge amount | Chi-squared | Risk ratio | 96.25% |
| Unused opioid tablets | Quantile regression | Median difference | 97.5% |
| Morphine milligram equivalents at 2 weeks | Quartile regression | Median difference | 98.75% |
| Worst pain severity at 2 weeks | Quantile regression | Median difference | 95% |
| Pain interference score ≥ 4 | Chi-squared | Risk ratio | 95% |
| Improved Global Impression of Change | Linear regression | Risk ratio | 95% |
| Opioid prescriptions filled | Chi-squared | Not reported | Not reported |
6. Primary Result: Moderate to Severe Pain at 1 Week
The primary endpoint compared the proportion of participants with moderate to severe worst pain at 1 week post hospital discharge. The registry reports a risk difference of 0.67, with a two-sided 95% confidence interval from −2.03 to 3.37.
Primary risk difference
95% CI: −2.03 to 3.37
Risk difference defined as fixed-group rate minus IOPP-group rate
Non-inferiority margin: −5 percentage points
The lower confidence limit, −2.03, is above the prespecified non-inferiority boundary of −5. Therefore, under the registry's stated non-inferiority rule, the confidence interval satisfies the criterion for declaring IOPP non-inferior to the fixed prescription with respect to the primary pain endpoint.
What the estimate means: The estimated risk difference was 0.67 percentage points in the registry's direction, meaning the fixed-group event rate minus the IOPP-group event rate was 0.67 percentage points.
What it does not mean: A risk difference of 0.67 does not mean that individual patients experienced a 0.67-point change in their pain score. The endpoint is binary: whether the participant's worst pain score met the threshold of 4 or higher.
What the confidence interval means: The two-sided 95% confidence interval extends from −2.03 to 3.37. It includes zero, so the result does not establish a statistically distinguishable difference in either direction at the conventional two-sided 5% level. For a non-inferiority question, however, the key comparison is with the prespecified lower boundary of −5, not simply with zero.
Why the p-value is not the main criterion: Non-inferiority is a margin-based hypothesis. The relevant question is whether the plausible range of the treatment difference remains above the clinically specified unacceptable-loss boundary. A conventional superiority p-value would answer a different question.
Important caution: The primary analysis included only participants with the required pain response at 1 week, and the registry reports missing primary endpoint data for 230 participants in the IOPP group and 219 in the fixed group. Missing data are therefore directly relevant to the interpretation of the primary comparison.
7. Understanding the Non-Inferiority Design
Non-inferiority trials differ from conventional superiority trials because the goal is not necessarily to demonstrate that one treatment produces a better outcome. Instead, the question is whether the new strategy is sufficiently close to the comparator that a prespecified amount of worse outcome can be ruled out.
The 5% margin
The registry specifies a non-inferiority margin of 5 percentage points. For the reported risk-difference convention, the lower confidence limit had to be at least −5.
The observed estimate
The estimated risk difference was 0.67, which is substantially closer to zero than the −5 percentage-point non-inferiority boundary.
The confidence interval
The interval from −2.03 to 3.37 remains entirely above the non-inferiority boundary, even though it crosses zero.
Why zero is different from −5
Zero represents no observed difference. The −5 boundary represents the largest loss in the IOPP direction that the trial design was willing to accept while still calling the strategy non-inferior.
This distinction is central to interpreting the PACT primary result. A confidence interval that crosses zero can still support non-inferiority when the entire interval remains within the prespecified acceptable range.
8. Secondary Result: Additional Opioid Prescriptions Beyond the Discharge Amount
The registry reports the number of participants who filled one or more opioid prescriptions beyond the amount prescribed at discharge, assessed after hospital discharge and through 90 days postpartum.
Risk ratio
96.25% CI: 0.99–1.51 · P = 0.046
Chi-squared analysis
The confidence interval was 96.25% rather than 95% because the registry states that false discovery rate adjustment was used.
A risk ratio of 1.22 means that the estimated event rate in the IOPP group was 1.22 times the event rate in the fixed group for this outcome. Equivalently, the estimated relative rate was 22% higher in the IOPP group, using the risk-ratio scale.
The confidence interval of 0.99 to 1.51 spans 1.00. Thus, the interval includes the possibility of essentially no relative difference as well as a higher event rate in the IOPP group. The reported P-value of 0.046 should be interpreted alongside the false discovery rate adjustment rather than as a standalone measure of effect size.
9. Secondary Result: Unused Opioid Tablets
The number of opioid tablets unused since discharge was assessed at 90 days postpartum. The registry reports a median difference using quantile regression.
Median difference
97.5% CI: −6.6 to −3.5 · P < 0.001
Quantile regression
The reported negative median difference indicates fewer unused tablets in the IOPP group relative to the fixed group under the registry's reported comparison. The entire 97.5% confidence interval is below zero, from −6.6 to −3.5 tablets.
A median difference is not the same as a difference in means. The reported estimate of −5 describes a difference in the median-oriented comparison produced by the quantile-regression analysis. It should not be read as saying that every participant used exactly five fewer tablets or that the mean difference was −5.
The confidence interval gives a range of values for the estimated median difference under the analysis framework. Its exclusion of zero indicates that the registry's reported statistical comparison is clearly separated from no difference on this scale.
The P-value of <0.001 describes evidence against the corresponding null comparison; it does not measure the clinical importance of a five-tablet median difference. The effect estimate and its confidence interval provide the more direct information about magnitude and precision.
10. Secondary Result: Morphine Milligram Equivalents at 2 Weeks
Morphine milligram equivalents used at 2 weeks post discharge were analyzed with quartile regression and summarized using a median difference.
Median difference
98.75% CI: −19.2 to −10.8 · P < 0.001
Quartile regression
The reported estimate indicates a lower median-oriented amount of morphine milligram equivalents used in the IOPP group relative to the fixed group under the registry's comparison.
The −15 estimate is expressed in morphine milligram equivalents, not in tablets. This distinction matters because different opioid products and doses can correspond to different morphine-equivalent quantities.
The 98.75% confidence interval extends from −19.2 to −10.8. The interval does not include zero, and the registry reports P < 0.001. The confidence level is deliberately higher than 95% because of the stated false discovery rate adjustment.
The estimate describes the central tendency captured by the quartile-regression analysis. It does not imply that all participants reduced opioid exposure by 15 morphine milligram equivalents.
11. Secondary Result: Worst Pain Severity at 2 Weeks
Worst pain severity score was assessed at 2 weeks post discharge. The registry reports a median difference from quantile regression.
Median difference
95% CI: 0–0 · P = 1.0
Quantile regression
The reported median difference was exactly 0, with a confidence interval from 0 to 0 and a P-value of 1.0. This indicates no observed difference on the reported median-difference scale for this endpoint.
12. Secondary Result: Pain Interference Score ≥ 4
Pain interference was evaluated at 2 weeks post discharge, with an endpoint defined as a Pain Interference Score of at least 4.
Risk ratio
95% CI: 0.85–1.09 · P = 0.52
Chi-squared analysis
A risk ratio of 0.96 indicates that the estimated event risk in the IOPP group was 0.96 times that in the fixed group, a relative difference close to one on the risk-ratio scale.
The 95% confidence interval from 0.85 to 1.09 includes 1.00. The reported P-value of 0.52 is consistent with the absence of strong statistical evidence for a difference between the groups on this endpoint.
The result does not establish that the two groups are exactly equivalent. A non-significant superiority test and a formal equivalence or non-inferiority conclusion are different statistical claims and require different prespecified hypotheses and margins.
13. Secondary Result: Improved Global Impression of Change
The registry reports the number of participants who indicated an improved Global Impression of Change at 2 weeks post discharge.
Risk ratio
95% CI: 0.98–1.00 · P = 0.15
Registry method: linear regression
The reported risk ratio is close to 1.00, indicating a small estimated relative difference between the groups on the reported scale.
The confidence interval from 0.98 to 1.00 is narrow and lies very close to the null value of 1.00. The P-value of 0.15 does not provide conventional evidence of a superiority difference.
Because the registry labels the analysis as linear regression while reporting a risk ratio, the exact model specification is not recoverable from the registry entry alone. The reported effect measure should therefore be interpreted as recorded rather than mapped onto a more specific regression parameterization.
14. Secondary Result: Number of Opioid Prescriptions Filled
The number of opioid prescriptions filled was evaluated through 90 days postpartum. The registry reports a chi-squared analysis and a P-value but does not list a corresponding effect estimate or confidence interval in the posted analysis entry.
Reported statistical test
Chi-squared analysis
Time frame: 90 days postpartum
The P-value provides evidence against the null comparison used for the categorical analysis, but without a posted risk difference, risk ratio, or corresponding confidence interval, the magnitude of the between-group difference cannot be described on an effect-size scale from this analysis entry.
15. Secondary Results in One View
| Endpoint | Estimate | Confidence interval | P-value | Method |
|---|---|---|---|---|
| Additional opioid prescription beyond discharge amount | RR 1.22 | 96.25% CI 0.99–1.51 | 0.046 | Chi-squared |
| Unused opioid tablets | Median difference −5 | 97.5% CI −6.6 to −3.5 | <0.001 | Quantile regression |
| Morphine milligram equivalents at 2 weeks | Median difference −15 | 98.75% CI −19.2 to −10.8 | <0.001 | Quartile regression |
| Worst pain severity at 2 weeks | Median difference 0 | 95% CI 0–0 | 1 | Quantile regression |
| Pain interference score ≥ 4 | RR 0.96 | 95% CI 0.85–1.09 | 0.52 | Chi-squared |
| Improved Global Impression of Change | RR 0.99 | 95% CI 0.98–1.00 | 0.15 | Linear regression |
| Opioid prescriptions filled | Not reported | Not reported | <0.001 | Chi-squared |
16. Serious Adverse Events
The registry reports serious adverse events separately for mothers and for neonates/infants. The reported values are presented as affected participants divided by participants at risk.
| Population | Individualized Opioid Prescription | Fixed Opioid Prescription |
|---|---|---|
| Mothers | 133 / 2,748 | 96 / 2,772 |
| Neonates / infants | 37 / 2,856 | 30 / 2,889 |
These figures should be interpreted as safety-event counts and denominators rather than as estimates of a causal treatment effect. The registry's reported serious-adverse-event information does not provide a corresponding confidence interval, risk ratio, or hypothesis test in the statistical analysis entries summarized here.
17. Missing Data and Analysis Populations
Missing outcome data are explicitly reported for several analyses. The primary endpoint had missing data for 230 participants in the IOPP group and 219 participants in the fixed group. Other endpoints had different amounts of missingness.
| Endpoint | IOPP missing | Fixed missing |
|---|---|---|
| Primary worst pain at 1 week | 230 | 219 |
| Additional opioid prescription beyond discharge amount | 94 | 76 |
| Unused opioid tablets | 207 | 180 |
| Worst pain severity at 2 weeks | 234 | 286 |
| Pain interference score ≥ 4 at 2 weeks | 234 | 286 |
| Opioid prescriptions filled | 94 | 75 |
The registry does not specify an imputation procedure for these missing outcome observations in the statistical-analysis entries. This matters particularly for a non-inferiority analysis because missing-data assumptions can influence the estimated difference and the resulting confidence interval.
The registry also identifies one participant in the fixed group who was enrolled for a second pregnancy. Only the first pregnancy was included in the analyses described for the primary endpoint and several secondary endpoints.
18. Statistical Methods Explained
Why use a risk difference for the primary non-inferiority endpoint?
The risk difference expresses the absolute difference in the proportion of participants experiencing the binary outcome. For this trial, the registry defines the direction as the fixed-group rate minus the IOPP-group rate. That makes the prespecified −5 percentage-point non-inferiority margin directly interpretable on the absolute risk scale.
Why can a confidence interval crossing zero still support non-inferiority?
Because non-inferiority is not the same hypothesis as superiority. A confidence interval that crosses zero means that no difference remains plausible under the chosen confidence level. But non-inferiority asks whether the unfavorable side of the interval extends beyond the prespecified acceptable-loss boundary. Here, the lower limit of −2.03 remains above −5.
Why is the non-inferiority margin more important than a conventional p-value?
A conventional superiority test evaluates a null hypothesis such as no difference. The non-inferiority question instead evaluates whether the new strategy could be worse by more than an amount considered acceptable. The margin therefore defines the decision boundary for the trial's primary hypothesis.
What does a risk ratio of 1.22 mean?
A risk ratio of 1.22 means that the estimated event probability in the IOPP group was 1.22 times the event probability in the fixed group for that endpoint. It is a relative measure. It does not tell us the absolute percentage-point difference without knowing the underlying event rates.
Why were quantile regression and median differences used?
Several opioid-use outcomes were summarized with median differences rather than means. Quantile regression directly models a conditional quantile and can be useful when an outcome's distribution is not well summarized by a mean and standard deviation. The reported estimate therefore describes a median-oriented contrast rather than an arithmetic mean difference.
What does the false discovery rate adjustment change?
The registry reports confidence levels of 96.25%, 97.5%, and 98.75% for selected secondary endpoints because of false discovery rate adjustment. The corresponding intervals should be interpreted at their stated confidence levels. A 97.5% interval, for example, is not interchangeable with a conventional 95% confidence interval.
Why is missing data especially important in a non-inferiority trial?
In a superiority trial, missing outcomes can reduce statistical power. In a non-inferiority trial, missing data can also complicate the direction of bias because incomplete follow-up may make groups appear more similar than they truly are. The PACT registry explicitly reports missing primary endpoint observations, so the analysis population and missing-data assumptions are important components of interpretation.
19. Risk Difference Versus Risk Ratio
PACT illustrates why absolute and relative effect measures should not be treated as interchangeable.
| Measure | What it describes | PACT example |
|---|---|---|
| Risk difference | Absolute difference between two event probabilities | Primary RD 0.67; 95% CI −2.03 to 3.37 |
| Risk ratio | Ratio of two event probabilities | Additional prescription outcome: RR 1.22; 96.25% CI 0.99–1.51 |
| Median difference | Difference in a median-oriented distributional measure | Unused tablets: −5; 97.5% CI −6.6 to −3.5 |
These measures answer different questions. The primary endpoint uses an absolute difference because the non-inferiority margin is specified in percentage points. Secondary binary endpoints use risk ratios, while continuous or count-like opioid-use outcomes use median differences.
20. Primary Endpoint: A Deeper Statistical Reading
The primary estimate of 0.67 should be read together with its entire confidence interval and the non-inferiority margin. Considering the point estimate alone would miss the central feature of the trial design.
The observed interval stretches from a modest difference favoring the fixed prescription under the registry's direction to a modest difference favoring IOPP. Most importantly for non-inferiority, its lower limit remains above the −5 percentage-point boundary.
Point estimate
0.67 is the single best estimate from the reported primary analysis, but it is not the full inferential result.
Uncertainty
The 95% CI spans −2.03 to 3.37, showing that the estimate is compatible with modest differences in either direction.
Non-inferiority criterion
The lower limit is above −5, satisfying the registry's stated criterion for non-inferiority.
Superiority is different
The interval includes zero, so the primary result should not be reframed as a superiority claim simply because the point estimate is positive.
21. What the Secondary Results Add
The secondary analyses address several different dimensions of postoperative opioid prescribing and pain rather than a single composite outcome.
- Additional prescriptions: the reported risk ratio was 1.22, with a 96.25% confidence interval of 0.99–1.51 and P = 0.046.
- Unused tablets: the reported median difference was −5 tablets, with a 97.5% confidence interval of −6.6 to −3.5 and P < 0.001.
- Morphine milligram equivalents: the reported median difference was −15, with a 98.75% confidence interval of −19.2 to −10.8 and P < 0.001.
- Worst pain at 2 weeks: the reported median difference was 0, with a 95% confidence interval of 0 to 0 and P = 1.0.
- Pain interference: the reported risk ratio was 0.96, with a 95% confidence interval of 0.85–1.09 and P = 0.52.
- Global Impression of Change: the reported risk ratio was 0.99, with a 95% confidence interval of 0.98–1.00 and P = 0.15.
- Opioid prescriptions filled: a chi-squared analysis was reported with P < 0.001, without a posted effect estimate in the registry analysis entry.
The pattern demonstrates why a trial should not be summarized by one P-value. Different endpoints use different scales, time frames, estimands, and statistical methods. Their results therefore need to be interpreted in their own statistical context.
22. Limitations
- Missing primary outcome data: 230 participants in the IOPP group and 219 in the fixed group were missing the primary pain outcome. The registry analysis entry does not specify an imputation procedure.
- Non-inferiority assumptions: non-inferiority conclusions depend on the prespecified 5 percentage-point margin and on the validity of the analysis population and missing-data assumptions.
- Confidence levels differ across endpoints: selected secondary endpoints use 96.25%, 97.5%, or 98.75% confidence intervals because of false discovery rate adjustment.
- Secondary endpoints are heterogeneous: the outcomes range from binary prescription events to opioid quantities and pain scores, so their effect measures are not directly interchangeable.
- Incomplete model specification: the registry does not provide a detailed statistical model for every posted analysis. In particular, the primary analysis method is not named in the analysis entry, and the Global Impression of Change analysis is labeled linear regression while reporting a risk ratio.
- Safety interpretation: maternal and neonatal/infant serious-adverse-event counts use different denominators and should not be collapsed into a single safety rate.
- Registry-level reporting: several statistical-analysis entries provide estimates and P-values without enough model detail to reproduce the full analysis from the registry record alone.
23. Why This Trial Matters Statistically
PACT is a useful statistical teaching case because it combines a non-inferiority primary question with several secondary endpoints that examine opioid use and pain from different perspectives.
| Concept | How it appears in PACT |
|---|---|
| Randomization | Randomized parallel-group phase 3 design with 5,521 participants enrolled. |
| Non-inferiority | Primary endpoint evaluated against a prespecified 5 percentage-point margin. |
| Risk difference | Primary effect measure, defined as fixed-group rate minus IOPP-group rate. |
| Confidence interval | Primary 95% CI used to compare the lower bound with the non-inferiority margin. |
| Chi-squared testing | Used for several binary secondary outcomes. |
| Risk ratio | Used for multiple binary secondary endpoints. |
| Quantile regression | Used for unused opioid tablets and worst pain severity at 2 weeks. |
| Quartile regression | Used for morphine milligram equivalents at 2 weeks. |
| Multiplicity | False discovery rate adjustment resulted in confidence levels above 95% for selected secondary endpoints. |
| Missing data | Different amounts of missingness were reported across primary and secondary outcomes. |
24. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
25. Related Statistical Calculators
26. Sources
- ClinicalTrials.gov: NCT04296396 — PACT, Opioid Prescription After Cesarean Trial.
- PubMed: PubMed record associated with PACT.
Continue through the Clinical Biostats statistical library
Connect the trial's endpoints and methods to deeper statistical tutorials and practical analysis tools.
27. Record Summary
PACT provides a clear example of how a clinical trial can combine a non-inferiority primary endpoint with secondary outcomes covering opioid prescribing, opioid exposure, pain severity, pain interference, and patient-reported change. The primary analysis used a risk difference with a prespecified 5 percentage-point non-inferiority margin. The reported estimate was 0.67 with a 95% confidence interval of −2.03 to 3.37, placing the lower confidence limit above the −5 percentage-point boundary specified by the registry.
The secondary analyses demonstrate the importance of matching an effect measure to the clinical question. Binary outcomes were analyzed with chi-squared methods and, where reported, risk ratios. Opioid quantities and pain scores were analyzed with quantile or quartile regression and summarized using median differences. Several secondary confidence intervals used higher confidence levels because of false discovery rate adjustment.
For statistical interpretation, the central lesson is that non-inferiority is a margin-based conclusion, not simply a non-significant superiority test. The primary confidence interval crosses zero but remains above the prespecified non-inferiority boundary. At the same time, missing primary endpoint observations and the registry's limited reporting of model specifications remain important considerations when interpreting the complete statistical evidence.