Pulsewatch: Smartwatch Monitoring for Atrial Fibrillation After Stroke: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Pulsewatch: Smartwatch Monitoring for Atrial Fibrillation After Stroke (NCT03761394), a randomized na study. The primary outcome is Usability of Pulsewatch System (System Usability Scale & Rating Scale) at 14 Days Post the First Randomization.

2. Clinical Question

3. Trial Design

The registry record describes a randomized, single-masked crossover study with an enrollment of 120 and a completed status. The listed start date is 2019-09-03, and primary completion is listed as 2021-09-20. The sponsor is University of Massachusetts, Worcester. Phase is recorded as not applicable. The identity of the people masked, the sequence of crossover assignments, and the detailed timing and conduct of the randomizations are not reported in the registry record.

The initial device comparison paired testing devices with Cardea Solo for the intervention group and Cardea Solo alone for the control group over a 14-day period. The record separately describes extended-use groups: testing-device use and Kardia Mobile ECG use for the extended-use intervention group, and no device use for the extended-use control group. The extended-use period is described as 30 days following completion of the initial period. The registry mentions a second randomization in the timing of some secondary outcomes, but the allocation procedure and crossover sequence are not reported in the registry record.

The primary usability assessment concerned only the first-phase intervention group. Although the record gives 83 as the number analyzed for that outcome, it gives no control-group value for usability. The registry’s safety record separately reports groups at risk for adverse events. Those group counts should not be taken as the analysis populations for every outcome. A detailed participant flow by phase and outcome is not reported in the registry record.

The study’s stated goal was to develop a smartphone application and cardiovascular surveillance system for monitoring paroxysmal atrial fibrillation in people with a stroke history or risk. That purpose provides context, but the posted primary outcome measures reported usability rather than the accuracy of atrial-fibrillation detection, stroke prevention, or another clinical benefit.

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: Patients were either randomized into the intervention or control arm of the study. The control patients were given a gold-standard cardiac monitor patch (2 patches over 14 days) but were not offered the use of the Pulsewatch system. Thereafter, we re-randomized the 120 participants into a 1:1 randomization of control group (n=60) and intervention group (n=60) to continue using the device for an additional 30 days. Control group did not receive any devices during this time.

14-day Intervention & Control Period

Reasons for not completing this period:

30-day Period (Extension 44 Days)

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Only the intervention group in the first phase had outcomes assessed for their Pulsewatch system usability. The overall number analyzed is 83 instead of 90 because 3 participants withdrew and the other 4 participants either had their device lost in transit or the data was not able to sync.

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.

Population: Participants in the "Intervention Group for Extended Use" and "Control Group for Extended Use" did not complete baseline assessments as they are a continuation of the arms

6. Primary Endpoint

The registry lists the primary outcome as Usability of Pulsewatch System (System Usability Scale & Rating Scale) at 14 Days Post the First Randomization, measured over Assessed 14 days post the first randomization.

The primary endpoint is a participant count tied to usability of the Pulsewatch system, assessed 14 days after the first randomization. The outcome description says that participants self-reported likes, dislikes, and problems with the smartphone application and smartwatch. Items about likes and dislikes used a five-point response scale, while problems had yes-or-no responses with space for explanation. The description also refers to Mobile Application Rating Scale questions, whose items used a five-point scale and whose subscales were summarized by an average mean.

The registry’s posted primary result is the number of participants with a System Usability Scale score > 68. It reports 31 such participants among 83 analyzed in the first-phase intervention group. A threshold count offers a straightforward summary of how many assessed participants crossed the specified score boundary. It does not show the distribution of all usability scores, the average score, or the specific problems and preferences reported by participants. Those details are not reported in the registry record.

Usability is relevant to a system intended for participant use because ease of use and experience can affect whether people engage with monitoring. That is a general reason to measure usability, not evidence that the system detected atrial fibrillation accurately or improved health outcomes. The primary endpoint does not itself compare the device system with Cardea Solo, since the registry record posts no usability result for the control group.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is a thresholded count rather than a continuous comparison. In general, a count above a prespecified score boundary can be easy to interpret as a number of participants meeting that boundary, but it discards information about how far scores lie above or below it. The registry reports 31 participants with a System Usability Scale score > 68 among 83 analyzed. It does not report the full score distribution or an uncertainty interval for this count. Because no control usability outcome is posted, a relative effect cannot be estimated from the primary result.

What do the secondary mean differences represent?

The secondary outcomes are changes in scores, reported as group means with standard deviations. In general, a mean difference compares the average outcome in one group with the average outcome in another. A positive or negative sign indicates direction according to the contrast’s ordering; it does not, by itself, indicate clinical importance. The recalculated comparisons use the posted summary statistics and are unadjusted. They are not sponsor-reported analyses, and the registry record lists no reported analyses for these secondary outcomes.

A standard deviation describes variation among individual observations around a group mean. It is not the same as the uncertainty in the mean or in a mean difference. In general, standard errors and confidence intervals describe estimate precision, while standard deviations describe outcome spread. The recalculated comparisons use the posted group means, standard deviations, and sample sizes. Some reported confidence-interval limits have more than three decimal places; the interpretation can be conveyed without reproducing that excess precision. Each reported interval for the secondary mean differences spans zero, indicating that the summary statistics are compatible with values on either side of no difference under the calculation used.

How should confidence intervals and p-values be read?

In general, a confidence interval represents a range of values reasonably compatible with the data and the calculation assumptions. A wide interval indicates limited precision; an interval spanning zero for a mean difference does not establish that the groups are the same. A p-value summarizes how unusual data at least as incompatible with a null hypothesis would be under that hypothesis and the model used. It is not the probability that the null hypothesis is true, nor is it a measure of clinical importance. The posted unadjusted p-values should therefore be read alongside the estimates and intervals, not as stand-alone conclusions.

The registry record does not report the sponsor’s inferential method, a covariate adjustment strategy, or the model used for the crossover design. The recalculations are unadjusted comparisons of group summaries and should not be mistaken for a model that accounts for within-participant correlation, period effects, or sequence effects. Whether such issues were addressed in the sponsor’s own analyses is not reported in the registry record.

What do the planning illustrations mean?

The sample-size illustrations are explicitly planning calculations based on observed effects, not the trial’s own design. The stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The rows describe planning for 100%, 75%, and 50% of the observed effect, with participants per arm for 80% and 90% power. These are hypothetical sample-size illustrations. They are not observed power, and they do not show that the trial was designed around those effects or assumptions. The trial’s sample-size justification is not reported in the registry record.

In general, smaller effects require larger sample sizes to distinguish them from random variation at a fixed significance level and target power. When observed estimates are imprecise, using them as the basis for planning can produce unstable illustrations. The displayed calculations therefore serve as sensitivity examples rather than reliable forecasts of what a future study would require. The record does not report assumptions about attrition, repeated measures, crossover correlation, or adjustment for multiple outcomes in its original design.

Why does over-dispersion matter for count outcomes?

In general, a count endpoint may vary more than a simple statistical model expects; this is called over-dispersion. If it is present and ignored, uncertainty can be understated. The primary usability endpoint is reported as a count of participants above a score threshold, but the registry record does not report a comparative count analysis or a dispersion estimate. There is therefore no basis in the posted primary result to assess over-dispersion or to infer that a particular count model was used.

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("Intervention Group")
n      <- c(83)
events <- c(31)
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  Intervention Group  83  31  37.3  27.0  48.7  27.7  48.1  27.7  48.1

11. Multiplicity, Missing Data and Analysis Populations

The registry record lists multiple secondary outcomes, including anxiety, general-health MCS and PCS, disease-management self-efficacy, symptom-management self-efficacy, patient activation, and medication adherence. It reports ten outcome measures overall, but the strategy for prioritizing outcomes or controlling the chance of false-positive findings across them is not reported in the registry record. The recalculated secondary comparisons are unadjusted. In general, when many outcomes are examined, an isolated small p-value can occur by chance more readily than when a single outcome is tested. This makes prespecified analysis plans and multiplicity procedures important for confirmatory interpretation.

There is also no reported analysis plan describing the treatment of missing outcomes, the handling of crossover periods, or the population used for each sponsor analysis. The registry explains that the primary usability analysis included 83 rather than 90 participants because 3 withdrew and 4 had a device lost in transit or data that could not sync. The reasons are reported, but the method for handling those missing usability observations is not. A missing-data method, such as imputation or complete-case analysis, is not reported in the registry record.

The record gives an enrollment of 120, but that enrollment figure should not be treated as the number analyzed for any specific endpoint. Group-specific analysis populations differ across posted outcomes, and their relationship to all randomized participants is not fully documented. The identity of the people masked and the extent to which outcome assessment was masked are also not reported in the registry record. These gaps limit assessment of attrition, analysis-population choices, multiplicity, and the possibility of bias. They do not by themselves show that the study’s analyses were inappropriate; they mean the relevant procedures cannot be evaluated from the registry record.

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.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: myocardial infarction and hospitalization (3.3%), covid-19 (3.3%), pancreatitis (3.3%), chest pain and hospitalization (1.1%).

The registry reports serious and other adverse-event counts separately across the initial and extended-use groups. For the initial intervention group, the record lists 1 serious event among 90 at risk and 1 other event among 90 at risk; deaths are listed as 0. For the initial control group, it lists 2 serious events among 30 at risk and 3 other events among 30 at risk; deaths are listed as 0. In the extended-use intervention group, it lists 1 serious event among 57 at risk and 0 other events among 57 at risk; deaths are listed as 0. In the extended-use control group, both event categories are listed as 0 among 58 at risk, with deaths listed as 0.

These are registry counts, not estimates of comparative risk. The group sizes at risk differ, and the record does not report a formal between-group safety analysis or uncertainty intervals. The listed serious-event terms include myocardial infarction and hospitalization, COVID-19, pancreatitis, and chest pain and hospitalization; the registry reports percentages of 3.3% for the first three terms and 1.1% for chest pain and hospitalization. Event counts and terms should be interpreted in the context of the separate group periods and not combined into a causal claim about the devices. The registry record does not provide enough information to determine attribution or comparative safety.

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

The primary result means that the registry recorded 31 participants with a System Usability Scale score > 68 among 83 participants assessed in the first-phase intervention group. It is a descriptive statement about that assessed group and endpoint. It does not say how the scores were distributed overall, which features participants liked or disliked, or how many encountered particular problems. The record does not provide a control-group usability result, so it cannot show a difference in usability between the intervention and control conditions.

The endpoint also does not establish that the system detects paroxysmal atrial fibrillation accurately, reduces stroke risk, or changes clinical care. Those are distinct questions from usability, and the posted primary outcome is not a clinical-effectiveness comparison. The trial’s stated development goal gives context for why usability was measured, but an intended use is not evidence of performance or benefit.

The recalculated secondary comparisons do not settle whether there are meaningful differences in anxiety or general-health scores. Their intervals span zero, and the analyses are unadjusted calculations from posted summary statistics rather than reported sponsor analyses. In general, a result that does not distinguish groups clearly is not equivalent to evidence that the groups are identical. Conversely, a difference in a score would not by itself establish a clinically important effect without context about the measure and an appropriate analysis.

The adverse-event figures are descriptive counts by registry group. They do not determine whether an event was caused by a device, and they are not a formal comparison of safety. Taken together, the record documents usability reporting, selected secondary score summaries, and adverse-event counts. Important details—including analysis procedures, missing-data methods, and sponsor-reported inferential analyses—are not reported in the registry record. Conclusions should therefore remain limited to the measurements and summaries actually posted.

14. Why This Trial Matters Statistically

This study illustrates why an endpoint’s form matters. A thresholded usability count communicates how many assessed participants crossed a score boundary, but it does not retain the detail available from the underlying score distribution. It also cannot support a comparative estimate when the corresponding control outcome is absent. Readers should distinguish an observed count from a treatment contrast and from a claim about clinical performance.

The crossover structure adds another statistical consideration. In general, crossover studies can have within-participant comparisons, and analysis may need to address period, sequence, and correlation between observations. The registry record describes randomization and a crossover model but does not report how these design features were incorporated into sponsor analyses. Recalculations from group summaries cannot reconstruct that information.

The posted secondary results also demonstrate the difference between descriptive summaries and confirmatory inference. Means and standard deviations help describe group outcomes, while unadjusted comparisons and confidence intervals communicate uncertainty under a particular calculation. With multiple outcomes, an analysis plan is important for interpreting isolated findings. Here the reported secondary comparisons are recalculated and unadjusted, and the multiplicity strategy is not reported in the registry record.

Finally, planning illustrations based on observed effects should not be confused with evidence about the original design’s adequacy. The registry’s illustrations state their assumptions and show hypothetical participants per arm for selected fractions of observed effects. They are planning scenarios, not observed power or proof that the study had a particular design target. This distinction is useful when reading registry results whose sample-size justification and analysis plan are not reported.

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.

Several limitations are specific to the information posted for this study. The primary endpoint is available only for the first-phase intervention group, with no control-group usability value. The result is a threshold count rather than a full summary of System Usability Scale scores, and the registry does not post the underlying rating distribution or detailed response information. The reported analysis population is smaller than the stated expected count because of withdrawals and device/data problems; how those missing outcomes were handled is not reported in the registry record.

The study is described as randomized, single-masked, and crossover, but the registry does not identify who was masked or explain the assignment sequence and period structure in enough detail to evaluate the design’s analysis implications. The sponsor’s analyses are not reported for the posted secondary outcomes. The available comparisons are recalculated from summary statistics and do not document adjustment for repeated observations, period effects, or other covariates.

Although several secondary outcomes are listed, numerical results are not posted for all of them. The multiplicity strategy, analysis populations, and missing-data approach are not reported in the registry record. The available secondary comparisons are unadjusted, and their interpretation should account for the number of outcomes considered. Adverse-event counts are presented across groups with differing at-risk counts, without a reported comparative safety analysis or attribution details.

These limitations describe what can and cannot be assessed from the registry record. They do not establish that unreported analyses were not conducted, nor do they indicate the direction of any unreported result. They restrict interpretation to the posted usability count, selected secondary summaries, and descriptive safety figures.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Pulsewatch: Smartwatch Monitoring for Atrial Fibrillation After Stroke — NCT03761394.
• Registry record last updated 2023-11-18; status last verified 2023-10.
• Results first posted 2023-11-18.
• 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

The registry record describes Pulsewatch: Smartwatch Monitoring for Atrial Fibrillation After Stroke as a completed, randomized, single-masked crossover study sponsored by University of Massachusetts, Worcester. Enrollment is recorded as 120. The listed study start is 2019-09-03 and primary completion is 2021-09-20. Its stated development goal concerned smartphone and smartwatch monitoring for paroxysmal atrial fibrillation in people with stroke history or stroke risk.

The primary outcome assessed self-reported usability 14 days after the first randomization. The posted count is 31 participants with a System Usability Scale score > 68 among 83 analyzed in the first-phase intervention group. No control usability result is posted. Selected numerical secondary outcomes cover anxiety, MCS, and PCS; the recalculated comparisons are unadjusted and not sponsor-reported analyses. Adverse-event counts are also reported by group. The registry record does not report the detailed analysis plan, multiplicity strategy, or missing-data methods.