Evaluating the Safety and Efficacy of Long-Acting Injectable Cabotegravir Compared to Daily Oral TDF/FTC for Pre-Exposure Prophylaxis in HIV-Uninfected Women: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Evaluating the Safety and Efficacy of Long-Acting Injectable Cabotegravir Compared to Daily Oral TDF/FTC for Pre-Exposure Prophylaxis in HIV-Uninfected Women (NCT03164564), a randomized phase 3 study. The primary outcome is Number of Pariicipants With Documented Incident HIV Infections.

2. Clinical Question

The study compares two pre-exposure prophylaxis approaches in HIV-uninfected women: long-acting injectable cabotegravir and daily oral TDF/FTC. The main outcome is a count of participants with a documented incident HIV infection, rather than a measure of viral load, adherence, or a time-to-event summary. The registry lists infections in both randomized arms and gives an unadjusted comparison of the outcome counts.

The comparison is relevant to the trial’s stated question about safety and efficacy, but the posted primary outcome addressed is specifically HIV infection. Safety outcomes are recorded separately. The figures describe the enrolled trial groups; they do not by themselves explain why any individual participant did or did not have an infection. Details about the participants’ characteristics beyond being HIV-uninfected women are not reported in the registry record.

3. Trial Design

This is a phase 3, randomized, parallel-group study with quadruple masking. The registry lists 3224 participants as enrolled. Arm A comprised 1614 participants and Arm B comprised 1610 participants. The registry status is active, not recruiting. The listed study start date is 2017-11-07, and the primary completion date is 2020-11-05.

The groups compared different active prophylaxis regimens with placebos intended to maintain masking: Arm A received CAB and CAB LA with placebo TDF/FTC, while Arm B received TDF/FTC with placebo CAB and placebo CAB LA. The arm descriptions state that participants began with a five-week oral period, followed by injection visits and continued oral medication or placebo. The available descriptions are incomplete, so further regimen details are not reported in the registry record.

Primary outcome testing was scheduled at enrollment, weeks 2, 4, and 5, then every 4 weeks through week 25, and every 8 weeks thereafter. The outcome was analyzed through week 185 or the date of the DSMB decision to unblind all participants, whichever was earliest. The registry does not report the allocation sequence method, the detailed masking procedures, the trial’s original sample-size justification, or whether any interim analyses were planned. Randomization and masking describe the study design, but do not substitute for those missing analysis details.

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:

Analysis population for the primary outcome, as defined in the registry: All enrolled participants

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: Overall enrolled participants

6. Primary Endpoint

The registry lists the primary outcome as Number of Pariicipants With Documented Incident HIV Infections, measured over HIV tests at enrollment, weeks 2, 4, and 5, then every 4 weeks through week 25, then every 8 weeks. Analyzed through week 185 or the date of DSMB decision to unblind all participants, whichever is earliest..

The primary endpoint is the number of participants with a documented incident HIV infection. An infection had to be confirmed by the HPTN Laboratory Center and Endpoint Adjudication Committee. The registry identifies the outcome population as all enrolled participants and gives the unit as participants, so the posted result is a count of people who met the endpoint rather than a count of all positive tests or a measure of infection duration.

The testing schedule was at enrollment, weeks 2, 4, and 5, then every 4 weeks through week 25 and every 8 weeks after that. The analysis period extended through week 185 or the date of the DSMB decision to unblind all participants, whichever was earliest. Scheduled testing and confirmation help define how incident infections were identified and adjudicated. The registry record does not describe additional endpoint definitions, such as how an infection’s date was assigned or how missing tests were handled.

A count endpoint is useful for reporting how many participants in each group experienced the defined event during the specified follow-up. Interpreting the counts in relation to group size requires the arm denominators, which are also posted. The outcome does not capture every possible aspect of prevention or safety, and it does not by itself describe the timing of infections within follow-up.

7. Results Posted in the Registry

• Arm B: TDF/FTC + Placebo CAB + Placebo CAB LA: 2.2% (36/1610) versus 0.2% (4/1614) on control; risk difference 2.0 percentage points (95% CI 1.2 to 2.8), risk ratio 9.02 (3.22–25.29); unadjusted two-proportion z-test p = <0.001.

8. Statistical Interpretation

9. Statistical Methods Explained

What kind of outcome is this?

The primary outcome is binary at the participant level: each participant either had a documented incident HIV infection during the specified analysis period or did not meet that endpoint. The registry reports the number of participants with an event in each arm. The arm denominators are essential context for those counts because the groups can differ in size. Here, the posted analysis compares 1614 participants in Arm A with 1610 in Arm B.

For a binary outcome, an event risk is the proportion of participants in a group who experience the event. The registry’s comparison expresses the difference between the groups using a risk difference and also gives relative measures, the risk ratio and odds ratio. These measures answer related but distinct questions. The risk difference describes an absolute contrast in event risk; the risk ratio compares the risks as a ratio; the odds ratio compares event odds. General statistical note: when an event is uncommon, odds and risks can be close, but they are not the same quantity. No new risk values need to be calculated to interpret the measures already posted.

How should the risk difference be read?

The posted risk-difference estimate is 0.020 for the comparison of Arm B with Arm A. The confidence interval is 0.012 to 0.028. A risk difference is signed: its direction depends on the group order used in the comparison. The registry identifies Arm B as the comparison arm and Arm A as the control, so the estimate is a contrast in that stated order. The estimate is an aggregate comparison, not an individual-level prediction.

General statistical note: an absolute difference is often useful because it describes the size of a contrast on the original risk scale. A relative measure such as a risk ratio can make the proportional contrast clearer, but it does not replace the absolute scale. Neither measure, by itself, indicates the reasons for an outcome or captures outcomes that were not measured.

What do the confidence interval and p-value contribute?

A confidence interval presents an estimate together with uncertainty under the statistical procedure used. General statistical explanation: a confidence interval procedure is designed to produce intervals that cover the target parameter at its stated confidence level over repeated applications under its assumptions. It is not a probability statement that the particular interval contains the true value. The registry posts the interval limits for the risk difference and for the ratio measures, but the confidence level and construction method are not reported in the registry record.

The posted unadjusted p-value is <0.001. General statistical explanation: a p-value measures how compatible the observed data, or data more extreme under a specified test statistic, are with a null model and its assumptions. It is not a measure of the size or practical importance of a difference, and it is not the probability that the null model is correct. The precise test used for the posted p-value is not reported in the registry record.

Why distinguish the posted analysis from other analyses?

The registry lists zero analyses and no secondary results. It supplies an unadjusted comparison for the primary count outcome, but does not report details of model specifications, adjustment variables, analysis populations other than all enrolled participants, handling of missing data, or sensitivity analyses. Those details can affect how uncertainty is calculated and how a result generalizes to the intended population. Their absence means readers should not assume a particular model or missing-data procedure.

The study’s design is randomized and parallel, with quadruple masking. Those design features describe how participants were assigned and masked, but do not disclose the statistical procedures used to produce each posted estimate. The registered outcome and arm counts are interpretable as observed group results; design details do not fill in unreported analysis methods.

Why can count data need special care?

General statistical note: counts of events over a period can be analyzed in different ways depending on whether the target is the number of participants with any event, the total number of events, or the time until an event. These outcomes are not interchangeable. This trial’s primary endpoint is the number of participants with an incident infection, so the reported measure is a participant count rather than a recurrent-event total or a time-to-event estimate. The registry’s stated testing schedule and analysis horizon define the period in which the primary endpoint was assessed.

General statistical note: over-dispersion matters when count models assume variability that is smaller than the variability observed in the data; failing to account for it can make standard errors too small and intervals too narrow. That issue is not an identified feature of the posted primary analysis because the endpoint is a participant-level binary outcome and no count-model specification is reported. The registry record does not report a statistical analysis plan, so no additional assumptions about the model, interim monitoring, or adjustment strategy can be inferred.

10. Confidence Intervals and Precision

The principal precision information for the posted risk difference is its estimate, 0.020, and confidence interval from 0.012 to 0.028. The interval is wholly on one side of zero, indicating that the interval procedure’s compatible values for the specified contrast do not include no difference. The reported unadjusted p-value is <0.001. These statements concern the registry’s particular comparison and do not establish that every reasonable analysis would produce the same estimate.

The interval has a lower limit of 0.012 and an upper limit of 0.028. The limits convey uncertainty in the estimated absolute contrast; they should not be read as the range of outcomes for individual participants. General statistical explanation: narrower intervals indicate greater precision for the parameter under the chosen method, while wider intervals indicate more uncertainty. Precision depends on factors such as the number of observations, event frequency, and analysis method. The record does not report the confidence level or the interval construction method.

For the risk ratio, the posted estimate is 9.022 and the interval limits are 3.219 and 25.29. For the odds ratio, the estimate is 9.206 and the limits are 3.269 and 25.92. Both intervals span a broad range of relative contrasts, so the exact magnitude on those scales is less precisely determined than a single point estimate might suggest. The relative intervals and absolute risk-difference interval describe different quantities and should be interpreted on their own scales.

Confidence intervals address sampling uncertainty under the relevant assumptions; they do not account automatically for unreported sources of uncertainty such as missing outcome data, deviations from the planned analysis, or multiplicity. Those methods and details are not reported in the registry record. The point estimates, intervals, and event counts therefore provide useful but bounded statistical information about the posted comparison.

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 events. They are illustrations computed from the observed effect, rather than a report of the trial’s original design calculation. The listed assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. Each scenario uses the stated fraction of the observed effect as the effect to detect. The original design assumptions and sample-size justification are not reported in the registry record.

For planning for 100% of the observed effect, the figures are 486 participants per arm for 80% power and 650 participants per arm for 90% power. For planning for 75% of the observed effect, they are 694 participants per arm for 80% power and 928 participants per arm for 90% power. For planning for 50% of the observed effect, they are 1,174 participants per arm for 80% power and 1,571 participants per arm for 90% power.

In general, the 80% power column indicates the planned number per arm for a design with an 80% chance of detecting the specified effect under the stated assumptions; the 90% column uses a higher target chance of detection. These are design targets, not guarantees. The examples show that smaller effects need more participants to maintain the same power target. This is not observed power: it is a planning illustration based on the observed effect, not a measure of the trial’s chance of success after results were observed.

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 = 0.002; treatment proportion p₁ = 0.022; average p̄ = 0.012.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.1566 = 0.3070.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.1560 = 0.1313.
4. n = (0.3070 + 0.1313)² / (0.0199)² = 485.9, rounded up to 486 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 <- 0.00247831474597274                          # control proportion
p1 <- 0.022360248447205                          # 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(3164564)
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 = 486 (80% power), 650 (90% power)
 75% of observed effect: n per arm = 694 (80% power), 928 (90% power)
 50% of observed effect: n per arm = 1,174 (80% power), 1,571 (90% power)
power.prop.test: 485.9 
Simulated power at n = 486 per arm: 0.854 (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. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome for the posted results: the number of participants with documented incident HIV infections. It lists two outcome measures overall, zero analyses, and zero posted secondary results. The record does not explain how the primary endpoint was selected relative to other outcome measures, nor does it describe a hierarchy for testing multiple outcomes.

The reported p-value is unadjusted. No multiplicity procedure is reported in the registry record. When several outcomes, time points, subgroups, or analytic choices are tested, the chance of at least one apparently unusual result can increase; multiplicity methods are sometimes used to control that issue. This is general statistical context, not a claim that a particular set of unreported tests was conducted in this trial. Without a reported analysis plan or multiplicity approach, the p-value should be interpreted as the posted unadjusted result for this comparison.

The registry identifies all enrolled participants as the primary outcome population. It does not report whether the analysis followed an intention-to-treat definition, whether participants were excluded after randomization, or how incomplete follow-up was handled. Missing outcome data and reasons for missingness are not reported in the registry record. General statistical note: missingness can affect an estimate if the probability of missing data is related to the outcome or differs across groups; the consequences depend on the pattern and handling of the missing information.

Other analysis details are also unavailable, including covariate adjustment, sensitivity analyses, subgroup methods, and the precise statistical test used. Interim-analysis plans and any stopping boundaries are not reported in the registry record. The DSMB decision to unblind all participants is identified as one possible endpoint for the follow-up window, but this does not specify whether interim efficacy analyses occurred or how decisions were made. These omissions limit how fully the inferential procedure can be reconstructed from the registry results.

14. 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.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: any event in soc (0.7%), malaria (0.3%), peptic ulcer (0.1%), pelvic inflammatory disease (0.1%), tonsillitis bacterial (0.1%).

The registry lists 33 participants with a serious adverse event in Arm A among 1614 at risk, and 33 in Arm B among 1610 at risk. It lists 1486 participants with other adverse events in each arm, among the respective at-risk groups of 1614 and 1610. Deaths are listed as 3 in Arm A and 0 in Arm B. These figures are descriptive counts; the registry record does not provide a causal assessment for individual events.

The most frequently listed serious-event term is “Any Event in SOC” at 0.7%. Other listed terms are malaria at 0.3%, peptic ulcer at 0.1%, pelvic inflammatory disease at 0.1%, tonsillitis bacterial at 0.1%, and blood creatine phosphokinase increased at 0.1%. These are the posted percentages for the terms; the record does not clarify in this summary whether individuals could contribute to more than one term.

Serious adverse events and other adverse events are separate categories in the record and should not be combined. The registry does not report event-level narratives, duration, severity grades beyond the serious classification, relatedness assessments, or comparative statistical testing for these safety figures. Counts alone do not establish that an event was caused by an assigned regimen.

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

The posted primary outcome shows 4 participants with documented incident HIV infections in Arm A and 36 in Arm B, with group sizes of 1614 and 1610, respectively. The registry’s risk-difference estimate is 0.020, with interval limits 0.012 and 0.028, and its unadjusted p-value is <0.001. This is a statistical description of the observed outcome contrast over the specified follow-up period.

The results do not, by themselves, explain the biological or behavioral reasons for the observed counts. They do not indicate how any particular participant would fare under either regimen, and they do not report individual adherence or exposure patterns. The aggregate figures also do not describe outcomes outside the defined primary endpoint. No medical conclusion or treatment recommendation follows from the counts alone.

The registry comparison is based on a randomized, parallel-group study with quadruple masking, features that are relevant to how a trial comparison is designed. Yet the registry does not report all details needed to independently reconstruct the analysis, including missing-data methods, the precise test, or any adjustment for multiplicity. Such limitations do not erase the posted results, but they define what can be concluded from the record.

The confidence interval communicates uncertainty around the estimated contrast under the method used. It does not incorporate every possible source of uncertainty, and the confidence level and construction method are not reported. The ratio estimates are presented on different scales and have their own interval limits. Taken together, the counts and estimates describe the comparison as posted; interpretation should remain tied to this endpoint, this analysis period, and the methods that are actually documented.

16. Why This Trial Matters Statistically

This trial illustrates how a randomized comparison of a binary clinical outcome can be reported using both event counts and effect estimates. Counts make the observed outcomes concrete: 4 participants in Arm A and 36 in Arm B met the primary endpoint. The denominators show the size of each randomized group, while the risk difference expresses the contrast on an absolute scale. Risk and odds ratios provide additional relative descriptions, each with a separate interpretation.

The distinction between absolute and relative measures is important for clinical research communication. A relative estimate alone does not state the absolute difference, and an absolute estimate alone does not convey the proportional comparison. Presenting both can help readers understand different aspects of the same observed outcome without treating one measure as a substitute for another. Confidence intervals add information about precision that a point estimate or p-value cannot provide alone.

The registry also offers a useful example of the boundary between design information and analysis information. Randomization, parallel groups, and quadruple masking describe the trial structure. They do not specify how missing data, multiple outcomes, or covariates were handled. The absence of posted analyses and secondary results makes transparent reporting especially important: readers need the analysis plan and implementation details to assess the full inferential context.

Finally, the outcome is participant-level incidence rather than recurrent-event counts or time to infection. Keeping the endpoint definition in view prevents a count of affected participants from being confused with a total event count or an estimate of when events occurred. This is a practical lesson in aligning the statistical measure with the clinical question and the registry’s stated outcome.

17. 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 reports the primary outcome counts and an unadjusted comparison, but not the full statistical analysis plan. The confidence level and construction method for the confidence intervals, the test underlying the p-value, and any covariate adjustment are not reported. The record lists zero analyses, so there are no posted analysis details to clarify these choices.

Missing data, loss to follow-up, exclusions after randomization, adherence, and deviations from assigned regimens are not described in the results presented. Although the primary outcome population is all enrolled participants, the registry does not specify how participants without complete testing were treated analytically. The results therefore cannot be used to assess the effect of a particular missing-data assumption or analysis population beyond the stated population label.

The registry reports two outcome measures but no secondary results, and does not report a multiplicity plan. Interim-analysis methods, stopping boundaries, and the original sample-size justification are also not reported. The endpoint period is specified as through week 185 or the DSMB decision to unblind all participants, whichever came first, but further details on the decision process are not reported.

The arm descriptions in the registry are truncated, limiting the available detail on treatment sequence beyond the stated oral lead-in and injection schedule fragments. Participant characteristics beyond the eligibility description are not reported in this record. Safety results are counts and selected serious-event terms; event narratives, relatedness assessments, and detailed severity information are not included in the summary. These limitations restrict independent reconstruction and broader interpretation of the posted comparison.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Evaluating the Safety and Efficacy of Long-Acting Injectable Cabotegravir Compared to Daily Oral TDF/FTC for Pre-Exposure Prophylaxis in HIV-Uninfected Women — NCT03164564.
• Registry record last updated 2025-10-09; status last verified 2025-09.
• Results first posted 2023-03-28.
• 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.

21. Record Summary

NCT03164564 is a phase 3 randomized, parallel-group, quadruple-masked study comparing long-acting injectable cabotegravir with daily oral TDF/FTC for pre-exposure prophylaxis in HIV-uninfected women. The registry lists 3224 enrolled participants, with 1614 in Arm A and 1610 in Arm B. Its status is active, not recruiting.

The primary outcome is the number of participants with a confirmed incident HIV infection, assessed through week 185 or the DSMB decision to unblind all participants, whichever was earliest. The record lists 4 events in Arm A and 36 in Arm B. The posted risk-difference estimate is 0.020, with a confidence interval from 0.012 to 0.028, and an unadjusted p-value of <0.001.

The record includes safety counts and planning illustrations based on the observed effect. It does not report the original sample-size justification, detailed statistical analysis plan, missing-data methods, or multiplicity procedures.