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
NADIA was a randomized phase 3 treatment trial with a factorial design. The registry identifies human immunodeficiency virus as the condition and lists a primary endpoint of plasma viral load below 400 copies/ml at 48 weeks.
| Feature | NADIA |
|---|---|
| Acronym | NADIA |
| Full title | Nucleosides And Darunavir/Dolutegravir In Africa |
| Phase | Phase 3 |
| Status | UNKNOWN |
| Condition | Human Immunodeficiency Virus |
| Allocation | Randomized |
| Design model | Factorial |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 465 |
| Number of arms | 4 |
| Lead sponsor | Makerere University |
| Sponsor type | OTHER |
| Start date | 2019-07-30 |
| Primary completion date | 2020-09-30 |
| ClinicalTrials.gov | NCT03988452 |
2. Clinical Question
The registry identifies NADIA as a phase 3 treatment trial in human immunodeficiency virus using a randomized factorial design. The intervention set includes darunavir, ritonavir, dolutegravir, zidovudine, tenofovir, and lamivudine.
Population
The registry lists human immunodeficiency virus as the condition under study. A more detailed eligibility description is not reported in the record summarized here.
Intervention set
The listed interventions are darunavir, ritonavir, dolutegravir, zidovudine, tenofovir, and lamivudine.
Comparator structure
Because the design model is factorial with four arms, the statistical comparison is structured around combinations of treatment factors rather than a single two-arm treatment-versus-control contrast.
Primary question
The registry's primary efficacy question is represented by plasma viral load below 400 copies/ml at 48 weeks.
3. Trial Design
A factorial design evaluates combinations of treatment factors within a single randomized experiment. Its statistical advantage is efficiency: when the factors are appropriately specified, one participant population can contribute information to more than one treatment comparison. The interpretation depends on the exact factorial structure and the assumptions used to separate main effects from possible interactions.
Four randomized arms
- The registry specifies four arms.
- The overall allocation is randomized.
- The design model is factorial.
- The record identifies no masking.
Listed treatment components
- Darunavir
- Ritonavir
- Dolutegravir
- Zidovudine
- Tenofovir
- Lamivudine
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Primary endpoint | Plasma viral load < 400 copies/ml at 48 weeks | 48 weeks |
The registry identifies one primary endpoint. It is a binary virologic outcome at a specified follow-up time: whether plasma viral load is below 400 copies/ml at 48 weeks.
This converts the endpoint into a binary response for statistical comparison. The most appropriate model depends on the factorial structure and the analysis population defined in the statistical analysis plan.
5. Statistical Methodology
Binary endpoint analysis
The primary endpoint is binary because each participant either satisfies the criterion of plasma viral load below 400 copies/ml at 48 weeks or does not. A natural analysis therefore compares proportions between randomized treatment groups or estimates treatment effects through a binary-response regression model.
For a simple two-group comparison, the response proportion can be written as:
where p is the probability that the participant has plasma viral load below 400 copies/ml at 48 weeks.
Possible measures of treatment effect include a risk difference, risk ratio, or odds ratio. In a factorial trial, however, the primary model should reflect the factorial allocation rather than reducing the design to arbitrary pairwise contrasts.
Factorial main effects
The defining statistical feature of NADIA is its factorial design. Suppose two treatment factors are represented by \(A\) and \(B\). A model for a binary endpoint can include both main effects and their interaction:
Here, βA describes the association with factor A, βB describes factor B, and βAB represents effect modification between the factors. The actual coding and model specification should follow the prespecified statistical analysis plan.
Interaction
An interaction asks whether the effect of one treatment factor depends on the level of another factor. This is different from asking whether each factor has a main effect. If the interaction is important, a single overall main-effect estimate may obscure clinically meaningful differences among the factorial combinations.
Risk difference
For a binary endpoint, the risk difference is the difference between two response probabilities:
A positive risk difference means that the probability of meeting the endpoint is higher in the first group under the chosen coding. Unlike an odds ratio, it is expressed directly on the probability scale.
Risk ratio
The risk ratio compares the probabilities of meeting the endpoint:
A risk ratio of 1 would indicate equal probabilities under the specified comparison. A value above 1 indicates a higher probability in the numerator group, while a value below 1 indicates a lower probability.
Confidence intervals
A treatment-effect estimate should be interpreted together with its confidence interval. The interval communicates statistical uncertainty around the estimated effect under the specified model and sampling framework. It does not describe the range of responses that individual participants will experience.
6. Statistical Methods Explained
Why is the factorial design statistically important?
A factorial design allows more than one treatment factor to be investigated within the same randomized experiment. This can be more efficient than running separate trials, but the interpretation depends on whether the factors can reasonably be analysed through main effects and whether an interaction is present.
What is a main effect?
A main effect describes the average association of one randomized treatment factor with the endpoint across levels of the other factor. In a factorial trial, it is an average effect over the other factor rather than necessarily the effect in every individual treatment combination.
What is an interaction?
An interaction means that the effect of one factor changes according to the level of another factor. For example, if the effect associated with factor A differs depending on whether factor B is present, a model containing only the two main effects may be inadequate.
Why is plasma viral load below 400 copies/ml a binary endpoint?
The registry defines the endpoint as whether plasma viral load is below a fixed threshold at a fixed follow-up time. That makes the endpoint suitable for analysis as a binary outcome rather than as a continuous viral-load measurement.
What does a risk difference tell us?
The risk difference expresses the absolute difference in the probability of meeting the endpoint between two specified groups. For example, a risk difference of 0.10 would correspond to a 10-percentage-point difference, although no such observed estimate is reported for NADIA in the registry.
Why should pairwise comparisons be interpreted cautiously?
Four arms can generate several possible pairwise comparisons. Testing many comparisons without a prespecified strategy can increase the probability of a false-positive finding. Factorial analysis can reduce this problem by focusing on the treatment factors and interactions that were defined in advance.
7. Planned Analysis
The registry identifies plasma viral load < 400 copies/ml at 48 weeks as the primary endpoint. For an endpoint of this type, a typical analysis would compare the proportion of participants meeting the virologic threshold between the randomized treatment groups using a method appropriate to the factorial structure.
Primary outcome scale
The endpoint is binary: below 400 copies/ml versus not below 400 copies/ml at 48 weeks.
Factorial analysis
The analysis would ordinarily reflect the factorial allocation, allowing prespecified main effects and, where appropriate, assessment of interaction.
Effect measures
Risk differences, risk ratios, or odds ratios can describe binary treatment effects. The prespecified analysis determines which measure is primary.
Uncertainty
Confidence intervals would quantify uncertainty around the treatment-effect estimates.
8. Analysis Population and Missing Data
The registry information summarized here does not specify an analysis population, a detailed missing-data strategy, or an imputation method for the 48-week endpoint.
That distinction matters for a binary endpoint measured at a fixed time point. Participants without an evaluable week-48 result may require a prespecified classification or missing-data procedure. Different approaches can change the estimated response proportion, particularly when the amount or mechanism of missingness differs between randomized groups.
| Statistical issue | Why it matters for NADIA |
|---|---|
| Randomized analysis | Randomization provides the basis for causal comparison of treatment strategies. |
| Week-48 availability | The primary endpoint is defined at a specific follow-up time, so missing week-48 measurements require prespecified handling. |
| Analysis population | The registry information summarized here does not specify the exact population used for the primary endpoint analysis. |
| Imputation | No imputation method is specified in the registry information summarized here. |
9. Multiplicity and Factorial Interpretation
Multiplicity is especially relevant to factorial studies because the design can produce several scientifically interesting comparisons. A four-arm trial can be described through individual arm comparisons, main effects of treatment factors, and interaction effects. These are not interchangeable hypotheses.
A prespecified analysis should identify which comparisons are confirmatory and how type I error is controlled when more than one hypothesis is tested. The registry information summarized here does not report an alpha-allocation or multiplicity-adjustment strategy.
Main-effect question
Does one randomized treatment factor affect the probability of achieving plasma viral load below 400 copies/ml at 48 weeks, averaged across the other factor?
Interaction question
Does the effect of one treatment factor differ according to the level of the other factor?
10. Non-Inferiority, Crossover, and Bayesian Methods
The registry information summarized here does not report a non-inferiority margin, crossover strategy, or Bayesian statistical methodology. These design features therefore cannot be used to characterize the statistical analysis of NADIA from the registry record.
| Design topic | Registry information |
|---|---|
| Non-inferiority margin | Not reported in the registry information summarized here |
| Crossover | Not reported in the registry information summarized here |
| Bayesian methods | Not reported in the registry information summarized here |
| Interim analysis | Not reported in the registry information summarized here |
| Multiplicity strategy | Not reported in the registry information summarized here |
11. Trial Timeline
Trial start
The registry lists 2019-07-30 as the study start date.
Primary completion
The registry lists 2020-09-30 as the primary completion date.
Primary endpoint time frame
The primary endpoint is plasma viral load < 400 copies/ml at 48 weeks.
12. Statistical Interpretation of the Primary Endpoint
Plasma viral load < 400 copies/ml at 48 weeks is a fixed-time binary endpoint. It asks whether each participant has reached the specified viral-load threshold at the specified time point.
A risk difference would quantify the absolute difference in the probability of meeting the week-48 endpoint. A risk ratio would describe the relative probability. An odds ratio would compare the corresponding odds. Each measure describes a different aspect of the same binary outcome.
A confidence interval would show the statistical precision of the estimated treatment effect. A narrow interval would indicate greater precision under the specified analysis framework, while a wider interval would indicate greater uncertainty.
If a p-value were reported, it would describe the evidence against a prespecified null hypothesis under the chosen statistical model. It would not tell us the magnitude of the treatment effect. Effect size and uncertainty should therefore be reported separately from hypothesis-testing results.
13. Factorial Design: A Deeper Statistical View
The four-arm structure is the most important design feature for understanding the statistical problem. A factorial trial can be represented conceptually as a table of treatment-factor combinations.
| Factor B: Level 1 | Factor B: Level 2 | |
|---|---|---|
| Factor A: Level 1 | Arm 1 | Arm 2 |
| Factor A: Level 2 | Arm 3 | Arm 4 |
This representation is conceptual rather than a reconstruction of the specific NADIA arm assignments. The registry establishes four arms and a factorial design, but the information summarized here does not provide the exact factor-to-arm coding needed to populate this table with the named treatment combinations.
Main effects versus individual arms
Suppose the response probabilities in the four factorial cells are \(p_{11}, p_{12}, p_{21}, p_{22}\). A main effect compares averages across one factor while averaging over the other factor. An interaction asks whether the contrast for factor A is different at the two levels of factor B.
A value near zero is consistent with similar absolute treatment differences across the two levels of the second factor. A nonzero value indicates that the treatment contrast changes across those levels. Formal inference requires an appropriate model and confidence interval or hypothesis test.
This is why a factorial analysis should be planned around the scientific questions rather than simply listing every possible pairwise comparison. The structure can be statistically efficient, but only when the estimands are clearly defined.
14. Safety
The registry information summarized here does not report serious adverse events by arm or other arm-specific safety results. No safety estimate is therefore presented as an observed result.
| Safety component | Registry information |
|---|---|
| Serious adverse events by arm | Not reported in the registry information summarized here |
| Arm-specific safety estimates | Not reported in the registry information summarized here |
| Comparative safety analysis | Not reported in the registry information summarized here |
For a randomized treatment trial, safety is generally summarized separately from the primary efficacy endpoint. Common descriptive measures include the number and proportion of participants experiencing adverse events, serious adverse events, treatment discontinuations, and clinically important event categories. The exact safety analysis should follow the prespecified safety population and definitions.
15. Limitations
- No posted statistical analyses: the registry does not report a formal statistical analysis for the primary endpoint in the record summarized here.
- No observed efficacy estimate: the registry information summarized here does not provide a treatment-effect estimate, confidence interval, or p-value for plasma viral load < 400 copies/ml at 48 weeks.
- Factorial specification: the registry identifies a factorial design but the information summarized here does not specify the exact factor coding needed to reconstruct the four treatment combinations statistically.
- Analysis population: the registry information summarized here does not identify the exact primary analysis population.
- Missing data: no specific missing-data or imputation strategy is reported in the registry information summarized here.
- Multiplicity: no multiplicity-adjustment strategy or alpha allocation is reported in the registry information summarized here.
- Additional design features: non-inferiority margins, crossover, Bayesian methods, and interim-analysis procedures are not reported in the registry information summarized here.
- Safety: arm-specific serious adverse-event results are not reported in the registry information summarized here.
16. Why This Trial Matters Statistically
NADIA is a useful teaching case because its central statistical challenge is not simply comparing two treatment groups. The randomized phase 3 study uses a four-arm factorial design and evaluates a binary virologic endpoint at a fixed follow-up time.
| Concept | How it appears in NADIA |
|---|---|
| Randomization | The allocation is randomized across four arms. |
| Factorial design | The registry identifies the design model as factorial. |
| Binary endpoint | Plasma viral load < 400 copies/ml at 48 weeks can be represented as a binary response. |
| Fixed-time assessment | The primary endpoint is evaluated at 48 weeks. |
| Main effects | Factorial analysis can estimate the effects of treatment factors across levels of the other factor when the model is appropriately specified. |
| Interaction | The factorial structure permits assessment of whether one treatment effect depends on another treatment factor. |
| Effect measures | Risk differences, risk ratios, and odds ratios are possible measures for a binary outcome. |
| Confidence intervals | Uncertainty around treatment-effect estimates should be presented alongside the estimates. |
| Multiplicity | Multiple possible factorial hypotheses make prespecification and error control important. |
| Missing data | A fixed-time endpoint requires a prespecified approach for participants without an evaluable week-48 result. |
17. Planned Analysis in Practical Terms
A statistically coherent analysis of NADIA would begin by defining the estimand for the week-48 virologic response. The analysis would then respect the factorial randomization, specify the treatment-factor contrasts of interest, and distinguish main effects from interaction effects.
For the binary endpoint, the analysis could report the response proportion in each relevant randomized group together with an absolute and/or relative treatment effect and its confidence interval. A regression model could incorporate the factorial factors directly. If interaction is included, the interpretation of the main effects changes because the effect of one factor is then allowed to vary according to the other factor.
Missing week-48 assessments would need to be addressed according to a prespecified rule. This is particularly important because a binary endpoint at a single time point does not naturally retain partial information from an earlier or later measurement in the same way as a longitudinal model.
18. Statistical Concepts in This Trial
The statistical structure of NADIA makes factorial design the central methodological concept. Understanding that design helps clarify how multiple treatment factors, binary endpoints, interaction terms, effect measures, and multiplicity fit together in a randomized clinical trial.
19. Related Tutorials
Learn more about the methods used in this trial:
20. Related Calculators
21. Sources
- ClinicalTrials.gov: NCT03988452 — NADIA.
- PubMed: PubMed record.
- PubMed: PubMed record.
Continue through the Clinical Biostats knowledge graph
Explore statistical tutorials and calculators that develop the design and analysis concepts used in randomized clinical trials.
22. Record Summary
NADIA is a randomized phase 3 trial with 465 participants and four arms using a factorial design. Its registered primary endpoint is plasma viral load < 400 copies/ml at 48 weeks. Statistically, the central issue is how to analyse that binary fixed-time endpoint while preserving the factorial structure of the randomization. Main effects, possible interactions, appropriate effect measures, confidence intervals, missing-data handling, and multiplicity all influence the interpretation of such a design.