Helping HIV Infected Patients in South Africa Adhere to Drug Regimens: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Helping HIV Infected Patients in South Africa Adhere to Drug Regimens (NCT00076804), a randomized phase 2 study. The primary outcome is Impact of DOT Compared to Self-administered Treatment as Measured by HIV Viral Load at 12 Months of Treatment.

2. Clinical Question

The statistical comparison is between the proportion meeting the stated viral-load threshold in the peer-supporter arm and the proportion in the self-administration arm. Because the outcome is whether a participant did or did not meet that threshold, the result is expressed as a binary outcome rather than a change in a continuous laboratory measurement. The observed result describes the trial participants’ recorded outcomes; it does not, by itself, establish that the intervention caused a difference. Interpretation also depends on uncertainty around the estimated difference, how outcomes were handled, and the trial’s conduct. Details about those aspects are limited in the registry record.

3. Trial Design

The registry describes a phase 2, randomized, parallel-group trial with no masking. It lists enrollment of 274 and identifies Johns Hopkins University as the sponsor. The study began in 2005-02, had primary completion in 2008-09, and is marked terminated. The registry record does not report why the trial was terminated.

Participants were assigned to a peer-supporter strategy or self-administration. In the peer-supporter arm, a participant-nominated peer observed the morning dose of antiretroviral drugs. In the comparator arm, participants self-administered their antiretroviral drugs. The intervention is therefore a difference in support and dose observation around treatment-taking, rather than a stated comparison between medicines. The registry record does not report the allocation sequence method, concealment procedures, masking of outcome assessors, or the intended duration of the intervention.

The posted primary outcome is the proportion of patients with HIV RNA levels below 400 at 12 months, described as an intention-to-treat analysis. The primary outcome title refers to viral load at 12 months, while its timeframe field states 12 and 24 months of treatment. The posted result described in the record is at 12 months; a result at 24 months is not reported in the registry record. The record’s study hypothesis also refers to mean change in CD4 count at 12 and 24 months, but no secondary results are posted in the listed outcome data.

The registry reports 136 participants in each outcome arm. The analysis is presented as a comparison of binary outcome counts, with a chi-squared test reported by the sponsor. More detail on the prespecified analysis plan, handling of missing outcomes, and any adjustments is 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

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 Impact of DOT Compared to Self-administered Treatment as Measured by HIV Viral Load at 12 Months of Treatment, measured over at 12 and 24 months of treatment.

The primary endpoint records whether a participant’s HIV RNA level was below 400 at 12 months. Each participant therefore contributes to one of two categories: meeting the threshold or not meeting it. The posted results give the number meeting the threshold and the number assessed in each arm. The endpoint is a measure of viral-load status at a specified follow-up point; it is not itself a measure of adherence, CD4 change, or the trajectory of viral load over time.

A threshold-based endpoint can make a clinical outcome straightforward to summarize and compare between groups. Statistically, it turns a continuous laboratory value into a binary outcome. That simplifies interpretation of the proportion reaching the stated threshold, but it also means the comparison does not describe how far above or below the threshold individual values were. Two participants on opposite sides of the cutoff may have values close to one another, while participants in the same category may have quite different values. The record does not report the distribution of the underlying HIV RNA measurements.

The endpoint description specifies an intention-to-treat analysis. That label ordinarily refers to analyzing participants according to their randomized arm, regardless of adherence to the assigned strategy. The registry record does not explain the operational handling of participants whose 12-month viral-load result was unavailable, so the label alone does not clarify how missing outcomes were treated. Although the outcome timeframe mentions 12 and 24 months, the posted primary result is the 12-month threshold outcome.

7. Results Posted in the Registry

• Self Administration: 68.4% (93/136) versus 72.8% (99/136) on control; risk difference -4.4 percentage points (95% CI -15.2 to 6.4), risk ratio 0.94 (0.81–1.10); unadjusted two-proportion z-test p = 0.425.

8. Statistical Interpretation

9. Statistical Methods Explained

The posted comparison is a binary-outcome analysis. A participant either had an HIV RNA level below 400 at 12 months or did not. The arm-level counts are 99 of 136 for Peer Supporter and 93 of 136 for Self Administration. This structure supports summaries based on proportions and comparisons of risks, where “risk” means the proportion experiencing the outcome being counted. In this case, the counted event is meeting the viral-load threshold, not an adverse event.

What does the risk difference describe?

The risk difference is the proportion in Self Administration minus the proportion in Peer Supporter. Its estimate is -0.044. This absolute measure expresses the contrast on the proportion scale, rather than as a relative change. A negative estimate means the recorded proportion meeting the threshold was lower in Self Administration. Its confidence interval, from -0.152 to 0.064, describes substantial uncertainty around that estimate and includes zero, the value corresponding to no absolute difference. The registry record does not report an adjusted risk difference.

How do relative measures differ?

The risk ratio compares the event proportions as a ratio, and the odds ratio compares the odds of the event. The posted risk ratio is 0.939, with a confidence interval from 0.806 to 1.095; the odds ratio is 0.808, with a confidence interval from 0.479 to 1.363. For these relative measures, the value corresponding to no contrast is 1 rather than zero. Both confidence intervals include that value. Relative measures can look different from an absolute risk difference because they use different scales and answer different questions. The odds ratio is not the same as a risk ratio, particularly when the outcome is not rare.

What does the chi-squared test assess?

The sponsor reports a chi-squared test with p = 0.42. This test assesses whether the observed distribution of the binary outcome across the arms is compatible with a null hypothesis of no association between arm and outcome, under the test’s assumptions. The unadjusted recalculation reports p = 0.425. A p-value does not measure the probability that the null hypothesis is true and does not quantify the size of the effect. It should be interpreted alongside the effect estimate and its confidence interval.

Why does the confidence interval matter?

A confidence interval represents uncertainty in an estimate under a specified statistical procedure. In repeated use of that procedure under its assumptions, intervals constructed in this way would cover the true parameter at the stated confidence level at the corresponding long-run rate. For an individual trial, the interval is best read as a range of estimates reasonably compatible with the observed data and model, not as a probability distribution for the true effect. Here, the risk-difference interval permits both a negative and a positive contrast. That is why the point estimate alone is not enough to characterize the result.

What assumptions are relevant?

These simple comparisons use the posted arm-level counts and do not account for covariates or other design features. General statistical considerations include independent observations, correctly classified outcomes, and a well-defined analysis population. If observations are correlated, standard errors calculated as if they were independent can be too small; over-dispersion, when present in a model, means variability exceeds what the model assumes and may also make uncertainty appear too narrow. These are general considerations, not claims that either problem occurred in this trial. The registry record does not report diagnostics or enough methodological detail to assess them.

The endpoint is described as intention-to-treat, but the registry record does not report a missing-data method, a full prespecified statistical analysis plan, or whether any sensitivity analyses were performed. Consequently, the posted summaries support interpretation of the displayed comparison, but not a reconstruction of every analysis decision. The study hypothesis mentions CD4 change at 12 and 24 months, while the listed results do not include a posted analysis of that measure. No conclusion about that hypothesis can be drawn from the viral-load result alone.

10. Confidence Intervals and Precision

The risk-difference estimate is -0.044, with a 95% confidence interval from -0.152 to 0.064. This interval spans zero, so the data do not pin down the direction of the absolute difference. Its lower end is compatible with a larger negative contrast for Self Administration minus Peer Supporter, while its upper end is compatible with a positive contrast. The interval therefore conveys more than a binary significant-or-not-significant label: it shows the range of effect sizes that remain compatible with the posted data under the calculation.

The interval is not a guarantee that the true difference lies within those exact limits, and it does not assign probabilities to individual possible values. Its interpretation depends on the statistical method and assumptions. The registry record reports the estimate and interval but does not provide enough detail to evaluate all assumptions or any adjustments. In particular, the listed result does not state how unavailable viral-load outcomes, if any, were handled.

For the risk ratio, the estimate is 0.939 and the interval runs from 0.806 to 1.095. For the odds ratio, the estimate is 0.808 and the interval runs from 0.479 to 1.363. These relative-effect intervals include 1, their no-contrast value. They should be read on their own scales and not directly compared numerically with the risk-difference interval.

The posted result is therefore imprecise about the size and direction of the contrast. That does not mean every possible effect is equally plausible, and it does not establish equivalence. It means the trial’s displayed data leave meaningful uncertainty around the point estimate. The confidence interval is a statement about statistical precision, not a measure of clinical importance; whether particular effects would matter clinically is not reported in the registry record.

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 and not total events. They are illustrations based on the observed effect, not the trial’s original design assumptions. The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. Each row uses a specified fraction of the observed effect as the effect to detect: 100% of the observed effect, 75% of the observed effect, or 50% of the observed effect. The original design assumptions are not reported in the registry record.

The 80% power column gives the planned participants per arm under those assumptions for a design with an 80% chance of detecting the selected effect if that effect were the true effect and the assumptions held. The 90% power column gives the corresponding per-arm planning figure for a 90% chance. For example, the row planning for 100% of the observed effect lists 1,674 participants per arm for 80% power and 2,240 for 90% power. For 75% of the observed effect, the figures are 2,943 and 3,939; for 50%, they are 6,547 and 8,763.

Smaller effects are harder to distinguish from random variation, so they require more participants under the same assumptions. The curve illustrates that planning relationship across effect sizes; it should be read as a projection, not as a record of trial enrollment or outcomes. These planning figures are not observed power.

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 = 0.728; treatment proportion p₁ = 0.684; average p̄ = 0.706.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.6444 = 1.2630.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.6436 = 0.5417.
4. n = (1.2630 + 0.5417)² / (-0.0441)² = 1,673.2, rounded up to 1,674 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 <- 0.727941176470588                          # control proportion
p1 <- 0.683823529411765                          # 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(76804)
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 = 1,674 (80% power), 2,240 (90% power)
 75% of observed effect: n per arm = 2,943 (80% power), 3,939 (90% power)
 50% of observed effect: n per arm = 6,547 (80% power), 8,763 (90% power)
power.prop.test: 1673.2 
Simulated power at n = 1,674 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

The sponsor’s reported analysis is a chi-squared test with p = 0.42. The unadjusted recalculation from the posted arm-level counts gives p = 0.425, alongside a risk difference of -0.044 and its confidence interval from -0.152 to 0.064. The small numerical difference between the reported p-values is consistent with the sponsor value being displayed to fewer decimal places. The registry record does not flag a discrepancy between the sponsor analysis and the recalculation.

The sponsor’s posted result gives a test and p-value but does not provide a sponsor-reported effect estimate or confidence interval in the analysis entry. The recalculated risk difference and relative measures are simple unadjusted summaries of the posted counts; they are not substitutes for any analysis that may have been performed under a fuller protocol or statistical plan. Such details are not reported in the registry record.

Both p-values indicate that the displayed test did not cross conventional evidence thresholds for a difference, but neither establishes that the groups are identical. The effect estimate and its interval are needed to understand the direction and range of the observed contrast. Because no discrepancy is flagged, the recalculation can be read as consistent with the sponsor’s reported test at its displayed precision.

14. Multiplicity, Missing Data and Analysis Populations

The registry record lists four outcome measures and four analyses, but the posted results include no secondary outcome results. It does not report whether the viral-load result was part of a prespecified hierarchy, whether one outcome was designated as the sole confirmatory primary endpoint in the full protocol, or how multiple outcomes and time points were handled statistically. The record’s study hypothesis refers to mean change in CD4 count at 12 and 24 months, whereas the posted primary outcome analyzed is the proportion below the HIV RNA threshold at 12 months. The registry does not explain how these statements were reconciled in the analysis plan.

Multiplicity is the general statistical issue that testing several outcomes, time points, subgroups, or comparisons creates more opportunities for a chance-positive result if each test is considered separately. Researchers may address this with a prespecified hierarchy or other adjustment, but the registry record does not report whether any approach was used. It is therefore not possible to determine from the posted information whether the reported p-value was adjusted for other analyses or should be interpreted as one of several tests.

The endpoint is labeled intention-to-treat, which generally means participants are analyzed according to randomized assignment. However, the record does not report the number of missing viral-load measurements, whether missing outcomes were imputed, or whether participants with unavailable measurements were classified in a particular way. It also does not describe sensitivity analyses for missing data. As a result, the effect of missing outcome information on the posted estimate cannot be assessed from the registry record.

Details about the exact analysis population, protocol deviations, adherence to assigned support, and exclusions are not reported in the registry record. Nor is a full statistical analysis plan available in the listed details. These omissions limit assessment of potential bias and the reproducibility of the analysis. They do not demonstrate that a particular problem occurred; they mean that the relevant methods cannot be verified from the posted registry information.

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 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: death (14.6%), hospitalization (5.1%), daids grade iv ae (3.6%).

The serious adverse event figures list 19 affected participants among 137 at risk in the Peer Supporter group and 29 among 137 at risk in the Self Administration group. The other adverse event figures list 71 among 137 at risk for Peer Supporter and 89 among 137 at risk for Self Administration. These are counts as recorded in the registry, not estimates of a causal effect of the intervention. The entries do not include confidence intervals or a reported statistical comparison.

The record lists Death, Hospitalization, and DAIDS Grade IV AE among the leading serious-event terms, with displayed percentages of 14.6%, 5.1%, and 3.6%, respectively. The registry record does not clarify the denominator or whether these terms apply to the combined groups or to each group. It lists death counts as null, so a separate number of deaths by arm is not reported in the registry record. The listed serious-event terms should not be treated as interchangeable with the affected-participant counts.

Interpretation is limited because definitions, event timing, severity adjudication, and attribution to the intervention are not reported in the registry record. The figures can be read as descriptive safety reporting only. They do not establish whether the groups differ in safety or whether an event was caused by the assigned strategy.

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

The posted results describe a randomized comparison of peer-supported, directly observed morning dosing with self-administration. For the 12-month endpoint, the recorded number meeting the HIV RNA threshold was 99 of 136 in Peer Supporter and 93 of 136 in Self Administration. The estimated risk difference is -0.044 for Self Administration minus Peer Supporter, and its 95% confidence interval includes zero. The reported chi-squared p-value is 0.42. Together, these results show a numerical contrast with considerable uncertainty, rather than a precise estimate establishing a difference.

They also do not establish equivalence. Showing that a conventional test is not statistically significant is not the same as demonstrating that any difference is small enough to be unimportant. That would require a defined equivalence or non-inferiority margin and an analysis designed for that question; such details are not reported in the registry record. Likewise, the point estimate alone cannot establish that peer support caused better viral-load outcomes.

The endpoint concerns whether a threshold was reached at a single stated follow-up time. It does not directly report adherence behavior, the amount of viral-load change, or mean CD4 change. The study hypothesis mentions CD4 change at 12 and 24 months, but a posted result for that hypothesis is not reported in the registry record. Although the primary outcome timeframe mentions 12 and 24 months, a 24-month viral-load result is not included among the posted data analyzed.

The trial is marked terminated. The reason is not reported in the registry record, and termination alone does not explain the pattern of results. The interpretation should remain confined to the posted 12-month comparison and its uncertainty. It cannot support conclusions about outcomes that were not posted, later follow-up, or specific reasons for the trial’s status.

17. Why This Trial Matters Statistically

This trial illustrates the distinction between a point estimate and the precision of that estimate. The observed direction of the risk difference favors Peer Supporter numerically for the threshold outcome, but the confidence interval extends across zero and allows effects on either side. The p-value alone would conceal that range of uncertainty. Reading the effect estimate, interval, and test together gives a fuller account of what the data do and do not resolve.

It also illustrates why the scale of an effect matters. The risk difference describes an absolute contrast in proportions, while risk ratios and odds ratios describe relative contrasts. Their no-difference values differ, and their numerical values are not directly interchangeable. Reporting multiple measures can aid interpretation, provided their definitions and directions are clear.

The sample-size illustration highlights a general planning principle: designs targeting smaller effects need more participants under otherwise unchanged assumptions. These projections use the observed effect and specified planning assumptions, not the trial’s original design basis. They should not be mistaken for evidence about the trial’s power after completion.

Finally, a registry record is useful for distinguishing an outcome’s stated definition from what was actually posted. Here, the study hypothesis mentions CD4 change at two follow-up times, while the posted result analyzed is a 12-month viral-load threshold. The absence of other posted results, missing-data details, and multiplicity information limits the conclusions that can be drawn. Statistical interpretation depends not only on the number and p-value, but also on the endpoint, analysis population, uncertainty, and reporting completeness.

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 registry record leaves important details about trial conduct and analysis unreported. It describes the study as randomized and unmasked but does not give the allocation-generation method, concealment procedures, or whether outcome assessors were masked. It does not report the reason for termination. These gaps prevent a detailed assessment of potential sources of bias or the reasons the trial ended.

The primary outcome title specifies HIV viral load at 12 months, while the timeframe states 12 and 24 months. The listed result gives the 12-month threshold outcome; a corresponding 24-month result is not reported in the registry record. The study hypothesis also refers to CD4 change, but no posted secondary results are listed. The record does not explain the relation between the hypothesis and the reported primary outcome or how outcomes and time points were prioritized.

The result is presented as an intention-to-treat analysis, but the registry does not report missing viral-load data, the method used to handle missingness, or sensitivity analyses. It also does not provide adjusted analyses, subgroup results, or a full statistical analysis plan. These limitations affect assessment of robustness and do not imply that any specific bias or analytical error occurred.

Safety reporting is also limited: definitions and event attribution are not described, death counts are null, and the denominators behind the displayed serious-event term percentages are not clarified. The posted counts therefore support descriptive reading only. Finally, the planning illustrations are based on the observed effect and listed assumptions, not on the trial’s original design. They cannot be used to infer why the enrollment was chosen or whether the trial met its original design target.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Helping HIV Infected Patients in South Africa Adhere to Drug Regimens — NCT00076804.
• Registry record last updated 2015-03-27; status last verified 2015-03.
• Results first posted 2010-07-02.
• 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

The registry describes a phase 2, randomized, parallel-group trial without masking, sponsored by Johns Hopkins University. It enrolled 274 and is marked terminated. Participants were assigned to a peer supporter observing the morning antiretroviral dose or to self-administration.

The posted primary result is the proportion with HIV RNA below 400 at 12 months: 99 of 136 in Peer Supporter and 93 of 136 in Self Administration. The unadjusted risk difference for Self Administration minus Peer Supporter is -0.044, with a 95% confidence interval from -0.152 to 0.064. The sponsor reports a chi-squared p-value of 0.42.

The registry record does not report the reason for termination, a 24-month result, results for the stated CD4 hypothesis, or detailed missing-data and multiplicity methods. The posted 12-month comparison is uncertain and does not establish either a difference or equivalence.