Delineation of Sensorimotor Subtypes Underlying Residual Speech Errors: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Delineation of Sensorimotor Subtypes Underlying Residual Speech Errors (NCT03736213), a randomized phase 1 study. The primary outcome is Normalized F3-F2 Distance, an Acoustic Measure That Correlates With Perceptual Accuracy of /r/, Measured From /r/ Sounds Produced in Treatment Sessions..

2. Clinical Question

This is a within-subject comparison: participants experienced both conditions, so the central question concerns differences in response between treatments within the study participants. It does not compare separate groups assigned to receive only one type of treatment. The trial therefore addresses the relative performance of the two approaches on the measured acoustic outcome under the study conditions; it does not establish how either approach compares with no treatment or with another form of care.

3. Trial Design

The registry identifies NCT03736213 as a randomized, single-masked, crossover study of visual-acoustic and ultrasound biofeedback for residual speech errors. It is listed as Phase 1, with enrollment of 7 and completed status. The study began on 2019-03-01, and primary completion was recorded as 2022-02-01. New York University is the sponsor.

The crossover structure is important for interpreting the comparison. Each participant received both treatments, and individual sessions were randomly assigned to feature one condition or the other. The endpoint description states that each participant completed 10 sessions of each type, administered over five weeks. Thus, the treatment conditions were compared within participants rather than by contrasting separate treatment-only groups. The registry identifies randomization, but further information about the sequence-generation procedure, allocation concealment, or how treatment order was handled is not reported in the registry record.

The record describes single masking but does not say who was masked or how masking was maintained. It also does not report whether any period, sequence, or carryover effects were assessed. Those details matter in crossover studies because a participant’s response in one condition could be related to the order or experience of the other condition; whether such effects were relevant or analyzed is not reported in the registry record.

There was one primary outcome measure and one posted analysis. The outcome summaries pool participants and sessions, while the stated paired t-test compares conditions across subjects. The record does not describe the participant-level analysis dataset, how session-level information was combined for each participant, or which analysis population was used. These omissions limit the detail with which the analysis can be independently reconstructed from 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.

Before assignment: This study used a within-subjects design. Each participant received both treatment conditions, with sessions randomly assigned to feature one condition or the other.

Overall Study

Analysis population for the primary outcome, as defined in the registry: Note that this study used a within-subjects design. Each participant received both treatment conditions, with individual sessions randomly assigned to feature one condition or the other. Each participant completed 10 sessions of each type.

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 Normalized F3-F2 Distance, an Acoustic Measure That Correlates With Perceptual Accuracy of /r/, Measured From /r/ Sounds Produced in Treatment Sessions., measured over Acoustic accuracy was measured in all ten sessions of each type of treatment, which were administered over five weeks..

The primary endpoint was normalized F3-F2 distance, measured from /r/ sounds produced during treatment sessions and expressed as a z-score. F2 and F3 are formants extracted from the speech signal; the endpoint is the distance between the second and third formants. The registry describes this acoustic measure as correlated with perceptual accuracy of /r/.

For measurement, one trial in each block of 10 was flagged, and the first three formants were extracted from the center of the /r/ interval. F3-F2 distances were converted to z-scores relative to normative data from age-matched children with typical speech. A z-score of 0 represents the mean F3-F2 distance for typical children. A z-score of 1 indicates one standard deviation of the normative sample above the sample mean. Because F3-F2 is small in perceptually accurate /r/, larger values indicate lower accuracy; values above 2 are considered clinically atypical.

The outcome was chosen to quantify acoustic accuracy during treatment rather than to report a direct measure of everyday communication or a broader clinical outcome. The registry reports condition means and standard deviations pooled across participants and sessions. The summary measure can support a comparison of acoustic production between conditions, but it does not by itself describe individual response patterns or establish how a measured acoustic change relates to communication outside treatment sessions.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a paired t-test compare?

In a paired design, the analysis uses the relationship between measurements from the same participant under each condition. Rather than treating the treatment conditions as unrelated groups, a paired test evaluates within-participant differences. This can account for stable participant characteristics shared across both measurements. The registry says that each participant received both treatments and that a two-tailed paired-samples t-test was used to compare mean normalized F3-F2 distance across subjects.

A paired t-test relies on the distribution of participant-level differences being suitable for the test. The record does not provide those differences or describe diagnostic checks, so readers cannot assess that assumption from the posted summary statistics. The reported condition means and standard deviations are pooled across participants and sessions; they do not reveal the distribution of paired differences. In general, a paired analysis is most interpretable when the unit of analysis and the construction of each participant’s condition-level value are clear. Those details are not reported in the registry record.

What does a two-tailed test mean?

A two-tailed test considers departures in either direction from the null hypothesis of no mean difference. The registry labels the analysis as a superiority comparison and reports p = .13. The p-value is not the probability that the null hypothesis is true, nor does it express the probability that the observed difference is due to chance. It describes the extremeness of the test result under the null model and the test’s assumptions. Without a confidence interval, the p-value does not convey the range of effect sizes compatible with the data.

How should the acoustic measure be understood?

The endpoint is a continuous z-score based on F3-F2 distance. In general, a continuous measure can retain more graded information than reducing observations to categories. Here, the registry defines the direction of the measure: larger values indicate lower accuracy, and values above 2 are considered clinically atypical. The z-score reference comes from age-matched children with typical speech. This helps anchor the acoustic measure to a normative distribution, but it does not make the endpoint a direct measure of every aspect of speech performance.

The registry states that summary statistics report means and standard deviations pooled across participants and sessions. These summaries describe the central tendency and spread of the measured outcome, but the pooled standard deviations are not the standard error of the paired treatment contrast. In a crossover analysis, precision depends on participant-level paired differences and their variability. Those differences and a confidence interval are not reported, so the precision of the contrast cannot be fully reconstructed from the condition summaries.

What does the crossover structure add?

Because each participant experienced both conditions, the design can compare treatments within the same participants. Random assignment of individual sessions to feature one treatment or the other is reported. General statistical considerations for crossover designs include treatment order, period effects, and possible carryover from one condition to another. Whether these were assessed, and how the randomized session assignments were implemented, is not reported in the registry record. The record also does not specify who was masked despite identifying the study as single-masked.

How should planning illustrations be read?

Sample-size calculations based on an observed effect are planning illustrations, not observed power and not evidence that the trial was adequately powered. The registry describes the planning approach as using a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation, with the effect to detect expressed as a fraction of the observed effect. In general, such calculations depend on their assumptions and on the effect size selected for planning. They do not change the uncertainty in the completed study’s observed result, and they are not a substitute for a confidence interval. The registry record does not report the trial’s own sample-size justification.

What can the posted result establish?

The posted analysis is a test of a mean difference in the acoustic endpoint between two treatment conditions in this crossover study. It does not, on its own, establish equivalence, noninferiority, or clinical importance. Those conclusions require appropriate designs, prespecified margins where relevant, and adequate precision. The registry record does not report an equivalence or noninferiority framework. The p-value should therefore be interpreted as the result of the stated two-tailed paired test, not as proof that the conditions produce the same outcome.

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("Visual-acoustic Biofeedback", "Ultrasound Biofeedback")
n   <- c(7, 7)
m   <- c(5.73, 6.27)
sp  <- c(6.33, 6.27)
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  Visual-acoustic Biofeedback  7  5.730  6.330  -0.124  11.584
CARM  Ultrasound Biofeedback  7  6.270  6.270  0.471  12.069

11. Analyses Reported by the Sponsor

The registry-posted analysis is a two-tailed paired t-test comparing final mean values for visual-acoustic and ultrasound biofeedback. It reports a mean difference of .54 and p = .13. A confidence interval is not reported in the registry record. The condition means and standard deviations are also posted, but the registry’s stated skipped recalculation means no separate two-arm recalculated contrast is available for comparison with the sponsor’s analysis.

The record does not flag a discrepancy between the sponsor’s result and a recalculated result. Since there is no recalculated contrast to compare, the posted paired analysis is the available inferential result. The registry does not state the subtraction order for the reported mean difference, so the numerical sign should not be assigned a direction beyond the condition-specific means.

The sponsor’s test should be read in light of the crossover design and the limited enrollment. The registry does not provide participant-level paired differences, a confidence interval, or enough detail to reproduce the test from the posted summaries. Thus, the reported p-value and mean difference can be described, but the precision and robustness of the comparison cannot be fully evaluated from the registry record alone.

12. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome measure, one analysis, and no posted secondary outcomes. This limits the multiplicity questions that can be assessed from the record, but it does not establish that no other outcomes or analyses were considered during the study. Whether the primary analysis was prespecified in a protocol or statistical analysis plan, and whether any multiplicity adjustment was planned or used, is not reported in the registry record.

The record does not describe missing-data handling. It does not state whether all scheduled treatment sessions contributed to the outcome summaries, whether any measurements were unavailable or excluded, or how incomplete participant records would have been handled. The posted enrollment and condition summaries alone do not answer those questions. The analysis population, including whether the reported analysis included all enrolled participants, is also not reported in the registry record.

For a crossover study, the treatment comparison depends on paired information from participants exposed to both conditions. In general, if an observation from one condition is missing, the availability of the other condition’s observation and the analysis method can affect which information contributes to a paired test. The record does not explain whether the analysis required complete paired data or how any incomplete pairs were treated. It also does not report sensitivity analyses for missing observations.

The registry provides no fuller statistical analysis plan or account of interim analyses. Whether interim analyses occurred is not reported in the registry record. These omissions do not show that a particular problem occurred; they mark limits on what can be determined about prespecification, missingness, and the analysis population from the posted information.

13. 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 no serious adverse events among those listed as at risk in either treatment condition. For visual-acoustic biofeedback, the record lists 7 at risk and 0 affected by a serious adverse event. For ultrasound biofeedback, it lists 7 at risk and 0 affected. Deaths are listed as 0 in both conditions, and no serious adverse-event terms are listed.

These are counts recorded for the treatment groups, not evidence that risk is absent. The study enrollment was 7, and each participant received both conditions, so the treatment-group figures should not be added as though they represented separate groups of participants. The registry record does not provide additional detail about adverse-event ascertainment, the observation period for safety reporting, or non-serious adverse events beyond the listed counts.

With no serious adverse events recorded in these groups, the posted safety summary describes what was reported during the study. It cannot rule out uncommon events or establish a broader safety profile. The figures should be interpreted alongside the study’s limited enrollment and the information available in the registry record, without treating zero reported events as proof that an event cannot occur.

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

The posted data describe a comparison of two biofeedback treatments on normalized F3-F2 distance measured from /r/ sounds during treatment sessions. The visual-acoustic condition mean was 5.73, and the ultrasound condition mean was 6.27. Larger values indicate lower accuracy on this acoustic measure. The paired analysis reports a mean difference of .54 and p = .13, with no confidence interval reported.

These results do not demonstrate that the treatments are equivalent. A non-significant p-value is not an equivalence test, and the registry does not report an equivalence margin or a noninferiority framework. Nor does the result establish that either treatment is superior. The analysis is based on a small within-subject study, and the uncertainty around the paired contrast cannot be fully judged without a confidence interval or participant-level paired data.

The endpoint is an acoustic measure that the registry says correlates with perceptual accuracy of /r/. It is not itself a comprehensive measure of intelligibility, communication in everyday settings, or longer-term outcomes. The record does not report follow-up results beyond the treatment-session outcome. It also does not report secondary outcomes.

Accordingly, the record supports a description of the measured acoustic outcomes and the posted paired-test result. It does not establish a treatment effect on broader communication outcomes, determine which approach is preferable for an individual, or show whether observed performance continues after the study period. Such conclusions would require evidence beyond the outcome and information reported in this record.

15. Why This Trial Matters Statistically

Statistically, this study is notable for comparing treatment conditions within participants rather than assigning separate participants to treatment-only groups. In general, a crossover structure can make individual participants their own comparison, which is useful when participant characteristics influence the outcome. The registry also reports random assignment of individual sessions to feature one condition or the other. These design features make the paired nature of the analysis central to interpreting the result.

The endpoint is a continuous acoustic measure with an explicit direction: larger normalized F3-F2 values indicate lower accuracy. That definition helps readers interpret the condition summaries without treating higher scores as better by default. The study also links the acoustic measure to a normative reference and identifies values above 2 as clinically atypical, although the posted condition means are pooled across participants and sessions.

The record also illustrates why effect estimates and their precision matter alongside p-values. The posted paired test reports p = .13, but the registry does not provide a confidence interval or participant-level paired differences. Without those details, it is difficult to assess the range of plausible treatment contrasts or how consistently participants differed. The absence of those details is especially relevant when interpreting a study with enrollment of 7.

Finally, the record leaves several design and analysis details unspecified, including missing-data handling, the analysis population, and whether order or carryover effects were assessed. These are not reasons to assume the analysis was inappropriate; they identify information needed for a fuller statistical appraisal. The study therefore offers a useful example of how an outcome’s direction, a within-participant design, and reporting of uncertainty all shape interpretation.

16. 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 specific to this study is the small enrollment of 7. With limited participant-level information, estimates of the average paired contrast may be sensitive to the outcomes of individual participants. The registry does not provide individual differences or a confidence interval, so the degree of uncertainty around the reported mean difference cannot be assessed directly.

The outcome summaries pool participants and sessions, while the analysis is described as a paired t-test across subjects. The record does not explain how session measurements were reduced to participant-level values for the paired comparison. It also does not report the number of measurements contributing to each condition summary or how missing session measurements, if any, were handled. This makes independent reconstruction of the analysis difficult.

The study is a crossover design with random assignment of individual sessions to feature either condition. The record does not report whether treatment order, period effects, or carryover were assessed. It identifies single masking but does not specify who was masked or how masking was achieved. These details could inform interpretation of a comparison in which participants experienced both treatments.

The primary measure is an acoustic proxy for perceptual accuracy of /r/, not a direct measure of all clinically relevant speech outcomes. The registry record does not report secondary outcomes or follow-up results beyond the treatment-session measurement described. It also does not report a confidence interval, a detailed analysis plan, missing-data methods, or an analysis population. These limitations constrain conclusions about precision, robustness, and broader relevance without implying that any particular bias or data problem occurred.

17. Related Tutorials

18. Related Statistical Calculators

19. Sources

• ClinicalTrials.gov: Delineation of Sensorimotor Subtypes Underlying Residual Speech Errors — NCT03736213.
• Registry record last updated 2023-12-11; status last verified 2021-12.
• Results first posted 2022-12-29.
• 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.

20. Record Summary

NCT03736213 is a completed randomized, single-masked crossover study sponsored by New York University. It enrolled 7 children with speech sound disorder and compared visual-acoustic biofeedback with ultrasound biofeedback, with both conditions targeting /r/ distortions. Each participant received both conditions, with individual sessions randomly assigned to feature one condition or the other.

The primary outcome was normalized F3-F2 distance from /r/ sounds produced during treatment sessions. The posted means were 5.73 for visual-acoustic biofeedback and 6.27 for ultrasound biofeedback. The registry reports a paired two-tailed t-test with a mean difference of .54 and p = .13; a confidence interval is not reported. The endpoint’s direction is such that larger values indicate lower accuracy.

The record lists no serious adverse events in either condition. It does not report secondary outcomes, missing-data methods, the analysis population, or several details needed to assess the precision and reproducibility of the comparison.