Efficacy of an Abstinence-Only HIV Risk-Reduction Intervention for Young African-American Adolescents: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Efficacy of an Abstinence-Only HIV Risk-Reduction Intervention for Young African-American Adolescents (NCT00640653), a randomized na study. The primary outcome is Self-report of Ever Having Sexual Intercourse.

2. Clinical Question

3. Trial Design

The registry lists this as a randomized, single-masked, parallel study with an enrollment of 662 and completed status. It identifies the sponsor as the University of Pennsylvania. The study dates are listed as 2001-09 for start and 2004-08 for primary completion. Phase is recorded as NA. The registry describes the study’s broader aim as developing and evaluating culturally appropriate HIV/STD risk-reduction interventions for young African-American adolescents.

Five behavioral interventions were assigned to separate arms: abstinence-only, safer-sex-only, long comprehensive, short comprehensive, and health-promotion control. The long and short comprehensive arms are described as 12-h and 8-h interventions, respectively. The registry identifies the control arm for the posted comparisons as the health-promotion control.

The primary outcome analysis concerns a subset rather than all enrolled participants: those reporting never having had sexual intercourse at baseline and with follow-up data on self-reported intercourse during post-intervention assessments. Arm counts for this outcome are 92 for comprehensive-long, 97 for comprehensive-short, 85 for safer-sex only, 95 for abstinence-only, and 88 for health-promotion control. These outcome counts should not be treated as the overall enrollment or as a description of all participants in each arm.

The registry record does not report which people were masked, the allocation sequence or concealment procedures, or the original sample-size justification. The intervention delivery procedures and the specific content of the control intervention are also not reported in the registry record. Although the study is listed as randomized and single-masked, those labels alone do not establish how the randomization or masking was implemented.

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 reporting never having sexual intercourse at baseline (i.e., virgins) with follow-up data on self-reported sexual intercourse during the post-intervention assessments.

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.

Population: Sociodemographic Characteristics and Self-Reported Sexual Behaviors at Baseline of Participating African American Students in Grades 6 and 7 by Intervention Arm

6. Primary Endpoint

The registry lists the primary outcome as Self-report of Ever Having Sexual Intercourse, measured over 24 months post-intervention.

The primary endpoint is self-report of ever having sexual intercourse, measured among participants who reported never having had sexual intercourse at baseline. The registry describes the measure as self-reported sexual initiation during the follow-up period and gives the timeframe as 24 months post-intervention. The posted arm values are counts of participants reporting the event, alongside the number of participants represented in each arm for this outcome.

Restricting this outcome to participants who reported no prior intercourse at baseline makes it an assessment of reported initiation during the follow-up period rather than a measure of whether participants had ever had intercourse at any point in life. Self-report is the recorded method, so the endpoint reflects participants’ reports; the registry does not describe an independent confirmation method.

This endpoint is useful for asking whether the proportions reporting initiation differed between intervention arms and the health-promotion control among the stated baseline subgroup. It is a binary outcome: each participant either contributes an event count or does not, according to the recorded measure. It is not an HIV or STD diagnosis. Although the trial concerns HIV and sexually transmitted disease prevention, the posted primary outcome is sexual behavior, and the registry does not report a primary infection outcome in these results.

The endpoint’s interpretation is limited to the stated population, self-report measure, and follow-up timeframe. The registry record does not report how missing follow-up responses were handled or whether the endpoint was collected at additional time points beyond the described post-intervention assessments.

7. Results Posted in the Registry

• Comprehensive-long: 42.4% (39/92) versus 46.6% (41/88) on control; risk difference -4.2 percentage points (95% CI -18.7 to 10.3), risk ratio 0.91 (0.66–1.26); unadjusted two-proportion z-test p = 0.571.
• Comprehensive-short: 41.2% (40/97) versus 46.6% (41/88) on control; risk difference -5.4 percentage points (95% CI -19.7 to 8.9), risk ratio 0.89 (0.64–1.23); unadjusted two-proportion z-test p = 0.464.
• Safer-sex Only: 51.8% (44/85) versus 46.6% (41/88) on control; risk difference 5.2 percentage points (95% CI -9.7 to 20.1), risk ratio 1.11 (0.82–1.51); unadjusted two-proportion z-test p = 0.496.
• Abstinence-only: 32.6% (31/95) versus 46.6% (41/88) on control; risk difference -14.0 percentage points (95% CI -28.0 to 0.1), risk ratio 0.70 (0.49–1.01); unadjusted two-proportion z-test p = 0.053.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is binary: participants are counted according to whether they reported ever having had sexual intercourse during the stated follow-up assessment. For a binary outcome, an event count is interpreted relative to the number of participants represented in that arm. The posted arm summaries give both counts and denominators, while the comparisons express the between-arm contrast as a risk difference. The registry’s sponsor analysis reports a risk ratio instead. These measures answer related but different questions, so their numerical values should not be compared as if they were the same scale.

What does a risk difference express?

A risk difference subtracts the event proportion in the comparator group from the event proportion in the intervention group. It is measured on the proportion scale. A value below zero means the observed proportion was lower in the intervention arm; a value above zero means it was higher. A value of zero represents equal proportions. The recalculated estimates use the posted event counts and denominators and are described as simple unadjusted recalculations. They do not incorporate covariates or other model features not reflected in those summary statistics.

What does a risk ratio express?

A risk ratio compares the event proportion in one group with the event proportion in another by division. A value of 1 represents equal proportions, while values below or above 1 indicate a lower or higher proportion in the numerator group, respectively. The sponsor-reported risk ratio is from generalized linear regression. The registry record does not report the regression link, covariates, or other details needed to describe the model beyond that method label. The sponsor estimate should therefore be reported as the registry reports it, rather than reverse-engineered from the arm counts.

How should the p-values be read?

A p-value is a measure of how compatible the observed data are with a specified null model, given the analysis assumptions. It is not the probability that the null hypothesis is true, and it does not quantify the size or practical importance of a difference. The unadjusted recalculated p-values accompany the risk differences. The sponsor’s p = .03 accompanies its generalized linear regression result. Because the sponsor and recalculated results differ for the abstinence-only comparison, the sponsor’s own analysis takes precedence for that reported analysis; the unadjusted calculation is a distinct summary comparison, not a replacement for it.

Why does the confidence interval matter?

A confidence interval gives a range of values compatible with the observed estimate under the interval method and its assumptions. A wide interval means the data leave substantial uncertainty about the size and possibly the direction of the effect. For a risk difference, an interval spanning zero includes no difference. For a risk ratio, an interval spanning 1 includes equal proportions. These reference values help readers assess precision, but an interval does not account for unreported sources of bias, measurement error, or missing data.

For this binary endpoint, the posted counts are the basis for the recalculated comparisons. No over-dispersion estimate is reported, and the endpoint is not a count of repeated events per participant. In general, over-dispersion matters when a model assumes less variability than the data contain: standard errors can then be too small and intervals too narrow. That general concern should not be interpreted as evidence that over-dispersion occurred in this trial.

What is known about the sponsor analysis?

The registry identifies generalized linear regression and reports a risk ratio, confidence interval, and p-value for the abstinence-only versus control comparison. It does not report the full statistical analysis plan, covariate set, link function, handling of missing values, or exact analysis population for that result. The study lists one reported analysis in the record, but the registry information does not establish whether that represents all analyses conducted by investigators. Details of interim analyses and any multiplicity procedure are not reported in the registry record.

The posted comparisons also include relative-risk and odds-ratio intervals, but the main recalculated contrast is the risk difference. An odds ratio compares the odds rather than the risks, and can be less intuitive when an outcome is not rare. A confidence interval’s reference value differs by measure: zero for differences, and 1 for ratios. The existence of multiple effect measures does not by itself show that investigators selected among them after seeing the results.

All interpretation remains tied to the recorded endpoint: self-reported sexual initiation among participants who reported no intercourse at baseline and had follow-up data. Statistical methods cannot resolve limitations in self-report or establish how missing outcomes would have changed the result. Those features, along with the unreported analysis details, constrain how far the numerical comparisons can be generalized.

10. Confidence Intervals and Precision

Precision is reflected in the width of a confidence interval: narrower intervals indicate less sampling uncertainty under the analysis assumptions, while wider intervals leave a broader range of plausible values. The recalculated risk-difference intervals are broad relative to their estimates. For comprehensive-long, the interval is -0.187 to 0.103; for comprehensive-short, -0.197 to 0.089; for safer-sex only, -0.097 to 0.201; and for abstinence-only, -0.280 to 0.001. Each interval includes zero or, for abstinence-only, extends to a value very close to zero.

These intervals indicate that the unadjusted estimates do not pin down the magnitude of the risk difference precisely. For the first three comparisons, plausible values under the interval method include differences in either direction. The abstinence-only interval extends from a more negative value to 0.001, so the unadjusted interval does not cleanly exclude no difference. Its associated unadjusted p-value is 0.053. This description is distinct from the sponsor’s generalized linear regression result, which reports a risk ratio of 0.67 with a 95% confidence interval from 0.48 to 0.96 and p = .03.

Confidence intervals reflect random uncertainty under a statistical method; they do not automatically include uncertainty from measurement error, selection, or missing follow-up. The registry record does not report enough information to assess those sources fully. The differing sponsor and unadjusted results further indicate that interpretation depends on which analysis is being discussed. For the sponsor-reported comparison, the sponsor analysis takes precedence, while the unadjusted interval remains a descriptive recalculation from the posted counts.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are participants per arm, not total participants across both arms and not total events. They are illustrative calculations based on the observed effect, rather than the trial’s original design assumptions. The listed assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, a normal approximation, and an effect to detect set to the stated fraction of the observed effect. The registry record does not report the trial’s own sample-size justification or original planning assumptions.

The 80% power column gives the number of participants per arm that the planning calculation estimates would be needed to have an 80% chance of detecting the stated effect under those assumptions. The 90% power column gives the corresponding per-arm number for a 90% chance. Power is a feature of a planned design under assumptions; it is not a measure of whether the observed result is true or important.

Rows for 100%, 75%, and 50% of the observed effect explore progressively smaller target effects. The figures increase as the target effect becomes smaller, because smaller differences are harder to distinguish from random variation and generally require more participants under otherwise unchanged assumptions. This pattern can be read across the comparisons, but the resulting numbers should not be mistaken for participant requirements established by the study investigators.

These figures are planning illustrations, not observed power. The registry record does not report whether the original trial used these assumptions or whether other planning considerations informed its enrollment.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.466; treatment proportion p₁ = 0.424; average p̄ = 0.445.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.7028 = 1.3775.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.7022 = 0.5910.
4. n = (1.3775 + 0.5910)² / (-0.0420)² = 2,197.0, rounded up to 2,197 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
p0 <- 0.465909090909091                          # control proportion
p1 <- 0.423913043478261                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(640653)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 2,197 (80% power), 2,941 (90% power)
 75% of observed effect: n per arm = 3,916 (80% power), 5,242 (90% power)
 50% of observed effect: n per arm = 8,829 (80% power), 11,819 (90% power)
power.prop.test: 2197 
Simulated power at n = 2,197 per arm: 0.800 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Analyses Reported by the Sponsor

For the abstinence-only and health-promotion control comparison, the registry reports the sponsor’s generalized linear regression analysis as a risk ratio of 0.67, with a 95% confidence interval from 0.48 to 0.96 and p = .03. The posted recalculation from the arm counts is an unadjusted risk difference of -0.140, with a 95% confidence interval from -0.280 to 0.001 and p = 0.053. The measures use different scales and the analyses are not identical.

Because the record flags a discrepancy between the sponsor-reported analysis and the recalculation, the sponsor’s own analysis takes precedence when describing the sponsor’s result. The unadjusted calculation is useful as a transparent comparison based on the posted counts, but it should not be used to overwrite or reinterpret the sponsor analysis. The registry record does not report enough model detail to explain why the results differ, including whether the analyses use the same population, missing-data handling, or model specifications. No explanation for the discrepancy should be inferred from the summary statistics alone.

14. Multiplicity, Missing Data and Analysis Populations

The trial has a primary outcome and lists four secondary outcomes. The registry record does not report a multiplicity adjustment, a prespecified hierarchy for interpreting these outcomes, or a complete analysis plan. When investigators test multiple outcomes or comparisons, the chance of at least one apparently unusual result can increase if no adjustment or other strategy is used. That is a general statistical concern, not evidence that a particular adjustment was or was not applied outside what the registry reports.

The registry lists one reported analysis, the sponsor’s comparison of abstinence-only with health-promotion control. It does not establish that this was the only analysis conducted. Interim analyses are not reported in the registry record. The record also does not explain whether the sponsor-reported analysis was prespecified or how the other intervention arms were incorporated into the statistical plan.

Missing-data handling is not reported in the registry record. The primary outcome description specifies participants with follow-up data, and the posted outcome denominators differ across arms, but the record does not explain why participants lack outcome data, whether any imputation was used, or how missingness was assessed. Therefore, the reported arm counts describe participants represented in the posted outcome results; they do not by themselves establish that every randomized participant contributed to the analysis.

The analysis population is described for the primary measure as participants who reported never having had intercourse at baseline and had follow-up data on self-reported intercourse during post-intervention assessments. The registry record does not state whether this is an intention-to-treat population, a per-protocol population, or another formally defined analysis set. It also does not report how baseline reports were verified or how deviations from intervention assignment were handled. These omissions limit evaluation of the sponsor analysis and the generalizability of its estimate.

These are reporting limitations, not proof that appropriate procedures were absent. A complete protocol or statistical analysis plan could contain details not included in the registry record, but those details are not reported in the registry record available for this analysis.

15. 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 lists serious adverse event counts of 0 in each of the five intervention groups. The numbers at risk for serious adverse events are 131 for comprehensive-long, 134 for comprehensive-short, 129 for safer-sex only, 134 for abstinence-only, and 134 for health-promotion control. It also lists 0 other adverse events in each group, with the same group-specific numbers at risk shown for that category.

These entries indicate that no serious or other adverse events are recorded in the posted group summaries. They should be read as registry-reported counts, not as proof that no unrecorded event occurred. The registry lists deaths as null for each group, so death counts are not reported in the registry record. No serious adverse event terms are listed.

The adverse-event risk sets are distinct from the primary outcome denominators, and the record does not explain the difference in the group counts. It also does not report additional detail about event definitions, collection procedures, or follow-up for safety. These figures are a description of the posted safety fields rather than a complete assessment of all possible harms.

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

The posted results describe how often participants in the stated baseline subgroup reported sexual initiation by the post-intervention follow-up assessments. The arm counts range from 31 of 95 in abstinence-only to 44 of 85 in safer-sex only; the health-promotion control count is 41 of 88. The recalculated comparisons quantify differences in these observed proportions, with uncertainty intervals that are broad and do not cleanly establish the direction of the unadjusted difference for any comparison.

For the abstinence-only versus control comparison, the registry separately reports a sponsor analysis with a risk ratio of 0.67, a 95% confidence interval from 0.48 to 0.96, and p = .03. Because the posted recalculation differs, the sponsor’s own analysis takes precedence for describing that analysis. The registry summary does not give enough details to explain the difference, and neither result should be substituted for the other without qualification.

The endpoint is a self-reported behavioral measure, not a laboratory-confirmed HIV or STD outcome. The record does not report results for the four listed secondary behavior outcomes. Accordingly, the primary endpoint results do not establish what happened for those other behaviors or for infections. The trial’s broader prevention focus does not change the specific meaning of the posted endpoint.

These group comparisons do not determine what would happen for an individual participant, and the estimates do not by themselves establish a causal mechanism. Randomized allocation is listed, but the registry record does not provide details of its implementation, analysis-population handling, or missing-data procedures. Those omissions limit assessment of possible bias and of the extent to which the posted analysis represents all participants enrolled.

In general, statistical significance and practical importance are different ideas. A p-value assesses compatibility with a null model under assumptions; it does not measure benefit, harm, or importance to participants. A confidence interval communicates precision under its method, but cannot remove uncertainty from unreported design and analysis features. The appropriate reading is therefore specific: these are reported behavioral outcomes and statistical comparisons for the described population and timeframe, with important reporting limits.

17. Why This Trial Matters Statistically

This trial is statistically notable for having five behavioral arms and a primary endpoint restricted to participants reporting no prior intercourse at baseline. That structure makes the target population for the primary comparison different from the full enrollment. It also requires care when reading the posted arm counts: the denominators describe the endpoint population, not everyone enrolled in the study.

The record offers both sponsor-reported regression results and recalculations from arm-level counts. Their difference in the abstinence-only comparison illustrates why an effect estimate should be interpreted together with its scale, model, analysis population, and confidence interval. The sponsor reports a risk ratio, while the recalculation reports a risk difference. Both are legitimate statistical descriptions in general, but they are not interchangeable, and the record does not include sufficient model detail to reconcile them.

The wide unadjusted intervals also highlight the value of precision alongside point estimates and p-values. A single estimate cannot convey how uncertain a comparison is, and a p-value alone does not state the range of effects compatible with the data. The planning illustrations show how assumed target effect size relates to required participants per arm under specified planning assumptions; they are not retrospective evidence of trial power.

Finally, the registry distinguishes a behavioral endpoint from HIV/STD diagnoses, and lists secondary behavioral outcomes without posting their results. This matters for statistical interpretation because conclusions should track the measured outcome rather than the broader motivation for the intervention. The record also leaves important questions about missingness, multiplicity, and analysis populations unanswered, underscoring how registry reporting detail affects reproducibility and appraisal.

18. Limitations

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

The primary endpoint is self-reported. The registry does not describe whether responses were validated through another source, or how the assessment was administered. Self-report may be affected by recall or reporting behavior in general; the record does not quantify whether either issue affected this trial.

The endpoint analysis is limited to participants reporting never having had intercourse at baseline who also had follow-up data. The posted arm denominators range from 85 to 97. The registry does not explain missing follow-up, describe imputation, or report a sensitivity analysis. It therefore cannot establish from these results alone how the outcome would look for participants without observed follow-up data.

The sponsor’s generalized linear regression result and the unadjusted recalculation differ for the abstinence-only comparison. The sponsor’s own analysis takes precedence, but model specifications and population details needed to explain the discrepancy are not reported in the registry record. Full adjustment variables, the analysis plan, and interim analyses are also not reported. The registry lists one reported analysis, but that does not establish the full set of analyses conducted.

Although the trial is listed as randomized and single-masked, details about allocation procedures and which parties were masked are not reported. The record also does not describe intervention adherence or the statistical handling of participants who did not receive their assigned intervention. The health-promotion control’s specific content is not reported, limiting interpretation of the contrast in intervention content.

The primary endpoint is behavior rather than HIV or STD diagnosis, and results for the listed secondary outcomes are not reported in the registry record. The posted safety fields list no adverse events, while deaths are null; the record gives no further details on safety event collection. These limits concern what can be interpreted from the registry entry and do not establish that additional study documentation is unavailable.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Efficacy of an Abstinence-Only HIV Risk-Reduction Intervention for Young African-American Adolescents — NCT00640653.
• Registry record last updated 2017-12-13; status last verified 2017-11.
• Results first posted 2017-12-13.
• 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.

22. Record Summary

NCT00640653 is listed as a completed, randomized, single-masked, parallel study with enrollment of 662, sponsored by the University of Pennsylvania. It evaluated five behavioral interventions for HIV/STD risk reduction in young African-American adolescents.

The primary endpoint was self-reported sexual initiation among participants who reported never having had intercourse at baseline, with a timeframe of 24 months post-intervention. The posted arm counts were 39 of 92 for comprehensive-long, 40 of 97 for comprehensive-short, 44 of 85 for safer-sex only, 31 of 95 for abstinence-only, and 41 of 88 for health-promotion control.

The sponsor reported a generalized linear regression risk ratio of 0.67 for abstinence-only versus control, with a 95% confidence interval from 0.48 to 0.96 and p = .03. An unadjusted recalculation differs, so the sponsor’s own analysis takes precedence. The registry record does not report results for the four listed secondary outcomes, or details on missing-data handling, multiplicity, and the full statistical plan.