Study of a PST-Trained Voice-Enabled Artificial Intelligence Counselor(SPEAC) for Adults With Emotional Distress (Phase 2): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Study of a PST-Trained Voice-Enabled Artificial Intelligence Counselor(SPEAC) for Adults With Emotional Distress (Phase 2) (NCT05603923), a randomized na study. The primary outcome is Change From Baseline Functional Magnetic Resonance Scan at 18 Weeks.

2. Clinical Question

The study is intended to examine an intervention-delivery question: whether PST delivered by a voice-enabled artificial-intelligence counselor can be evaluated alongside human-coached PST and a delayed-treatment option. It is not a direct comparison of all participants receiving treatment at the same time, because the delayed group is offered Lumen PST only after the 18-week follow-up assessment. The stated primary outcome is a brain-imaging measure, rather than a symptom score. Anxiety, depression, worry, affect, problem-solving, disability, work productivity, and daily mood are named among the secondary outcomes. Their presence does not establish that any group improved or differed from another; the registry record does not report their comparative results.

3. Trial Design

This is a randomized, double-masked, parallel-group study listed as Phase 2 in its title, while its phase field is recorded as NA. The registry lists enrollment as 200. The brief description says approximately 200 participants are assigned among Lumen-coached PST, human-coached PST, and optional delayed Lumen-coached PST. It describes planned group sizes of 100, 50, and 50, respectively; those allocation figures should not be mistaken for a report of how many participants completed follow-up or contributed to each outcome.

The Lumen group is described as completing eight PST sessions using a study iPad at home over 12 weeks, with the first four sessions occurring weekly. The human-coached group begins with an in-person session and completes the remaining sessions by videoconference or phone, according to the brief description. The delayed group can choose to complete Lumen PST after its 18-week follow-up assessment. All participants are described as completing assessments at baseline and 18 weeks after randomization.

The study began on 2023-01-23, had a primary completion date of 2025-04-15, and is listed as completed. The registry names the sponsor as University of Illinois Chicago. It does not report which people were masked, how randomization was generated or concealed, or whether the allocation groups were balanced on baseline characteristics. These details matter when assessing how the design limits bias, but they cannot be inferred from the labels randomized and double-masked alone.

The primary fMRI outcome has data-specific availability: the registry says that 71 Lumen participants, 36 human-coached participants, and 34 delayed-Lumen participants had quality fMRI data at 18 weeks. These figures refer to participants with quality imaging data, not necessarily all people enrolled or all people with any follow-up data.

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.

Overall Study

Analysis population for the primary outcome, as defined in the registry: The number of participants in the Participant Flow module reflects number of participants who provided any data (e.g., HADS data) at 18 weeks. Among them, 71 Lumen Intervention participants, 36 Human coached, and 34 Delayed Lumen coached participants had quality fMRI data at 18 weeks.

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 Change From Baseline Functional Magnetic Resonance Scan at 18 Weeks, measured over Baseline, 18 weeks.

The primary outcome is change from baseline in functional magnetic resonance imaging (fMRI) at 18 weeks, expressed as a beta weight. The registry defines beta weight as a standardized regression coefficient: positive values mean greater activation, and negative values mean less activation in the monitored brain region. The primary neural target is the right dorsolateral prefrontal cortex (DLPFC), a region the study associates with cognitive control. The left DLPFC and bilateral amygdala are designated secondary neural targets; the amygdala is associated in the description with nonconscious threat-related emotional reactivity.

The imaging tasks include viewing facial expressions representing threat-related emotions, loss-related emotion, reward-related emotion, and neutral expressions, as well as a Go-NoGo paradigm. These tasks provide standardized contexts for measuring brain activation. The registry identifies the right DLPFC as the primary target, so a broad description of fMRI change should not be read as though every measured brain region were equally primary.

fMRI is used as a measure of neural activity during specified tasks, not as a direct measure of symptoms or daily functioning. A difference in beta weight would describe an imaging contrast, and its interpretation depends on the task, region, and analysis. The registry record does not report the group-level changes, contrasts, uncertainty intervals, or statistical tests for this endpoint. It therefore does not establish the direction, size, or precision of any group difference.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Statistical methods are described only to the extent that the registry record states them. It identifies the study as randomized, double-masked, and parallel-group, and it identifies a continuous primary outcome measured at baseline and 18 weeks. A detailed statistical analysis plan, including the model used for the primary comparison, is not reported in the registry record. The explanations below describe general statistical concepts; they do not assert that a particular unreported method was used in this study.

What does a change-from-baseline outcome represent?

In general, a change score compares a participant’s measurement at follow-up with that participant’s baseline measurement. A group mean change describes the average change among participants included in that analysis. A comparison between groups asks whether their average changes differ. Such a comparison requires the actual estimates and an account of their uncertainty. The registry names change from baseline at 18 weeks as the endpoint but does not report the group changes or comparative estimates.

For continuous outcomes, a mean is a measure of the centre of the observed values, while a standard deviation describes their spread around the mean. The primary outcome’s dispersion is identified as standard deviation. In general, a standard deviation is not the same as a standard error: the former describes variation among measurements, while the latter describes uncertainty in an estimated mean. A confidence interval communicates a range of values compatible with an estimate under the assumptions of the analysis. No confidence interval for the primary outcome is reported in the registry record.

How should the beta weight be understood?

The registry describes beta weight as a standardized regression coefficient and states that positive values mean greater activation and negative values mean less activation in the monitored region. This sign convention gives direction within the measurement definition, but it does not by itself establish clinical importance. Interpretation also depends on the brain region, task contrast, and comparison being estimated. The right DLPFC is the primary neural target; the left DLPFC and bilateral amygdala are secondary neural targets. The registry does not report the estimated beta weights or the exact group contrasts used to summarize them.

What does randomization contribute?

In general, random assignment is intended to make treatment groups comparable, on average, with respect to measured and unmeasured characteristics at the point of assignment. It supports an unbiased comparison when allocation is implemented as planned and follow-up and analysis are handled appropriately. It does not guarantee identical group characteristics in a particular study, nor does the label alone explain how assignments were generated or concealed. Those implementation details, along with baseline characteristics, are not reported in the registry record.

The parallel design means the groups are assigned to different study strategies rather than each participant cycling through every strategy. The delayed group’s timing differs from the active intervention groups because Lumen PST is offered after the 18-week assessment. This makes the timing and definition of the comparison important: a contrast at 18 weeks would compare the groups as assigned during that period, not compare simultaneous receipt of PST in all groups.

Why does the amount of usable imaging data matter?

In general, fewer usable measurements mean less information for estimating a group mean or contrast, and usually less precision. The registry reports the number of participants in each group with quality fMRI data at 18 weeks. It does not report why other scans were unavailable, whether included participants differed from those without usable scans, or whether missing-data methods were applied. If data availability is related to outcomes or participant characteristics, a complete-case analysis can be biased; the direction and extent of any bias depend on the missingness process and cannot be determined from the counts alone.

General approaches to missing data include analyzing available observations under explicit assumptions, using model-based methods, or conducting sensitivity analyses that examine how conclusions might change under alternative assumptions. The registry record does not report which approach, if any, was used for the fMRI endpoint. It also does not report whether the primary analysis followed an intention-to-treat population, a per-protocol population, or another definition.

What do the planning assumptions mean?

The registry describes sample-size calculations as planning illustrations based on the observed effect, with a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. It says the target effect is a stated fraction of the observed effect. These assumptions describe an illustration, not the trial’s original design justification, and the registry record does not report the trial’s own sample-size rationale. Planning calculations are not observed power and do not show that the study was adequately powered for its primary question. The numerical outputs for those illustrations are not reported in the registry record.

In general, sample-size calculations estimate how much information a future design might need under specified assumptions. Their usefulness depends on the reliability of the effect used, the outcome’s variability, the analysis plan, and the comparison being planned. An illustration based on an observed effect should not be interpreted as an independent confirmation of that effect. It also does not substitute for the observed estimate and its confidence interval.

How should significance and uncertainty be read?

In general, a p-value measures how incompatible the observed data are with a specified null model, under the assumptions of the analysis. It is not the probability that the null hypothesis is true and does not measure effect size or clinical importance. A confidence interval gives information about both the estimated effect and its precision. Neither a p-value nor an interval for the primary comparison is reported in the registry record, so statistical significance and precision cannot be assessed.

10. Multiplicity, Missing Data and Analysis Populations

The study names a primary fMRI outcome and multiple secondary outcomes, and the fMRI description distinguishes a primary neural target from secondary targets. When a study examines several outcomes, brain regions, tasks, or contrasts, the chance of at least one apparently unusual result can increase if each is tested without accounting for the collection of tests. General approaches to multiplicity include designating a primary comparison in advance, adjusting significance thresholds, or interpreting secondary findings as exploratory. The registry record does not report a multiplicity strategy, a hierarchy among all outcome tests, or which contrasts were formally tested.

The title of the primary outcome refers broadly to functional MRI change, while the description specifies the right DLPFC as the primary neural target and names the left DLPFC and bilateral amygdala as secondary targets. It also describes more than one task. The record does not report how task- and region-specific results were combined, which exact contrasts defined the primary analysis, or whether the planned primary outcome was narrowed further in an analysis plan.

Missing-data handling is also not reported. The registry says the participant-flow figures reflect participants who provided any data at 18 weeks, and separately reports the numbers with quality fMRI data at that time. It does not explain missing assessments, scan exclusions, imputation, sensitivity analyses, or whether the analysis included all randomized participants. These details affect interpretation because the analyzed group may differ from the group originally assigned.

It is not reported in the registry record whether interim analyses were planned or conducted, what stopping rules applied, or how protocol deviations were handled. The statistical analysis plan and the precise definition of the analysis populations are not reported in the registry record. These omissions limit assessment of selective reporting, multiplicity, and the assumptions behind any eventual estimate.

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 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pulmonary/respiratory (2.0%).

The registry lists serious adverse-event figures for the Lumen-coached group as 0 among 100 at risk, for the human-coached group as 1 among 50 at risk, and for the delayed-Lumen group as 0 among 50 at risk. Deaths are listed as 0 in all three groups. The leading serious-event term is reported as Pulmonary/Respiratory, 2.0%. The registry record does not explain how that term relates to the group-level serious-event count or provide additional details about the events.

For other adverse events, the registry reports figures of 20 among 100 at risk in the Lumen group, 10 among 50 at risk in the human-coached group, and 5 among 50 at risk in the delayed-Lumen group. These are descriptive registry figures. The record does not report event definitions, severity categories beyond the serious/non-serious distinction, timing, duration, relationship to study interventions, or a statistical comparison. Therefore, the counts alone do not establish that an intervention caused an event or changed risk. They should be read as reported safety information, not as a causal comparison or a complete account of harms.

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

The study addresses a timely methodological question: how to evaluate PST delivered by a voice-enabled virtual coach in comparison with human coaching and a delayed-treatment option. Its randomized, parallel-group design provides a framework for comparing assigned strategies. The primary endpoint is distinctive because it focuses on task-related brain activation, with right DLPFC activity designated as the primary neural target. Secondary outcomes cover symptoms and functioning, which are different kinds of measures and should not be treated as interchangeable with fMRI.

The record does not report the primary comparative estimates, uncertainty intervals, or statistical tests. It therefore cannot establish whether groups differed in fMRI change, how large any difference was, or how precisely it was estimated. The same limitation applies to the listed secondary outcomes: their names indicate planned measurement domains, not findings. A completed study status does not itself supply the missing results.

Likewise, the reported adverse-event figures are descriptive and do not establish causation. The registry does not give enough information to evaluate event timing, relatedness, or comparative statistical uncertainty. The group counts with quality fMRI data describe availability of imaging observations; they are not estimates of intervention effect and do not by themselves explain missingness.

These limits do not make the design irrelevant. The study record identifies the population, delivery approaches, follow-up timing, primary imaging target, and several secondary domains. Those details help readers understand what the trial set out to compare. But assessing what the observed data mean requires outcome estimates, analysis methods, and uncertainty information. Until those are reported, the defensible interpretation is about the question and design, not whether any treatment was effective, ineffective, or equivalent to another.

13. Why This Trial Matters Statistically

Clinical studies increasingly evaluate not only whether an intervention is delivered, but also how its delivery method can be compared with another approach. This trial’s three-group structure distinguishes Lumen-coached PST, human-coached PST, and a delayed Lumen option. That distinction matters statistically because the delayed group has a different treatment timing at the 18-week assessment. The interpretation of any comparison depends on which groups are contrasted and what each group has received by that time.

The study also pairs a neural primary outcome with secondary measures of symptoms, affect, problem-solving, disability, work productivity, and mood. These outcomes represent different domains. A change in brain activation should not automatically be interpreted as a change in symptoms or everyday functioning; each outcome requires its own estimate and interpretation. Naming several domains makes transparent that the study’s assessment is broader than imaging alone, while also raising questions about the planned analysis hierarchy and multiplicity.

For biostatistical interpretation, the reported quality-fMRI counts are important because usable imaging data determine how much information is available for the primary endpoint. The registry record does not explain missing scans or analysis populations, so data availability cannot be equated with complete follow-up or unbiased comparisons.

Finally, the record illustrates why trial status, design labels, and outcome results must be distinguished. A randomized and completed study may still have no comparative effect estimates reported in its registry entry. Without estimates and measures of uncertainty, readers cannot judge the size or precision of the primary result. Clear reporting of those elements, alongside analysis and missing-data methods, is necessary for statistical appraisal.

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.

The main limitation for interpreting outcomes is that the registry record does not report comparative results for the primary fMRI endpoint or the listed secondary outcomes. It therefore offers no estimates, confidence intervals, or statistical tests with which to assess direction, size, or precision. The completed status cannot fill that gap.

The primary outcome description identifies a primary brain region and secondary regions, and describes more than one imaging task. The exact primary contrast and the handling of task- and region-specific measures are not reported in the registry record. A multiplicity plan is also not reported. These omissions make it difficult to determine how a single primary result was distinguished from a wider set of possible imaging comparisons.

Although the registry reports quality fMRI data at 18 weeks for participants in each group, it does not explain why other scans were unavailable, whether those with usable scans differed from those without, or how missing outcome data were handled. The analysis populations and any sensitivity analyses are not reported in the registry record. These issues are especially relevant when the endpoint depends on successful imaging and quality review.

Masking is listed as double, but the people masked and the procedures used to maintain masking are not reported. Details of sequence generation, allocation concealment, baseline group comparability, adherence, and protocol deviations are also not reported in the registry record. The safety entries provide counts but not event timing, relatedness, or fuller definitions. These limitations leave important parts of both the outcome and safety interpretation unresolved.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Study of a PST-Trained Voice-Enabled Artificial Intelligence Counselor(SPEAC) for Adults With Emotional Distress (Phase 2) — NCT05603923.
• Registry record last updated 2026-09-23; status last verified 2026-08.
• Results first posted 2026-09-23.
• 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

NCT05603923 is a completed randomized, double-masked, parallel-group study sponsored by University of Illinois Chicago. It concerns adults with mild-to-moderate, untreated depression and/or anxiety and compares Lumen-coached PST, human-coached PST, and an optional delayed Lumen-coached group. The registry lists enrollment as 200 and assessment at baseline and 18 weeks.

The primary endpoint is change from baseline in fMRI beta weight at 18 weeks, with right DLPFC activation identified as the primary neural target. The record reports which participants had quality fMRI data but does not report comparative outcome estimates or their uncertainty. Multiple secondary outcomes are listed, but their results are not reported in the registry record.

Safety entries include serious and other adverse-event figures and no listed deaths. The registry does not report enough detail to interpret causality or statistical differences in those figures. Analysis plans, missing-data methods, and multiplicity procedures are not reported in the registry record.