Persistence of Effect and Safety of Valbenazine for the Treatment of Tardive Dyskinesia: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Persistence of Effect and Safety of Valbenazine for the Treatment of Tardive Dyskinesia (NCT03891862), a randomized phase 4 study. The primary outcome is Change From Randomization (Week 8) in Abnormal Involuntary Movement Scale (AIMS) Dyskinesia Total Score at Week 16.

2. Clinical Question

3. Trial Design

This Phase 4 study is described as a randomized, double-blind, placebo-controlled randomized withdrawal study. It used a parallel-group design, and the registry records triple masking. The study enrolled 135 participants and is listed as completed. Its stated purpose was to evaluate persistence of effect and safety of valbenazine for tardive dyskinesia.

The treatment sequence matters when interpreting the comparison. Participants first received open-label valbenazine for 8 weeks: 40 mg once daily for 1 week, followed by 80 mg once daily for the remainder of that period. Randomization occurred after this open-label treatment. During the subsequent 8-week placebo-controlled period, one randomized arm continued valbenazine 80 mg once daily and the other received matching placebo once daily.

Accordingly, the randomized comparison concerns what happened during the period after randomization among participants who had already received open-label valbenazine. It is not a comparison of starting valbenazine against never receiving valbenazine. The distinction is central to the study’s focus on persistence of effect after withdrawal versus continued treatment.

The primary outcome was assessed from Week 8, the point of randomization, to Week 16. The posted primary endpoint results use least-squares means and 95% confidence intervals. Secondary outcomes measure changes in health-state index and visual analogue scale scores from baseline to Week 16. Safety results are presented for the open-label and placebo-controlled treatment groups; those group counts have different at-risk denominators and should be read separately. A detailed account of the randomization procedure, allocation concealment, and the reasons for the selected sample size is not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: After receiving valbenazine 40 mg for the first week followed by valbenazine 80 mg for 7 weeks during the open-label period, participants were randomly assigned to either placebo or to continue to receive valbenazine 80 mg.

Open-label Period

Reasons for not completing this period:

Placebo-controlled Period and Follow-up

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Persistence of effect analysis set, which includes all randomized participants who received at least one dose of randomized study drug and have at least one AIMS assessment during the placebo-controlled period. Participants who do not have an AIMS assessment at a scheduled or mapped early termination visit are not included in the analysis.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: Persistence of effect analysis set, which includes all randomized participants who received at least one dose of randomized study drug and have at least one AIMS assessment during the placebo-controlled period.

6. Primary Endpoint

The registry lists the primary outcome as Change From Randomization (Week 8) in Abnormal Involuntary Movement Scale (AIMS) Dyskinesia Total Score at Week 16, measured over Week 8, Week 16.

The primary endpoint is change from randomization at Week 8 to Week 16 in the AIMS Dyskinesia Total Score. The AIMS rates 10 items of involuntary movement, covering facial, oral, extremity, and trunk movements, as well as self-awareness of abnormal movements. Each item ranges from 0, indicating no dyskinesia, to 4, indicating severe dyskinesia. The total score sums items 1–7 and ranges from 0 to 28; higher scores indicate more severe dyskinesia.

Using the randomization visit as the starting point makes the endpoint relevant to the withdrawal design: it describes score change during the randomized period after the open-label treatment phase. The endpoint is a continuous measure, so the reported arm-level least-squares means describe average modeled change rather than the experience of every participant. The registry states that these means were estimated using a mixed-effects model for repeated measures.

The endpoint captures a scored measure of movement symptoms, not every aspect of health or daily life. The registry also lists health-status outcomes, but those are distinct from the primary AIMS measure. A detailed explanation of why this particular endpoint and assessment interval were selected is not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a least-squares mean represent?

The primary endpoint is reported as a least-squares mean, estimated with a mixed-effects model for repeated measures. In general, a least-squares mean is a model-based estimate of an average. It is not necessarily the same as the raw arithmetic average, because it comes from a statistical model that uses the outcome data and the model’s structure. The registry identifies the model type but does not report its full specification, including the covariates, covariance structure, or exact handling of incomplete observations.

A repeated-measures model is generally used when outcomes are assessed at multiple times. It can account for the fact that observations from the same participant are related rather than independent. The primary outcome’s timeframe is Week 8 and Week 16, and the reported change is anchored at randomization. Details beyond the stated mixed-effects model for repeated measures are not reported in the registry record.

How should the confidence intervals be read?

The primary results give a 95% confidence interval for each arm’s modeled change. In general, a confidence interval communicates uncertainty around an estimate: narrower intervals indicate greater precision, while wider intervals indicate less precision. It does not give the probability that a particular fixed value is true. The intervals in the primary results describe the arm-level estimates; they are not intervals for a between-arm contrast.

To evaluate a randomized comparison directly, the relevant quantity is an estimated difference between arms with its uncertainty, and often a test result under the analysis plan. Those primary comparison results are not reported in the registry record. It would be incorrect to infer a direct comparison solely from the two separate arm estimates or their separate intervals.

Why does the study sequence matter?

In general, a randomized withdrawal design first exposes participants to a treatment and then randomizes them to continue it or switch to a comparator. That design addresses outcomes after randomization among participants who have already experienced the initial treatment period. It does not estimate the same contrast as a trial that randomizes participants before any treatment exposure. In this study, randomization followed 8 weeks of open-label valbenazine, and the placebo-controlled period lasted 8 weeks.

This sequence also means that the open-label safety group and the randomized treatment groups represent different periods and treatment contexts. Safety counts should not be treated as if they arose from one single, parallel comparison lasting throughout the study. The registry lists at-risk counts separately for the open-label and placebo-controlled groups.

What do the secondary comparisons show statistically?

The two secondary outcomes are continuous measures of health status, reported as mean changes from baseline to Week 16 with standard errors. The registry lists arm means and standard errors, along with unadjusted comparisons. In general, a standard error describes the estimated precision of a mean; it is not the spread of individual participant outcomes. The unadjusted p-values describe each comparison without accounting for the presence of other secondary tests.

The posted secondary comparisons have mean differences of 0.07 for the health-state index and 2.0 for the visual analogue scale. Their unadjusted p-values are 0.099 and 0.594, respectively. The confidence interval for the health-state index comparison crosses zero, and the interval for the visual analogue scale comparison also crosses zero. These comparisons are not reported as adjusted analyses, and the registry lists no reported analyses for either secondary outcome.

What are the planning illustrations?

The planning illustrations estimate participants per arm under stated assumptions; they are not observed trial results and are not observed power. They use the observed effect as a reference and consider planning for 100%, 75%, and 50% of that effect. The stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The calculations use effects defined as the stated fraction of the observed effect.

For the health-state index, the illustration gives 168 participants per arm for 80% power and 224 participants per arm for 90% power when planning for 100% of the observed effect. For 75% of that effect, the corresponding illustrations are 298 and 399 participants per arm. For 50%, they are 670 and 896 participants per arm. These values are hypothetical planning results based on the observed effect, not a recommendation or the trial’s original design target.

For the visual analogue scale, the corresponding illustrations are 1,611 and 2,156 participants per arm for 100% of the observed effect; 2,863 and 3,833 for 75%; and 6,442 and 8,624 for 50%. Smaller target effects generally require more participants per arm to detect under a fixed set of assumptions. This is a general statistical relationship, not evidence about what effect is clinically important. The registry record does not report the trial’s own sample-size assumptions or justification.

These planning calculations depend on the observed effects, which can be uncertain, and on the assumptions used in the calculation. They should not be read as showing that the completed study was adequately or inadequately powered. The registry’s primary outcome has no posted between-arm contrast from which to interpret a primary planning calculation, and no primary planning illustration is listed in the record.

10. Multiplicity, Missing Data and Analysis Populations

The registry record lists one primary outcome and two secondary health-status outcomes. It provides unadjusted comparisons for the secondary outcomes but does not report a multiplicity adjustment, a prespecified testing hierarchy, or an order in which the outcomes should be interpreted. Whether an adjustment or hierarchy was specified elsewhere is not reported in the registry record.

Multiplicity matters because examining several outcomes can increase the chance of a seemingly noteworthy result arising by chance. In general, methods that control this issue can reduce that risk, though they may also change how much evidence is needed for an individual outcome. Without a reported approach, the secondary p-values should be treated as unadjusted and interpreted cautiously. The record lists no reported analyses for the secondary outcomes.

The primary outcome is described as using a mixed-effects model for repeated measures, but the model’s full specification is not reported in the registry record. The exact missing-data method, assumptions about missing observations, sensitivity analyses, and any imputation procedure are also not reported. The primary analysis-population description begins with randomized participants who received at least one dose of randomized study drug and had at least one AIMS assessment during the placebo-controlled period, but the registry text cuts off before the full definition is complete.

Those details matter because the participants included in an analysis and the way missing measurements are handled can influence an estimated treatment effect. A model can use available repeated measurements under assumptions, but the registry’s brief description is not enough to identify those assumptions or assess their robustness. The record does not report interim analyses, a statistical analysis plan, or a sample-size justification. Their absence from the registry record does not establish whether such material exists elsewhere; it means the details are not reported in the registry record.

11. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 3% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: cellulitis (1.7%), sepsis (1.7%), psychotic disorder (1.7%), knee arthroplasty (1.7%), accidental overdose (0.8%).

The registry lists serious adverse events and other adverse events separately for the open-label period and the two placebo-controlled arms. In the open-label valbenazine group, 3 of 132 participants at risk had a serious adverse event, 17 of 132 had an other adverse event, and 1 death was recorded. In the placebo-controlled placebo group, the corresponding counts were 2 of 59 for serious adverse events, 11 of 59 for other adverse events, and 0 deaths. In the placebo-controlled valbenazine group, the counts were 1 of 59, 7 of 59, and 0 deaths.

These are counts over the registry’s stated groups, not estimates adjusted for time at risk or evidence of a causal relationship. The open-label group covers the earlier treatment period, while the placebo-controlled groups follow randomization after open-label treatment; the groups should not be combined or compared as if they represented identical exposure periods. The registry record does not report the relationship of the death or adverse events to study treatment, the event durations, or detailed event-level circumstances.

The listed serious-event terms include cellulitis, sepsis, psychotic disorder, and knee arthroplasty at 1.7% each, and accidental overdose and dehydration at 0.8% each. These figures describe the posted serious-event terms; they do not establish that treatment caused them.

12. What the Result Does, and Does Not, Mean

The primary posted results describe modeled AIMS change during the randomized period after an 8-week open-label valbenazine phase. They give one least-squares mean and confidence interval for each randomized arm, but no direct between-arm estimate, confidence interval for the contrast, or primary p-value. The result therefore communicates arm-specific change and its precision, not a complete statistical answer to whether the randomized groups differed.

The endpoint itself is a scored measure in which higher AIMS Dyskinesia Total Scores indicate more severe dyskinesia. The registry describes the scale range and the model used, but the posted primary analysis does not include a direct comparison. The sign of a change is most responsibly interpreted alongside the exact change convention, which is not stated in the registry record.

The secondary outcomes provide additional health-status information, but they are not substitutes for the primary outcome. Their comparisons are unadjusted, and the registry does not report a multiplicity strategy or secondary analyses. Their results should be viewed as exploratory descriptions rather than as decisive evidence about the primary question.

More generally, an estimate does not establish clinical importance by itself. Statistical precision, the scale’s interpretation, the study population, and the design all matter. The registry record does not provide a threshold for a clinically important AIMS change or a sample-size justification. Nor do the reported results establish how outcomes would apply to people who were not represented in the analysis population or to treatment situations outside the study sequence.

The safety figures are descriptive counts with different group contexts and denominators. They do not establish causation, and the registry does not report event attribution. The findings can be read as the outcome and safety information posted for this study, with the important limits imposed by the missing primary contrast and incomplete analysis details.

13. Why This Trial Matters Statistically

This study is statistically distinctive because it uses a randomized withdrawal design. Participants received open-label valbenazine before randomization to continued valbenazine or matching placebo. The randomized period therefore asks about outcomes following prior treatment exposure, rather than comparing treatment initiation with no prior exposure. That sequence shapes both the population to which the comparison applies and the interpretation of the Week 8 to Week 16 primary outcome.

The primary endpoint is a repeated-measures, continuous score, and the registry reports model-based arm estimates with confidence intervals. This is useful for understanding the estimated change and its uncertainty in each arm. However, the registry does not report the key between-arm contrast. That omission illustrates why arm-specific estimates, even when accompanied by confidence intervals, are not equivalent to the comparison that randomization is designed to support.

The record also illustrates the distinction between observed outcomes and planning calculations. The secondary outcome illustrations use observed effects and specified assumptions to show hypothetical participants per arm for target power levels. They are not observed power and do not reveal the study’s original design rationale. The trial’s own sample-size justification is not reported in the registry record.

Finally, secondary outcomes and safety counts need their own context. Unadjusted secondary comparisons have less confirmatory weight when multiple outcomes are assessed, while safety counts span different study periods and at-risk groups. Reading the design, estimand, precision, multiplicity, and analysis-population details together prevents the reported estimates from being given a broader interpretation than the registry record supports.

14. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

A central limitation of the posted primary results is that they do not include a direct randomized-arm contrast, its confidence interval, or a p-value. The arm-level least-squares means describe changes within each arm, but they do not quantify the difference between those changes. This restricts interpretation of the main randomized comparison.

The primary analysis-population description is incomplete in the registry record: the available text ends after beginning to describe how participants without an AIMS assessment at a scheduled or mapped early termination were handled. The complete rule is not reported in the registry record. Further details of the repeated-measures model, missing-data assumptions, sensitivity analyses, and outcome coding are also not reported.

The study’s withdrawal sequence is another important limitation on generalization. Participants were treated with open-label valbenazine before randomization, so the randomized comparison pertains to a post-exposure period. It does not directly represent a comparison between initial treatment and no treatment from the start of care. The registry record does not provide enough information to assess how the randomized population compares with people who might not enter or complete the open-label period.

Both secondary comparisons are unadjusted, with no reported multiplicity strategy or secondary analyses. Their interpretive weight is therefore limited. Safety counts are descriptive and relate to groups in different study periods; the registry does not report treatment attribution or event-level detail. The record also does not report a sample-size justification, interim analyses, or a full statistical analysis plan. These are limits of the information reported in the registry, not claims about whether additional study documentation exists.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Persistence of Effect and Safety of Valbenazine for the Treatment of Tardive Dyskinesia — NCT03891862.
• Registry record last updated 2021-04-20; status last verified 2021-03.
• Results first posted 2021-03-10.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

18. Record Summary

The registry record describes NCT03891862, a completed Phase 4, randomized, double-blind, placebo-controlled valbenazine study for tardive dyskinesia. It enrolled 135 participants and used a randomized withdrawal design: an 8-week open-label valbenazine period followed by an 8-week randomized period of continued valbenazine or matching placebo.

The primary outcome was change in AIMS Dyskinesia Total Score from randomization at Week 8 to Week 16, reported as arm-level least-squares means with 95% confidence intervals. The posted results do not include a direct primary between-arm comparison. Two secondary health-status outcomes and descriptive safety counts are also recorded. Details of multiplicity handling, the complete primary analysis-population definition, missing-data procedures, and the study’s sample-size justification are not reported in the registry record.