Adaptive Hip Exoskeleton for Stroke Gait Enhancement: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Adaptive Hip Exoskeleton for Stroke Gait Enhancement (NCT05536739), a na na study. The primary outcome is Temporal Convolutional Network (TCN) Model Performance (Joint Moment Accuracy).

2. Clinical Question

The study concerns people who have previously had a stroke and have gait impairment. The registry describes walking and other movement tasks with a robotic hip exoskeleton, including treadmill walking and tasks such as over-ground walking, stairs, and ramps. It does not provide further eligibility details in the registry record.

The registry also lists walking tests, balance and confidence measures, and a stroke impact measure as secondary outcomes. These measures broaden the outcomes being assessed, but the registry record does not report their numerical results. The primary endpoint is measured only while the exoskeleton is being worn, so it is a device-model performance measure rather than a direct comparison of that endpoint between the device and baseline conditions.

3. Trial Design

The registry identifies this as a completed, single-group study with enrollment of 12. The sponsor is Georgia Institute of Technology. The recorded study start is 2025-05-21, and primary completion is recorded as 2025-08-29. The phase is listed as NA. Allocation is listed as NA, masking as NONE, and the model as SINGLE GROUP.

The study description says participants are tested with the hip exoskeleton and at baseline. It describes walking while wearing the device on a specialized treadmill, as well as other movement tasks. Although baseline and device conditions appear in the arm description, the registry explicitly identifies one study arm and does not describe a randomized, like-for-like control comparison. The order of conditions, any washout period, and the procedures used to standardize or compare those conditions are not reported in the registry record.

The primary endpoint is the temporal convolutional network model's joint-moment accuracy, expressed as a coefficient of determination. Its timeframe is 5 Days, and the endpoint description says it is measured only during the hip exoskeleton intervention. The registry lists 13 outcome measures and 12 analyses. It also names secondary measures including the 10 Meter Walk Test (Self-selected), the Timed up and go (TUG), the 6 Minute Walk Test, the Modified Stroke Impact Scale, Modified Activities-specific Balance Confidence, and Fast Self-selected Walking Speed.

The registry record does not describe a randomized allocation process or masking. It also does not report the planned analysis sequence, the timing of individual assessments beyond the primary timeframe, or the rationale for the enrollment target. Those details should not be inferred from the completed status or from the single-group description.

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: This outcome measure was only measured during the hip exoskeleton intervention as it can only be measured if wearing a hip exoskeleton as it is representative of the error with which the deep learning model embedded into the hip exoskeleton microprocessor predicts hip joint moments in stroke patients.

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 Temporal Convolutional Network (TCN) Model Performance (Joint Moment Accuracy), measured over 5 Days.

The primary endpoint is Temporal Convolutional Network (TCN) Model Performance (Joint Moment Accuracy), measured in coefficient of determination. It evaluates how closely the hip exoskeleton's embedded deep learning model predicts hip joint moments compared with ground-truth measurements. The registry describes ground truth as coming from a laboratory-grade force plate system and inverse dynamics calculations.

The coefficient of determination, written as R² in the registry description, expresses the correspondence between predicted and observed values in this context. The record states that values closer to 1.0 indicate better correlation between predicted and actual hip joint moments. This provides a model-performance measure during the device intervention: it concerns the accuracy of joint-moment estimates, not a direct measure of walking ability or a comparison of clinical outcomes between randomized groups.

The endpoint is used because the model's estimates are part of the device's operation, and their agreement with laboratory measurements can characterize its predictive accuracy during the assessed tasks. The registry says improved values represent better assistive capabilities, but it does not report a numerical result for this endpoint. The timeframe is 5 Days. The record also states that this outcome is measured only while participants wear the exoskeleton; therefore, it cannot serve as a device-versus-baseline endpoint comparison. The registry record does not report the observed mean, standard deviation, confidence interval, or a test of a treatment contrast for the primary outcome.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry identifies the primary outcome as continuous and specifies a mean parameter with standard deviation as the dispersion measure. For a continuous outcome, a mean describes the central level of the measured values, while a standard deviation describes how spread out individual observations are around that mean. These are general statistical interpretations. The registry record does not report the primary outcome's mean or standard deviation, so neither the measured level nor its variability can be assessed from the posted results.

What does the coefficient of determination represent?

The endpoint compares model-predicted hip joint moments with ground-truth measurements from a laboratory-grade force plate system and inverse dynamics calculations. In general, a coefficient of determination summarizes how well a set of predictions corresponds to observed values under the specified calculation. The registry describes higher R² values, closer to 1.0, as indicating better correlation between predicted and actual hip joint moments. It also frames this measure as an assessment of the model embedded in the exoskeleton, rather than as a direct patient-centered walking outcome. The precise implementation of the calculation, including any task-specific aggregation, is not reported in the registry record.

How would uncertainty be assessed?

In general, a confidence interval gives a range of values compatible with the data and the statistical method used, making it useful for judging the precision of an estimate. A narrow interval indicates greater precision than a wide interval, all else being equal. A p-value addresses the compatibility of data with a specified statistical null model, but it does not by itself measure the size or importance of an effect. The registry record does not report a confidence interval, p-value, standard error, or other uncertainty estimate for the primary endpoint. No conclusion about precision or statistical evidence can be drawn from those quantities because they are absent.

What does the single-group structure mean?

The study is listed as SINGLE GROUP, with a single arm that includes hip exoskeleton and baseline testing. In general, repeated conditions within a group are not equivalent to a randomized comparison between separate treatment and control groups. Interpretation of condition differences can depend on the order of testing, changes over time, task conditions, and how repeated measurements are handled. Those design and analysis details are not reported in the registry record. The primary endpoint is explicitly limited to the exoskeleton condition, so the registered primary measure does not itself estimate a baseline-to-device contrast.

How should planning calculations be read?

The registry's sample-size assumptions describe a planning illustration based on a fraction of 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, such illustrations are expressed as participants per arm. Such calculations are general planning aids: they indicate sample sizes under stated assumptions for detecting a specified effect. They are not observed power, and they do not show that a study achieved a particular probability of detecting an effect. The observed effect and numerical planning estimates are not reported in the registry record, so specific planning rows cannot be interpreted from the information available.

The registry lists 13 outcome measures and 12 analyses, but does not provide the statistical methods for those analyses. The analysis model, handling of repeated measures, covariate adjustment, distributional checks, and any planned sensitivity analyses are not reported in the registry record. The record also does not report a sample-size justification or interim analyses. These omissions prevent assessment of whether the statistical procedures were prespecified, how uncertainty was handled, or how the analysis addressed the structure of measurements from the same participants.

In general, over-dispersion matters when observed variability is greater than a statistical model assumes: standard errors may then be understated, and uncertainty may appear smaller than it is. That concept is relevant only if a model's distributional assumptions call for it. The registry identifies the primary outcome as continuous and names a mean and standard deviation, but does not describe a model for the primary analysis or report evidence about its distribution. It would therefore be inappropriate to assert that over-dispersion was present or that a particular adjustment was used.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Hip Exoskeleton for Stroke Gait Assistance")
n   <- c(12)
m   <- c(0.7228)
sp  <- c(0.0846)
sd  <- sp      # posted standard deviations
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Hip Exoskeleton for Stroke Gait Assistance  12  0.723  0.085  0.669  0.777

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 13 outcome measures and 12 analyses, including a primary model-accuracy endpoint and secondary walking, balance-confidence, and stroke-impact measures. When multiple outcomes or analyses are considered, general statistical practice recognizes that the chance of finding at least one apparently notable result can increase if each is assessed independently without a plan for multiplicity. Adjustment procedures, hierarchical testing, or a prespecified focus on selected outcomes can address different aspects of that issue. Which approach is appropriate depends on the study question and analysis plan.

The registry record does not report whether multiplicity was addressed, whether any outcomes were designated as exploratory, or whether a hierarchy or adjustment procedure was prespecified. The presence of multiple measures alone does not establish that an adjustment was required or omitted; the relevant analysis plan is not reported in the registry record. Since numerical outcome results are also absent, the effect of any multiplicity approach cannot be evaluated from the posted information.

Missing-data handling is not reported in the registry record. The record does not state whether all enrolled participants completed each assessment, how unavailable measurements were handled, or whether any imputation or sensitivity analysis was used. The enrollment figure of 12 does not establish that every outcome was observed for all participants.

Analysis populations are also not reported in the registry record. There is no description of whether the primary analysis included all enrolled participants, only those completing a particular condition, or another defined group. The record gives no participant flow by analysis or outcome. These missing details limit interpretation of representativeness, completeness, and the relationship between the enrolled sample and any analysis set.

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 record lists 12 participants at risk in the baseline condition and 12 at risk in the hip exoskeleton condition. For each condition, the record lists 0 participants with serious adverse events, 0 participants with other adverse events, and 0 deaths. These are the counts recorded in the registry, not a comparison of event rates between randomized groups. The conditions are recorded within the study's single-group structure.

Counts of participants affected do not describe event severity beyond the serious-versus-other classification, the nature of any events, their timing, or how events were assessed. The record lists no top serious adverse-event terms. The posted counts therefore describe what the registry reports, but do not give detail about safety monitoring procedures, duration of follow-up, or whether event collection was complete across all tasks.

Zero recorded events should be read as zero participants reported with those event categories in the registry record. It does not establish that no event could occur, nor does it support a comparison with another population or study. The record does not report additional information that would clarify exposure time or ascertainment methods.

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

The registry record establishes that a completed, single-group study enrolled 12 people and evaluated a hip exoskeleton for stroke gait assistance. Its primary outcome was defined as the coefficient of determination for predicted versus ground-truth hip joint moments during the exoskeleton intervention. The record also identifies several secondary measures spanning walking performance, balance confidence, and stroke impact.

Those details explain what the study set out to assess, but the registry record does not report numerical primary or secondary outcome results. It therefore does not allow readers to estimate the model's accuracy, determine the size or precision of any change, or judge whether the secondary measures differed between conditions. No reported numerical result should be inferred from the study's completed status or from the endpoint descriptions.

The study design also limits what the record can establish. Baseline and exoskeleton testing are described within a single-group study, not as a randomized, like-for-like comparison between separate groups. The primary model-accuracy outcome is only measured while the device is worn, so it is not a direct baseline comparison. The registry does not report condition order, detailed methods for comparing conditions, or the analysis plan. These gaps prevent strong interpretation of any potential differences even if numerical results were available.

The adverse-event figures record 0 participants affected by serious events, 0 affected by other events, and 0 deaths in each listed condition. These registry counts are informative as recorded, but they do not establish that risk is absent or characterize event ascertainment in detail. Overall, the posted record supports a description of the study's questions, design, outcome definitions, and recorded safety counts. It does not establish the magnitude, precision, or clinical meaning of outcome effects.

14. Why This Trial Matters Statistically

This study is statistically notable because its primary endpoint assesses a technical feature of a wearable robotic device: how accurately an embedded model predicts hip joint moments against laboratory-derived ground truth. This is distinct from asking whether participants walk faster, farther, or with greater confidence. Keeping those outcome types separate helps readers understand what each measure can support.

The registry also describes device and baseline testing within a single-group study. That distinction matters when interpreting comparisons: a baseline condition within one group is not the same design as a randomized control group. The primary model-accuracy measure is only assessed while wearing the exoskeleton, which further means the primary endpoint is a device-performance measure rather than a direct comparison against baseline.

The list of secondary measures shows that the study encompasses more than a model metric. Walking tests, balance confidence, and stroke impact address different domains and may not move together. A statistical interpretation would need the actual estimates, their uncertainty, and a clear account of how the multiple measures were analyzed. Those results and methods are not reported in the registry record.

The record's completed status and enrollment figure help place the study in context, while the absence of outcome values limits what can be concluded. Reporting estimates and uncertainty is important because a point estimate alone cannot convey precision, and a completed study without posted numerical outcomes cannot answer how large or consistent the observed findings were. The record is therefore useful for identifying the methodological questions that matter, even though it does not provide enough results to resolve them.

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 most important limitation for interpreting the registry record is that it does not report numerical results for the primary endpoint or the listed secondary outcomes. Without estimates and measures of uncertainty, the accuracy of the model and the performance of participants on the clinical measures cannot be quantified. Completion status does not substitute for results reporting.

The primary endpoint is measured only during the hip exoskeleton intervention. It cannot directly compare device use with baseline, and the record does not report a primary endpoint contrast between conditions. Although the study summary describes comparing device performance with not using the device, the specific method for that comparison is not reported in the registry record.

The design is listed as single-group, allocation as NA, and masking as NONE. The registry record does not report condition order, whether task procedures were standardized across conditions, or how possible order and practice effects were addressed. These details matter for interpreting repeated condition assessments, but should not be presumed.

Further limitations include the lack of a reported statistical analysis plan, sample-size justification, multiplicity approach, missing-data procedure, or analysis population. Detailed eligibility criteria and outcome-level completion information are also not reported in the registry record. The registry lists 13 outcome measures and 12 analyses, but does not explain how those measures were selected or prioritized. The adverse-event counts are recorded, but event definitions, monitoring procedures, and duration of observation are not described. Together, these omissions constrain assessment of precision, bias, completeness, and generalizability.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Adaptive Hip Exoskeleton for Stroke Gait Enhancement — NCT05536739.
• Registry record last updated 2026-07-01; status last verified 2026-05.
• Results first posted 2026-07-01.
• 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

NCT05536739 is a completed, single-group study sponsored by Georgia Institute of Technology, with enrollment of 12. The registry describes a hip exoskeleton for stroke gait assistance and baseline testing. The primary outcome is TCN model performance in predicting hip joint moments, measured by coefficient of determination during the exoskeleton intervention, with a timeframe of 5 Days.

The registry lists 13 outcome measures and 12 analyses, including walking, balance-confidence, and stroke-impact measures. Numerical outcome results and the statistical analysis details are not reported in the registry record. The adverse-event record lists 0 participants with serious adverse events, 0 with other adverse events, and 0 deaths in each condition, with 12 at risk in each.