This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Reported results are kept separate from statistical interpretation, which is labeled as such throughout.
1. Trial at a Glance
SOLO-CONTINUE was a randomized, parallel-group, triple-masked phase 3 trial. Adults with atopic dermatitis who had been treated in two parent studies (NCT02277743 and NCT02277769) were randomized to continue dupilumab 300 mg at one of three dosing frequencies or to switch to placebo. The trial measured how much of the improvement achieved in the parent studies was retained after 36 further weeks.
| Feature | SOLO-CONTINUE |
|---|---|
| Official title | A Study to Confirm the Efficacy and Safety of Different Dupilumab Dose Regimens in Adults With Atopic Dermatitis (AD) |
| Phase | Phase 3 |
| Condition | Atopic dermatitis |
| Design | Randomized, parallel-group, triple-masked, treatment purpose |
| Arms | 4: placebo weekly (QW); dupilumab 300 mg every 8 weeks (Q8W); every 4 weeks (Q4W); every 2 weeks or weekly (Q2W/QW) |
| Enrollment | 422 |
| Parent studies | NCT02277743 and NCT02277769 |
| Co-primary endpoints | Percent change in EASI (from parent-study baseline) between current-study baseline and Week 36; proportion maintaining EASI-75 at Week 36 |
| Multiplicity control | 2-sided hierarchical testing at the 0.05 level for co-primary and key secondary endpoints |
| Dates | Start 2015-03-25; primary completion 2016-07 |
| Sponsor | Regeneron Pharmaceuticals (industry) |
2. Clinical Question
Most atopic dermatitis trials ask whether a treatment induces improvement. SOLO-CONTINUE asks a different question: once patients have improved on dupilumab, how much of that improvement is maintained if dosing continues at the same or a reduced frequency, compared with stopping active treatment? This makes it a trial about durability and dose-interval reduction rather than initial efficacy.
Population
Adults with atopic dermatitis who entered from the parent studies NCT02277743 and NCT02277769, in which they had received a dupilumab regimen.
Intervention
Dupilumab 300 mg continued every 2 weeks or weekly, or given less often every 4 weeks or every 8 weeks.
Comparator
Placebo given weekly (withdrawal of active treatment).
Primary question
Compared with placebo, does each dupilumab regimen reduce loss of EASI improvement and increase the proportion who maintain EASI-75 at Week 36?
3. Trial Design
| Arm | Regimen | Participants at risk (safety) |
|---|---|---|
| Placebo QW | Placebo weekly | 82 |
| Dupilumab 300 mg Q8W | Dupilumab every 8 weeks | 84 |
| Dupilumab 300 mg Q4W | Dupilumab every 4 weeks | 87 |
| Dupilumab 300 mg Q2W/QW | Dupilumab every 2 weeks or weekly | 167 |
The Q2W/QW arm has roughly twice as many participants at risk as each other arm, consistent with unequal allocation favoring continuation of the most frequent regimen. The arm label combines two dosing frequencies; the stratification factor "dupilumab regimen received in parent studies" suggests that participants in this arm continued the regimen they had previously received. Because the placebo arm is labeled "QW" and the active arms have dosing intervals up to 8 weeks, weekly injections in every arm, with placebo filling the gaps between active doses, would be the natural way to keep the triple masking intact; the registry does not describe the injection schedule in more detail.
4. Analysis Populations, Stratification and Missing Data
| Endpoint | Analysis population (registry) | Handling of rescue and missing data |
|---|---|---|
| EASI percent change difference | Full analysis set: all randomized participants | Values after first rescue treatment set to missing, then multiple imputation (MI) |
| EASI-75 maintained at Week 36 | FAS participants with EASI-75 at current-study baseline | Values after first rescue set to missing; missing at Week 36 = non-responder |
| IGA response within 1 point of baseline | FAS participants with IGA 0 or 1 at baseline (from IVRS) | Analyzed by stratified CMH |
| IGA 0 or 1 at Week 36 | FAS participants with IGA 0 or 1 at baseline (from IVRS) | Analyzed by stratified CMH |
| Pruritus NRS increase ≥3 points | FAS participants with NRS ≤7 at baseline | Analyzed by stratified CMH |
The full analysis set includes all randomized participants, which follows the intention-to-treat principle for the continuous co-primary endpoint. The binary endpoints are restricted to participants who met the relevant response threshold at current-study baseline. This restriction is logical, since one can only "maintain" a response that was present, but it means each binary endpoint rests on a different, smaller subset of randomized participants.
5. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Co-primary: EASI percent change difference | Difference between current-study baseline and Week 36 in percent change in EASI from parent-study baseline (NCT02277743 and NCT02277769). EASI measures erythema, infiltration, excoriation and lichenification over head, trunk, upper and lower extremities; total 0 to 72, higher = worse. Values after first rescue set to missing before multiple imputation. | Baseline (parent study), baseline (current study) and Week 36 (current study) |
| Co-primary: EASI-75 maintained | Percentage of participants with EASI-75 at current-study baseline maintaining EASI-75 (≥75% overall improvement in EASI) at Week 36. Values after first rescue set to missing; missing at Week 36 = non-responder. | Week 36 |
| Secondary: IGA within 1 point | Percentage of participants maintaining Investigator Global Assessment response within 1 point of baseline at Week 36 | Baseline, Week 36 |
| Secondary: IGA 0 or 1 | Percentage of participants maintaining IGA response at 0 or 1 at Week 36 | Week 36 |
| Secondary: pruritus worsening | Percentage of participants with weekly average of peak daily pruritus NRS increased by 3 or more points from baseline to Week 35 | Baseline up to Week 35 |
Understanding the continuous co-primary endpoint
The EASI endpoint is unusual and worth reading carefully. Improvement is anchored to the parent-study baseline, before any dupilumab. At entry to SOLO-CONTINUE, each participant already had some percent improvement from that original baseline. The endpoint measures how that percent improvement changed between entry and Week 36. A value near zero means improvement was held steady; a positive value means improvement was lost. A negative between-group difference therefore means less loss of improvement with dupilumab than with placebo.
The registry labels each comparison as "Placebo QW vs Dupilumab", but the signs across endpoints (negative for EASI loss and pruritus worsening, positive for maintained responses) are consistent with differences computed as dupilumab minus placebo, and they are interpreted that way here.
6. Results: EASI Percent Change Difference (Co-Primary)
All three comparisons used ANCOVA on the full analysis set after multiple imputation and were tested within the prespecified 2-sided hierarchy at the 0.05 level.
| Comparison | Difference (percentage points of EASI % change) | 95% CI (two-sided) | P-value |
|---|---|---|---|
| Dupilumab 300 mg Q8W vs placebo | −14.83 | −22.34 to −7.33 | 0.0001 |
| Dupilumab 300 mg Q4W vs placebo | −17.83 | −25.33 to −10.34 | <0.0001 |
| Dupilumab 300 mg Q2W/QW vs placebo | −21.61 | −28.36 to −14.87 | <0.0001 |
Dupilumab 300 mg Q2W/QW vs placebo
95% CI: −28.36 to −14.87 · P < 0.0001
Q4W: −17.83 (−25.33 to −10.34) · Q8W: −14.83 (−22.34 to −7.33)
What the estimate means. For the Q2W/QW regimen, the estimated change in percent EASI improvement between entry and Week 36 was 21.61 percentage points more favorable than with placebo. Put simply, participants withdrawn to placebo lost more of their earlier improvement, and continuing dupilumab reduced that loss by an estimated 21.61 points on the percent-improvement scale. The corresponding figures for Q4W and Q8W were 17.83 and 14.83 points.
What it does not mean. The estimate is not a 21.61% reduction in EASI score, nor a 21.61-point change on the 0 to 72 EASI scale. It is a difference in percent change, a ratio-based quantity anchored to each participant's parent-study baseline. It also does not show that each dupilumab participant retained their improvement fully; it compares group averages.
Precision. All three intervals exclude zero and are of broadly similar width. The ordering of point estimates follows dosing frequency, but the intervals overlap substantially: the Q8W interval (−22.34 to −7.33) and the Q2W/QW interval (−28.36 to −14.87) share a wide common range. No formal between-regimen comparison was posted.
P-value vs effect size. A p-value of 0.0001 for Q8W and <0.0001 for the other regimens indicates strong incompatibility with no difference, under the model. It does not indicate how clinically large the difference is; that information comes from the estimate and its interval.
Cautions. The result depends on the rescue strategy and on multiple imputation, which assumes that missing values can be predicted from observed data. If participants who needed rescue would have done worse than predicted, the estimates could be affected. The ANCOVA covariates are not listed in the registry.
7. Results: EASI-75 Maintained at Week 36 (Co-Primary)
These comparisons used Cochran-Mantel-Haenszel (CMH) methods stratified by region, baseline disease severity and parent-study dupilumab regimen, among participants with EASI-75 at current-study baseline. Each test was performed only if the preceding endpoint in the hierarchy was significant.
| Comparison | Difference in % maintaining EASI-75 | 95% CI (two-sided) | P-value |
|---|---|---|---|
| Dupilumab 300 mg Q8W vs placebo | 24.5 | 9.70 to 39.29 | 0.0040 |
| Dupilumab 300 mg Q4W vs placebo | 28 | 13.32 to 42.58 | 0.0004 |
| Dupilumab 300 mg Q2W/QW vs placebo | 41.2 | 28.93 to 53.52 | <0.0001 |
Dupilumab 300 mg Q2W/QW vs placebo
95% CI: 28.93 to 53.52 · P < 0.0001
Q4W: 28.0 (13.32 to 42.58) · Q8W: 24.5 (9.70 to 39.29)
What the estimate means. Among participants who entered with EASI-75, the estimated proportion still meeting EASI-75 at Week 36 was 41.2 percentage points higher with Q2W/QW dupilumab than with placebo, 28.0 points higher with Q4W, and 24.5 points higher with Q8W. These are absolute, stratum-adjusted differences in proportions.
What it does not mean. A 41.2-point difference does not mean 41.2% of patients maintained response; it is the gap between two proportions. It also does not mean that participants who lost EASI-75 returned to their pre-treatment severity, since the endpoint is a threshold and says nothing about how far below it a participant fell.
Precision. These intervals are wide, reflecting the smaller subset of participants with EASI-75 at entry. The Q8W lower limit of 9.70 is compatible with a modest benefit, while the Q2W/QW lower limit of 28.93 lies above the Q8W point estimate, a pattern suggestive of better maintenance with more frequent dosing, though not a formal test of it.
P-value vs effect size. The p-values (0.0040, 0.0004, <0.0001) decline as the estimates grow, but the p-value reflects both effect size and precision. A smaller p-value should not be read as proof that one regimen is better than another.
Cautions. Non-responder imputation for missing Week 36 data and for post-rescue values counts both dropout and rescue as failure. This is a demanding, composite-style definition of maintenance. Because the CMH analysis is confined to a subset defined by baseline response status, the randomization balance within that subset is expected but not guaranteed.
8. Secondary Endpoint Results
Key secondary endpoints were analyzed with the same stratified CMH approach and tested within the hierarchy.
| Endpoint | Comparison vs placebo | Difference (percentage points) | 95% CI | P-value |
|---|---|---|---|---|
| IGA response maintained within 1 point of baseline at Week 36 | Q8W | 21.4 | 4.86 to 38.00 | 0.0130 |
| Q4W | 33.5 | 17.38 to 49.72 | 0.0003 | |
| Q2W/QW | 42.1 | 28.36 to 55.76 | <0.0001 | |
| IGA response maintained at 0 or 1 at Week 36 | Q8W | 18.5 | 4.14 to 32.91 | 0.0209 |
| Q4W | 29.7 | 14.89 to 44.42 | 0.0007 | |
| Q2W/QW | 39.7 | 27.42 to 51.95 | <0.0001 | |
| Peak pruritus NRS increased ≥3 points from baseline to Week 35 | Q8W | −14.4 | −29.21 to 0.32 | 0.1048 |
| Q4W | −20.6 | −35.32 to −5.89 | 0.0107 | |
| Q2W/QW | −36.1 | −48.40 to −23.74 | <0.0001 |
For the IGA endpoints, higher values favor dupilumab (more participants maintaining response). For the pruritus endpoint, the outcome is worsening, so negative values favor dupilumab. The pruritus comparison for Q8W is the only posted comparison whose confidence interval includes zero (−29.21 to 0.32; P = 0.1048).
9. Safety: Serious Adverse Events
| Arm | Participants with SAEs | Participants at risk |
|---|---|---|
| Placebo QW | 1 | 82 |
| Dupilumab 300 mg Q8W | 3 | 84 |
| Dupilumab 300 mg Q4W | 4 | 87 |
| Dupilumab 300 mg Q2W/QW | 6 | 167 |
Serious adverse events were infrequent in every arm, with single-digit counts. With so few events, between-arm differences are highly uncertain, and no formal statistical comparison of SAEs was posted. The Q2W/QW arm also has about twice as many participants at risk, so raw counts should be compared only alongside their denominators. The safety denominators reflect treated participants and need not match the 422 enrolled.
10. Statistical Methodology
ANCOVA for the continuous co-primary endpoint
Analysis of covariance compares group means of an outcome while adjusting for one or more baseline covariates. Adjustment for prognostic baseline variables reduces residual variance, which narrows confidence intervals without introducing bias in a randomized trial.
Yi is the EASI percent-change difference for participant i, Xi represents baseline covariates, and τ is the adjusted treatment difference. The registry does not list the covariates used in the SOLO-CONTINUE model.
Multiple imputation with ANCOVA
Because post-rescue values were set to missing, some participants lacked a usable Week 36 value. Multiple imputation creates several completed datasets by drawing plausible values from a model fitted to the observed data, analyzes each with ANCOVA, and combines the results using Rubin's rules, so that the final standard error reflects both within-imputation and between-imputation variability.
τ̂k is the estimate in imputed dataset k of m, W̄ is the average within-imputation variance, and B is the between-imputation variance. The registry does not describe the imputation model or the number of imputations.
Stratified Cochran-Mantel-Haenszel analysis
For binary endpoints, the CMH approach combines 2×2 tables (treatment by response) across strata defined by region, baseline disease severity and parent-study dupilumab regimen. It tests whether treatment is associated with response after accounting for these factors, and a weighted combination of stratum-specific differences yields an adjusted difference in proportions with its confidence interval.
Each stratum h contributes its own treatment difference, weighted by its size. Stratifying by factors that predict response removes their influence from the comparison and can improve precision.
Hierarchical (fixed-sequence) testing
The registry describes a 2-sided hierarchical procedure for the co-primary and key secondary endpoints in a prespecified order, with each test performed at the 0.05 level only if the previous one was significant. This controls the familywise type I error at 0.05 without splitting alpha, at the cost of making later tests conditional on earlier successes.
Intention-to-treat framing
The full analysis set, defined as all randomized participants, is the intention-to-treat population for the continuous co-primary endpoint. Participants are analyzed in the arm to which they were randomized, preserving the comparability that randomization provides.
11. Statistical Methods Explained
Why was ANCOVA used for the EASI co-primary endpoint?
The EASI endpoint is continuous and participants arrived with different degrees of prior improvement. ANCOVA adjusts the treatment comparison for baseline characteristics, which increases precision relative to an unadjusted comparison of means and accounts for chance baseline imbalance.
Why was a Cochran-Mantel-Haenszel test used for the responder endpoints?
Response rates may differ across regions, disease severity levels and parent-study regimens. The CMH test respects that structure by comparing treatment groups within strata and then combining the evidence, giving a stratum-adjusted difference in proportions rather than a crude one.
What does a difference of −21.61 in EASI percent change mean?
It means that, relative to placebo, participants on Q2W/QW dupilumab lost an estimated 21.61 fewer percentage points of the EASI improvement they had achieved since the parent-study baseline. It is a difference in percent improvement retained, not a change on the 0 to 72 EASI scale.
How does hierarchical testing control false positives?
Each hypothesis is tested at the full 0.05 level, but only if every earlier hypothesis in the sequence was rejected. Because no test can "succeed" unless all preceding tests did, the chance of any false-positive claim across the sequence stays at or below 0.05. The trade-off is that a single non-significant result, such as the Q8W pruritus comparison (P = 0.1048), stops confirmatory testing of everything placed after it.
Why were rescue-treated values set to missing?
Rescue therapy can improve skin disease independent of the randomized regimen. Using post-rescue values as if they reflected the assigned regimen would make a failing regimen look better. Setting them to missing, then using multiple imputation or non-responder imputation, keeps the estimate focused on how well the randomized regimen maintained control on its own.
Do the gradients across Q8W, Q4W and Q2W/QW prove a dose-response relationship?
Not formally. Across the co-primary and secondary endpoints, estimates tend to be larger with more frequent dosing, which is suggestive. However, no trend test or head-to-head comparison between regimens was posted, and the confidence intervals of adjacent regimens overlap. A dose-response claim would require a prespecified comparison between regimens.
12. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
All six co-primary comparisons favored dupilumab with 95% intervals excluding zero, within a prespecified hierarchy at the 0.05 level. Among the posted secondary comparisons, only Q8W for pruritus worsening did not reach significance. The intervals are wide, particularly for the responder endpoints analyzed in subsets.
Clinical interpretation
In participants who had already improved on dupilumab, continuing treatment was associated with better maintenance of skin improvement and less itch worsening than switching to placebo. Point estimates were generally largest with the most frequent regimen and smallest with every-8-week dosing. Serious adverse events were infrequent in all arms.
13. Important Limitations and Interpretation Issues
- Enriched population: participants entered after treatment in the parent studies, so the results apply to maintenance in people already exposed to dupilumab, not to treatment initiation.
- Complex endpoint scale: the continuous co-primary endpoint is a change in percent improvement anchored to an earlier baseline, which is less intuitive than a change in EASI score and sensitive to how low the parent-study baseline was.
- Subset analyses for binary endpoints: each responder endpoint is restricted to participants meeting its baseline threshold, reducing sample size and widening intervals.
- Missing-data assumptions: multiple imputation assumes missingness is predictable from observed data; non-responder imputation assumes every dropout or rescue is a failure. Both are assumptions rather than facts.
- Hierarchy position not fully described: the registry does not list the order of every dose-regimen comparison, so it is not possible to say exactly which results after the non-significant Q8W pruritus test are nominal.
- No between-regimen tests: all posted comparisons are against placebo; dose-interval differences are descriptive.
- Limited safety precision: SAE counts are small in every arm.
- Registry detail: the ClinicalTrials.gov record does not report baseline characteristics, ANCOVA covariates, or the imputation model in the fields summarized here.
14. Why This Trial Matters Statistically
SOLO-CONTINUE is a strong teaching example for maintenance trials: it combines a randomized-withdrawal structure, a continuous endpoint analyzed with ANCOVA after multiple imputation, stratified CMH analyses of responder endpoints in baseline-defined subsets, and a formal hierarchical testing strategy in which one comparison failed.
| Concept | How it appears in SOLO-CONTINUE |
|---|---|
| Randomized withdrawal | Randomization after prior dupilumab exposure in parent studies |
| ANCOVA | Adjusted mean difference for the EASI percent-change co-primary endpoint |
| Multiple imputation | Post-rescue values set to missing and imputed before ANCOVA |
| Cochran-Mantel-Haenszel test | Stratified differences in proportions for EASI-75, IGA and pruritus endpoints |
| Risk difference | Responder effects expressed as percentage-point differences |
| Intention-to-treat analysis | Full analysis set of all randomized participants |
| Multiplicity adjustment | 2-sided fixed-sequence testing at 0.05 across co-primary and key secondary endpoints |
| Confidence intervals | Wide intervals in baseline-defined subsets; one interval includes zero |
| P-values | Range from <0.0001 to 0.1048; hierarchy determines which are confirmatory |
| Dose-interval comparison | Three regimens vs placebo without formal between-regimen tests |
15. Related Tutorials
Learn more about the methods used in this trial:
16. Related Calculators
17. Sources
- ClinicalTrials.gov registry record: NCT02395133 — A Study to Confirm the Efficacy and Safety of Different Dupilumab Dose Regimens in Adults With Atopic Dermatitis (AD).
- Related publications indexed in PubMed: PMID 39969783; PMID 39588375; PMID 31876900.
Explore the methods behind the results
Each trial page links its endpoints and analyses to a deeper statistical tutorial and to a calculator for working through the numbers.
18. Record Summary
SOLO-CONTINUE randomized 422 adults with atopic dermatitis, previously treated with dupilumab in two parent studies, to three dupilumab maintenance regimens or placebo. Against placebo, the EASI percent-change differences were −14.83 (Q8W), −17.83 (Q4W) and −21.61 (Q2W/QW), and the differences in the proportion maintaining EASI-75 were 24.5, 28.0 and 41.2 percentage points, all with intervals excluding zero. A careful reading combines these adjusted effect estimates with their interval widths, the rescue and imputation rules, the baseline-defined analysis subsets, and the hierarchical testing structure that determines which results are confirmatory.