Men Together Making a Difference: Reducing HIV/STD Risk Behavior Among South African Men: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Men Together Making a Difference: Reducing HIV/STD Risk Behavior Among South African Men (NCT01490359), a randomized na study. The primary outcome is Self-reported Consistent Condom Use During Vaginal Intercourse in the Past 3 Months.

2. Clinical Question

The trial addresses a behavioral prevention question: whether an intervention directed at men’s sexual risk behavior is associated with a difference in consistent condom use compared with another health-promotion program. Consistent use was defined using self-reports about vaginal intercourse and condom-protected acts during the past 3 months. The endpoint therefore concerns reported behavior, not HIV or sexually transmitted disease diagnoses.

The registry reports an odds ratio of 1.32, with a 95% confidence interval from 1.03 to 1.71 and p = .008, from a generalized estimating equations analysis. However, the registry does not identify the group coding or contrast direction for this estimate. The number alone cannot establish which group had higher odds. The intended population, exact analysis population, and repeated assessment schedule also matter when interpreting the result. The posted result should be read as an estimate for the specified behavioral endpoint, not as a direct estimate of infection prevention.

3. Trial Design

This was a randomized, parallel-group behavioral trial sponsored by the University of Pennsylvania. The registry lists enrollment of 1181 and a single-masking design. It does not state who was masked, so the masking description cannot be interpreted more specifically. The study is marked completed. Its start is recorded as 2007-11 and primary completion as 2009-12.

Participants were assigned to Men Making a Difference HIV/STD Risk Reduction Intervention or Health Promotion Control. The comparison is therefore between two behavioral programs, rather than between an intervention and no program. The HIV/STD risk-reduction intervention was designed to reduce sexual risk behaviors. The control program focused on physical activity, healthful diet, and other behaviors aimed at noncommunicable disease risks, including diabetes, hypertension, and cancers. The registry lists the phase as NA.

The primary outcome was assessed at baseline and at 6 months and 12 months post-intervention. The analysis population is described as participants with data at baseline and at least one follow-up. The registry does not report how participants were randomized, whether allocation was concealed, or how the assigned interventions were delivered. It also does not report the planned sample-size justification or detailed protocol procedures.

An analysis note flags the design as cluster-randomized and cautions that a simple individual-level re-analysis would be misleading. That warning is important: outcomes from participants in the same cluster may be correlated, and treating them as independent can give misleading uncertainty estimates. The registry’s displayed design information also identifies randomized allocation and a parallel model, but details of the clusters and their handling are not reported. The sponsor’s own analysis is therefore the relevant posted estimate; no individual-level recalculation is presented.

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.

Overall Study

Analysis population for the primary outcome, as defined in the registry: Participants with data at baseline and at least one follow-up.

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 Self-reported Consistent Condom Use During Vaginal Intercourse in the Past 3 Months, measured over Baseline, 6 months, 12 months post-intervention.

The primary endpoint is self-reported consistent condom use during vaginal intercourse in the past 3 months. A participant was coded as practicing consistent, or 100%, condom use if he reported at least one vaginal intercourse act and the number of condom-protected vaginal intercourse acts equaled the number of vaginal intercourse acts. Someone reporting at least one act but fewer condom-protected acts was coded as not practicing consistent use. Separate binary variables also represented consistent use with primary partners and casual partners.

This endpoint captures a specific behavior relevant to the intervention’s stated aim of reducing sexual risk. A binary outcome makes it possible to compare the odds of meeting the stated consistency criterion between groups. Because it relies on self-report, it describes reported behavior rather than an objectively verified behavior or a biological outcome. The registry does not report validation procedures for these reports.

Assessments were scheduled at baseline and at 6 months and 12 months post-intervention. The endpoint description does not by itself establish how measurements at those times were combined in the reported model. The registry reports a generalized estimating equations analysis, but details of the time terms and outcome aggregation are not reported. The endpoint is distinct from the listed secondary outcomes, which include other self-reported condom-use measures and sexual behaviors.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry reports generalized estimating equations (GEE) and an odds ratio for the primary comparison. The statistical explanation below is general; details of the model beyond the reported method and effect measure are not reported in the registry record.

What does GEE do?

In general, generalized estimating equations are used to analyze outcomes that may be correlated within a person or group. Repeated measurements from one participant are not necessarily independent: a person’s reports at baseline and follow-up can be related. GEE provides a framework for estimating a group-level association while accounting for a specified correlation structure. The registry lists assessments at baseline, 6 months, and 12 months post-intervention, making the repeated-measure context relevant.

GEE does not automatically resolve every design issue. The model’s working correlation structure, variance estimator, covariates, time specification, and handling of clusters all affect how the analysis is implemented. Those details are not reported. In particular, an analysis note warns that the trial is cluster-randomized and that a simple individual-level recalculation would be misleading. If participants within a cluster share characteristics or experiences, their outcomes may be correlated beyond the repeated measurements within a participant. Ignoring that structure can make standard errors and confidence intervals unreliable. The sponsor’s reported analysis takes precedence over a simple re-analysis.

How should the odds ratio be read?

In general, an odds ratio compares the odds of an outcome between two groups. An odds ratio of 1 represents equal odds under the stated contrast; values above or below 1 indicate a difference whose direction depends on which group is in the numerator. The posted estimate is 1.32, but the registry does not document the contrast coding. That omission prevents a secure translation of the estimate into a statement about which named intervention had higher odds of consistent condom use.

Odds are not the same quantity as probabilities or percentages. An odds ratio does not directly state the percentage-point difference between groups. Translating it into an absolute probability difference would require additional information, including a baseline probability and a clearly defined comparison. Those quantities are not reported, and no such translation is made.

What does the confidence interval convey?

In general, a confidence interval gives a range of estimates compatible with the observed data and the analysis assumptions at the stated confidence level. The registry reports a 95% confidence interval from 1.03 to 1.71. The interval expresses uncertainty in the estimated odds ratio; it is not a range containing a specified proportion of individual participants’ outcomes. Its interpretation depends on the model being appropriate, including how the repeated observations and clustered design are handled.

The interval is also useful for understanding precision. A narrower interval would indicate greater precision on the odds-ratio scale than a wider one, assuming comparable methods and data. The posted interval supplies more information than the point estimate alone, but it does not resolve the missing contrast direction or the unreported details of model specification.

What does the p-value add?

In general, a p-value describes how unusual data at least as incompatible with a specified null model would be if that model and the analysis assumptions were correct. The posted p-value is .008. It is not the probability that the null hypothesis is true, and it is not a measure of effect size. It also does not address whether a behavioral difference is important, sustained, or associated with infection outcomes. Its interpretation depends on the reported model and on the context of other outcomes and analyses.

The registry labels the analysis as superiority/other, but that label does not fill in the absent group coding. The estimate, interval, and p-value are reported together; the record does not report a separate adjusted analysis or a detailed statistical analysis plan. The analysis should therefore be described as posted, without inventing covariates, model terms, or assumptions not documented in the record.

How do the planning illustrations differ from trial results?

The sample-size section contains planning illustrations based on the observed effect, rather than the trial’s own design justification. It describes a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; the effect to detect is expressed as a fraction of the observed effect. In general, such calculations illustrate the sample size that a hypothetical design might require under specified assumptions. They are not observed power, and they do not explain why the original trial enrolled its participants. The registry’s own sample-size justification is not reported.

The tables and plots present the calculated displays, so this explanation does not repeat them. No additional values are calculated. Given the warning about cluster randomization, the sponsor’s own analysis remains the basis for interpretation rather than a simple individual-level reconstruction.

10. Analyses Reported by the Sponsor

The posted odds ratio, confidence interval, and p-value are from the sponsor’s generalized estimating equations analysis. The record’s analysis note specifically warns that the design is cluster-randomized and that a simple individual-level re-analysis would be misleading. For that reason, this page does not set a recalculated result alongside the sponsor’s estimate or use an unadjusted calculation to challenge it. The sponsor’s own analysis takes precedence.

This does not mean that every relevant analytical detail is available. The registry does not report the model’s correlation structure, covariates, cluster-level variance method, or the precise contrast coding. In particular, the group order for the odds ratio is not identified. The estimate of 1.32, its 95% confidence interval from 1.03 to 1.71, and p = .008 can be quoted as the posted result, but the direction between the named study groups cannot be established from the record alone.

The planning illustrations elsewhere on the page are separate from the observed trial analysis. They use the observed effect under stated hypothetical assumptions and are not a replacement for the sponsor’s analysis or a retrospective measure of observed power.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists eight outcome measures and one reported analysis. The primary endpoint is consistent condom use during vaginal intercourse in the past 3 months; the listed secondary outcomes cover other condom-use measures and sexual behaviors. The record does not explain how outcome priority was set, whether all outcomes were tested, or whether the analysis was selected from a larger set of analyses.

It is not reported in the registry record whether a multiplicity adjustment was planned or applied. When several outcomes or comparisons are examined, the chance of seeing a nominally notable result somewhere among them can differ from the chance associated with a single pre-specified test. This is a general statistical concern, not evidence that any particular adjustment was required or omitted in the trial. The reported p-value should be read in light of this unreported context rather than treated as a summary of all possible comparisons.

The primary-outcome population is described as participants with baseline data and at least one follow-up. That description identifies a data-availability condition, but does not tell readers how many participants met it, how missing follow-up observations were handled, or whether any imputation or sensitivity analyses were performed. Those details are not reported in the registry record. It is also not reported whether participants were analyzed according to original assignment or whether any other analysis population was used.

The scheduled assessments were baseline, 6 months, and 12 months post-intervention. The registry does not specify how participants with only one follow-up assessment contributed to the GEE analysis, or how incomplete outcome reports were treated. GEE can use repeated observations under particular missing-data assumptions, but those assumptions and their plausibility cannot be evaluated from the posted description. The registry also does not report a detailed statistical analysis plan or any interim analyses. These gaps limit assessment of outcome selection, missing-data sensitivity, and the full analysis process.

12. Safety Results

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

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

The registry’s serious-adverse-event summary lists 609 participants at risk in the HIV/STD Risk-reduction group, with 0 affected, and 572 at risk in the Health Promotion Control group, also with 0 affected. For other adverse events, it likewise lists 609 at risk and 0 affected in the risk-reduction group, and 572 at risk and 0 affected in the control group. These are the posted figures; they should not be read as proof that no event could occur outside the registry’s reporting scope.

No top serious adverse-event terms are listed. The death entries are null, so death counts are not reported in the registry record. The adverse-event summary does not give event definitions, severity grading beyond the serious category, event-collection procedures, or details about follow-up completeness. Without those details, the figures are best read narrowly as the registry’s reported counts of affected participants and participants at risk.

These safety entries do not change the interpretation of the primary endpoint, which is self-reported condom-use behavior. Nor do the zero affected counts establish the absence of all harms or provide evidence about infection outcomes. They describe only what is posted in the adverse-event summary.

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

The result concerns a defined behavioral endpoint: self-reported consistent condom use during vaginal intercourse in the past 3 months. The registry reports an odds ratio of 1.32, a 95% confidence interval from 1.03 to 1.71, and p = .008 from a GEE analysis. These figures summarize the association estimated by the sponsor’s model. The confidence interval describes uncertainty on the odds-ratio scale, while the p-value describes the reported test under the model’s assumptions.

The record does not identify the odds-ratio contrast direction, so it does not establish which intervention group had higher odds. The result also does not directly measure HIV or STD incidence. Although the trial’s stated rationale concerns reducing HIV/STD risk-associated behavior, a reported behavioral endpoint cannot by itself establish fewer infections. Nor does one primary result establish effects on the several listed secondary behaviors.

The endpoint is based on self-reported sexual behavior, with a specific rule for classifying consistent use among participants who reported at least one vaginal intercourse act. As a result, the finding does not describe all participants irrespective of reported activity, nor does it establish how accurate each report was. The analysis population also required baseline data and at least one follow-up, while the handling of missing observations is not reported.

Statistical interpretation depends on the analysis and design. The registry note flags cluster randomization and cautions against simple individual-level re-analysis. The sponsor’s GEE result is therefore the posted basis for discussion, but model details and group coding remain unavailable. The record also does not describe multiplicity procedures for the eight outcome measures.

In short, the posted result is an estimate for a self-reported behavioral outcome in this trial, with stated statistical uncertainty. It does not, on its own, establish a directional comparison between the named arms, an effect on infection, or a broader conclusion about the intervention beyond the outcome and analysis recorded.

14. Why This Trial Matters Statistically

This trial is statistically instructive because its primary outcome is a repeated behavioral measure rather than a single biological event. The registry lists baseline, 6-month, and 12-month post-intervention assessments. Repeated measures can contain information about behavior over time, but observations from the same participant may be correlated. A method such as GEE is intended to address correlated outcomes within a model, though the details of this implementation are not reported.

The record also highlights the importance of design-level dependence. An analysis note identifies cluster randomization and explicitly cautions against a simple individual-level re-analysis. This matters because the unit of assignment and the unit of observation need not be identical. When observations share a cluster, treating them as independent can understate uncertainty. The warning explains why the sponsor’s analysis, rather than a mechanical recalculation, is the appropriate posted estimate to discuss.

Another statistical lesson is that a point estimate is not enough to interpret a comparison. The registry reports an odds ratio, confidence interval, and p-value, but not the group coding for the contrast. Even a numerical estimate with an interval and test result can be difficult to translate into a plain-language direction if the comparison is undocumented. The reporting gap also illustrates why the endpoint scale matters: an odds ratio is not a percentage-point difference.

Finally, the registry’s outcome list, analysis count, missing-data description, and adverse-event summary each answer different questions. A single reported analysis does not describe whether multiplicity was considered across listed outcomes. A stated analysis population does not explain missing-data handling. Zero affected counts in the posted safety summary do not establish the absence of every possible adverse event. Reading these pieces separately helps keep the statistical claim aligned with what the record actually documents.

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.

Several trial-specific limitations of interpretation arise from the registry record. The most consequential is the analysis note flagging cluster randomization and warning against simple individual-level re-analysis. The record does not describe cluster construction, the number of clusters, or the details of how clustering was handled in the sponsor’s model. This prevents independent assessment of the analysis implementation and is why no simple recalculation is presented.

The reported odds ratio’s contrast direction is not identified. Without knowing which group is the numerator, the estimate cannot be translated confidently into a statement that one named intervention had higher odds of the endpoint. The model’s correlation structure, covariates, time specification, and variance method are also not reported. The record gives the method label, effect measure, interval, and p-value, but not enough detail to reconstruct the model.

The primary endpoint relies on self-report and applies only to participants reporting at least one vaginal intercourse act under the stated coding rule. The record does not describe validation of self-reports. It defines the analysis population as participants with baseline data and at least one follow-up, but does not report the number with complete data, the extent of missingness, or the method for handling incomplete observations.

The record lists eight outcome measures but does not report whether multiplicity was addressed, how the primary outcome was chosen, or whether a protocol or detailed analysis plan prespecified the analysis. It also does not identify who was masked under the single-masking designation. The adverse-event summary lacks event-collection details, and death counts are not reported. These are limits of what can be established from the registry record, not proof that the corresponding procedures or information did not exist elsewhere.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Men Together Making a Difference: Reducing HIV/STD Risk Behavior Among South African Men — NCT01490359.
• Registry record last updated 2017-12-19; status last verified 2017-11.
• Results first posted 2017-12-19.
• 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

NCT01490359 is the completed South African Men Health Promotion Project, sponsored by the University of Pennsylvania. The registry lists randomized parallel allocation, single masking, and enrollment of 1181. The study compared Men Making a Difference HIV/STD Risk Reduction Intervention with a behavioral Health Promotion Control.

The primary outcome was self-reported consistent condom use during vaginal intercourse in the past 3 months, assessed at baseline, 6 months, and 12 months post-intervention. The sponsor’s GEE analysis reports an odds ratio of 1.32, a 95% confidence interval from 1.03 to 1.71, and p = .008. The contrast direction is not identified, and the registry warns that the design is cluster-randomized, making a simple individual-level re-analysis misleading.

The record lists eight outcome measures and one analysis. Missing-data procedures, multiplicity handling, detailed model specifications, and death counts are not reported in the registry record. The posted result should therefore be read as a model-based estimate for the defined behavioral endpoint, not as a direct estimate of infection outcomes.