← Clinical Trials
HIV Infections Phase 3 Non-Inferiority NCT04021290

SALSA: Complete Statistical Analysis of Dolutegravir/Lamivudine in HIV-1 Infection

An independent statistical analysis of the randomized phase 3 SALSA trial evaluating a regimen switch to dolutegravir/lamivudine fixed-dose combination from a current antiretroviral regimen in HIV-1 infected and virologically suppressed adults.

SALSA  ·  Phase 3  ·  Randomized parallel design  ·  Enrollment 493
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

SALSA was a randomized, open-label, phase 3, parallel-group trial evaluating whether switching virologically suppressed adults with HIV-1 infection to once-daily dolutegravir/lamivudine fixed-dose combination (DTG/3TC FDC) maintained antiviral activity compared with continuing a current antiretroviral regimen (CAR).

493
Enrollment
2 randomized arms
48
Primary time point
Week 48
−0.8
Primary ADP
95% CI −2.4 to 0.8
<5%
NI criterion
Upper 95% CI bound
FeatureSALSA
Trial nameSALSA
NCT identifierNCT04021290
PhasePhase 3
ConditionHIV Infections
AllocationRandomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Enrollment493
InterventionsDTG/3TC FDC; CAR
Lead sponsorViiV Healthcare
Sponsor typeIndustry
StatusCompleted

The trial started on November 11, 2019, and the primary completion date was April 23, 2021. ClinicalTrials.gov reports 27 outcome measures and 6 statistical analyses for the posted results.

2. Clinical Question

The primary statistical question was whether switching from a current antiretroviral regimen to once-daily DTG/3TC FDC maintained non-inferior antiviral activity over 48 weeks. The registered primary endpoint counted participants with plasma HIV-1 RNA ≥50 copies/mL using the FDA Snapshot category at Week 48.

Population

HIV-1 infected and virologically suppressed adults enrolled in the SALSA trial.

Intervention

DTG/3TC fixed-dose combination, identified in the registry as DTG/3TC FDC.

Comparator

Continuation of the current antiretroviral regimen (CAR).

Primary question

Does switching to DTG/3TC FDC have non-inferior antiviral activity compared with continuing CAR over 48 weeks?

3. Trial Design

01
Enroll 493 participants
02
Randomize 2 parallel groups
03
Receive regimen DTG/3TC FDC or CAR
04
Assess Virologic and health-status outcomes
05
Week 48 Primary endpoint
Allocation
Randomized
Design model
Parallel
Masking
None
Primary purpose
Treatment
RANDOMIZED PHASE

DTG/3TC FDC

  • Intervention identified in the registry as DTG/3TC FDC.
  • Compared with continuation of the current antiretroviral regimen.
  • Primary comparison evaluated the FDA Snapshot category at Week 48.
RANDOMIZED PHASE

CAR

  • Comparator identified in the registry as CAR.
  • Participants continued their current antiretroviral regimen.
  • Primary comparison used a stratified analysis adjusting for baseline third-agent class.

The registry also reports a continuation phase. The posted serious-adverse-event data include a continuation-phase group with 1 affected participant among 96 participants at risk.

4. Endpoints

Endpoint roleRegistered endpointTime frameType
Primary Number of Participants With Plasma HIV-1 RNA ≥50 Copies/Milliliter (c/mL) as Per FDA Snapshot Category at Week 48 Week 48 Binary
Secondary Number of Participants With Plasma HIV-1 RNA <50 c/mL Using the FDA Snapshot Algorithm at Week 48 Week 48 Binary
Secondary Change From Baseline in Health Status by HIV Treatment Satisfaction Questionnaire (TSQ) at Week 24 Baseline (Day 1) and Week 24 Continuous
Secondary Change From Baseline in Health Status by HIV TSQ at Week 48 Baseline (Day 1) and Week 48 Continuous
Secondary Change From Baseline in Health Status by Symptom Distress Module (SDM) at Week 24 Baseline (Day 1) and Week 24 Continuous
Secondary Change From Baseline in Health Status by SDM at Week 48 Baseline (Day 1) and Week 48 Continuous

The registry's primary endpoint definition states that participants with plasma HIV-1 RNA ≥50 c/mL were evaluated using the FDA Snapshot algorithm at Week 48 to demonstrate the non-inferior antiviral activity of switching to DTG/3TC FDC once daily compared with continuation of CAR over 48 weeks.

Why the endpoint is binary: the primary outcome classifies participants according to whether the Week 48 FDA Snapshot category meets the specified virologic threshold. The primary treatment effect is therefore expressed as a difference in proportions rather than as a mean change or a time-to-event hazard ratio.

5. Statistical Methodology

Cochran-Mantel-Haenszel analysis

The primary endpoint was analyzed using a Cochran-Mantel-Haenszel (CMH) test. The reported adjusted difference in percent (ADP) was based on a CMH stratified analysis adjusting for the baseline third-agent class: protease inhibitor (PI), non-nucleoside reverse transcriptase inhibitor (NNRTI), and integrase inhibitor (INI).

This approach is appropriate for a binary outcome when the comparison is intended to account for a prespecified categorical stratification factor. Rather than treating all participants as belonging to a single unstratified population, the analysis incorporates the baseline third-agent class into the treatment comparison.

Primary effect measure
ADP = proportion with HIV-1 RNA ≥50 c/mL in DTG/3TC FDC − proportion with HIV-1 RNA ≥50 c/mL in CAR

The registry reports the primary effect measure as an adjusted difference in percent. A negative value therefore corresponds to a lower adjusted proportion meeting the ≥50 c/mL endpoint in the DTG/3TC FDC group.

Non-inferiority framework

The primary analysis was explicitly a non-inferiority comparison. The registry states that non-inferiority would be concluded if the upper bound of the two-sided 95% confidence interval for the CMH-adjusted difference in the proportion of participants with plasma HIV-1 RNA ≥50 c/mL between DTG/3TC FDC and CAR was less than 5%.

This is a particularly important statistical distinction. The question is not simply whether a conventional superiority test produces a small p-value. Instead, the confidence interval is compared with a prespecified non-inferiority margin.

MMRM for longitudinal health-status outcomes

The secondary health-status outcomes were analyzed with mixed model repeated measures (MMRM). The registry reports that the model adjusted for treatment, visit, baseline third-agent class, age as a continuous variable, sex, race, baseline value as a continuous variable, treatment-by-visit interaction, and baseline-value-by-visit interaction, with visit as the repeated factor.

MMRM allows measurements collected at multiple visits to be analyzed jointly rather than treating each visit as an unrelated analysis. The model can estimate adjusted treatment differences at specified visits while accounting for the longitudinal structure of the observations.

Conceptual longitudinal model
Outcome = treatment + visit + baseline value + covariates + treatment × visit + baseline value × visit + repeated-measures structure

The expression is conceptual rather than a reconstruction of the complete covariance specification, which is not provided in the ClinicalTrials.gov record.

Analysis populations

The primary analysis population was the Intent To Treat-Exposed (ITT-E) population, defined as all randomized participants who received at least one dose of study medication, either DTG/3TC or CAR.

The secondary TSQ and SDM analyses were also conducted in the ITT-E population, with the registry specifying that only participants with data available at the specified time points were analyzed for those outcomes.

6. Primary Endpoint Result

The primary endpoint was the number of participants with plasma HIV-1 RNA ≥50 copies/mL according to the FDA Snapshot category at Week 48. The posted analysis compared DTG/3TC FDC with CAR using a CMH-adjusted difference in percent.

Adjusted difference in percent

−0.8%

95% CI: −2.4% to 0.8%   ·   Two-sided 95% CI

Analysis: Cochran-Mantel-Haenszel test

Population: ITT-E

Direction of the adjusted primary effect
DTG/3TC FDC − CAR
−0.8%

The visual represents the direction and reported estimate; it is not a reconstruction of the underlying participant-level proportions.

Clinical Biostats interpretation

The reported adjusted difference was −0.8 percentage points, comparing DTG/3TC FDC with CAR for the proportion of participants classified with plasma HIV-1 RNA ≥50 c/mL at Week 48. Because the difference is DTG/3TC FDC minus CAR, the negative estimate points toward a lower adjusted proportion meeting the endpoint in the DTG/3TC FDC group.

The estimate does not mean that the probability of virologic failure was exactly 0.8% lower for every individual participant. It is a population-level adjusted treatment comparison produced by the CMH analysis.

The two-sided 95% confidence interval extends from −2.4% to 0.8%. This interval describes the statistical uncertainty around the adjusted treatment difference under the analysis framework. It is not a range containing the individual effects experienced by patients.

For a non-inferiority question, the critical comparison is with the prespecified margin. The registry states that non-inferiority would be concluded when the upper bound of the two-sided 95% CI is below 5%. Here, the upper bound is 0.8%, which is below the stated 5% margin.

The non-inferiority conclusion therefore comes from the relationship between the confidence interval and the margin, rather than from a p-value. The posted primary analysis does not provide a p-value for this endpoint in the ClinicalTrials.gov record.

Understanding the non-inferiority logic

QuantityReported valueRole in interpretation
Adjusted difference in percent−0.8Estimated DTG/3TC FDC minus CAR difference
95% CI lower bound−2.4Lower end of statistical uncertainty interval
95% CI upper bound0.8Upper end used for the non-inferiority decision
Non-inferiority margin5%Upper-bound criterion specified by the registry

In this setting, the null concern is that switching could result in an unacceptably larger proportion of participants meeting the ≥50 c/mL endpoint. The non-inferiority margin defines how large that adverse difference could be while still being considered compatible with the prespecified non-inferiority objective.

Non-inferiority is not the same as equivalence. A non-inferiority design asks whether the data exclude a treatment disadvantage larger than the prespecified margin. It does not establish that the two regimens have exactly identical effects.

7. Secondary Endpoint Results

HIV-1 RNA <50 c/mL at Week 48

The secondary virologic endpoint counted participants with plasma HIV-1 RNA <50 c/mL using the FDA Snapshot algorithm at Week 48. The posted analysis used the ITT-E population and a CMH-adjusted difference in percent.

Adjusted difference in percent

1.6%

95% CI: −2.8% to 5.9%   ·   Two-sided 95% CI

Analysis: Cochran-Mantel-Haenszel test

Non-inferiority margin: −12%

Clinical Biostats interpretation

The estimated adjusted difference was 1.6 percentage points for the proportion with HIV-1 RNA <50 c/mL, comparing DTG/3TC FDC with CAR. The positive estimate points toward a higher adjusted success proportion in the DTG/3TC FDC group, although the confidence interval spans both negative and positive values.

The 95% confidence interval ranges from −2.8% to 5.9%. That interval represents uncertainty around the adjusted difference; it does not mean that individual participants have a probability somewhere between those values.

The registry specifies a non-inferiority criterion based on the lower bound of a two-sided 95% confidence interval being greater than −12%. The reported lower bound is −2.8%, which is above −12%.

No p-value is reported for this analysis in the ClinicalTrials.gov record. More importantly, the non-inferiority interpretation is driven by the prespecified margin and confidence interval rather than by treating a conventional superiority p-value as the principal decision rule.

HIV TSQ: Week 24

Adjusted mean difference

1.4

95% CI: 0.7 to 2.2   ·   P <0.001

MMRM analysis, ITT-E population

Clinical Biostats interpretation

The reported adjusted mean difference was 1.4 score units for change from baseline in HIV Treatment Satisfaction Questionnaire status at Week 24. The confidence interval extends from 0.7 to 2.2, indicating the range of values supported by the reported model and sampling framework at the 95% confidence level.

The p-value of <0.001 is evidence against a null hypothesis of no adjusted mean difference under the model used. It does not measure the size or clinical importance of the difference. The magnitude is represented by the estimate and its confidence interval.

The result is adjusted for treatment, visit, baseline third-agent class, age, sex, race, baseline value, treatment-by-visit interaction, and baseline-value-by-visit interaction, with visit as the repeated factor.

HIV TSQ: Week 48

Adjusted mean difference

1.4

95% CI: 0.7 to 2.2   ·   P <0.001

MMRM analysis, ITT-E population

Clinical Biostats interpretation

At Week 48, the reported adjusted mean difference was again 1.4 score units, with a 95% confidence interval from 0.7 to 2.2 and a p-value of <0.001.

The statistical interpretation is similar to the Week 24 analysis: the estimate describes an adjusted group difference, while the confidence interval describes precision and the p-value addresses evidence against a null difference. None of these quantities alone establishes how important a one-point change is to an individual participant.

Because the analysis is longitudinal and adjusted for multiple baseline and visit-related variables, the result should be understood as a model-based adjusted comparison rather than a simple difference between two unadjusted sample means.

Symptom Distress Module: Week 24

Adjusted mean difference

−1.6

95% CI: −2.9 to −0.2   ·   P = 0.021

MMRM analysis, ITT-E population

Clinical Biostats interpretation

The Week 24 SDM analysis estimated an adjusted mean difference of −1.6 score units, with a 95% confidence interval from −2.9 to −0.2.

The negative sign identifies the direction of the adjusted difference as DTG/3TC FDC minus CAR. The ClinicalTrials.gov record does not define the clinical meaning of a one-unit change on the SDM, so the statistical direction should not be converted into a clinical-benefit claim beyond what the score itself establishes.

The p-value of 0.021 indicates evidence against a null adjusted mean difference under the specified model. It is not a measure of effect magnitude and does not by itself establish clinical importance.

Symptom Distress Module: Week 48

Adjusted mean difference

−0.5

95% CI: −1.8 to 0.7   ·   P = 0.398

MMRM analysis, ITT-E population

Clinical Biostats interpretation

The Week 48 SDM analysis estimated an adjusted mean difference of −0.5 score units. The 95% confidence interval ranges from −1.8 to 0.7 and therefore includes zero.

The p-value of 0.398 does not provide evidence against a null adjusted mean difference at conventional significance thresholds. This should not be interpreted as proof that the two groups are identical: a nonsignificant p-value can reflect uncertainty, and the confidence interval describes the range of adjusted differences supported by the data.

8. Secondary Results at a Glance

EndpointEstimate95% CIP-valueMethod
HIV-1 RNA <50 c/mL at Week 48 1.6% −2.8% to 5.9% Not reported CMH
HIV TSQ change at Week 24 1.4 0.7 to 2.2 <0.001 MMRM
HIV TSQ change at Week 48 1.4 0.7 to 2.2 <0.001 MMRM
SDM change at Week 24 −1.6 −2.9 to −0.2 0.021 MMRM
SDM change at Week 48 −0.5 −1.8 to 0.7 0.398 MMRM

The ClinicalTrials.gov record contains six posted statistical analyses: one primary endpoint analysis and five secondary endpoint analyses. The table above summarizes all six.

9. Safety Results

The registry provides serious adverse-event counts by study phase and arm. These data are reported as affected participants divided by participants at risk.

Phase / groupSerious adverse events
Randomized phase — first reported group 7/246
Randomized phase — second reported group 16/247
Continuation phase 1/96

The ClinicalTrials.gov record identifies the randomized-phase denominators as 246 and 247, but the serious-adverse-event field does not associate the two randomized-phase entries with DTG/3TC FDC versus CAR in a way that can be safely reconstructed from the registry-reported text. Accordingly, the counts are presented in the registry's reported order rather than assigning them to treatment labels.

Interpretation caution: serious adverse events are a safety outcome and should not be combined with the efficacy estimates into a single statistical measure. The serious-adverse-event counts also use phase-specific populations at risk, so their denominators should not automatically be treated as the original enrollment total.

10. Statistical Methods Explained

Why was a Cochran-Mantel-Haenszel test used?

The primary endpoint was binary: participants were classified according to the FDA Snapshot category at Week 48. The CMH approach provides a way to compare the treatment groups while adjusting for the baseline third-agent class used for stratification.

In this trial, the baseline third-agent class consisted of PI, NNRTI, and INI categories. The adjusted difference therefore incorporates the prespecified stratification factor rather than relying solely on an unadjusted comparison of the two randomized groups.

Why is the confidence interval central to the non-inferiority analysis?

Non-inferiority asks whether the data exclude an unacceptable loss of efficacy. A point estimate alone cannot answer that question because the estimate is uncertain. The confidence interval supplies an upper bound on the adverse treatment difference that can be compared directly with the non-inferiority margin.

For the primary endpoint, the reported upper bound was 0.8%, while the registry specifies a 5% non-inferiority margin. The relevant comparison is therefore between 0.8% and 5%, not between a p-value and an arbitrary significance threshold.

What does an adjusted difference of −0.8% mean?

The estimate is the reported CMH-adjusted difference in percent between DTG/3TC FDC and CAR for the proportion of participants with plasma HIV-1 RNA ≥50 c/mL at Week 48. Because the comparison is DTG/3TC FDC minus CAR, a negative number indicates a lower adjusted proportion in the DTG/3TC FDC group.

It is not a relative risk, odds ratio, hazard ratio, or individual-level probability. It is an adjusted absolute difference between group proportions.

Why was MMRM used for the TSQ and SDM outcomes?

The TSQ and SDM endpoints involve measurements at multiple time points, including baseline and later visits. MMRM is designed for longitudinal data and can estimate adjusted treatment differences while accounting for the repeated nature of measurements within participants.

The SALSA registry analysis adjusted for treatment, visit, baseline third-agent class, age, sex, race, baseline value, treatment-by-visit interaction, and baseline-value-by-visit interaction. This makes the reported adjusted mean different from a simple arithmetic difference in observed means.

What does a p-value of <0.001 mean for the TSQ result?

Under the statistical model and null hypothesis used for that analysis, a p-value <0.001 indicates strong evidence against a null adjusted mean difference. It does not mean that there is a <0.1% probability that the treatment effect is real, nor does it measure clinical importance.

The estimate of 1.4 and its 95% confidence interval of 0.7 to 2.2 provide the effect-size and precision information. The p-value answers a different question.

Why can a nonsignificant p-value still leave uncertainty?

The Week 48 SDM analysis had an estimate of −0.5, a 95% CI of −1.8 to 0.7, and a p-value of 0.398. Because the confidence interval includes zero, the analysis does not distinguish clearly between a negative, near-zero, or modest positive adjusted difference within that interval.

A nonsignificant result therefore should not be translated into a claim that the treatments have exactly the same effect. It indicates that the analysis did not provide sufficient statistical evidence against the specified null hypothesis.

Why is the analysis population called ITT-E?

The registry defines ITT-E as all randomized participants who received at least one dose of study medication, either DTG/3TC or CAR. This preserves the randomized treatment comparison among participants who were actually exposed to study medication, while distinguishing the analysis population from a broader set that might include randomized participants who never received treatment.

11. Covariate Adjustment and Stratification

The primary analysis adjusted for the baseline third-agent class. The registry identifies three categories: protease inhibitor (PI), non-nucleoside reverse transcriptase inhibitor (NNRTI), and integrase inhibitor (INI).

Statistical elementSALSA implementation
Primary endpointBinary FDA Snapshot category at Week 48
Primary methodCochran-Mantel-Haenszel test
Stratification variableBaseline third-agent class
Third-agent categoriesPI, NNRTI, INI
Primary effect measureAdjusted Difference in Percent

Stratification is useful when a baseline categorical variable is considered relevant to the treatment comparison. Instead of ignoring that variable, the CMH framework combines information across its strata to produce an adjusted comparison.

Conceptual CMH structure
Adjusted treatment comparison = information combined across prespecified baseline strata

The registry's analysis text specifically identifies baseline third-agent class as the adjustment factor. The ClinicalTrials.gov record does not provide the stratum-specific participant counts or cell counts, so the page does not reconstruct the underlying CMH calculation.

12. Intention-to-Treat Analysis in SALSA

The ITT-E population is central to interpreting the posted efficacy results. Participants were randomized before treatment exposure, and the registry defines ITT-E as all randomized participants who received at least one dose of DTG/3TC or CAR.

Why randomization matters

Randomization creates the basis for a treatment-group comparison in which baseline differences are not intentionally selected by investigators.

Why exposure matters here

The registry's ITT-E definition requires at least one dose of study medication, distinguishing the analysis population from all randomized participants regardless of exposure.

Why this is not per-protocol

The ClinicalTrials.gov record explicitly identify the primary analysis population as ITT-E rather than a per-protocol population.

Interpretation

The primary result should therefore be understood as the treatment comparison specified for the ITT-E population, not as an analysis restricted to participants who completed every assessment.

13. Missing Data and Repeated Measures

The registry explicitly states that the TSQ and SDM secondary analyses were conducted in the ITT-E population and that only participants with data available at the specified time points were analyzed.

For these outcomes, MMRM was used with visit as the repeated factor. This is different from simply analyzing only the final observed score for every randomized participant. A repeated-measures model uses the longitudinal data structure and the covariates specified in the analysis.

What the ClinicalTrials.gov record does not establish: the ClinicalTrials.gov record does not specify a particular missing-data imputation algorithm, such as last observation carried forward, multiple imputation, or a specific missing-not-at-random sensitivity analysis. The page therefore does not attribute one to SALSA.

14. Multiplicity and Multiple Endpoints

The posted registry results include one primary endpoint analysis and five secondary endpoint analyses. The ClinicalTrials.gov record identifies the hypothesis type for the primary endpoint as non-inferiority and identify the secondary health-status analyses as "Other / not stated."

Endpoint familyNumber of posted analysesReported inferential framework
Primary virologic endpoint1Non-inferiority
Secondary virologic endpoint1Non-inferiority
HIV TSQ2Other / not stated
SDM2Other / not stated

The ClinicalTrials.gov record does not provide a multiplicity-adjustment strategy across the secondary endpoints. Consequently, the reported secondary p-values should be interpreted as the p-values associated with their respective posted analyses, without assuming an unreported familywise error-control procedure.

Why this distinction matters: multiple endpoint analyses can increase the chance of observing a statistically notable result by chance if each is interpreted as an independent confirmatory test. Without a documented multiplicity strategy in the ClinicalTrials.gov record, the secondary results should be kept distinct from the primary non-inferiority conclusion.

15. Non-Inferiority Trial Design

SALSA provides a useful example of why non-inferiority trials require a different statistical reading from conventional superiority trials.

FeaturePrimary SALSA analysis
EndpointParticipants with plasma HIV-1 RNA ≥50 c/mL at Week 48
Effect measureCMH adjusted difference in percent
ComparisonDTG/3TC FDC − CAR
Confidence intervalTwo-sided 95%
Non-inferiority margin5%
CriterionUpper confidence-limit <5%
Observed upper confidence limit0.8%

For the primary endpoint, the reported upper confidence limit of 0.8% is below the prespecified 5% margin. Statistically, this means the uncertainty interval excludes an adjusted increase as large as the stated non-inferiority margin.

This framework does not require the point estimate to be exactly zero. A non-inferiority analysis can produce an estimate favoring either treatment while still meeting its predefined criterion, provided the confidence interval remains on the acceptable side of the margin.

16. Crossover, Factorial Design, and Interim Analysis

The ClinicalTrials.gov record does not report a crossover design or a factorial design. The registered design model is parallel, with two arms.

The registry-reported statistical-analyses data also do not identify an interim analysis or an alpha-spending procedure. Therefore, this page does not infer one from the phase 3 designation or from the presence of a non-inferiority hypothesis.

Design topicWhat the ClinicalTrials.gov record supports
Parallel designYes — registered design model is parallel
RandomizationYes
MaskingNone
Factorial designNot reported in the ClinicalTrials.gov record
CrossoverNot reported in the ClinicalTrials.gov record
Interim analysisNot reported in the ClinicalTrials.gov record
Bayesian analysisNot reported in the ClinicalTrials.gov record

17. What the Primary Effect Measure Does — and Does Not — Mean

Effect estimate

The primary adjusted difference of −0.8% is an absolute difference in the percentage of participants meeting the plasma HIV-1 RNA ≥50 c/mL endpoint, adjusted using the CMH framework for baseline third-agent class.

Confidence interval

The 95% confidence interval of −2.4% to 0.8% describes uncertainty around the adjusted treatment difference. It does not give a range of outcomes for individual participants.

Non-inferiority interpretation

The registry's criterion focuses on the upper bound of the two-sided 95% confidence interval. Because the upper bound is 0.8%, the reported interval remains below the 5% non-inferiority margin.

What it is not

The primary estimate is not a hazard ratio, odds ratio, relative risk, or percentage reduction in an individual participant's risk. It is an adjusted difference in percentages.

18. Interpreting the MMRM Estimates

The secondary health-status outcomes illustrate a different statistical language from the primary binary endpoint. The MMRM estimates are adjusted mean differences rather than risk differences.

OutcomeTime pointEstimate95% CIP-value
HIV TSQWeek 241.40.7 to 2.2<0.001
HIV TSQWeek 481.40.7 to 2.2<0.001
SDMWeek 24−1.6−2.9 to −0.20.021
SDMWeek 48−0.5−1.8 to 0.70.398

The signs of the estimates should be interpreted using the direction of the underlying questionnaire score. The ClinicalTrials.gov record identifies the instruments and statistical estimates but do not provide a clinical threshold for what constitutes a meaningful change on either scale. Statistical significance therefore should not be substituted for a clinical-importance threshold that is not provided here.

19. Limitations

20. Why This Trial Matters Statistically

SALSA is a useful teaching case because its primary endpoint and secondary outcomes require several different statistical ideas within the same randomized trial. The primary analysis is a stratified non-inferiority comparison of a binary virologic endpoint, while the health-status outcomes use longitudinal mixed models with extensive covariate adjustment.

ConceptHow it appears in SALSA
RandomizationRandomized phase 3 parallel-group design
Binary endpointPlasma HIV-1 RNA ≥50 c/mL at Week 48
Non-inferiorityPrimary endpoint evaluated against a 5% margin
Cochran-Mantel-Haenszel testPrimary stratified analysis
Risk differenceNormalized form of the reported adjusted difference in percent
Confidence intervalTwo-sided 95% interval used directly in the non-inferiority criterion
StratificationBaseline third-agent class: PI, NNRTI, INI
ITT-E analysisPrimary analysis population
MMRMTSQ and SDM longitudinal health-status outcomes
Covariate adjustmentAge, sex, race, baseline value and other prespecified factors in MMRM
Repeated measuresVisit used as the repeated factor in MMRM
MultiplicityMultiple secondary analyses are posted; adjustment strategy not reported

21. Related Tutorials

Learn more about the methods used in this trial:

22. Related Statistical Calculators

23. Sources

Source boundary: Numerical trial results and design facts on this page are restricted to the registry-reported SALSA trial data. The PubMed links are provided as source navigation; no additional numerical results have been introduced from those publications.

Continue through the Clinical Biostats statistical pathway

Use the trial's methods as a starting point for deeper study of stratified categorical analysis, non-inferiority designs, confidence intervals, risk differences, and longitudinal mixed models.

24. Record Summary

SALSA provides a compact example of several important clinical-trial statistical principles. The primary endpoint was binary and evaluated at Week 48 using a CMH-adjusted difference in percent, with baseline third-agent class as the stratification factor. The trial specified a non-inferiority margin of 5%, and the reported primary estimate was −0.8% with a two-sided 95% confidence interval of −2.4% to 0.8%.

The secondary analyses demonstrate a different class of methods. Virologic suppression below 50 c/mL was analyzed using the same general CMH framework, while TSQ and SDM outcomes were evaluated using MMRM with treatment, visit, baseline third-agent class, age, sex, race, baseline value, treatment-by-visit interaction, and baseline-value-by-visit interaction included in the reported model structure.

The most important statistical lesson is that the different endpoints require different interpretations. The primary non-inferiority result is governed by the relationship between its confidence interval and the prespecified margin. The MMRM outcomes are adjusted mean differences, where confidence intervals and p-values describe statistical uncertainty but do not automatically establish clinical importance. Safety results should be interpreted separately from efficacy, and secondary p-values should not be treated as though an unreported multiplicity procedure had been performed.

Clinical Biostats methodology: A trial-results page should distinguish the reported statistical evidence from the interpretation of that evidence. For SALSA, that means keeping the non-inferiority margin, adjusted risk difference, confidence interval, analysis population, stratification factor, MMRM covariates, and safety denominators explicit rather than collapsing them into a single summary statistic.