Real-time fMRI and Neurofeedback of Brain Networks Mediating Trauma Memory Recall in PTSD: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Real-time fMRI and Neurofeedback of Brain Networks Mediating Trauma Memory Recall in PTSD (NCT02500719), a non-randomized na study. The primary outcome is Patient Emotional Response to Volitional Engagement and Disengagement of Emotional Arousal as Measured Using Support Vector Machine Decodings When the Decoding is Provided as Real-time Neurofeedback Guidance or Not..

2. Clinical Question

The study’s stated aim was to improve understanding of brain mechanisms involved in psychological treatments for PTSD. Its initial healthy-participant pilot was intended to refine the method before implementation with participants with PTSD symptoms. That staged purpose matters when interpreting the comparison: the healthy group and PTSD group represent different phases and populations, rather than randomized groups designed to isolate a treatment effect.

3. Trial Design

The registry describes a completed, non-randomized, parallel-design study with no masking. Enrollment was 30. The study concerned post-traumatic stress disorder and used a computational model based on a real-time support vector machine. The sponsor was the University of Arkansas. The record lists the study start as 2015-08-15 and primary completion as 2019-04.

The study had two participant groups and a staged purpose. Healthy participants were enrolled first in a pilot phase to refine application of the real-time fMRI algorithm. Participants with symptoms of PTSD were enrolled in an implementation phase to evaluate guidance of brain networks thought to mediate emotional arousal, including whether participants could learn volitional control of those networks. This sequencing is central to understanding why the groups should not be interpreted as interchangeable treatment and control arms.

Allocation is listed as non-randomized, and the registry explicitly cautions that the arms are not a like-for-like control comparison. No masking was used. The record does not report the assignment procedure, detailed session schedule, or whether every participant experienced both visible and hidden feedback conditions. Those details should not be inferred from the broad description of the outcome.

The registry lists one primary outcome measure and one analysis. It does not report a protocol-based sample-size justification, a planned interim analysis, or enough statistical-plan detail to determine the model used for that analysis. The record includes planning assumptions for an illustrative sample-size calculation based on an observed effect, but that illustration is not the study’s design justification and is not evidence of observed power.

Although the study is marked completed, completion status alone does not establish that every planned measurement was obtained or that every enrolled participant contributed to the primary analysis. The number completing the endpoint and the handling of incomplete observations are 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.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Completing subjects.

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 Patient Emotional Response to Volitional Engagement and Disengagement of Emotional Arousal as Measured Using Support Vector Machine Decodings When the Decoding is Provided as Real-time Neurofeedback Guidance or Not., measured over Real-time within the measurement of functional MRI (within 10 seconds of functional MRI volume acquisition and reconstruction).

The primary endpoint is a continuous measure: the mean support vector machine decoding, expressed as a standard score. The decoding represents the Euclidean distance and direction of an fMRI data volume relative to the support vector machine decision hyperplane. Positive distances denote engagement of emotional arousal; negative distances denote disengagement. The magnitude of the distance represents the magnitude of the decoded engagement or disengagement.

The endpoint is measured during volitional engagement or disengagement of emotional arousal. The registry describes the measurement as occurring in real time, within 10 seconds of fMRI volume acquisition and reconstruction. Decodings could be shown to participants as a visual representation of distance or hidden, with the visual display remaining stationary when feedback was hidden.

Using the decoding as an endpoint connects the measurement to the study’s central methodological question: whether participants can exercise volitional control over the targeted brain-network signal, and whether visible feedback provides information during that process. A mean summarizes the average decoded signal for the endpoint population, while the standard deviation describes variability in values. Those summaries do not by themselves establish why the values differ, whether any difference is attributable to feedback, or how precisely an estimated contrast is known.

The population for the endpoint is described as completing subjects. The registry does not report the endpoint means, standard deviations, contrast estimate, confidence interval, or test result. Therefore, the observed direction, size, and precision of the primary outcome cannot be described from this record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry describes a continuous primary outcome summarized by a mean and standard deviation. A continuous measure can take values along a scale rather than being limited to categories. Here, the scale is a standard score based on support vector machine decodings of fMRI data. The mean describes the average decoding in the analysis population; the standard deviation describes the spread of individual values around that average. These are descriptive summaries, not by themselves tests of a difference or measures of its uncertainty.

What does the decoding represent?

A support vector machine uses a decision hyperplane to distinguish patterns in data. In this study, each fMRI volume was represented by its Euclidean distance and direction relative to that hyperplane. The registry assigns a substantive interpretation to the direction: positive values denote engagement of emotional arousal and negative values denote disengagement. The distance’s magnitude is described as representing the magnitude of volitional engagement or disengagement. This is an operational interpretation of the model output; it should not be confused with a direct measurement of a clinical outcome.

Real-time neurofeedback refers to returning information about the decoding to participants while the measurement process is underway. The record describes a visual representation of distance when feedback was shown. When decodings were hidden, the visual representation remained stationary. This contrast makes the feedback display part of the study procedure, but the registry does not provide enough detail to establish the exact sequence or number of exposure conditions for each participant.

What can a statistical comparison establish?

A comparison can estimate how an outcome differs between conditions or groups. The estimate describes the observed contrast; a confidence interval, when reported, conveys uncertainty around that estimate. A p-value, when reported, describes how compatible the data are with a specified null model under the assumptions of the analysis. Neither measure alone establishes clinical importance or proves that a difference was caused by the intervention. The registry does not report an effect estimate, confidence interval, p-value, or the statistical test or model used for the listed analysis.

The non-randomized design is especially relevant. Random allocation, when used, is intended to make groups comparable on average with respect to factors that could influence outcomes. This study is explicitly listed as non-randomized, and the registry cautions against treating its arms as a like-for-like control comparison. The staged design further distinguishes the healthy-participant pilot from implementation with participants with PTSD symptoms. Thus, any contrast between groups would require careful interpretation and could not automatically be attributed to neurofeedback.

How should sample-size illustrations be read?

The registry describes a planning illustration based on the observed effect, using a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. For non-time-to-event outcomes, its planning quantities are expressed as participants per arm; the record notes that time-to-event quantities would instead refer to total events. The calculation is an illustration of sample-size requirements under stated assumptions, not the trial’s own design justification. It is not observed power, and it does not show that the study had a specified probability of detecting an effect. Numerical planning results are not reported in the registry record.

In general, over-dispersion means that observed variability is greater than a statistical model assumes. If it is present and not accommodated, standard errors and uncertainty estimates may be too small, making results appear more precise than they are. Whether over-dispersion is relevant depends on the outcome and model. The registry does not report the analysis model or an assessment of over-dispersion, so it is not possible to say whether it arose or how it was handled in this study.

Interpretation also depends on which participants and measurements enter the analysis. The primary population is specified as completing subjects, but the number of completers and the rules for including observations are not reported. Without those details, neither the available information per condition nor the influence of incomplete measurements can be evaluated. These are distinct questions from the overall enrollment figure.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome measure and one analysis. It does not describe a multiplicity adjustment, a hierarchy of outcomes, or a prespecified strategy for controlling the chance of false-positive findings across multiple tests. The record also does not report whether additional secondary or exploratory outcomes were analyzed. With only the listed information, the scope of any multiplicity issue cannot be established.

Missing-data handling is not reported in the registry record. The endpoint population is described as completing subjects, but the record does not give the number completing the endpoint, reasons for non-completion, or the amount of missing outcome data. It does not state whether incomplete observations were excluded, imputed, or handled through a model using available observations. These distinctions can affect both the target population represented by an analysis and the uncertainty of its estimates.

The enrollment of 30 should not be read as the endpoint analysis count. Enrollment records how many participants entered the study, whereas the endpoint description refers to subjects who completed the relevant measurement. The registry does not report whether all enrolled participants completed the endpoint or how many contributed to each feedback condition.

The analysis population beyond the phrase “completing subjects” is not specified. The record does not describe an intention-to-treat population, a per-protocol population, or criteria for excluding participants or scans. Nor does it state how data from the pilot phase and implementation phase were handled in the primary analysis. Without such information, the population to which a result would apply cannot be reconstructed from 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 0% reporting threshold.

The registry reports serious adverse events and other adverse events separately for the healthy-participant and PTSD-participant groups. For healthy participants, 11 were listed at risk for serious events and 11 at risk for other events; zero serious events and zero other events were reported. For PTSD participants, 9 were listed at risk in each category; zero serious events and zero other events were reported. Deaths were reported as zero in both groups.

These figures describe the events recorded in the registry for the participants listed at risk. They do not establish that no symptoms, discomfort, or other experiences occurred, because the record does not provide event terms or a detailed account of monitoring and reporting. The registry lists no top serious-event terms. The number of participants at risk is not a measure of exposure time, and these group counts should not be used to compare safety between the non-randomized study arms.

The safety figures also do not establish that an intervention is safe or unsafe in other populations or settings. They are a descriptive account of the registry’s adverse-event reporting for this study. The record does not report further detail about event definitions, collection procedures, or the duration of safety follow-up.

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

The record describes a study of a real-time fMRI decoding method and feedback display, with an initial healthy-participant pilot followed by implementation with participants with PTSD symptoms. Its primary measurement was a support vector machine decoding intended to reflect engagement or disengagement of emotional arousal. The study therefore addresses a measurement and volitional-control question, not solely a clinical treatment outcome.

The registry does not report the primary outcome values, an estimated difference, or uncertainty around any estimate. It is not possible to determine from this record whether decodings differed when feedback was shown or hidden, how large a difference might have been, or how consistent measurements were across participants. The word “completed” describes study status, not the direction or success of an outcome.

The registry’s stated longer-term aim was to improve understanding of whether volitional control of brain areas might improve therapeutic outcomes. That aim should not be mistaken for a result: therapeutic outcomes are not reported as the primary endpoint described in the record. A decoding value is a model-derived signal, and a change in that signal would not, on its own, demonstrate improvement in PTSD symptoms or treatment benefit.

The non-randomized design and staged groups further limit causal interpretation. Healthy participants took part in a pilot phase, whereas participants with PTSD symptoms took part in an implementation phase. Those groups are not a randomized comparison, and the registry explicitly cautions that they are not like-for-like controls. The record can therefore support an account of the study’s purpose, procedures, and reported adverse-event counts, but not a causal claim about the effect of neurofeedback or a clinical claim about treatment effectiveness.

13. Why This Trial Matters Statistically

This study is statistically interesting because it uses a model-derived, continuous signal as an outcome while testing a real-time feedback procedure. The decoding is not simply a symptom score: it is a signed distance from a support vector machine decision hyperplane, with direction and magnitude assigned meanings related to emotional arousal. Understanding how such a measure is constructed is essential before interpreting an average or comparing conditions.

The staged structure also illustrates a common distinction between refining a method and evaluating it in a target population. Healthy participants were enrolled first so the method could be refined before implementation in participants with PTSD symptoms. That sequence can be informative for development, but it does not produce randomized, like-for-like comparison groups. The registry’s explicit warning about the arms is important when considering any apparent contrast.

Another statistical issue is the distance between a study objective and the endpoint actually reported. The broad aim concerns brain mechanisms and the possible relationship between volitional control and therapeutic outcomes, while the registered primary measure is a real-time fMRI decoding. A model signal may help investigate a proposed mechanism, but it is not equivalent to a clinical outcome. Separating these questions avoids interpreting a technical measurement as evidence of patient benefit.

Finally, this record shows why completion status and enrollment do not substitute for results or analysis details. The registry does not report the endpoint estimates, their precision, the analysis method, or the number of endpoint completers. Without those elements, statistical interpretation must remain limited to the design and measurement description. Reporting those details would be necessary to judge the observed signal and its uncertainty.

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 principal limitation for statistical interpretation is that the registry record does not report numerical results for the primary endpoint. Means, standard deviations, a contrast estimate, a confidence interval, and a p-value are not reported. The method used for the listed analysis is also not described, so its assumptions and handling of repeated measurements or participant-level variation cannot be evaluated.

The study is non-randomized and uses groups enrolled at different stages: healthy participants in a pilot and participants with PTSD symptoms in an implementation phase. The registry cautions that the arms are not a like-for-like control comparison. Therefore, group differences would not by themselves identify an effect of feedback. The assignment process and detailed exposure sequence are not reported.

The endpoint is defined for completing subjects, but the number of completers and the amount of missing endpoint data are not reported. The registry does not describe missing-data methods, analysis-population criteria, or exclusions. Overall enrollment of 30 does not resolve those uncertainties.

The listed outcome is a model-derived decoding rather than a reported clinical symptom endpoint. The record does not report whether a change in decoding corresponded to therapeutic outcomes. In addition, the protocol’s sample-size justification, interim-analysis plan, and detailed statistical plan are not reported in the registry record. The planning illustration described in the record should not be treated as the study’s original design justification or as observed power.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Real-time fMRI and Neurofeedback of Brain Networks Mediating Trauma Memory Recall in PTSD — NCT02500719.
• Registry record last updated 2021-08-20; status last verified 2021-04.
• Results first posted 2021-08-20.
• 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

NCT02500719 is a completed, non-randomized, unmasked parallel study sponsored by the University of Arkansas, with enrollment of 30. It investigated real-time fMRI support vector machine decodings and neurofeedback related to emotional arousal. Healthy participants took part in an initial pilot phase, followed by an implementation phase with participants experiencing PTSD symptoms. The registry lists one continuous primary outcome and one analysis.

The outcome represents the mean decoding value, with positive and negative directions interpreted as engagement and disengagement of emotional arousal. The registry does not report numerical primary-outcome results, an effect estimate, or its precision. It also does not report the analysis method, endpoint completer count, missing-data approach, or protocol-based sample-size justification. Adverse-event figures list zero serious events, zero other events, and zero deaths in both participant groups. These are registry descriptions; they do not establish a treatment effect or a clinical benefit.