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 appears in clearly labelled boxes.
1. Trial at a Glance
MermaiHD was a randomized, parallel-group, quadruple-masked phase 3 trial sponsored by Teva Branded Pharmaceutical Products R&D, Inc. It asked whether pridopidine (ACR16) at 45 mg or 90 mg improved motor function in people with Huntington's disease compared with placebo, measured as the change in the Modified Motor Score (mMS) from baseline to week 26.
| Feature | MermaiHD |
|---|---|
| Phase | Phase 3 |
| Condition | Huntington's disease |
| Design | Randomized, parallel-group, quadruple-masked; primary purpose treatment |
| Arms | 3 randomized arms: placebo, ACR16 45 mg, ACR16 90 mg; followed by an open-label phase |
| Enrollment | 437 |
| Primary endpoint | Change from baseline in Modified Motor Score (mMS) at week 26 |
| Primary analysis | ANCOVA with treatment, baseline mMS, gender and antipsychotic use; 97.5% two-sided CIs |
| Status | Completed; results posted |
| ClinicalTrials.gov | NCT00665223 |
| Lead sponsor | Teva Branded Pharmaceutical Products R&D, Inc. (industry) |
2. Clinical Question
The central question was whether pridopidine, given at either of two doses, reduced motor impairment in Huntington's disease relative to placebo over 26 weeks of randomized treatment. Because two active doses were compared against a shared placebo group, the trial effectively posed two related questions at once, which has direct consequences for how its error rate was controlled.
Population
Participants with Huntington's disease enrolled in a phase 3 treatment trial.
Intervention
Pridopidine (ACR16) at 45 mg or 90 mg during the randomized phase.
Comparator
Matching placebo during the randomized phase.
Primary question
Does either dose of ACR16 produce a smaller increase (or a larger decrease) in mMS from baseline to week 26 than placebo?
3. Trial Design
Placebo
- Placebo during the randomized phase
- Registry label: Placebo/ACR16 90 mg
ACR16 45 mg
- Pridopidine 45 mg during the randomized phase
- Registry label: ACR16 45 mg/90 mg
ACR16 90 mg
- Pridopidine 90 mg during the randomized phase
- Registry label: ACR16 90 mg/90 mg
The arm labels in the registry pair the randomized-phase assignment with a second term (for example, "Placebo/ACR16 90 mg"), and the safety tables list a separate open-label phase. The primary comparisons, however, are explicitly described as comparing each ACR16 dose with placebo during the randomization phase. The open-label period is therefore relevant to safety, not to the randomized efficacy estimates.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Primary: Change from baseline in Modified Motor Score (mMS) at week 26 | The mMS is a subscale of the UHDRS total motor score comprising 13 responses from items 4–10 and 13–15 of the UHDRS motor assessment: dysarthria, tongue protrusion, finger taps (right and left), pronate/supinate hands (right and left), Luria first-hand-palm sequencing, arms rigidity (right and left), body bradykinesia, gait, tandem walking and retropulsion pull test. Each item is rated 0 (normal) to 4 (marked impairment). Total score ranges from 0 to 52; higher scores indicate more severe motor impairment. | Baseline, week 26 |
Reading the direction of the scale
Because higher mMS scores indicate worse motor impairment, a negative change from baseline means improvement and a negative treatment difference (ACR16 minus placebo) favours ACR16. This matters when reading the results below: both reported estimates are negative, so both point in the direction of benefit, and the question is whether they are large enough and precise enough to exclude zero.
The mMS uses a defined subset of items from the UHDRS motor assessment rather than the full total motor score. Its range of 0 to 52 provides the frame for judging the size of any between-group difference.
5. Analysis Population and Covariates
| Element | Registry description |
|---|---|
| Analysis population | Full Analysis Set (FAS): all randomized participants who received at least 1 dose of study drug and had at least 1 clinical assessment after randomization; participants analyzed were those evaluable for the outcome measure |
| Model | ANCOVA main-effects model |
| Treatment term | Randomized treatment group |
| Covariates | Baseline mMS, gender, use of antipsychotic medication |
| Effect measure | Mean difference (final values) |
| Confidence interval | 97.5%, two-sided |
| Hypothesis | Superiority |
The FAS is a modified intention-to-treat population. It keeps participants in their randomized groups but excludes anyone who was never dosed or never assessed after randomization. Such exclusions are usually small and, when caused by events unrelated to treatment, unlikely to bias the comparison, but they mean the analysis is not a strict analysis of every randomized participant.
6. Primary Results: ACR16 90 mg vs Placebo
Adjusted mean difference in mMS change at week 26
97.5% CI: −2.08 to 0.10 · P = 0.042
ANCOVA; ACR16 90 mg vs placebo during the randomization phase; FAS
| Comparison | Estimate | 97.5% CI (two-sided) | P-value | Method |
|---|---|---|---|---|
| ACR16 90 mg vs placebo | −0.99 | −2.08 to 0.10 | 0.042 | ANCOVA, superiority |
What the estimate means. After adjustment for baseline mMS, gender and antipsychotic use, the mean change in mMS at week 26 was estimated to be 0.99 points lower (better) with ACR16 90 mg than with placebo. On a scale that runs from 0 to 52, this is a difference of about one point.
What it does not mean. It does not mean that every participant on 90 mg improved by about one point, nor that ACR16 reversed motor impairment. It is a difference in group means, and individual changes would have varied around it.
Precision. The 97.5% interval runs from −2.08 to 0.10. It includes zero, so at the confidence level the trial chose for this comparison, a true difference of zero (or a very small difference favouring placebo) remains compatible with the data. The interval also includes differences of around two points in favour of ACR16.
Why P = 0.042 is not the effect size. The P-value measures how surprising the observed difference would be if the true difference were zero; it says nothing directly about how large or clinically meaningful the difference is. A one-point difference could produce a small P-value in a large trial and a large P-value in a small one.
Caution: the P-value and interval must be read together. A P-value of 0.042 is below the conventional 0.05, but the reported 97.5% interval includes zero. These two facts are consistent only if the comparison was judged against a threshold stricter than 0.05. A 97.5% two-sided interval corresponds to a two-sided significance level of 0.025, and 0.042 does not fall below 0.025. Read against the level implied by its own reported interval, this comparison did not reach statistical significance.
7. Primary Results: ACR16 45 mg vs Placebo
Adjusted mean difference in mMS change at week 26
97.5% CI: −1.44 to 0.72 · P = 0.456
ANCOVA; ACR16 45 mg vs placebo during the randomization phase; FAS
| Comparison | Estimate | 97.5% CI (two-sided) | P-value | Method |
|---|---|---|---|---|
| ACR16 45 mg vs placebo | −0.36 | −1.44 to 0.72 | 0.456 | ANCOVA, superiority |
What the estimate means. The adjusted mean change in mMS was estimated to be 0.36 points lower with ACR16 45 mg than with placebo, a difference in the direction of benefit but smaller than that seen with 90 mg.
What it does not mean. A non-significant result is not evidence that 45 mg has no effect. It means the data do not distinguish the observed difference from zero with the chosen level of confidence.
Precision. The interval from −1.44 to 0.72 is centred close to zero and spans modest benefit and modest harm. It is roughly the same width as the 90 mg interval, which is expected when group sizes and variability are similar.
P-value. P = 0.456 indicates that a difference at least this large would be quite common if pridopidine 45 mg and placebo had the same true mean effect. It is well above both 0.05 and the stricter 0.025 level implied by the reported interval.
Caution. The two dose comparisons share the same placebo group, so their estimates are not statistically independent. The 90 mg and 45 mg results should not be treated as two separate trials confirming or contradicting one another.
Both comparisons on one scale
The vertical line marks zero. Both intervals cross it, but the 90 mg interval only barely does so (upper limit 0.10), while the 45 mg interval is centred much closer to zero. The pattern is consistent with a possible dose-related effect, but the trial did not report a formal test of dose response, so this is a descriptive reading rather than a tested conclusion.
8. Multiplicity: Two Doses, One Placebo Group
With two active doses each compared with placebo, the trial had two chances to declare a positive result. If each comparison were tested at a two-sided 0.05 level, the probability of at least one false-positive finding under the null hypothesis would exceed 0.05. The registry reports 97.5% two-sided confidence intervals for both comparisons, which matches an approach in which the overall two-sided 0.05 level is divided equally between the two doses (0.025 each), as in a Bonferroni adjustment.
| Comparison | P-value | Below 0.05? | Below 0.025? | 97.5% CI excludes 0? |
|---|---|---|---|---|
| ACR16 90 mg vs placebo | 0.042 | Yes | No | No (−2.08 to 0.10) |
| ACR16 45 mg vs placebo | 0.456 | No | No | No (−1.44 to 0.72) |
9. Safety: Serious Adverse Events
The registry reports the number of participants with serious adverse events (SAEs) in each group, separately for the randomized and open-label phases.
| Phase | Group | Participants with SAEs / at risk |
|---|---|---|
| Randomized | Placebo | 11 / 144 |
| Randomized | ACR16 45 mg | 10 / 148 |
| Randomized | ACR16 90 mg | 9 / 145 |
| Open label | Placebo | 6 / 113 |
| Open label | ACR 16 | 15 / 240 |
During the randomized phase, the counts of participants with SAEs were similar across the three groups, with no numerical excess in either ACR16 group relative to placebo. The at-risk denominators for the randomized phase (144, 148 and 145) sum to 437, the total enrollment.
10. Statistical Methodology
Analysis of covariance (ANCOVA)
The primary endpoint was analysed with an ANCOVA main-effects model. The outcome was modelled as a linear function of treatment group plus three covariates: baseline mMS, gender and use of antipsychotic medication. The treatment coefficient is the adjusted mean difference between groups.
βT is the treatment effect for a given dose versus placebo, estimated at −0.99 for 90 mg and −0.36 for 45 mg. A "main-effects" model contains no treatment-by-covariate interaction terms, so it assumes the treatment effect is the same across levels of the covariates.
Why adjust for baseline mMS
Baseline motor score is almost certainly the strongest predictor of week-26 motor score. Including it as a covariate removes a large share of the between-participant variability, which narrows the confidence interval for the treatment effect compared with an unadjusted comparison of means. In a randomized trial, this adjustment is primarily about precision rather than bias, because randomization already balances baseline scores in expectation.
Final values vs change scores
The registry describes the effect measure as a "mean difference (final values)" for an endpoint named as a change from baseline. When baseline is included as a covariate, an ANCOVA on final values and an ANCOVA on change scores give the same treatment-effect estimate and standard error, because the change score is simply the final value minus a term already in the model. The labelling difference therefore does not alter the interpretation of the treatment difference.
Superiority hypothesis
Each comparison was framed as a superiority test: the null hypothesis was no difference between ACR16 and placebo, and the alternative was a difference. There was no non-inferiority margin.
Confidence level
Two-sided 97.5% confidence intervals were reported for both dose comparisons. As discussed in Section 8, this corresponds to a two-sided 0.025 level per comparison.
11. Statistical Methods Explained
Why was an ANCOVA used rather than a simple t-test?
The outcome is a continuous score measured at baseline and week 26. ANCOVA compares groups at week 26 while adjusting for each participant's starting score and prespecified prognostic factors. This gives a more precise estimate than a t-test on raw week-26 scores or on unadjusted change scores, and it accounts for regression to the mean in baseline values.
What does an adjusted mean difference of −0.99 mean?
It means that, for participants with the same baseline mMS, gender and antipsychotic use, the model estimates the week-26 mMS to be 0.99 points lower on 90 mg than on placebo. Since lower is better, this favours ACR16. It is an average group-level difference on a 0 to 52 scale, not a guaranteed individual change.
Why were the confidence intervals 97.5% rather than 95%?
Two doses were each tested against placebo. Splitting the overall two-sided 0.05 error rate equally gives 0.025 per comparison, and the matching confidence interval is 97.5%. A wider interval is the price of protecting against a false-positive finding across two tests.
How can P = 0.042 and a confidence interval that includes zero both be correct?
A P-value below α corresponds to a (1 − α) confidence interval that excludes the null value. P = 0.042 is below 0.05, so a 95% interval would exclude zero, but it is above 0.025, so the 97.5% interval includes zero. The two numbers are consistent; they are simply being judged against different thresholds.
Why does the Full Analysis Set matter?
The FAS includes randomized participants who took at least one dose and had at least one post-randomization assessment. It stays close to the intention-to-treat principle by analysing participants in their assigned groups, but it is a modified version: participants never treated or never assessed are left out. How participants without a week-26 value were handled is not described in the registry.
Why include gender and antipsychotic use as covariates?
Antipsychotic medications can affect motor function in Huntington's disease, and gender may be associated with disease presentation. Prespecifying these covariates adjusts for chance imbalances that could otherwise add noise to the treatment comparison. Because they were specified in the analysis model, they do not introduce the risks of selecting covariates after seeing the data.
12. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
Neither dose produced a 97.5% confidence interval that excluded zero. The 90 mg comparison was nominally significant at 0.05 (P = 0.042) but not at the 0.025 level implied by the reported interval; the 45 mg comparison was not significant at either level (P = 0.456).
Clinical interpretation
Both point estimates favoured ACR16, with a larger estimate for 90 mg, but the estimated differences were around one point or less on a 52-point scale. Whether a difference of this size is meaningful to patients requires an externally justified minimal clinically important difference, which the registry does not report.
13. Important Limitations and Interpretation Issues
- Multiplicity: with two dose comparisons, the per-comparison threshold is stricter than 0.05. Describing the 90 mg result as "significant" without reference to the 97.5% interval would overstate the evidence.
- Shared control group: both comparisons use the same placebo participants, so the two results are correlated and cannot be treated as independent replications.
- Modified analysis population: the FAS excludes randomized participants who were not dosed or not assessed after randomization, and the handling of missing week-26 values is not described in the registry.
- Main-effects assumption: the ANCOVA model assumes a common treatment effect across baseline severity, gender and antipsychotic use; differential effects were not assessed in the posted analyses.
- Short duration: the primary assessment was at week 26. Huntington's disease is progressive, and a 26-week result does not address longer-term motor trajectories.
- Clinical relevance: differences of about one point on the mMS need an external benchmark for clinical importance; statistical testing alone cannot supply it.
- Limited registry detail: the registry does not report arm-level means, standard deviations, the number analysed per arm, or subgroup results, which limits independent assessment of variability and consistency.
- Safety precision: with roughly 10 participants with SAEs per randomized arm, the trial has limited ability to detect differences in serious events.
14. Why This Trial Matters Statistically
MermaiHD is a compact teaching case for a common and often misread situation: a multi-dose trial in which one comparison crosses the conventional 0.05 line but not the multiplicity-adjusted threshold.
| Concept | How it appears in MermaiHD |
|---|---|
| Randomization and masking | Three parallel arms with quadruple masking of a rater-scored motor endpoint |
| ANCOVA | Adjusted mean difference with baseline mMS, gender and antipsychotic use as covariates |
| Covariate adjustment | Baseline score adjustment to improve precision in a randomized comparison |
| Confidence intervals | 97.5% two-sided intervals of −2.08 to 0.10 and −1.44 to 0.72 |
| P-values | P = 0.042 and P = 0.456, read against a per-comparison 0.025 level |
| Multiplicity | Two dose comparisons against one placebo group |
| Analysis population | Full Analysis Set as a modified intention-to-treat population |
| Scale direction | Negative differences favour treatment because higher mMS means worse impairment |
15. Trial Timeline
Study start
Enrollment begins in the randomized, quadruple-masked phase.
Primary completion
Final data collection for the primary endpoint, change in mMS at week 26.
Results posted
ANCOVA results for both dose comparisons and serious adverse events by phase and arm are available in the registry record.
16. Related Tutorials
Learn more about the methods used in this trial:
17. Related Calculators
18. Sources
- ClinicalTrials.gov: NCT00665223 – A Study of Treatment With Pridopidine (ACR16) in Participants With Huntington's Disease.
- PubMed: PMID 22071279.
- PubMed: PMID 36811812.
Explore more trial analyses
Each trial page connects its endpoints and methods to a deeper statistical tutorial and to an appropriate calculator.
19. Record Summary
MermaiHD randomized 437 participants with Huntington's disease to placebo, pridopidine 45 mg or pridopidine 90 mg and analysed change in Modified Motor Score at week 26 with a covariate-adjusted ANCOVA. Both adjusted mean differences favoured pridopidine (−0.99 for 90 mg and −0.36 for 45 mg), but neither 97.5% confidence interval excluded zero. The 90 mg comparison (P = 0.042) illustrates how nominal significance, multiplicity-adjusted significance, effect size and clinical relevance are distinct questions that must be answered separately.