Virtual Reality Behavioral Activation: An Intervention for Major Depressive Disorder: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Virtual Reality Behavioral Activation: An Intervention for Major Depressive Disorder (NCT04268316), a randomized na study. The primary outcome is Number of Participants Who Dropped Out of Each Study Arm.

2. Clinical Question

3. Trial Design

This was a randomized, parallel-group study sponsored by Stanford University. The registry lists the study phase as NA and masking as none. Its title is “Virtual Reality Behavioral Activation: An Intervention for Major Depressive Disorder,” and the stated primary purpose was to test the safety and feasibility of virtual reality technology for behavioral activation in major depressive disorder. A secondary purpose was to examine whether there was evidence of clinical efficacy.

The study has three arms: virtual reality behavioral activation, behavioral activation in real life, and waitlist control. The first two arms involved behavioral activation delivered in different settings. Their descriptions specify clinician meetings once a week for three weeks (4 sessions). The waitlist arm description says participants would not receive an intervention during the observation period and would complete the PHQ-9 once a week for three weeks (4 sessions); behavioral activation would be offered after four weeks. These descriptions do not fully detail how treatment was delivered or monitored.

The registry lists a start date of 2020-05-18, a primary completion date of 2021-01-15, and a completed status. Enrollment is listed as 20. For the primary outcome, the reported arm groups contain 5 participants in virtual reality behavioral activation, 4 in real-life behavioral activation, and 4 in waitlist control. The registry record does not explain how those outcome group sizes relate to the enrollment figure, nor does it report the number randomized to each arm separately from the primary outcome group sizes.

The primary endpoint was a count of participants who discontinued the study for any reason, compared across arms over a three-week timeframe. Change in PHQ-9 depression scores from baseline to Session 4, compared across the three arms, is listed as a secondary outcome. The registry record does not report the randomization procedure, allocation concealment, or a detailed analysis plan.

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: 20 participants signed consent, but only 13 participants were randomized to a study arm.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: This outcome measure was assessed in the Virtual Reality Behavioral Activation group only.

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.

6. Primary Endpoint

The registry lists the primary outcome as Number of Participants Who Dropped Out of Each Study Arm, measured over 3-weeks.

The primary endpoint counts participants who dropped out of each study arm. The registry defines treatment dropout as discontinuing the study for any reason and gives the timeframe as three weeks. The posted counts are 2 participants in virtual reality behavioral activation, 1 in real-life behavioral activation, and 0 in waitlist control. Arm group sizes reported with these outcomes are 5, 4, and 4, respectively.

Dropout is relevant to a feasibility study because it describes whether participants remained in the study through the specified observation period. A count is a direct description of the number of observed discontinuations. It does not, by itself, explain why participants left, whether they completed particular treatment activities, or whether dropout differed reliably between arms. The registry record does not report reasons for individual withdrawals or a separate adherence measure.

Counts also depend on the number of participants observed in each arm. They should therefore be read alongside the corresponding group sizes, rather than as standalone comparisons. A count is not a rate, and no dropout rates or uncertainty intervals are reported in the registry record. The endpoint is about study discontinuation, not change in depression symptoms. PHQ-9 change from baseline to Session 4 is listed separately as a secondary outcome; its result is not included in the posted outcome information described for the primary endpoint.

Using dropout as a feasibility measure can help characterize retention in a study, but a small set of counts does not establish why retention was as observed or how well the approach would work in another setting. The registry record does not report a prespecified threshold for judging feasibility.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is a binary event recorded for each participant: dropout or no dropout during the specified timeframe. The registry presents the number of participants with the event as a count for each arm. In the virtual reality arm, the count is 2 among a group size of 5; in the real-life arm, it is 1 among 4; and in the waitlist arm, it is 0 among 4. The counts are the posted observations, not estimates from a statistical model.

What does a count comparison tell us?

A count states how many participants experienced the event. It does not adjust for group size, account for differences in participant characteristics, or quantify uncertainty. A common way to compare binary outcomes is to estimate an event proportion in each group and then calculate a contrast, such as a risk difference or ratio. Those estimates require a clearly specified denominator and an analysis plan. No comparative estimates of that kind are reported in the registry record, and none should be inferred solely from the displayed counts.

For a formal comparison, analysts also need to specify the statistical method, the contrast of interest, and how uncertainty will be represented. A confidence interval is a range of values compatible with the data under the method’s assumptions; it communicates precision as well as the direction and scale of an estimate. A p-value, when used, describes how unusual data at least as extreme would be under a specified null model. Neither is a substitute for the estimated effect, and neither can be reconstructed from the posted primary outcome information without making additional choices. No confidence interval or p-value for the primary outcome is reported in the registry record.

Why is the comparison not estimated?

The registry record indicates that values are missing or not estimable for a two-arm comparison of this binary endpoint. The study has three arms, and no sponsor analysis or arm-to-arm inferential result is listed for the primary endpoint. The registry also lists zero analyses. Consequently, the primary outcome supports a descriptive account of the reported counts, not a reported statistical test or effect estimate. The sponsor’s own analysis does not take precedence in this case because no such analysis is reported for the comparison.

Counts in small groups can have substantial sampling uncertainty. General statistical reasoning cautions against treating a difference between observed counts as proof of a difference in the populations from which participants were drawn. The uncertainty is especially important when groups are small, because a change in the outcome of an individual participant can noticeably alter a group’s observed count. This general point does not quantify the uncertainty in these particular groups; the registry record gives no interval estimate for that purpose.

What about over-dispersion?

Over-dispersion is a general concern when a model assumes less variability than the data contain. It is most relevant to models for repeated or clustered observations, where outcomes may vary more than the model expects. The primary endpoint is reported as arm-level participant counts, without a fitted model or repeated-event analysis. The registry record does not report a model that would allow over-dispersion to be assessed, so it is not possible to determine whether it affected the analysis.

How should planning figures be read?

Sample-size calculations are planning illustrations, not evidence that a trial had a particular probability of detecting an effect after it was conducted. The registry’s stated planning assumptions describe a two-sided significance level of 0.05, equal allocation between two arms, a normal approximation, and an effect expressed as a fraction of the observed effect. Those assumptions do not describe the original trial design. The primary comparison is not estimable as a two-arm binary comparison in the registry record, and no participant-per-arm planning figures are reported there. Any planning illustration based on an observed effect would also inherit uncertainty from that observed effect; it should not be called observed power.

General statistical practice distinguishes a prespecified analysis from a recalculation based on posted summary statistics. A recalculation is an unadjusted derivation from the posted data and may not match a sponsor’s analysis, which could use a different population, model, or adjustment. No recalculated arm comparison is reported for this endpoint. The record does not identify an adjustment strategy or provide a sponsor model that could be compared with a simple calculation.

Finally, the PHQ-9 outcome is distinct from the dropout endpoint. The registry lists change in PHQ-9 scores from baseline to Session 4 as a secondary outcome, but a result for that outcome is not reported alongside these primary counts. No symptom estimate, standard error, or interval can therefore be interpreted from the posted primary endpoint data. The distinction matters: retention counts cannot substitute for symptom results, and symptom change would not by itself explain why participants discontinued.

10. Exact Estimates for Each Arm

No two-arm comparison is shown for this trial, but each arm's result can still be summarized with exact interval estimates computed from the posted counts. The Clopper-Pearson interval is exact and conservative; the Wilson interval is a common closer-to-nominal alternative; the Bayesian credible interval uses a uniform Beta(1, 1) prior. For small groups the three can differ noticeably.

R code and output

arm    <- c("Virtual Reality Behavioral Activation", "Behavioral Activation in Real-life", "Waitlist Control")
n      <- c(5, 4, 4)
events <- c(2, 1, 0)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  Virtual Reality Behavioral Activation  5  2  40.0  5.3  85.3  11.8  76.9  11.8  77.7
ARM  Behavioral Activation in Real-life  4  1  25.0  0.6  80.6  4.6  69.9  5.3  71.6
ARM  Waitlist Control  4  0  0.0  0.0  60.2  0.0  49.0  0.5  52.2

11. Multiplicity, Missing Data and Analysis Populations

The registry lists one secondary outcome: change in depression scores on the PHQ-9 from baseline to Session 4, compared across three study arms. It also lists 8 outcome measures and zero analyses. The posted primary outcome is dropout, while the PHQ-9 measure addresses symptom change. These are different questions and should not be treated as interchangeable evidence.

When a study evaluates multiple outcomes or makes multiple comparisons, the chance of at least one apparently notable result can increase. Statistical methods can address this issue by defining a primary comparison in advance or adjusting inference across a family of tests. This is a general statistical concern. The multiplicity strategy, if any, is not reported in the registry record. The record also does not specify which outcomes were treated as confirmatory, whether arm-to-arm comparisons were planned for each outcome, or whether any adjustment for multiple testing was used.

The analysis population is not reported in the registry record. It is not specified whether the outcome counts include all randomized participants, only participants with observed data, or another defined group. The record does not report an intention-to-treat approach or a per-protocol approach. The enrollment figure is 20, while the outcome group sizes are 5, 4, and 4; the record does not explain this relationship.

Missing-data handling is also not reported in the registry record. There is no description of how participants with unknown dropout status were treated, whether any outcome observations were imputed, or how missing PHQ-9 assessments were handled. Reasons for withdrawal are not reported, so the outcome count does not distinguish among causes of discontinuation. The record gives no sensitivity analysis for missing outcomes and no analysis plan describing how incomplete follow-up would affect inference.

These omissions limit interpretation of both the primary and secondary outcomes. They do not demonstrate that an inappropriate method was used; they mean the registry record does not document the relevant methods. Without the analysis population, missing-data approach, and multiplicity plan, readers cannot determine how the posted counts or any future outcome estimates relate to all randomized participants or to the study’s full set of planned questions.

12. 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 0% reporting threshold.

The adverse-event information is reported separately for the three study groups. For virtual reality behavioral activation, 0 serious adverse events affected participants among 5 at risk, and deaths were 0. For real-life behavioral activation, 0 serious adverse events affected participants among 4 at risk, and deaths were 0. For waitlist control, 0 serious adverse events affected participants among 4 at risk, and deaths were 0. The record also reports 0 other adverse events in each group, using the same group sizes at risk.

These entries describe the events recorded in the registry. A zero count means no event in that category was reported for the listed group; it does not establish that there was no possibility of harm or that every relevant event was captured. The registry does not list serious-event terms, and no such terms appear in the adverse-event information. The record does not explain how adverse events were solicited, defined, monitored, or attributed to an intervention.

Safety counts should be read with their group sizes and alongside the study’s design and follow-up period. The reported groups are small, and the adverse-event information alone cannot characterize uncommon events or establish comparative safety. The registry record does not report a formal statistical comparison of adverse events between arms. These figures describe recorded events; they are not a medical recommendation or a complete assessment of safety.

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

The primary result is a set of observed dropout counts: 2 participants in virtual reality behavioral activation, 1 in real-life behavioral activation, and 0 in waitlist control. The associated group sizes are 5, 4, and 4. These observations describe study discontinuation for any reason over the three-week timeframe. The registry record does not report an inferential comparison, an uncertainty interval, or an estimate of the size of an arm-to-arm effect.

The result does not establish that one study arm had a higher or lower underlying dropout probability than another. Observed counts can differ even when the underlying probabilities are similar, particularly in small groups. Conversely, the absence of a reported statistical test does not establish that the arms were similar. No equivalence framework or prespecified threshold for an important difference is reported in the registry record.

The study’s stated aim included safety and feasibility, and dropout can inform retention. But retention is only one dimension of feasibility. The record does not report whether participants completed the planned activities, how acceptable they found the intervention, or why anyone discontinued. It also does not give a predefined criterion for declaring the intervention feasible.

The result does not answer the secondary clinical-efficacy question. The registry lists PHQ-9 change from baseline to Session 4 as a secondary outcome, but the primary dropout counts do not show symptom change. A PHQ-9 result for that comparison is not reported in the registry record. The posted data therefore do not establish whether depression scores changed or whether any change differed between arms.

Similarly, the safety entries report zero serious adverse events and zero deaths in the listed groups, but those observations do not prove that an intervention is risk-free. The record does not describe adverse-event ascertainment or provide a statistical safety comparison. Overall, the available results support a narrow description of recorded discontinuation and adverse events. They do not settle comparative efficacy, comparative safety, or broader feasibility.

14. Why This Trial Matters Statistically

This study is statistically notable because its stated purpose combines feasibility and safety questions with an exploratory question about clinical efficacy, while the posted primary endpoint is a count of study dropouts. These outcomes address different aspects of a trial. Retention can describe whether participants remained enrolled, whereas PHQ-9 change is intended to describe symptom change. Keeping those questions distinct prevents a retention result from being interpreted as an efficacy result.

The study also illustrates why descriptive counts and inferential comparisons are not the same. The registry reports arm-specific counts and group sizes, but no comparative estimate or measure of precision. A reader can describe what was observed, but cannot assess how uncertain an arm contrast would be from the registry record. This distinction is especially important for a study with small reported outcome groups, where a small number of observations can strongly influence the counts.

Random allocation is listed, and the parallel design places the interventions in separate arms. Yet a randomized design does not by itself communicate the analysis population, missing-data procedure, or statistical method. The registry record does not report those details. Transparent reporting of these choices helps readers understand what population an estimate represents and how incomplete observations affect its interpretation.

The study’s three-arm structure also calls for clarity about the comparisons of interest. The record lists a secondary PHQ-9 comparison across the three arms but does not report an analysis result or multiplicity strategy. Prespecifying which comparisons are primary and explaining how multiple outcomes are handled can help readers distinguish planned inference from exploratory findings.

These statistical considerations matter for interpreting early-stage research. The posted counts may contribute to understanding retention in this study, but they do not answer every question motivating the trial. Clear separation of descriptive observations, inferential estimates, and planning assumptions is essential when drawing conclusions from a small randomized study.

15. 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.

The primary outcome information has important reporting limits. The registry gives dropout counts and group sizes, but no reasons for discontinuation, arm comparison, uncertainty interval, or statistical test. It does not report a criterion for judging whether the observed retention met a feasibility target. The counts therefore describe the recorded outcome without resolving why participants left or whether the arms differed beyond the observed data.

Enrollment is listed as 20, while the primary outcome groups contain 5, 4, and 4 participants. The registry record does not explain the relationship between the total enrollment and those arm-specific outcome groups. It also does not report a separate randomized count for each arm or clarify whether the outcome groups reflect all randomized participants. This makes the completeness of the primary outcome accounting difficult to assess.

Additional trial details are limited in the registry record. It does not report participant characteristics, baseline PHQ-9 values, intervention adherence, treatment fidelity, or reasons for missing assessments. The arm descriptions specify clinician meetings once a week for three weeks (4 sessions), but do not fully describe the intervention content or implementation. The masking field is listed as none; details about how outcomes were assessed are not reported.

The record lists PHQ-9 change as a secondary outcome but does not report its result in the primary outcome information. It also does not describe an analysis population, missing-data strategy, multiplicity adjustment, or a detailed statistical plan. No comparative safety analysis is reported. These limits constrain conclusions about symptom change, treatment delivery, retention, and adverse events. They are limits of what the registry record documents and do not establish what analyses or procedures may have occurred outside that record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Virtual Reality Behavioral Activation: An Intervention for Major Depressive Disorder — NCT04268316.
• Registry record last updated 2021-06-02; status last verified 2021-05.
• Results first posted 2021-05-10.
• 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.

19. Record Summary

NCT04268316 is a completed randomized, parallel-group study sponsored by Stanford University. It compares virtual reality behavioral activation, real-life behavioral activation, and waitlist control for major depressive disorder. The registry lists enrollment of 20, no masking, and a study phase of NA. Its stated purposes were to examine safety and feasibility and to explore whether evidence of clinical efficacy existed.

The primary outcome counts participants who discontinued for any reason over three weeks. Reported counts are 2 among 5 in virtual reality behavioral activation, 1 among 4 in real-life behavioral activation, and 0 among 4 in waitlist control. No inferential comparison or uncertainty interval is reported. PHQ-9 change from baseline to Session 4 is listed as a secondary outcome, but a result is not reported with the primary outcome information.

The registry also reports zero serious adverse events, zero other adverse events, and zero deaths in each listed group. The analysis population, missing-data methods, multiplicity strategy, and detailed statistical plan are not reported in the registry record. The posted primary results support a descriptive account of dropout counts, not a conclusion about comparative efficacy or safety.