← Clinical Trials
HIV-1 Infection Phase 3 Non-Inferiority NCT01263015

SINGLE: Complete Statistical Analysis of Dolutegravir in HIV-1 Infection

An independent statistical analysis of the randomized phase 3 SINGLE trial comparing dolutegravir 50 mg plus abacavir/lamivudine once daily with Atripla in participants with human immunodeficiency virus type 1 infection, focusing on virologic response, non-inferiority methodology, and longitudinal CD4+ cell-count analysis.

Trial status: Completed  ·  Enrollment: 844  ·  Primary completion: May 14, 2012
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.

1. Trial at a Glance

SINGLE was a randomized, parallel-group, quadruple-masked phase 3 trial comparing dolutegravir 50 mg plus abacavir/lamivudine once daily with an Atripla regimen in participants with human immunodeficiency virus type 1 infection. The registry reports a Week 48 binary primary endpoint based on plasma HIV-1 RNA below 50 copies/mL and identifies the primary comparison as a non-inferiority or equivalence analysis.

844
Enrollment
Randomized trial
2
Arms
Parallel design
7.3
Week 48 Difference
95% CI 2.3–12.2
0.003
Primary P-value
Two-sided analysis
FeatureSINGLE
PhasePhase 3
ConditionInfection, Human Immunodeficiency Virus I
DesignRandomized, parallel-group, quadruple-masked
Primary purposeTreatment
Enrollment844
Arms2
Primary endpoint typeBinary
Primary endpoint time frameWeek 48
Results postedYes
Statistical analyses posted4
Lead sponsorViiV Healthcare
ClinicalTrials.govNCT01263015

2. Clinical Question

The central statistical question was whether a once-daily regimen containing dolutegravir 50 mg plus abacavir/lamivudine could demonstrate an acceptable virologic response compared with EFV/TDF/FTC, the active regimen contained in Atripla, using a prespecified non-inferiority framework.

Population

Participants with infection caused by human immunodeficiency virus type 1 enrolled in the phase 3 SINGLE trial.

Intervention

Dolutegravir 50 mg plus abacavir/lamivudine 600/300 mg once daily.

Comparator

EFV/TDF/FTC 600/200/300 mg once daily, corresponding to the Atripla comparison regimen.

Primary question

What is the difference between treatment groups in the percentage of participants with plasma HIV-1 RNA below 50 copies/mL at Week 48, and is the lower confidence bound above the prespecified non-inferiority margin?

3. Trial Design

01
Randomize844 participants
02
2 ArmsParallel allocation
03
Quadruple MaskMasked treatment assignment
04
AssessVirologic response
05
FollowLater virologic and CD4 outcomes
ARM A

Dolutegravir + Abacavir/Lamivudine

  • Dolutegravir 50 mg once daily
  • Abacavir/lamivudine 600/300 mg once daily
  • Matching placebo components were part of the intervention structure
ARM B

EFV/TDF/FTC

  • EFV/TDF/FTC 600/200/300 mg once daily
  • Atripla is identified in the trial brief title as the comparator
  • Matching placebo components were part of the intervention structure

The registry classifies allocation as randomized and the design model as parallel. Masking is classified as quadruple. These design features are important statistically because randomization establishes the basis for comparing assigned treatment groups, while masking is intended to reduce the influence of treatment knowledge on trial conduct and assessment.

4. Trial Timeline and Registry Status

February 1, 2011

Trial start

The registry lists February 1, 2011 as the study start date.

May 14, 2012

Primary completion

The registry lists May 14, 2012 as the primary completion date.

Completed

Current registry status

the ClinicalTrials.gov record identifies the study as completed, with results posted on ClinicalTrials.gov.

5. Analysis Population

The primary statistical analysis uses the Intent-to-Treat-Exposed (ITT-E) population, defined in the registry as all randomized participants who received at least one dose of study medication. The same ITT-E definition is given for the reported secondary virologic analyses, while the CD4 analysis is identified as using the ITT-E population.

PopulationRegistry definition / role
Intent-to-Treat-Exposed (ITT-E)All randomized participants who received at least one dose of study medication; used for the primary Week 48 analysis and the reported virologic secondary analyses.
ITT-E for CD4 analysisThe registry identifies the CD4+ cell-count analysis population as ITT-E.
Why this matters: the ITT-E population preserves much of the randomized comparison while restricting the analysis to participants who were exposed to at least one dose of study medication. It is therefore important not to silently substitute a different population when interpreting the reported estimates.

6. Endpoints

EndpointRegistry definition / time frameType
Primary Proportion of Subjects Responding Based on Plasma HIV-1 RNA <50 c/mL at Week 48. The percentage of participants with plasma HIV-1 RNA <50 c/mL at Week 48 was assessed using the Missing, Switch, or Discontinuation equals Failure (MSDF) algorithm, as codified by the FDA Snapshot algorithm. Binary
Secondary virologic endpoint Percentage of Participants With Plasma HIV-1 RNA <50 Copies/Milliliter (c/mL) at Week 96 and Week 144. Binary
Secondary CD4 endpoint Change From Baseline in CD4+ Cell Counts at Week 144. Continuous

The registry reports 11 outcome measures overall, but the ClinicalTrials.gov record identifies the primary endpoint and the specific secondary analyses summarized on this page. The endpoint definitions and time frames above are retained in registry terminology where possible.

7. Statistical Methodology

Primary binary endpoint and nonparametric comparison

The primary analysis compared the two treatment groups using the Wilcoxon (Mann-Whitney) method. The registry normalizes this as a Wilcoxon / Mann-Whitney nonparametric analysis and reports the effect measure as a difference in percentage, equivalently described here as a risk difference for the binary response endpoint.

The estimated difference is explicitly defined in the analysis notes as the percentage responding on dolutegravir plus abacavir/lamivudine minus the percentage responding on EFV/TDF/FTC. Thus, positive values favor the first treatment group in terms of the reported response proportion.

Non-inferiority framework

The registry identifies the hypothesis type as non-inferiority or equivalence. For the primary endpoint, non-inferiority could be concluded if the lower bound of the two-sided 95% confidence interval for the treatment difference was > -10%.

Prespecified non-inferiority criterion
Lower bound of two-sided 95% CI > −10%

The relevant quantity is the difference in percentages: DTG + ABC/3TC minus EFV/TDF/FTC. The registry's criterion therefore asks whether the plausible lower end of the treatment difference remains above a loss of 10 percentage points.

MSDF and the Snapshot algorithm

The primary endpoint uses a Missing, Switch, or Discontinuation equals Failure (MSDF) algorithm, codified by the FDA Snapshot algorithm. The registry states that plasma samples were collected for quantitative assessment of HIV-1 RNA and that participants without HIV-1 RNA data at the visit of interest are treated according to this algorithm.

This is statistically important because a binary virologic-response endpoint is not simply a count of observed laboratory measurements. The handling of missing measurements, treatment changes, and discontinuation is incorporated into the endpoint classification itself.

Longitudinal CD4 analysis

Interpretation boundary: the ClinicalTrials.gov record provides the MMRM method and a P-value of 0.003 for the Week 144 CD4 analysis, but do not provide the adjusted mean changes, a between-group estimate, or a confidence interval. Those quantities are therefore not reported here.

8. Primary Result: Week 48 Virologic Response

The primary endpoint was the proportion of participants with plasma HIV-1 RNA <50 c/mL at Week 48. The analysis used the ITT-E population and compared dolutegravir 50 mg plus abacavir/lamivudine 600/300 mg once daily with EFV/TDF/FTC 600/200/300 mg once daily.

Difference in percentage responding

7.3 percentage points

95% CI: 2.3 to 12.2   ·   P = 0.003

Difference defined as DTG + ABC/3TC minus EFV/TDF/FTC.

Primary treatment difference
Estimated difference
7.3%
95% CI lower
2.3%
95% CI upper
12.2%
Clinical Biostats interpretation

The estimated treatment difference of 7.3 percentage points means that the percentage of ITT-E participants meeting the Week 48 virologic-response definition was estimated to be 7.3 percentage points higher with dolutegravir plus abacavir/lamivudine than with EFV/TDF/FTC, using the registry's specified endpoint and analysis.

It does not mean that every participant had a 7.3% greater probability of response, nor does it describe an individual patient's treatment effect. It is a group-level difference in the binary endpoint under the specified MSDF/Snapshot framework.

The two-sided 95% confidence interval of 2.3 to 12.2 percentage points describes statistical uncertainty around the estimated group difference. Importantly for the non-inferiority question, the entire interval is above the prespecified -10% margin. The registry therefore provides an analysis compatible with the stated non-inferiority criterion.

The P-value of 0.003 is evidence against the null hypothesis associated with the reported comparison; it is not a measure of the size or clinical importance of the 7.3-point difference. The estimate and its confidence interval provide the information about magnitude and precision.

The result should also be understood in the context of the ITT-E population and the MSDF/Snapshot classification rules. Changing the analysis population or missing-data convention could change the numerical result.

Why the non-inferiority margin is central

In a non-inferiority framework, the question is not simply whether the observed treatment difference is positive or whether a P-value crosses a conventional threshold. The design specifies how much worse the new treatment could plausibly be while still satisfying the trial's non-inferiority criterion.

Here, the registry specifies a -10% margin. Because the lower bound of the reported 95% confidence interval is 2.3%, it is above -10%. The statistical logic is therefore based directly on the relationship between the confidence interval and the prespecified margin.

9. Secondary Virologic Results

The registry contains two statistical analyses for the secondary endpoint concerning the percentage of participants with plasma HIV-1 RNA <50 copies/mL at Week 96 and Week 144. Both analyses use the ITT-E population and the Wilcoxon (Mann-Whitney) method.

Week 96

Difference in percentage responding

7.1 percentage points

95% CI: 1.2 to 13.1   ·   P = 0.016

Registry analysis note identifies this estimate as the Week 96 comparison.

Clinical Biostats interpretation

The estimated difference of 7.1 percentage points represents the percentage of participants with HIV-1 RNA <50 c/mL in the dolutegravir plus abacavir/lamivudine group minus the corresponding percentage in the EFV/TDF/FTC group at Week 96.

The 95% confidence interval of 1.2 to 13.1 percentage points indicates uncertainty around that estimated difference. The interval remains above the registry's stated -10% non-inferiority margin.

The P-value of 0.016 is evidence associated with the reported statistical comparison; it does not quantify the size of the treatment effect. The 7.1-point estimate and its confidence interval provide that effect-size and precision information.

As with the primary analysis, interpretation depends on the ITT-E population and the endpoint classification rules. This is a later virologic assessment and should not be treated as though it were a separate primary endpoint.

Week 144

Difference in percentage responding

8.3 percentage points

95% CI: 1.9 to 14.6   ·   P = 0.010

Difference defined as the percentage on DTG + ABC/3TC minus the percentage on EFV/TDF/FTC.

Clinical Biostats interpretation

The estimated 8.3 percentage-point difference indicates a higher estimated proportion meeting the plasma HIV-1 RNA <50 c/mL definition in the dolutegravir plus abacavir/lamivudine group than in the EFV/TDF/FTC group at the Week 144 analysis.

The two-sided 95% confidence interval extends from 1.9 to 14.6 percentage points. This expresses uncertainty around the estimated group difference and does not represent the range of individual treatment effects.

The P-value of 0.010 should be read as evidence from the specified statistical test, not as a percentage probability that the treatment difference is real and not as a measure of effect size.

The confidence interval remains above the registry's -10% non-inferiority margin. Nevertheless, this is a secondary analysis and should be interpreted in the hierarchy and context of the overall trial rather than treated as a replacement for the primary Week 48 analysis.

Virologic endpointTime frameEffect estimate95% CIP-value
Plasma HIV-1 RNA <50 c/mLWeek 487.3 percentage points2.3 to 12.20.003
Plasma HIV-1 RNA <50 c/mLWeek 967.1 percentage points1.2 to 13.10.016
Plasma HIV-1 RNA <50 c/mLWeek 1448.3 percentage points1.9 to 14.60.010

10. Secondary CD4+ Cell-Count Analysis

The registry also reports a secondary endpoint measuring change from baseline in CD4+ cell counts at Week 144. This is a continuous longitudinal outcome rather than a binary virologic-response outcome.

FeatureRegistry-reported analysis
EndpointChange From Baseline in CD4+ Cell Counts at Week 144
Time frameBaseline and Week 144
Outcome unitcells per millimeters cubed (cells/mm3)
PopulationITT-E
Groups comparedDTG 50 mg Plus ABC/3TC 600/300 mg Once Daily vs EFV/TDF/FTC 600/200/300 mg Once Daily
MethodRepeated Measure Mixed Model / MMRM
95% confidence interval95%
P-value0.003

The registry-reported statistical analysis does not provide the corresponding adjusted mean changes or between-group difference. It therefore would be inappropriate to infer an effect estimate from the P-value alone.

Clinical Biostats interpretation

An MMRM is suited to repeated measurements because it models observations across visits rather than treating the Week 144 measurement as an isolated value. The reported model includes treatment and visit as well as baseline covariates and treatment-by-visit interaction terms.

The reported P-value of 0.003 indicates evidence from the specified longitudinal model for a treatment-group difference under that model. It does not tell us the magnitude of the CD4 difference. Without the adjusted mean changes or the estimated between-group contrast, effect size cannot be reconstructed from the ClinicalTrials.gov record.

11. Statistical Methods Explained

Why was the Wilcoxon / Mann-Whitney test used?

The registry reports the Wilcoxon (Mann-Whitney) method for the binary virologic-response analyses. This is a nonparametric method and does not require the outcome to follow a normal continuous distribution. For this page, the important point is to preserve the method actually reported in the registry rather than substitute a different test simply because the endpoint is binary.

The effect measure reported alongside the test is a difference in percentages. The analysis notes explicitly define the direction as the first treatment group's percentage minus the comparator group's percentage.

What does a 7.3 percentage-point difference mean?

A 7.3 percentage-point difference means that the estimated proportion meeting the specified Week 48 response definition differed by 7.3 percentage points between the two randomized treatment groups, with the treatment difference defined as DTG + ABC/3TC minus EFV/TDF/FTC.

It is an absolute difference, not a hazard ratio and not a relative risk. It also does not mean that each individual participant has a 7.3 percentage-point change in their personal probability of response.

Why is non-inferiority judged against the margin?

Non-inferiority trials are designed around a clinically specified tolerance for how much worse the investigational treatment could be while still being considered statistically acceptable under the trial design. In SINGLE, the registry specifies a margin of -10%.

The relevant comparison is therefore between the lower confidence bound and -10%. The reported lower bound for the Week 48 treatment difference is 2.3%, which lies above the margin.

Why does the confidence interval matter more than the P-value for understanding effect size?

The P-value addresses evidence against a statistical null hypothesis under the specified testing framework. It does not say whether an effect is large, small, clinically important, or precisely estimated.

The confidence interval provides both the point estimate and a range reflecting statistical uncertainty. For the primary endpoint, the estimate is 7.3 percentage points and the 95% confidence interval is 2.3 to 12.2 percentage points.

Why use an MMRM for CD4+ cell counts?

CD4+ cell counts are continuous measurements that can be observed repeatedly over time. An MMRM allows treatment and visit to be modeled jointly and can incorporate baseline covariates and treatment-by-visit interactions. The registry specifically reports this type of repeated-measures model for the Week 144 CD4 endpoint.

The model is especially useful conceptually because it asks a longitudinal question rather than treating every visit as an unrelated analysis. However, the ClinicalTrials.gov record does not provide enough numerical detail to reproduce the fitted treatment contrast.

What does the MSDF/Snapshot rule contribute?

The primary endpoint is defined using a rule that classifies participants based on HIV-1 RNA measurements while incorporating missing data, treatment switching, and discontinuation into the response determination. The registry calls this the Missing, Switch, or Discontinuation equals Failure algorithm, codified by the FDA Snapshot algorithm.

This means the observed response proportion is partly a function of the prespecified endpoint algorithm. It should not be interpreted as simply the percentage of laboratory measurements that happened to be below 50 c/mL among participants with an observed value.

12. Non-Inferiority Logic in Detail

The SINGLE registry analysis provides an unusually clear example of why the direction of the effect measure matters in a non-inferiority trial.

Treatment difference
DTG + ABC/3TC response − EFV/TDF/FTC response

A positive estimate indicates a higher response percentage in the dolutegravir plus abacavir/lamivudine group. The registry's non-inferiority criterion is based on whether the lower bound of the two-sided 95% CI is greater than −10%.

ComponentPrimary analysis
EndpointPlasma HIV-1 RNA <50 c/mL at Week 48
PopulationITT-E
Effect measureDifference in percentage
DirectionDTG + ABC/3TC minus EFV/TDF/FTC
Estimate7.3
Two-sided confidence interval95%; 2.3 to 12.2
Non-inferiority margin-10%
Reported P-value0.003

The key statistical relationship is straightforward: the entire reported 95% confidence interval lies above the -10% margin. This is the central piece of evidence for the registry's stated non-inferiority framework.

Do not reverse the subtraction: if the treatment difference were defined in the opposite direction, the sign of the estimate and the interpretation of the non-inferiority margin would also reverse. Correct interpretation therefore begins with the exact effect-measure definition.

13. Confidence Intervals and Precision

The three reported virologic comparisons provide an opportunity to see how confidence intervals complement point estimates.

Time frameEstimate95% CIWidth of CI
Week 487.32.3 to 12.29.9 percentage points
Week 967.11.2 to 13.111.9 percentage points
Week 1448.31.9 to 14.612.7 percentage points

The confidence intervals do not all have the same width. That is expected: statistical precision depends on the information contributing to a particular analysis, including the effective amount of information available for the endpoint. The interval should therefore be considered alongside the estimate rather than treating the point estimate as exact.

For the primary endpoint, the interval from 2.3 to 12.2 percentage points tells a more complete story than the estimate of 7.3 alone. It shows that the reported estimate is compatible with a range of treatment differences under the statistical model and sampling framework.

14. Results by Statistical Method

EndpointTypePopulationMethodEffect measureReported result
HIV-1 RNA <50 c/mL at Week 48 Binary ITT-E Wilcoxon / Mann-Whitney Difference in percentage 7.3; 95% CI 2.3–12.2; P=0.003
HIV-1 RNA <50 c/mL at Week 96 Binary ITT-E Wilcoxon / Mann-Whitney Difference in percentage 7.1; 95% CI 1.2–13.1; P=0.016
HIV-1 RNA <50 c/mL at Week 144 Binary ITT-E Wilcoxon / Mann-Whitney Difference in percentage 8.3; 95% CI 1.9–14.6; P=0.010
Change from baseline in CD4+ cell counts at Week 144 Continuous ITT-E MMRM Adjusted mean framework 95% CI reported; P=0.003; numerical treatment estimate not reported

15. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm. These counts are presented as affected participants over the corresponding at-risk populations.

Safety measureDTG 50 mg + ABC/3TC 600/300 mg once dailyEFV/TDF/FTC 600/200/300 mg once daily
Serious adverse events65/41460/419

These figures should be read as affected participants / participants at risk, rather than as an effect estimate from a formal comparative safety analysis. The ClinicalTrials.gov record does not provide a confidence interval or P-value for the between-group comparison of serious adverse events.

Why safety is separate

Efficacy and safety answer different statistical questions. A virologic-response difference does not summarize the serious-adverse-event experience of the two treatment groups.

Why denominators matter

Reporting 65/414 and 60/419 preserves the affected counts and the corresponding at-risk populations reported by the registry.

16. What the Primary Estimate Does — and Does Not — Mean

Statistical interpretation

The primary estimate of 7.3 percentage points is an absolute difference in the percentages meeting the Week 48 HIV-1 RNA <50 c/mL endpoint, calculated as dolutegravir plus abacavir/lamivudine minus EFV/TDF/FTC.

It does not mean that treatment increased an individual's probability of response by exactly 7.3 percentage points. It is a treatment-group comparison based on the specified ITT-E population and endpoint-classification algorithm.

Why the confidence interval matters

The 95% CI of 2.3 to 12.2 percentage points communicates uncertainty around the estimated difference. It does not describe the distribution of treatment effects among individual participants.

Why the P-value is not an effect size

The primary P=0.003 is evidence from the reported statistical test. It does not indicate that the treatment has a 99.7% probability of being effective, and it does not distinguish between a clinically large and clinically small effect by itself.

Why the margin matters

The non-inferiority criterion is based on the lower confidence bound relative to -10%. The lower bound of 2.3% is above that margin, which is the key numerical relationship in the registry's stated non-inferiority analysis.

17. Missing Data and Endpoint Classification

The primary endpoint's use of the MSDF/Snapshot algorithm is one of the most important methodological details in the ClinicalTrials.gov record.

Observed measurement

A participant has a plasma HIV-1 RNA measurement available for the Week 48 assessment.

Missing measurement

The registry states that participants without HIV-1 RNA data at the visit of interest are handled according to the specified algorithm rather than simply being removed from the endpoint calculation.

Switch or discontinuation

The endpoint algorithm incorporates treatment switching or discontinuation into the failure classification framework.

Statistical consequence

The final binary response proportion reflects both virologic measurements and the prespecified rules governing missing, switched, or discontinued participants.

This distinction is important when comparing results across studies. Two trials can measure the same biological quantity but obtain different response proportions if they use different rules for missing observations, treatment changes, or discontinuation.

18. Randomization and Masking

the ClinicalTrials.gov record identifies the allocation as randomized and the design model as parallel. Masking is classified as quadruple.

Randomization provides the design basis for comparing treatment groups because treatment assignment is not chosen by the participant or investigator. In a parallel-group trial, participants remain associated with their randomized treatment group for the primary comparison framework.

Quadruple masking adds another layer of protection against treatment-knowledge effects. The ClinicalTrials.gov record does not specify the four masked parties, so this page does not assign particular roles to them.

19. Limitations

20. Why This Trial Matters Statistically

SINGLE is a useful teaching case because the registry data bring several important clinical-trial concepts together in one randomized phase 3 study: a binary virologic endpoint, a prespecified non-inferiority margin, a confidence-interval-based decision rule, ITT-E analysis, an explicit missing/switch/discontinuation algorithm, nonparametric testing, and an MMRM for a continuous longitudinal endpoint.

ConceptHow it appears in SINGLE
RandomizationAllocation is randomized.
Parallel designThe trial is classified as a parallel-group design with 2 arms.
BlindingMasking is classified as quadruple.
ITT-E analysisPrimary and reported secondary efficacy analyses use the Intent-to-Treat-Exposed population.
Binary endpointWeek 48 virologic response is defined by plasma HIV-1 RNA <50 c/mL.
Risk differenceThe primary effect measure is a difference in percentage, defined as treatment minus comparator.
Confidence intervalThe primary treatment difference is reported with a two-sided 95% CI.
Non-inferiorityThe lower confidence bound is compared with a prespecified -10% margin.
Wilcoxon / Mann-WhitneyReported for the binary virologic-response analyses.
Missing-data rulesMSDF/Snapshot rules are incorporated into the primary endpoint.
MMRMRepeated-measures modeling is used for change from baseline in CD4+ cell counts at Week 144.

21. A Practical Reading of the Statistical Story

The statistical story can be read in a sequence rather than as a collection of isolated numbers.

01
QuestionCompare virologic response
02
EndpointHIV-1 RNA <50 c/mL
03
PopulationITT-E
04
Estimate7.3 percentage points
05
MarginLower CI above −10%

First, the trial defines a binary virologic-response endpoint at Week 48. Second, the primary analysis uses the ITT-E population and a specified endpoint-classification algorithm. Third, the treatment difference is estimated as 7.3 percentage points. Fourth, uncertainty is summarized by a two-sided 95% confidence interval from 2.3 to 12.2 percentage points. Finally, that lower bound is evaluated against the prespecified -10% non-inferiority margin.

This sequence illustrates why statistical interpretation should begin with the estimand and endpoint definition, continue through the analysis population and method, and only then interpret the numerical estimate, confidence interval, and P-value.

22. Related Statistical Concepts

Learn more about the methods used in this trial:

23. Related Statistical Calculators

24. Sources

Source discipline: numerical trial results on this page are restricted to the ClinicalTrials.gov record. The linked PubMed records are provided as the publications associated with the registry record, but no numerical result has been imported from those publications into this analysis.

Continue through the Clinical Biostats statistical pathway

Explore the statistical concepts behind randomized trials, non-inferiority margins, binary endpoints, confidence intervals, and repeated-measures analysis.

25. Record Summary

SINGLE provides a compact example of how a randomized phase 3 trial can combine a binary virologic endpoint with a non-inferiority framework and a longitudinal continuous endpoint. The primary Week 48 analysis used the ITT-E population, an MSDF/Snapshot endpoint algorithm, and a Wilcoxon / Mann-Whitney analysis, producing a treatment difference of 7.3 percentage points with a two-sided 95% confidence interval of 2.3 to 12.2 and a P-value of 0.003.

The central non-inferiority question is determined by comparing the lower confidence bound with the prespecified -10% margin. The reported lower bound of 2.3% lies above that margin. Later registry-posted virologic analyses report differences of 7.1 percentage points at Week 96 and 8.3 percentage points at Week 144, while the CD4+ cell-count analysis uses an MMRM and reports P=0.003 without supplying a numerical treatment estimate in the ClinicalTrials.gov record.

The most useful statistical lesson is that the treatment estimate cannot be separated from its definition: the endpoint, analysis population, missing-data algorithm, effect-measure direction, confidence interval, and non-inferiority margin all determine how the reported number should be read.

Clinical Biostats methodology: A trial-results page should distinguish the reported statistical evidence from educational interpretation. For SINGLE, that means preserving the registry's endpoint definitions and methods, reporting the available numerical estimates exactly, and avoiding unsupported reconstruction of results that are not contained in the ClinicalTrials.gov record.