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
PEOPLE was a randomized, open-label, parallel phase 3 prevention trial evaluating three approaches to post-exposure prophylaxis for curbing transmission of leprosy. The registered primary endpoint was the number of new leprosy cases per 1000 person-years over 45 months.
| Feature | PEOPLE |
|---|---|
| Phase | Phase 3 |
| Condition | Leprosy |
| Primary purpose | Prevention |
| Design | Randomized, parallel |
| Masking | None |
| Enrollment | 144000 |
| Number of arms | 4 |
| Intervention | Rifampicin |
| Primary endpoint | Compare effectiveness in curbing transmission of leprosy of three different approaches of post exposure prophylaxis |
| Primary endpoint time frame | 45 months |
| ClinicalTrials.gov | NCT03662022 |
| Study status | Completed |
2. Clinical Question
The central statistical question was whether different approaches to post-exposure prophylaxis were associated with different rates of newly diagnosed leprosy cases during 45 months of follow-up.
Population
Participants enrolled in the PEOPLE phase 3 prevention trial addressing leprosy transmission in the Comoros and Madagascar.
Intervention
Post-exposure prophylaxis approaches involving rifampicin, including household PEP, PEP 100m, and PEP 100m among participants positive for Anti-PGL-I IgM antibody.
Comparator
No post-exposure prophylaxis, identified in the statistical analyses as arm 1.
Primary question
How do incidence rates of newly diagnosed leprosy compare between each post-exposure prophylaxis approach and no PEP?
3. Trial Design
No PEP
- Comparator arm used for all three primary incidence rate ratio calculations.
- Incidence rates were measured between the first and fourth household surveys.
Household PEP
- Post-exposure prophylaxis approach evaluated against no PEP.
- Primary comparison: no PEP versus Household PEP.
PEP 100m
- Post-exposure prophylaxis approach evaluated against no PEP.
- Primary comparison: no PEP versus PEP 100m.
PEP 100m + Positive for Anti-PGL-I IgM Ab
- PEP 100m approach among participants positive for Anti-PGL-I IgM antibody.
- Primary comparison: no PEP versus PEP 100m + Positive for Anti-PGL-I IgM Ab.
4. Primary Endpoint
| Endpoint | Definition | Time frame | Analysis |
|---|---|---|---|
| Compare Effectiveness in Curbing Transmission of Leprosy of Three Different Approaches of Post Exposure Prophylaxis | Number of new leprosy cases per 1000 person-years. Incident cases were cases newly diagnosed while under follow-up. | 45 months | Three incidence rate ratios comparing arm 1 with each intervention arm |
The registered analysis specifies that the three incidence rate ratios are based on incidence rates measured between the first and fourth household survey in each intervention arm, divided by the corresponding incidence rate in the comparator arm.
A rate ratio below 1 indicates a lower estimated incidence rate in the intervention group relative to the comparator. A rate ratio of 1 represents equal incidence rates.
5. Statistical Methodology
Mixed-effects model
The registry identifies Mixed Models Analysis as the method used for all three primary comparisons and normalizes this to a mixed-effects model. This is a longitudinal modeling framework that can account for the fact that observations may be correlated when participants are examined repeatedly over time.
That feature is important here because the registry explicitly notes that some participants are revisited across the years. Treating every observation as completely independent could underestimate the amount of statistical dependence in the data. A mixed-effects framework can incorporate dependence through model structure, allowing repeated observations from the same underlying participants or other clustered observations to contribute without requiring them to be treated as unrelated measurements.
Fixed effects represent systematic differences associated with the comparison of interest, while random effects provide a way to represent correlation or heterogeneity associated with repeated or clustered observations.
Incidence rate ratio
The effect measure reported by the registry is the incidence rate ratio. Unlike a simple proportion, an incidence rate incorporates the amount of observation time. The primary endpoint therefore compares the rate at which newly diagnosed cases accumulated rather than simply comparing the percentage of participants who developed leprosy.
Rate ratio below 1
The estimated incidence rate is lower in the intervention group than in the comparator group.
Rate ratio equal to 1
The estimated incidence rates are equal between the two groups.
Rate ratio above 1
The estimated incidence rate is higher in the intervention group than in the comparator group.
What it is not
A rate ratio is not a risk ratio and does not directly state the absolute probability that an individual participant will develop leprosy.
Analysis population
The registry identifies the analysis population for each primary statistical analysis as participants examined. The endpoint definition also makes clear that incidence rates were based on cases newly diagnosed while participants were under follow-up.
Superiority framework
All three primary comparisons are identified in the registry as superiority hypotheses. Thus, the statistical question is whether the incidence rate differs in the direction consistent with a beneficial effect of the intervention, rather than whether an intervention is merely no worse than the comparator within a prespecified non-inferiority margin.
6. Primary Results
The registry reports three formal statistical analyses for the primary endpoint. Each compares no PEP with one of the three intervention approaches using a mixed-effects model and an incidence rate ratio.
No PEP vs Household PEP
Incidence rate ratio
98.3% CI: 0.40–2.23 · P < 0.017
Analysis population: participants examined · 45 months
| Measure | Reported result |
|---|---|
| Comparison | No PEP vs Household PEP |
| Effect measure | Incidence rate ratio |
| Estimate | 0.95 |
| Confidence interval | 98.3% CI 0.40–2.23 |
| P-value | <0.017 |
| Hypothesis | Superiority |
| Method | Mixed Models Analysis |
An incidence rate ratio of 0.95 means that the estimated incidence rate in the compared intervention group was approximately 95% of the comparator rate under the fitted analysis. Expressed descriptively, that corresponds to an estimated rate that was about 5% lower.
The estimate alone does not mean that 5% fewer participants developed leprosy, nor does it describe an individual participant's probability of developing disease. It is a relative comparison of incidence rates.
The reported 98.3% confidence interval of 0.40–2.23 is wide and spans 1.00. Thus, the point estimate should not be interpreted in isolation: the interval reflects substantial uncertainty about the magnitude and even the direction of the underlying rate ratio within the stated confidence framework.
The registry also reports P < 0.017. A p-value addresses the compatibility of the observed data with the statistical null hypothesis under the specified testing procedure; it does not measure the size or clinical importance of the effect. The reported confidence interval and p-value should therefore be considered together rather than treating either as a standalone measure of effect.
No PEP vs PEP 100m
Incidence rate ratio
98.3% CI: 0.34–1.87 · P = 0.017
Analysis population: participants examined · 45 months
| Measure | Reported result |
|---|---|
| Comparison | No PEP vs PEP 100m |
| Effect measure | Incidence rate ratio |
| Estimate | 0.80 |
| Confidence interval | 98.3% CI 0.34–1.87 |
| P-value | 0.017 |
| Hypothesis | Superiority |
| Method | Mixed Models Analysis |
An incidence rate ratio of 0.80 means that the estimated incidence rate in the PEP 100m comparison was 80% of the comparator rate under the fitted model. Descriptively, the point estimate corresponds to an estimated 20% lower incidence rate.
This is a rate comparison, not a statement that 20% of cases were prevented or that an individual participant had exactly a 20% lower probability of disease.
The reported 98.3% confidence interval of 0.34–1.87 is substantially wider than the point estimate and includes 1.00. The interval therefore communicates considerable uncertainty around the estimated rate ratio. It allows values corresponding to materially lower incidence as well as values corresponding to higher incidence.
The registry reports P = 0.017. Because the registry simultaneously reports a 98.3% confidence interval that crosses 1.00, the confidence interval and p-value should be read exactly as reported rather than substituted for one another or converted into an ordinary 95% confidence-interval interpretation.
No PEP vs PEP 100m + Positive for Anti-PGL-I IgM Ab
Incidence rate ratio
98.3% CI: 0.22–1.56 · P < 0.017
Analysis population: participants examined · 45 months
| Measure | Reported result |
|---|---|
| Comparison | No PEP vs PEP 100m + Positive for Anti-PGL-I IgM Ab |
| Effect measure | Incidence rate ratio |
| Estimate | 0.58 |
| Confidence interval | 98.3% CI 0.22–1.56 |
| P-value | <0.017 |
| Hypothesis | Superiority |
| Method | Mixed Models Analysis |
An incidence rate ratio of 0.58 means that the estimated incidence rate in the compared intervention group was 58% of the comparator rate under the fitted model. As a descriptive translation of the point estimate, that corresponds to an estimated 42% lower incidence rate.
It does not mean that 42% of participants avoided leprosy, that 42% of cases were prevented in absolute terms, or that every individual participant experienced a 42% reduction in risk.
The 98.3% confidence interval of 0.22–1.56 is wide and crosses 1.00. The point estimate is therefore much more precise as a description of the fitted estimate than it is as a statement about the range of plausible population-level rate ratios.
The registry reports P < 0.017. As with the other analyses, the p-value does not quantify effect size. The reported p-value and the 98.3% confidence interval should be interpreted within the statistical framework used for the three comparisons rather than by replacing the registry's stated confidence level with the conventional 95% level.
7. Comparing the Three Primary Estimates
| Comparison | Rate ratio | 98.3% CI | P-value |
|---|---|---|---|
| No PEP vs Household PEP | 0.95 | 0.40–2.23 | <0.017 |
| No PEP vs PEP 100m | 0.80 | 0.34–1.87 | 0.017 |
| No PEP vs PEP 100m + Positive for Anti-PGL-I IgM Ab | 0.58 | 0.22–1.56 | <0.017 |
The point estimates differ across the three intervention comparisons: 0.95, 0.80, and 0.58. Taken descriptively, each point estimate is below 1, with the third comparison showing the lowest estimated incidence rate ratio. However, point estimates alone do not establish that the three approaches have different effects from one another.
The confidence intervals are all wide and all include 1.00. This is particularly important when interpreting the size of the estimated effects. A point estimate can suggest a direction while the corresponding interval indicates that substantial uncertainty remains about the underlying effect.
8. Understanding the 98.3% Confidence Intervals
The registry reports 98.3% confidence intervals rather than the more commonly encountered 95% intervals. This distinction matters. A confidence interval is constructed at a specified confidence level, and changing that level changes the width of the interval.
The interval describes statistical uncertainty around the estimated effect under the specified model and confidence framework. It is not a range containing the effects experienced by individual participants.
For an incidence rate ratio, 1.00 is the reference value corresponding to equal incidence rates. An interval that includes 1.00 therefore does not exclude equality of the incidence rates at that confidence level.
The three reported intervals are 0.40–2.23, 0.34–1.87, and 0.22–1.56. Their breadth indicates that the point estimates are accompanied by considerable statistical uncertainty.
9. Why a Mixed-Effects Model Fits the Longitudinal Structure
The registry states that part of the same participant population was revisited over the years. This creates a central statistical issue: repeated measurements from the same participant are generally correlated rather than independent.
Repeated observations
The same participant can contribute information at more than one household survey.
Correlation
Measurements from the same participant may be more similar to one another than measurements from unrelated participants.
Random effects
A mixed model can represent participant-level or cluster-level variation through random effects.
Rate estimation
The model provides a framework for estimating relative incidence while accounting for the longitudinal structure specified in the analysis.
The use of a mixed-effects approach does not eliminate uncertainty. Instead, it provides a model-based way to account for dependence and heterogeneity in longitudinal data. The resulting rate ratios remain estimates subject to sampling variation and model assumptions.
10. Statistical Methods Explained
Why was a mixed-effects model used?
The registry notes that some participants were revisited across the years. Repeated observations create within-participant correlation. A mixed-effects model is designed to accommodate such dependence rather than assuming that every observation is statistically independent.
What does a rate ratio of 0.80 mean?
A rate ratio of 0.80 means that the estimated incidence rate in the intervention comparison is 80% of the comparator rate under the fitted model. The corresponding descriptive translation is a 20% lower estimated incidence rate. It is not a 20-percentage-point reduction in disease risk.
Why is 1.00 the important reference value?
For an incidence rate ratio, 1.00 represents equal incidence rates in the two compared groups. Values below 1 indicate a lower estimated rate in the intervention group, while values above 1 indicate a higher estimated rate.
Why does the confidence interval matter as much as the point estimate?
The point estimate is only one estimate of the underlying treatment effect. The confidence interval shows how uncertain that estimate is under the specified statistical framework. A wide interval means that the data are compatible with a broad range of effect sizes.
Why does a p-value not measure effect size?
A p-value addresses evidence against a statistical null hypothesis under a specified testing procedure. It does not say whether an effect is large or small, and it does not tell us how clinically important an effect is. The rate ratio itself describes the estimated magnitude, while the confidence interval describes uncertainty around it.
Why should the three rate ratios not be treated as a direct ranking?
Each reported rate ratio compares one intervention approach with the same comparator. The analyses do not constitute direct pairwise comparisons among the three intervention approaches. Differences between their point estimates therefore do not automatically establish differences between their underlying effects.
11. Primary Statistical Interpretation
All three reported incidence rate ratios are below 1: 0.95 for Household PEP, 0.80 for PEP 100m, and 0.58 for PEP 100m + Positive for Anti-PGL-I IgM Ab. Descriptively, each point estimate corresponds to a lower estimated incidence rate than the comparator.
These rate ratios do not directly provide absolute numbers of cases prevented, individual-level probabilities, or a direct comparison among the three intervention strategies. They are model-based relative incidence measures.
The 98.3% confidence intervals are wide and cross 1.00 in all three comparisons. Consequently, the numerical point estimates should not be interpreted as precise estimates of the underlying incidence-rate differences.
The registry reports P < 0.017, 0.017, and P < 0.017 for the three analyses. These values are inferential quantities associated with the specified statistical tests; they do not quantify the magnitude of the estimated rate ratio. The unusual combination of reported p-values and 98.3% confidence intervals should be retained exactly as reported rather than translated into a different confidence level.
12. Safety Results
The registry reports serious adverse events by arm for three of the intervention/comparator groups for which affected and at-risk counts are provided.
| Group | Serious adverse events | At risk |
|---|---|---|
| Household PEP | 0 | 731 |
| PEP 100m | 0 | 15203 |
| PEP 100m + Positive for Anti-PGL-I IgM Ab | 0 | 2850 |
For the three groups listed in the registry, no serious adverse events were reported among the stated numbers at risk. The denominators differ substantially across these reported groups, so the counts should be interpreted alongside the corresponding at-risk populations.
13. Trial Timeline
Trial start
The ClinicalTrials.gov record lists January 2, 2019 as the study start date.
Primary outcome assessment
The registered primary endpoint was evaluated over 45 months, using the incidence of newly diagnosed leprosy cases per 1000 person-years.
Primary completion
The ClinicalTrials.gov record lists January 17, 2023 as the primary completion date.
14. Important Limitations and Interpretation Issues
- Wide uncertainty: all three reported 98.3% confidence intervals are wide and include 1.00, indicating substantial uncertainty around the corresponding point estimates.
- Repeated participants: the registry explicitly notes that some participants were revisited over the years. This longitudinal structure affects the statistical independence of observations and is relevant to the choice of a mixed-effects model.
- Different analysis denominators: the registry notes that repeated visits may explain discrepancies in the overall number of participants analyzed.
- Indirect comparison of interventions: the three primary estimates compare each intervention approach with the no-PEP comparator. They are not direct pairwise comparisons between the three intervention approaches.
- Rate ratio versus risk: an incidence rate ratio incorporates follow-up time and should not be interpreted as an individual-level risk ratio.
- Confidence level: the registry reports 98.3% confidence intervals. These should not be silently converted to 95% intervals because the confidence level changes the interpretation of the interval.
- Reported p-values and intervals: the registry reports p-values alongside 98.3% confidence intervals that cross 1.00. These quantities should be reported as stated and interpreted within their stated statistical framework rather than harmonized by assuming a different confidence level.
- Safety denominators: the registry reports serious adverse event counts for three groups with different numbers at risk; those values should not be generalized to groups for which a corresponding denominator is not reported.
15. Why This Trial Matters Statistically
PEOPLE is a useful statistical teaching case because it combines randomized comparison, a rate-based endpoint, repeated observations, mixed-effects modeling, multiple intervention comparisons, confidence intervals, and longitudinal follow-up. These features create several layers of interpretation beyond simply asking whether an estimated rate ratio is below 1.
| Concept | How it appears in PEOPLE |
|---|---|
| Randomization | The trial is registered as randomized with a parallel design. |
| Rate ratio | The primary effect measure is the incidence rate ratio for newly diagnosed leprosy cases. |
| Mixed-effects model | The primary analyses use Mixed Models Analysis, normalized as a mixed-effects model. |
| Longitudinal data | Some participants are revisited across the years. |
| Confidence intervals | Each primary estimate is accompanied by a 98.3% two-sided confidence interval. |
| P-values | The three primary analyses report p-values of <0.017, 0.017, and <0.017. |
| Multiple comparisons | Three incidence rate ratios compare the comparator arm with three intervention arms. |
| Superiority testing | All three primary analyses are identified as superiority hypotheses. |
| Repeated participants | The registry specifically notes revisits of some participants across years. |
| Safety analysis | Serious adverse event counts are reported for three groups with stated at-risk denominators. |
16. Statistical Methods Explained in Practice
The most important lesson from PEOPLE is that an effect estimate cannot be separated from the structure of the data that produced it. A rate ratio of 0.58 and a rate ratio of 0.95 are not simply numbers to compare visually; they must be considered alongside their confidence intervals, the repeated-observation structure, and the fact that each comparison uses the same comparator.
The mixed-effects framework is particularly relevant because the registry does not describe a single cross-sectional assessment. Participants were followed through household surveys, and some participants were revisited. The statistical analysis therefore has to distinguish variation between participants from variation within participants over time.
The use of an incidence rate also changes the interpretation. If two groups have different amounts of observation time, comparing raw case counts alone would not adequately characterize their experience. An incidence rate standardizes the occurrence of new cases by person-time, after which the rate ratio provides a relative comparison.
Finally, the confidence intervals demonstrate why statistical interpretation should not focus solely on the point estimate. The estimated rate ratios are all below 1, but the corresponding 98.3% intervals extend above 1.00. That distinction is central to communicating uncertainty accurately.
17. Clinical Biostats Interpretation Framework
Effect size
The three reported incidence rate ratios are 0.95, 0.80, and 0.58. Each point estimate is below 1.
Precision
The corresponding 98.3% confidence intervals are wide: 0.40–2.23, 0.34–1.87, and 0.22–1.56.
Model structure
The primary analyses use mixed models, which are relevant to the repeated longitudinal observations described in the registry.
Comparative interpretation
Each intervention is compared with no PEP. The three intervention estimates are not direct pairwise comparisons with each other.
A statistically careful reading therefore separates four questions: what is the estimated effect, how precise is that estimate, what model generated it, and what comparison does it actually represent?
18. Key Takeaways
- The PEOPLE trial is a randomized phase 3 prevention study with 144000 enrolled participants and four parallel arms.
- The registered primary endpoint measures newly diagnosed leprosy cases per 1000 person-years over 45 months.
- Three primary incidence rate ratios compare the no-PEP comparator with Household PEP, PEP 100m, and PEP 100m + Positive for Anti-PGL-I IgM Ab.
- The reported point estimates are 0.95, 0.80, and 0.58, respectively.
- All three primary analyses use Mixed Models Analysis, normalized as a mixed-effects model.
- The registry reports 98.3% two-sided confidence intervals of 0.40–2.23, 0.34–1.87, and 0.22–1.56.
- The corresponding reported p-values are <0.017, 0.017, and <0.017.
- The registry specifically notes that some participants were revisited across the years, an important reason to account for longitudinal correlation in the statistical analysis.
- The point estimates should not be treated as direct rankings of the three intervention approaches because each analysis compares an intervention with the same comparator rather than comparing interventions directly with one another.
- Serious adverse event counts of 0 are reported for Household PEP, PEP 100m, and PEP 100m + Positive for Anti-PGL-I IgM Ab, with respective at-risk populations of 731, 15203, and 2850.
19. Related Tutorials
Learn more about the methods used in this trial:
20. Related Statistical Calculators
21. Sources
- ClinicalTrials.gov: PEOPLE — NCT03662022.
- Linked publication: PubMed record — PMID 38762282.
- Linked publication: PubMed record — PMID 35850123.
- Linked publication: PubMed record — PMID 31805862.
Continue through the Clinical Biostats knowledge graph
Connect the statistical concepts used in PEOPLE to deeper tutorials, calculators, and clinical-trial methodology.
22. Record Summary
PEOPLE provides a useful example of statistical analysis for a large randomized prevention trial in which the primary outcome is an incidence rate rather than a simple binary proportion. The registry reports three primary incidence rate ratios, all analyzed with mixed models, together with 98.3% confidence intervals and p-values. The longitudinal structure is especially important because the registry notes that some participants were revisited across the years. Interpreting the results therefore requires attention to the rate-based estimand, repeated observations, model structure, confidence-interval width, and the fact that each intervention is compared with the same no-PEP comparator.