Evaluation of the Trans Women Connected Mobile App for Changes in Sexual Health-related Behavior Among Transgender Women: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Evaluation of the Trans Women Connected Mobile App for Changes in Sexual Health-related Behavior Among Transgender Women (NCT03897049), a randomized na study. The primary outcome is Self-reported HIV or STI Testing in Past 3 Months.

2. Clinical Question

This question concerns a reported behavior, not a laboratory-confirmed diagnosis or test result. It is one part of a broader study of sexual health-related outcomes. The registry also lists outcomes concerning condom use, health care visits, social support, and mentoring. The primary comparison reported for testing uses logistic regression and an odds ratio. The odds ratio describes the relative odds of the recorded outcome between the two randomized groups, but the registry record does not specify the group ordering for the reported estimate. Accordingly, the estimate should be interpreted as the reported comparison rather than assigned a treatment-favoring direction.

3. Trial Design

The registry describes a randomized, parallel-group, cluster randomized controlled trial with no masking. It lists the study status as completed and reports enrollment of 574. The brief summary describes a planned study with 450 trans women; that planned figure and the reported enrollment are different record details and should not be treated as interchangeable.

One group was assigned to Trans Women Connected, a mobile app designed as a sexual health promotion program for transgender women. The other group was assigned to a general health app described as TWC COVID only. The brief summary says participants in both arms were asked to download and use their assigned app during a month period. The intervention app included interactive activities, resource maps, mentoring opportunities, and communication forums; the comparator app included activities focused on COVID and COVID resources.

Assessments were described at baseline, immediately following the app-use period, and at later follow-ups. The primary endpoint was assessed 3-months post intervention and was defined as self-reported HIV or STI testing in the past 3 months. The listed primary analysis population comprises participants with baseline and 3-month follow-up surveys. The registry record does not report how cluster assignment was implemented, the unit of randomization in practical terms, or how clustering was represented in the statistical model.

The primary analysis reported is logistic regression with an odds ratio. Because the record explicitly identifies a cluster randomized design, an individual-level re-analysis that ignores clustering would not be an appropriate substitute. The reported analysis is therefore the relevant estimate available in the registry. Other details of the allocation process, masking beyond the designation of none, and the statistical plan are not reported in the registry record.

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: All participants with baseline and 3-month follow-up survey.

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 HIV or STI Testing in Past 3 Months, measured over 3-months post intervention.

The primary endpoint is self-reported HIV or STI testing in the past 3 months, measured 3-months post intervention. The registry defines it as the number of participants reporting that they had tested for HIV or STI. The analysis population is described as participants with both baseline and 3-month follow-up surveys.

Testing is used as a behavioral outcome because the intervention is a sexual health promotion program and the trial aims to assess sexual health-related behavior. A self-reported measure can capture whether participants say they obtained testing during the specified period. It does not, by itself, establish that testing was clinically documented, identify the test result, or distinguish HIV testing from STI testing in the reported primary outcome.

The outcome is binary at the participant level: a participant either reports testing within the specified period or does not. Logistic regression is a method for modeling a binary outcome, and the reported odds ratio compares the odds of the outcome between groups. This endpoint therefore addresses the occurrence of reported testing, rather than the number of tests, the timing or type of test, or subsequent health outcomes. The registry record does not report further detail about how self-reports were verified or how missing responses were handled.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary result uses logistic regression, a statistical method for outcomes recorded in two categories. Here the outcome is whether a participant reported HIV or STI testing in the past 3 months. Rather than model the outcome as a continuous measurement, logistic regression models the odds of the reported event. The registry reports the group contrast as an odds ratio.

What does the odds ratio express?

An odds ratio compares the odds of an outcome between two groups. A value of 1 corresponds to equal odds. Values below or above 1 indicate that the odds differ in the direction determined by which group is placed in the numerator. Because the registry record does not state the ordering of the groups for this estimate, the odds ratio of 0.961 should not be attributed to a specific arm as a directional effect. It is safest to describe it as the reported group comparison.

An odds ratio is not the same as a difference in the percentage of participants reporting testing. It is also not a direct measure of how many participants were tested. The record provides an odds ratio, confidence interval, and p-value, but does not report the group-specific event counts or proportions for this endpoint in the registry record used for this interpretation. No absolute difference should be inferred from the odds ratio alone.

How should the confidence interval be read?

The reported 95% confidence interval runs from 0.463 to 1.994. A confidence interval describes uncertainty in the estimated contrast under the statistical model and its assumptions. The interval includes 1, the value corresponding to equal odds. It also spans values below and above 1, so the result is compatible with different directions of association. Its breadth indicates limited precision about the magnitude of the group contrast. The confidence interval is not a guarantee that the true effect lies within these particular endpoints, nor does it assign a probability to each possible effect.

The reported p-value is 0.914. A p-value is a measure of how compatible the observed data are with a specified null hypothesis under the model and its assumptions. It is not the probability that the null hypothesis is true, and it does not measure the clinical importance of a result. Read alongside the confidence interval, it indicates that the reported analysis does not provide evidence against equal odds at conventional significance levels. It does not prove equivalence or rule out meaningful differences.

Why does the cluster design matter?

In a cluster randomized trial, assignment is organized around groups or clusters rather than treating every participant as an independent randomization unit. Participants in the same cluster can share circumstances that make their outcomes more alike. This within-cluster similarity is called intracluster correlation. General statistical principle: when outcomes are correlated within clusters, an analysis that treats all participants as independent can understate uncertainty. Appropriate methods generally need to account for the design to produce valid standard errors and confidence intervals.

The registry record labels the trial cluster randomized and reports logistic regression, but does not describe the cluster structure or how the model accounts for it. The registry entries therefore do not establish which adjustment was used. The sponsor's reported analysis is retained as the relevant result; an aggregate individual-level recalculation would not reproduce or replace a cluster-aware analysis. The registry record does not report additional model details such as covariates, estimation approach, or diagnostics.

What is not established by this result?

The estimate concerns self-reported testing during the specified recall period, in the stated follow-up population. It does not estimate the effect on test results, diagnose HIV or STI, or establish changes in other outcomes. The record lists many outcome measures and several secondary outcomes, but the multiplicity procedure for interpreting those analyses is not reported in the registry record. Missing-data procedures, a prespecified statistical analysis plan, and details about the primary model are also not reported in the registry record.

General statistical principle: estimates and tests are conditional on the data included and the model assumptions. In this trial, the stated analysis population requires baseline and 3-month survey data, while the approach to incomplete follow-up is unspecified. The reported confidence interval and p-value should therefore be interpreted as outputs of the sponsor's stated logistic-regression analysis, not as a complete account of every possible source of uncertainty.

10. Exact Estimates for Each Arm

No two-arm comparison is shown for this trial, but each arm's result can still be summarized with exact interval estimates computed from the posted counts. The Clopper-Pearson interval is exact and conservative; the Wilson interval is a common closer-to-nominal alternative; the Bayesian credible interval uses a uniform Beta(1, 1) prior. For small groups the three can differ noticeably.

R code and output

arm    <- c("TWC App", "General Health App")
n      <- c(98, 109)
events <- c(76, 88)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  TWC App  98  76  77.6  68.0  85.4  68.3  84.7  68.3  84.7
ARM  General Health App  109  88  80.7  72.1  87.7  72.3  87.0  72.3  87.0

11. Analyses Reported by the Sponsor

The registry reports the sponsor's own logistic-regression analysis for the primary outcome: an odds ratio of 0.961, a 95% confidence interval from 0.463 to 1.994, and p = 0.914. The analysis concerns self-reported HIV or STI testing in the past 3 months, assessed 3-months post intervention, among participants with baseline and 3-month follow-up surveys.

No independent recalculation is presented. The registry identifies the trial as cluster randomized, and a simple participant-level re-analysis from summary counts could be misleading because it may fail to account for clustering. The sponsor's reported analysis is therefore the result to interpret. The registry record does not report enough detail about the cluster structure or model specification to independently reproduce that analysis.

The interval includes 1 and extends across values below and above it. The estimate should be read as uncertain, and the group ordering is not specified in the record. The reported result does not establish equivalence or a particular directional effect. No discrepancy between a sponsor analysis and a recalculation is assessed because a recalculation is not shown.

12. Multiplicity, Missing Data and Analysis Populations

The registry lists 26 outcome measures and 26 analyses, including the primary measure and outcomes concerning condom use, health care visits, social support, and mentoring. When several outcomes or analyses are considered, the chance of at least one apparently notable result can increase. Statistical methods for addressing multiple comparisons can help control that risk, but the multiplicity procedure is not reported in the registry record. The primary outcome is identified, but the record does not explain whether other analyses were treated as confirmatory, exploratory, or adjusted for multiplicity.

The registry describes the primary analysis population as all participants with baseline and 3-month follow-up survey data. This defines who is included in the stated endpoint analysis, but it does not explain what happened to participants without follow-up data. The amount or pattern of missing data, the reasons for missingness, and any imputation or other missing-data approach are not reported in the registry record. It is therefore not possible to determine from the record whether missingness could affect the comparison or how the analysis addressed it.

The registry record also does not report a statistical analysis plan, the handling of protocol deviations, or further details about the analysis population beyond the stated survey requirement. It does not describe whether the model included covariates or how clustering was incorporated. These omissions do not invalidate the reported estimate, but they constrain evaluation of the analysis choices and their implications.

General statistical principle: multiplicity and missing data concern different sources of uncertainty. Multiplicity is about interpreting a family of analyses; missing data is about which observations are available and whether the observed cases represent the intended population. The registry information does not allow either issue to be fully assessed for this result.

13. 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 0% reporting threshold.

The registry lists no serious adverse events, no other adverse events, and no deaths in either reported group. For the TWC App group, the record reports 275 at risk and 0 affected for serious adverse events, other adverse events, and deaths. For the General Health App group, it reports 299 at risk and 0 affected for those categories.

These figures describe the adverse-event categories recorded in the registry. They do not show that no adverse experiences of any kind occurred, nor do they establish that the apps are free of risk. The record does not report further details about event definitions, collection procedures, follow-up completeness for safety reporting, or whether events were assessed as related to an intervention. The figures should therefore be read narrowly as the registry's reported counts in the listed categories.

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

The primary result is a logistic-regression comparison of self-reported HIV or STI testing in the past 3 months, assessed 3-months post intervention. The reported odds ratio is 0.961, with a 95% confidence interval from 0.463 to 1.994 and p = 0.914. The estimate is close to equal odds, but the confidence interval is wide and includes values on both sides of 1. The result is not precise enough to identify a clear direction or magnitude of difference from this estimate alone.

This does not mean that the groups have been shown to be equivalent. Demonstrating equivalence requires an appropriate design and a prespecified range of differences considered acceptably small; such a range is not reported in the registry record. A non-significant result also does not establish that the intervention has no effect. It means the reported analysis did not provide evidence against the null hypothesis at conventional significance levels.

The endpoint is self-reported testing within a specified recall period, not test results or diagnoses. The finding does not address all sexual health outcomes, all outcomes listed in the record, or every follow-up time. It is tied to the stated analysis population of participants with baseline and 3-month follow-up survey data. The registry record does not report how missing follow-up data were handled.

The trial is described as cluster randomized. The available record does not explain how clustering was accounted for in the logistic regression, so the model's implementation cannot be fully assessed from the information reported. The odds ratio's group ordering is also not stated, preventing attribution of a direction to one app arm. The result is best understood as an uncertain reported comparison for this endpoint and population, not a general conclusion about app-based programs or clinical outcomes.

15. Why This Trial Matters Statistically

This trial raises statistical questions that are important when evaluating behavioral interventions. The primary outcome is a participant-reported behavior rather than a clinical test result, so its interpretation depends on what was asked, the recall period, and who completed follow-up. The registry identifies the outcome and timing, but details about verification and missing responses are not reported.

The cluster randomized design is also consequential. Statistical analyses need to reflect the unit of randomization and potential similarities among participants within clusters. A model that ignores that structure can give misleading uncertainty estimates. The registry reports a logistic-regression result but does not describe the cluster adjustment, making it important to avoid treating an aggregate individual-level recalculation as equivalent to the sponsor's analysis.

The estimate illustrates why a point estimate, confidence interval, and p-value should be read together. The odds ratio is close to 1, while its confidence interval is broad and includes values in either direction. The p-value does not establish equivalence, and the interval leaves substantial uncertainty about the size of the contrast. Precision, not only statistical significance, matters when deciding what a result can support.

Finally, the record lists multiple outcomes and analyses, yet the approach to multiplicity is not reported. Alongside the missing information about follow-up and modeling, this limits assessment of how the primary result fits within the broader set of analyses. These considerations are relevant to interpreting the posted result without extending it beyond the endpoint and analysis that were reported.

16. 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 registry record gives limited detail about the statistical implementation. Although it identifies a cluster randomized design and logistic regression, it does not report the cluster structure, the method used to account for clustering, or further model specifications. This makes it difficult to independently assess the standard errors and confidence interval or reproduce the reported analysis.

The group ordering for the odds ratio is not specified, so the estimate cannot be attributed to a directional contrast favoring either app. The record also does not report group-specific event counts or proportions for the primary outcome in the registry record used for this interpretation. As a result, the odds ratio cannot be translated into an absolute difference from the reported information alone.

The endpoint relies on self-report of testing during the past 3 months. The registry does not report validation procedures, the wording of the survey question, or details about whether participants distinguished HIV from STI testing. The analysis population requires baseline and follow-up survey data, but the handling and extent of missing follow-up are not reported.

The record lists 26 outcome measures and 26 analyses without reporting a multiplicity approach. It also does not report a statistical analysis plan or enough information to distinguish confirmatory from exploratory analyses. These omissions limit assessment of selective interpretation across outcomes. The stated results should therefore be interpreted as the reported primary comparison, with the uncertainty and reporting gaps described, rather than as a complete account of every design and analysis decision.

17. Related Tutorials

18. Related Statistical Calculators

19. Sources

• ClinicalTrials.gov: Evaluation of the Trans Women Connected Mobile App for Changes in Sexual Health-related Behavior Among Transgender Women — NCT03897049.
• Registry record last updated 2024-05-06; status last verified 2023-12.
• 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.

20. Record Summary

The registry record describes a completed, randomized, parallel-group cluster trial with no masking, comparing the Trans Women Connected app with a general health app. It reports enrollment of 574 and identifies self-reported HIV or STI testing in the past 3 months, assessed 3-months post intervention, as the primary outcome.

The reported logistic-regression estimate is an odds ratio of 0.961, with a 95% confidence interval from 0.463 to 1.994 and p = 0.914. The interval includes 1 and is wide. The record does not specify the group ordering for the estimate or describe how clustering was handled in the model. It also does not report the approach to multiplicity or missing data. The result is an uncertain comparison for the stated self-reported endpoint and analysis population.