Preventing Sexual Transmission of HIV With Anti-HIV Drugs: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Preventing Sexual Transmission of HIV With Anti-HIV Drugs (NCT00074581), a randomized phase 3 study. The primary outcome is Linked Partner HIV Infection Rates in Early-ART and Delayed-ART Arms.

2. Clinical Question

The trial asked whether starting antiretroviral therapy (ART) in an HIV-positive partner, alongside HIV primary care, was associated with a different rate of sexually transmitted, genetically linked HIV infection in the couple’s initially HIV-negative partner than HIV primary care with ART delayed until specified clinical or CD4-count criteria were met.

The primary endpoint was not every HIV infection in a partner. For inclusion in the primary analysis, infection had to be confirmed by genotyping, with the viral envelope sequence in the index partner matching that of the infected partner. This restriction aimed to count infections attributed to the enrolled index partner rather than infections from other sources. The registry record describes the study as a phase 3 randomized trial with a parallel design and single masking; it does not report who was masked.

3. Trial Design

The registry record describes a randomized, single-masked, parallel-group phase 3 trial, marked completed. Overall enrollment was 3526. The study began in 2005-02, and primary completion was recorded as 2015-05. The sponsor was the National Institute of Allergy and Infectious Diseases (NIAID). The registry record does not report the randomization ratio or the identity of the masked parties.

In the early-ART arm, participants began ART in addition to receiving HIV primary care. In the delayed-ART arm, participants received HIV primary care and were to initiate ART when the CD4 count reached 200 to 250 cells/mm3, dropped below 200 cells/mm3, or an AIDS-defining illness developed. A note attached to that arm description refers to a Data and Safety and Monitoring Board recommendation and “as of May 1,” but the subsequent details are not reported in the registry record.

The listed drug interventions include atazanavir, didanosine, efavirenz, emtricitabine/tenofovir disoproxil fumarate, lamivudine, lopinavir/ritonavir, nevirapine and stavudine. The registry record does not specify in the posted arm descriptions how these agents were selected or assigned within the ART strategy, or provide a complete regimen-by-regimen account.

The primary endpoint data list 903 participants in the Early-ART arm and 890 in the Delayed-ART arm. These endpoint arm counts are distinct from the overall enrollment figure; the registry record does not explain the difference between them. The primary endpoint was assessed throughout the study. The registry record does not report the trial’s original sample-size justification, its planned statistical analysis details beyond the posted analysis, or whether the study used interim analyses.

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

Reasons for not completing this period:

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 Linked Partner HIV Infection Rates in Early-ART and Delayed-ART Arms, measured over Throughout study.

The primary endpoint was the event rate of incident HIV infection in partners who were HIV-negative at enrollment, reported per 100 person-years. The endpoint was assessed by randomized arm throughout the study. The key qualification is that only infections attributed to the HIV-positive index partner were included in the primary analysis. The registry record says attribution was confirmed by genotyping: the viral envelope sequence in the index case had to match that of the infected partner.

An exposure-adjusted rate expresses the number of events in relation to the time participants contribute to observation. In general, this can be useful when participants have different lengths of follow-up, because a rate uses accumulated person-time rather than treating every participant as if observed for an identical duration. A rate is not the same quantity as the proportion of participants who experienced an event.

Genetic confirmation makes the endpoint more specific to transmission within the enrolled couple, according to the registry’s stated definition. It also means the endpoint does not count all HIV infections among partners: infections not confirmed as linked to the index partner were excluded from the primary analysis. The registry record does not report additional details about follow-up completeness, the timing of individual infections, or how missing or indeterminate genotyping results were handled.

7. Results Posted in the Registry

• Delayed-ART: 1.03 versus 0.07 (event rate per 100 person-yr) on control; rate ratio 14.71 (95% CI 3.19–67.95, derived from the posted confidence limits); p ≈ <0.001.

8. Statistical Interpretation

9. Statistical Methods Explained

The registry record reports an endpoint expressed as an event rate per 100 person-years and also posts a Cox regression analysis. These methods address related but distinct aspects of event occurrence. The event rate uses accumulated exposure time as its denominator, while a Cox analysis compares the instantaneous event rates over follow-up through a hazard ratio. In general, the hazard is the event occurrence rate at a particular point in time among those who have not yet experienced the event. A hazard ratio compares hazards between groups; it is not simply a ratio of the final numbers of events.

What does an event rate describe?

In general, a rate divides event occurrence by time at risk, then expresses the result using a stated time unit. Here the unit is events per 100 person-years. A rate can account for unequal observation time, but it does not itself show how many individuals experienced an event or when events occurred. Two groups can have a similar rate and different total event counts if their accumulated person-time differs. The registry record does not post person-time totals or raw primary-event counts by arm.

The posted arm estimates are 0.07 for Early-ART and 1.03 for Delayed-ART, each per 100 person-years. The corresponding 95% confidence intervals are 0.01 to 0.2 and 0.74 to 1.38. A confidence interval describes the uncertainty in an estimate under the method used to construct it; it is not a range that contains a known true value with a stated probability after the data have been observed. Its width reflects the precision of the estimate, and its interpretation depends on the statistical assumptions.

How should the rate ratio be read?

The unadjusted recalculation compares the Delayed-ART rate with the Early-ART rate, producing a rate ratio of 14.71. A ratio of 1 represents equal rates; a ratio above 1 means the rate in the numerator group is higher than the rate in the denominator group. The 95% confidence interval is 3.186 to 67.95, and the recalculated p-value is p < 0.001. The interval is wide, so although the estimated direction is clear in this calculation, the precise size of the rate ratio is uncertain.

The recalculated comparison uses the posted arm-level rate estimates and their uncertainty with a Poisson-based approach. Such a calculation is unadjusted: it does not account for participant-level covariates or make the adjustments that may be part of a model using individual follow-up data. The registry record does not report enough detail to specify any covariates used in the sponsor’s Cox regression. The recalculation’s time basis is assumed to be 1 year because the time frame was not parseable for that calculation; it should not be mistaken for a recovered description of the trial’s actual follow-up duration.

Why can over-dispersion matter?

In general, a Poisson model assumes that the event-count variation is consistent with its mean, conditional on the model. Over-dispersion occurs when the data vary more than this assumption allows. If present but not accounted for, uncertainty can be understated and confidence intervals can be too narrow. The registry summary does not report a diagnostic assessment of over-dispersion or a specific adjustment for it. This is a general consideration for count-based rate analyses, not evidence that over-dispersion occurred in this trial.

How does the Cox result differ?

The posted Cox regression analysis reports an HR of 0.07, with a 95% confidence interval from 0.02 to 0.22 and p < 0.001. The arm ordering for that HR is Early-ART relative to Delayed-ART, while the recalculated rate ratio uses the reverse ordering. A hazard ratio also incorporates event timing in a way that a single overall rate ratio does not. The two estimates therefore answer related, but not identical, statistical questions. Their numerical values should not be compared as though one were a direct check of the other. The registry record does not report the Cox model’s detailed specification or its handling of assumptions such as proportional hazards.

In general, a p-value is calculated under a null hypothesis and a statistical model. It measures how incompatible the observed data, or more extreme data, are with that null under the model; it is not the probability that the null hypothesis is true. The posted p-values are small, but they do not quantify the magnitude of the effect or the certainty of its exact size. The estimates and their confidence intervals remain important for interpretation.

10. Confidence Intervals and Precision

The posted rate estimates are 0.07 events per 100 person-years for Early-ART and 1.03 for Delayed-ART. Their 95% confidence intervals are 0.01 to 0.2 and 0.74 to 1.38, respectively. The intervals represent statistical uncertainty around each rate estimate under the registry’s reported presentation. They are not ranges of individual participant outcomes.

The recalculated delayed-to-early rate ratio is 14.71, with a 95% confidence interval from 3.186 to 67.95. The interval excludes 1, the value corresponding to equal rates under this ratio definition, while extending across a broad range of possible relative magnitudes. This means the recalculated comparison indicates a clear direction under its assumptions, but does not pin down the size of the relative difference precisely. The p-value is p < 0.001.

The sponsor’s reported HR is 0.07, with a 95% confidence interval from 0.02 to 0.22 and p < 0.001. This interval also lies below 1, but it describes a hazard ratio rather than the recalculated rate ratio. The HR uses the Early-ART versus Delayed-ART orientation; the recalculated rate ratio reverses the group order. Confidence intervals for these measures should therefore be interpreted separately, not combined or treated as interchangeable.

In general, a narrower confidence interval indicates greater precision on the scale of the estimate, while a wider interval indicates more uncertainty. Precision is affected by the amount of information and the method used. The registry record does not report the number of primary events by arm, total person-time, or detailed model diagnostics, so those sources of uncertainty cannot be evaluated from the posted summary alone.

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 sample-size illustrations are planning calculations based on the observed effect, not the trial’s original design justification. For this rate endpoint, the figures are participants per arm, not total events. The listed assumptions are a two-sided significance level of 0.05, equal allocation between arms, a normal approximation, and an effect to detect set at the stated fraction of the observed effect. The original trial’s sample-size assumptions are not reported in the registry record.

The 80% power column gives the number of participants per arm that the planning calculation estimates would be needed for an 80% chance of detecting the specified effect under its assumptions, if that effect were the true one. The 90% power column gives the corresponding planning figure for a 90% chance. These are design illustrations; they are not evidence about the probability that the completed trial detected an effect.

The rows consider planning for 100%, 75% and 50% of the observed effect. A smaller effect is harder to distinguish from random variation, so the illustrations require more participants per arm as the effect to detect becomes smaller. The figures change with the chosen target effect and power level, and should be read as conditional planning estimates rather than as a recommended sample size for another study.

This is not observed power. It is a planning illustration calculated from the observed effect and stated assumptions. The registry record does not report whether its original design used these assumptions or this method.

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, λ₀ and λ₁ are event rates per patient-year, and T is follow-up per participant in years. This assumes Poisson counts and equal follow-up.

Worked calculation for this trial

1. Control rate λ₀ = 0.0007; treatment rate λ₁ = 0.0103 events per patient-year; follow-up T = 1.00 years.
2. n = (1.96 + 0.842)² × (1/0.0007 + 1/0.0103) / (1.00 × [ln(14.714)]²) = 1,656.3, rounded up to 1,657 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
lambda0 <- 0.0007                      # control rate, events per patient-year
lambda1 <- 0.0103                      # treatment rate
tt      <- 1                       # follow-up per participant, years
n_per_arm <- function(power, frac = 1) {
  l1 <- lambda0 * (lambda1 / lambda0)^frac   # smaller effect = fraction of the observed log rate ratio
  ceiling(round((z_a + qnorm(power))^2 * (1 / lambda0 + 1 / l1) / (tt * log(l1 / lambda0)^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 = ",")))

# Simulation 1: Poisson counts. Simulation 2: over-dispersed counts (variance = 1.5 x mean), analysed with the matching variance.
set.seed(74581)
n    <- n_per_arm(0.80)
reps <- 20000
m0 <- lambda0 * n * tt; m1 <- lambda1 * n * tt
x0 <- rpois(reps, m0); x1 <- rpois(reps, m1)
z  <- log(x1 / x0) / sqrt(1 / x0 + 1 / x1)
cat(sprintf("Simulated power, Poisson counts, n = %s per arm: %.3f\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE)))
phi <- 1.5
y0 <- rnbinom(reps, mu = m0, size = m0 / (phi - 1)); y1 <- rnbinom(reps, mu = m1, size = m1 / (phi - 1))
z2 <- log(y1 / y0) / sqrt(phi * (1 / y0 + 1 / y1))
cat(sprintf("Simulated power, over-dispersed counts (phi = 1.5), same n: %.3f\n", mean(abs(z2) > z_a, na.rm = TRUE)))

Output from running the code

100% of observed effect: n per arm = 1,657 (80% power), 2,218 (90% power)
 75% of observed effect: n per arm = 3,125 (80% power), 4,183 (90% power)
 50% of observed effect: n per arm = 7,821 (80% power), 10,470 (90% power)
Simulated power, Poisson counts, n = 1,657 per arm: 0.976
Simulated power, over-dispersed counts (phi = 1.5), same n: 0.782

The first simulation draws Poisson counts and should land close to the target power. The second allows extra variability (variance 1.5 times the mean, a purely illustrative value), which lowers the power at the same sample size, so rate trials usually need more participants than the Poisson formula suggests.

13. Analyses Reported by the Sponsor

The registry’s reported analysis is a Cox regression result: HR 0.07, 95% confidence interval 0.02 to 0.22, and p < 0.001. The unadjusted recalculation from the posted arm-level rates is a delayed-to-early rate ratio of 14.71, with a 95% confidence interval from 3.186 to 67.95 and p < 0.001. The registry record does not flag a discrepancy between the sponsor’s analysis and the recalculation.

These results should not be expected to match numerically. The sponsor’s HR compares hazards over time and is oriented Early-ART relative to Delayed-ART. The recalculated rate ratio compares overall rates in the reverse order and is derived from posted summary statistics with a Poisson-based approach. Different estimands, group ordering and analysis methods can yield different numbers while indicating a consistent direction in the posted results.

The sponsor’s analysis is the trial’s reported analysis and takes precedence as the sponsor’s own statistical result. The recalculation is an unadjusted check from summary statistics, not a replacement for the trial analysis or a reconstruction of its individual-level model. The registry record does not report the full Cox model specification, so the extent of any adjustment cannot be assessed from the posted description.

14. Multiplicity, Missing Data and Analysis Populations

The registry record lists 2 outcome measures and 2 analyses, while the posted results identify no secondary outcomes. One reported analysis is described as regression, Cox, with an HR of 0.07, a 95% confidence interval from 0.02 to 0.22 and p < 0.001. The primary endpoint comparison also has a recalculated unadjusted p-value of p < 0.001. The registry record does not report a multiplicity adjustment, a hierarchy of hypotheses, or a prespecified strategy for interpreting multiple outcomes or analyses.

Multiplicity matters in general when several hypotheses are tested: conducting more tests can increase the chance of at least one apparently unusual result under a setting where all null hypotheses hold. This is a general statistical consideration, not a claim about the trial’s actual error rate. Without a reported multiplicity plan, it is not possible to determine from the registry record whether or how the reported p-values were adjusted for other planned comparisons.

The registry record does not report how missing outcome information was handled, whether participants were excluded for missing data, or how incomplete or indeterminate genotyping was treated. It also does not report the analysis population definition, such as whether the reported comparison followed an intention-to-treat principle or another population rule. The endpoint arm counts are posted, but their relationship to overall enrollment and any exclusions is not explained.

The record does not report interim efficacy analyses, stopping boundaries, or other monitoring rules. A note in the delayed-ART arm description mentions a Data and Safety and Monitoring Board recommendation, but the available note is incomplete and does not establish the details of any interim statistical plan. These omissions limit assessment of how the analysis was planned and how missing information or multiple statistical questions were managed.

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.

Most frequent serious adverse event terms

Serious events with the highest arm-level rate: blood phosphorus decreased (9.8%), neutrophil count decreased (9.1%), abnormal loss of weight (8.5%), blood bilirubin increased (7.7%), neutropenia (7.5%).

The posted serious adverse event figures list 886 participants at risk in the Early-ART group, with 526 affected, and 877 at risk in the Delayed-ART group, with 522 affected. The registry also lists 886 and 877 at risk for the other adverse-event grouping and 0 affected in each. Death counts are not reported in the registry record.

The listed leading serious adverse event terms and percentages are Blood phosphorus decreased, 9.8%; Neutrophil count decreased, 9.1%; Abnormal loss of weight, 8.5%; Blood bilirubin increased, 7.7%; Neutropenia, 7.5%; and Hypophosphataemia, 4.9%. The posted list does not give arm-specific percentages for these terms, so it does not support a direct comparison of each listed event between arms. The registry record does not specify in this summary whether individuals could contribute to more than one term.

These registry figures describe reported adverse-event information; they do not, by themselves, establish that an event was caused by a particular intervention. The figures should be interpreted in light of the posted at-risk counts and the limitations of the available breakdown. The registry record does not provide further detail about event severity grades, timing, duration, or the criteria used to classify events as serious.

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

The posted results show lower estimated linked-HIV event rates in the Early-ART arm than in the Delayed-ART arm: 0.07 versus 1.03 events per 100 person-years. The arm-specific 95% confidence intervals are 0.01 to 0.2 and 0.74 to 1.38. The registry’s reported Cox analysis gives an HR of 0.07, with a 95% confidence interval from 0.02 to 0.22 and p < 0.001. The unadjusted recalculated rate ratio, oriented Delayed-ART relative to Early-ART, is 14.71, with a 95% confidence interval from 3.186 to 67.95 and p < 0.001.

These estimates describe the randomized comparison for the registry-defined endpoint: incident infections in initially HIV-negative partners that were genetically confirmed as linked to the HIV-positive index partner. They do not describe all HIV infections in partners, because infections not confirmed as linked were excluded from the primary analysis. The rate estimates also do not directly report the proportion of participants who became infected.

The point estimates indicate the direction of the observed difference, while the confidence intervals show uncertainty about its magnitude. The relatively broad recalculated rate-ratio interval is especially important when considering the exact size of the relative difference. Statistical evidence against a null comparison does not itself establish clinical importance, explain the mechanisms behind the results, or guarantee the same estimates in other settings.

The registry summary does not provide detailed participant-level data, complete follow-up information, the full analysis plan, or the handling of missing and indeterminate results. It therefore cannot resolve every question about model assumptions, exclusions or generalizability. The findings should be read as the results of this trial’s posted endpoint and analyses, not as a claim about circumstances outside the trial.

17. Why This Trial Matters Statistically

This trial offers a useful statistical example of how the choice of endpoint and estimand shapes interpretation. The primary outcome is an exposure-adjusted rate of genetically linked infections, whereas the sponsor’s posted analysis is a Cox-model hazard ratio. Both compare the randomized arms, but they summarize event occurrence differently. Reading the measure, group ordering and time basis alongside the estimate is necessary to understand what a comparison means.

The endpoint also illustrates how outcome definition affects the target of inference. Genotyping was required so that only infections linked to the HIV-positive index partner entered the primary analysis. This makes the endpoint more specific to transmission within the enrolled couple, while limiting its scope to infections meeting that confirmation rule. The endpoint is not equivalent to counting every infection among partners.

Precision is another central lesson. The posted rates differ, and the recalculated rate ratio is far from the equal-rate value, but its confidence interval spans a wide range of relative magnitudes. A small p-value and an uncertain effect size can coexist. Reporting the interval alongside the point estimate helps distinguish evidence about direction from certainty about magnitude.

Finally, the registry presents a sponsor-reported time-to-event analysis and a separate unadjusted recalculation from summary rates. These should be treated as related but non-interchangeable analyses, not as competing values for one identical estimand. The record’s missing details on model specification, missing data, multiplicity and original design assumptions also demonstrate why interpretation depends on transparent reporting, not only on a point estimate.

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.
• Rate comparisons use a Poisson approximation unless the sponsor posted confidence limits; real event counts are often over-dispersed.

The primary outcome is restricted to incident infections in partners who were HIV-negative at enrollment and whose infection was confirmed by genotyping as linked to the index partner. It therefore does not represent every HIV infection in the partner population. The registry record does not report how many infections were unlinked, unconfirmed, or indeterminate, or how such cases were managed in the analysis.

The posted primary arm counts are 903 for Early-ART and 890 for Delayed-ART, while overall enrollment is 3526. The registry record does not explain how these endpoint counts relate to the overall enrollment, describe the number of primary events, or provide person-time by arm. This limits independent assessment of the rate calculations and the amount of information contributing to each rate.

The recalculated comparison is based on arm-level summary statistics rather than individual participant data. It uses an assumed 1 year time basis because the time frame was not parseable for the calculation; that assumption is not a reported description of the trial’s follow-up. The recalculation is unadjusted and cannot reproduce a Cox model from the posted summary alone. The registry record does not report the full Cox model specification, including covariates or diagnostics.

The original sample-size justification, missing-data methods, analysis population, multiplicity strategy and interim analysis plan are not reported in the registry record. The sponsor’s posted analysis and the recalculation therefore have different levels of detail available for scrutiny. Safety information is also limited: the serious-event terms are not broken down by arm in the posted list, and death counts are not reported in the registry record.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Preventing Sexual Transmission of HIV With Anti-HIV Drugs — NCT00074581.
• Registry record last updated 2021-11-05; status last verified 2016-11.
• Results first posted 2016-10-31.
• 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

NCT00074581 is a completed, randomized, single-masked, parallel-group phase 3 trial sponsored by the National Institute of Allergy and Infectious Diseases (NIAID). It enrolled 3526 participants, began in 2005-02, and had primary completion recorded in 2015-05.

The trial compared early ART plus HIV primary care with HIV primary care and delayed ART in HIV-serodiscordant couples. Its primary endpoint was the rate of incident HIV infections in initially HIV-negative partners that were genotypically confirmed as linked to the HIV-positive index partner. The posted rates were 0.07 events per 100 person-years for Early-ART and 1.03 for Delayed-ART.

The registry’s reported Cox analysis gives an HR of 0.07, with a 95% confidence interval from 0.02 to 0.22 and p < 0.001. An unadjusted recalculation gives a delayed-to-early rate ratio of 14.71, with a 95% confidence interval from 3.186 to 67.95 and p < 0.001. The registry record does not report several details needed to fully assess the original design assumptions, missing-data handling, multiplicity plan or model specification.