Predictors of Response in Chronic Stroke: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Predictors of Response in Chronic Stroke (NCT04283253), a na na study. The primary outcome is A Change in the Fugl-Meyer Upper Extremity Assessment Score.

2. Clinical Question

3. Trial Design

This was a completed, single-group study registered as NCT04283253, with the title “Predictors of Response in Chronic Stroke.” The record lists one enrolled subject and a single intervention group, Robot + TTT Exercise. It does not describe a randomized allocation or a separate control arm; allocation is recorded as NA. Masking is recorded as none. The phase is recorded as NA.

The intervention combined wrist and shoulder-elbow robot training with transition to task training arm exercises. The group description specifies a 60 minute study intervention. The study also planned MRI and transcranial magnetic stimulation (TMS), alongside information on demographics, functional status, neurophysiology, neuroanatomy, and other potential biomarkers. The stated purpose was to validate a predictive model for clinically significant change in impairment following robot-assisted therapy.

The registry record gives a start date of 2019-10-22 and a primary completion date of 2020-03-31, and lists the study status as completed. These dates and status do not establish that all planned assessments were completed. The record specifically says that only one subject was enrolled and that data collection was incomplete due to the Covid-19 pandemic. Evaluations following the 12-week intervention were not performed.

Three outcome measures are listed, but the record gives no statistical analyses. The primary outcome is described as change in the Fugl-Meyer Upper Extremity Assessment score. No results from that assessment or predictive-model validation are reported. With one group and no like-for-like control comparison, the record cannot provide a controlled estimate of intervention effects. The incomplete data also prevent the planned assessment of prediction performance from being evaluated in the posted 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

Analysis population for the primary outcome, as defined in the registry: Only one subject was enrolled and data collection was incomplete due to the Covid-19 pandemic. The evaluations following the 12-week intervention were not performed.

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 A Change in the Fugl-Meyer Upper Extremity Assessment Score, measured over Immediately following the 12-week intervention.

The primary endpoint was change in the Fugl-Meyer Upper Extremity Assessment score, expressed in units on a scale. The registry describes the Fugl-Meyer as a stroke-specific assessment of upper-extremity impairment. It uses direct observation of volitional movement and assesses motor impairment related to reflexes, sensation, and abnormal synergies. Each item is rated on an ordinal scale from 0, meaning cannot perform, to 2, meaning performs fully. The full scale ranges from 0-66, with higher scores representing less motor impairment.

The measure was selected to assess change in arm use from baseline through the study’s training assessments and after training. The description refers to mid-training assessments after visits 12 and 24, post-training assessment after visit 36, and a 12 week retention period. The primary timeframe is listed as immediately following the 12-week intervention. The registry states that the Fugl-Meyer has been shown to be valid and reliable, with high inter-rater and test-retest reliability.

Those properties explain why the measure can be useful for assessing impairment, but they do not supply an outcome for this study. The record says the evaluations following the 12-week intervention were not performed. It therefore reports no primary endpoint result, change estimate, or uncertainty interval.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The study’s stated statistical aim was to validate a prediction model, not to compare randomized treatment groups. The proposed predictors included demographics, functional status, neurophysiology, neuroanatomy, and other potential biomarkers. The intended target was the likelihood of a clinically significant change in impairment at the end of a robot-assisted therapy intervention. The registry record does not describe the model form, predictor selection, definition of clinically significant change, validation procedure, or statistical analysis plan. Those details are not reported in the registry record.

What does model validation require?

General statistical explanation: a prediction model uses characteristics of participants to generate predictions about an outcome. Validation assesses how well those predictions correspond to outcomes in data not used simply to construct the model. Assessment can address discrimination, meaning whether predictions distinguish people with different outcomes, and calibration, meaning whether predicted risks or outcomes agree with what is observed. The appropriate measures depend on the outcome and model. The registry record does not report a validation method or any validation results, so no claim about discrimination, calibration, or predictive performance can be made.

In this study, validation would depend on having completed outcome assessments linked to the planned predictor information. The record states that only one subject was enrolled, data collection was incomplete, and the post-intervention evaluations were not performed. No statistical analyses are listed. Thus, the information needed to report observed prediction performance is not present in the registry record. This is not a question of interpreting a weak or strong model result: no model result is reported.

What does the single-group structure allow?

General statistical explanation: a single-group design can describe measurements within participants when the same outcome is measured at different times. It does not, on its own, show what would have happened over the same period without the intervention. A separate concurrent comparison group can help address that question, subject to the design and analysis. Here, the record lists one intervention group and no separate comparator. The primary endpoint was not collected after the intervention, so the record does not support a completed within-group change analysis either.

The intervention group description specifies a 60 minute study intervention combining robot training and TTT exercises. The record does not provide information on the number or spacing of sessions, adherence, or details of treatment delivery beyond that description. Those aspects are not reported in the registry record and should not be inferred from the stated intervention duration.

How should uncertainty be considered?

General statistical explanation: an estimate describes the result observed in a study, while a confidence interval communicates uncertainty around that estimate under specified assumptions. A small amount of completed data typically supports less precise estimates than a larger amount, but precision depends on the outcome and analysis as well as the amount of information. In this case, no primary outcome estimate or interval is reported. There is therefore no precision measure to interpret.

The listed sample-size assumptions describe a planning illustration based on an observed effect, with a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The illustration describes the effect to detect as a stated fraction of the observed effect. These are planning assumptions, not the trial’s own design justification and not observed power. The registry record does not report the study’s sample-size justification. Because no primary endpoint result or observed effect is reported, the planning illustrations should not be taken as evidence about the study’s achieved precision or its ability to detect an effect.

General statistical explanation: power is a design property calculated under assumptions about the effect, variability, significance level, and analysis. It is not a measure of whether a completed study found a true effect. A retrospective calculation based on an observed estimate is especially dependent on that estimate and should not be presented as observed power. The record’s planning illustration is not a substitute for completed outcome data or an uncertainty interval.

Why can over-dispersion matter?

General statistical explanation: over-dispersion occurs when count data vary more than a chosen statistical model expects. If an analysis assumes less variability than is actually present, uncertainty can be understated and inference can become unreliable. This general issue can matter in some count-outcome analyses, but the primary outcome in this record is a scale score, and no statistical analysis of that outcome is reported. The record does not provide a count-outcome analysis or enough information to say that over-dispersion was assessed.

The record lists three outcome measures but does not supply their full analysis results in the information available for interpretation. It lists no analyses. There is no reported method for handling repeated assessments, missing observations, or multiple outcomes. Those choices can affect estimates and uncertainty, but their use cannot be reconstructed from the registry record. Any discussion of such methods must therefore remain general rather than attributing a method to this study.

10. Multiplicity, Missing Data and Analysis Populations

The registry record lists three outcome measures, but it does not report a statistical analysis plan or any analyses. It therefore does not describe how the primary outcome was distinguished analytically from other outcomes, whether testing across outcomes was planned, or whether any adjustment for multiple testing was intended. Multiplicity methods are not reported in the registry record. No claim can be made about whether the reported findings were protected against inflated false-positive risk, because no findings or analysis procedures are reported.

Missing data are directly relevant: the record says data collection was incomplete due to the Covid-19 pandemic and that evaluations following the 12-week intervention were not performed. It does not describe a statistical method for handling missing endpoint data, such as imputation or a model using incomplete repeated measurements. It also does not report reasons for missingness beyond the stated pandemic-related incompleteness. No missing-data assumptions or sensitivity analyses are reported in the registry record.

The record does not define an analysis population, such as an intention-to-treat or per-protocol population, and it does not report which participants contributed to any outcome analysis. No analyses are listed. With the planned post-intervention evaluations absent, the registry record contains no primary endpoint analysis population or completed primary analysis to interpret. These omissions limit conclusions about both the planned analytical approach and the consequences of missing outcome data.

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 posted adverse-event figures list the Robot + TTT Exercise group with one subject at risk, zero serious adverse events, zero other adverse events, and zero deaths. These are counts recorded for the group, not estimates of risk in a wider population. The registry record does not report event terms for serious adverse events.

Zero recorded events in a group this small cannot establish that events would not occur in other participants or under other circumstances. General statistical explanation: an event count is a description of what was recorded during the relevant observation, while an event-rate estimate requires a suitable denominator and information about the observation period. The figures alone do not describe the completeness of safety surveillance. The record also says data collection was incomplete, but does not specify how that affected safety follow-up.

The record does not provide additional details about event definitions, severity grading, relatedness, or timing. Those details are not reported in the registry record. The figures should therefore be read narrowly as the adverse-event counts posted for this group, rather than as a comprehensive estimate of safety.

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

The registry record establishes that the study was intended to validate a model for predicting clinically significant impairment change in chronic stroke, using demographic, functional, neurophysiological, neuroanatomical, and other potential biomarker information. It also establishes that the planned intervention combined robot-assisted exercise and TTT arm exercises, and that the primary outcome was change in the Fugl-Meyer Upper Extremity Assessment score.

What the record does not establish is whether the intervention changed impairment, how much any change might have been, or whether the proposed model could predict response. Only one subject was enrolled, data collection was incomplete due to the Covid-19 pandemic, and the evaluations following the 12-week intervention were not performed. No primary outcome result or statistical analysis is reported. The absence of a result is not evidence of no effect; it is an absence of the planned endpoint information in the registry record.

The single-group design also does not offer a separate comparator. Even with completed before-and-after measurements, the design would not provide a controlled treatment contrast. A prediction objective and a treatment-effect comparison are different questions: model validation concerns how well predictions match outcomes, whereas a controlled comparison estimates differences between intervention conditions. The record reports neither a completed validation nor a controlled comparison.

Statistically, the record is therefore informative about the study’s aim, intervention structure, intended primary measure, and important data-collection limitation. It does not offer an estimate with a confidence interval, a hypothesis test, or evidence about prediction performance. The adverse-event counts are limited descriptive counts and likewise cannot establish population-level safety. Any stronger conclusion would go beyond the results reported in the registry record.

13. Why This Trial Matters Statistically

This study matters statistically because it illustrates the distinction between a well-defined research aim and evidence that can answer that aim. The record describes a clinically relevant prediction objective and identifies a scale-based primary outcome, but the necessary post-intervention outcome information was not collected. A planned model cannot be evaluated from its intended predictors alone; its predictions need to be compared with observed outcomes.

The record also shows why completion status should be interpreted alongside endpoint availability. The study is listed as completed, yet the registry says data collection was incomplete and that evaluations following the 12-week intervention were not performed. Completion in the record does not mean that the primary analysis or model validation was achieved.

For researchers planning predictive studies, the case highlights the need to distinguish enrollment, measurement completion, and analytical evaluation. These are separate stages. Statistical interpretation depends on the information available at each stage, not solely on the stated objective or study status. The record does not report a sample-size justification, model specification, validation strategy, missing-data approach, or analysis plan, so it cannot inform those design choices directly.

It also underscores the limits of a single-group structure. A prediction study may have a different primary purpose from a treatment comparison, but the absence of a separate control group still limits causal interpretation of change. Here, endpoint incompletion further means that neither change nor predictive performance can be estimated from the posted results. The most defensible statistical conclusion is that the registry record documents an intended study question but does not provide the completed data needed to answer it.

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 is that the planned primary outcome was not obtained after the intervention. The registry record states that only one subject was enrolled, data collection was incomplete due to the Covid-19 pandemic, and evaluations following the 12-week intervention were not performed. No primary result, estimate of change, or interval is reported.

The study had a single intervention group and no separate comparator in the registry record. This means it was not a like-for-like control comparison and could not provide a between-group estimate. The information reported also does not establish whether the intended predictive model was fitted or validated; no statistical analyses are listed.

Further limitations are the lack of reported detail on the model specification, clinically significant change threshold, predictor handling, sample-size justification, missing-data methods, analysis populations, and multiplicity approach. These details are not reported in the registry record. The record lists three outcome measures but does not provide an analysis plan or results for them. The adverse-event counts are descriptive and do not include reported event terms or fuller safety-follow-up context. Together, these limitations prevent assessment of efficacy, model performance, outcome precision, and broader safety from the posted record.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Predictors of Response in Chronic Stroke — NCT04283253.
• Registry record last updated 2022-03-03; status last verified 2022-02.
• Results first posted 2022-03-03.
• 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

NCT04283253, “Predictors of Response in Chronic Stroke,” was a completed single-group study sponsored by VA Pittsburgh Healthcare System. The record lists one enrolled subject, no masking, allocation recorded as NA, and a Robot + TTT Exercise intervention. The study aimed to validate a model predicting clinically significant impairment change using demographic, functional, neurophysiological, neuroanatomical, and other potential biomarker information.

The primary outcome was change in the Fugl-Meyer Upper Extremity Assessment score. The record states that data collection was incomplete due to the Covid-19 pandemic and that evaluations following the 12-week intervention were not performed. No statistical analyses or primary outcome results are reported. The posted adverse-event figures list zero serious adverse events, zero other adverse events, and zero deaths in the group.