Study 2: Learning New Words From Overhearing in Children With Autism Spectrum Disorder (ASD): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Study 2: Learning New Words From Overhearing in Children With Autism Spectrum Disorder (ASD) (NCT05192109), a non-randomized na study. The primary outcome is Proportion of Time During Which Children's Eye Gaze is Directed Toward the Target Object.

2. Clinical Question

The stated research goal is to explore whether children with ASD can learn word meanings from conversations they overhear, with TD children serving as a control. The specific aim for this study is to determine whether children with ASD can learn from both addressed and overheard teaching delivered through videoconferencing. The listed behavioral interventions are Overheard Speech and Addressed Speech.

The primary outcome is gaze toward an image of the target word, recorded while children look at images. This offers an observable measure related to attention to the target object after the teaching interaction. It is not, by itself, a direct report of a child's knowledge or a complete account of word learning. The registry record does not report a result that would establish how gaze differed between teaching conditions or groups.

The clinical question therefore concerns a behavioral measure after two types of video-based teaching, in children with ASD and TD children. The available record establishes the intended comparison and outcome, but not the magnitude, precision, or statistical significance of any difference.

3. Trial Design

The registry record identifies this as Study 2, “Learning New Words From Overhearing in Children With Autism Spectrum Disorder (ASD),” sponsored by New York University. The study is marked completed, with a listed enrollment of 342. Its recorded start date is 2020-12-01 and its primary completion date is 2024-12-01. The phase is listed as NA.

The allocation is non-randomized, the masking is single, and the model is crossover. These labels describe broad design features, but the registry record does not identify who was masked, how the sequence of teaching conditions was set, whether order was balanced, or whether any washout interval was used. Those details should not be inferred from the crossover label alone.

The two behavioral conditions are Addressed Speech and Overheard Speech. In the addressed condition, children participate by video in an interaction with an experimenter who uses the target word. In the overheard condition, children watch by video as an experimenter uses the target word while interacting with another experimenter. The study summary says that both types of teaching take place on video.

The registry record describes children with ASD and TD children. It does not provide the allocation of participants to these groups, the number analyzed in each condition, or participant characteristics. It states that exclusions occurred, for example because of experimenter error or insufficient attention, but does not report the number or distribution of those exclusions.

Because allocation is non-randomized, the condition comparison is not a randomized, like-for-like control comparison. The crossover designation indicates that the design involves repeated conditions, but without details on order and analysis it is not possible to assess how those features were handled. The statistical plan, sample-size justification, interim analyses, and masking procedures 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.

First Behavioral Task (3 Minutes)

Reasons for not completing this period:

Second Behavioral Task (3 Minutes)

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Not all enrolled participants were analyzed because of exclusions due to, e.g., experimenter error or children paying insufficient attention.

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 Proportion of Time During Which Children's Eye Gaze is Directed Toward the Target Object, measured over immediately after viewing 3-minute interaction.

The primary endpoint is the proportion of time during which children's eye gaze is directed toward the target object. The registry record describes gaze being recorded as children look at images; the dependent measure is the proportion of time gaze is directed to an image of the targeted word. The stated timeframe is immediately after viewing a 3-minute interaction.

This endpoint is a continuous measure: it can take different values across a range rather than representing only an event or category. A mean can describe the average proportion of time directed toward the target object, while a standard deviation describes how much individual measurements vary around that mean. The registry record identifies the planned summary as a mean with standard deviation.

Gaze can be useful in this research context because it gives an observable behavioral response following a teaching interaction. Attention to the target image may be relevant to a question about word learning, but it should not be treated as equivalent to a direct demonstration that a child knows or can use the word. The endpoint is specifically about where gaze is directed and for what proportion of time.

The recorded timeframe is immediate, so the outcome addresses behavior after viewing the interaction rather than later retention. The registry record does not report results for this endpoint, the measurement procedure beyond its brief description, or details such as how images were presented and how gaze data were processed. These omissions limit the assessment of measurement precision and interpretation.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: it is a proportion of time spent looking toward the target object. For a continuous measure, a mean describes the average observed value, and a standard deviation describes the spread of values around that mean. A mean can be useful for comparing condition summaries, but its interpretation depends on how measurements were collected and how participants contributed data. The registry record names the mean and standard deviation as the summary format but does not report their values.

What does a confidence interval tell a reader?

In general, a confidence interval gives a range of values compatible with an estimate under the statistical method and assumptions used. A narrower interval indicates greater precision in that estimate, while a wider interval indicates greater uncertainty. The interval does not remove possible bias from allocation, measurement, exclusions, or an unsuitable model. The registry record does not report a confidence interval for the primary endpoint.

A p-value, in general, describes how compatible the observed data are with a specified null model, assuming that model and the analysis assumptions are appropriate. It does not measure the size or importance of an effect, and it does not give the probability that the null hypothesis is true. No p-value for the primary endpoint is reported in the registry record. Without an estimate and uncertainty interval, a p-value alone would also be an incomplete account of the result.

Why does the crossover design matter?

In general, crossover studies measure participants under more than one condition. Measurements from the same child may be correlated because they share characteristics of that child. An analysis that accounts for within-participant correlation can use the paired structure; an analysis that ignores it may estimate uncertainty incorrectly. The registry record labels the model as crossover but does not report the order of conditions, the number of observations per child, the handling of period or order effects, or the statistical model. It is not possible to verify how the crossover design was represented in the analysis.

Non-randomized allocation is a separate interpretive consideration. In general, when condition assignment is not random, differences in participant characteristics, order, setting, or other factors may coincide with the comparison of interest. Adjustment can sometimes address measured factors, but cannot guarantee removal of unmeasured confounding. The registry record does not report the allocation procedure, covariate adjustment, or a causal analysis plan. The posted design should therefore not be described as a randomized comparison.

What does the planning illustration represent?

The registry record describes a planning illustration computed from the observed effect, using a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is described as a stated fraction of the observed effect. For non-time-to-event outcomes, the illustration is in participants per arm for 80% or 90% power; time-to-event calculations are described as total event counts. These are planning quantities, not observed outcomes, and not observed power.

In general, a planning sample-size calculation asks what enrollment might be needed to detect an assumed effect under specified design and analysis assumptions. Its result depends on the effect size chosen, variability, allocation, significance level, power target, and calculation method. A calculation based on an effect estimated from trial data inherits uncertainty in that estimate. It should not be interpreted as evidence that the trial was designed around that effect or that the analysis had a particular probability of detecting it. The registry record does not report the study's own sample-size justification or statistical plan.

Why can over-dispersion matter?

Over-dispersion is a general issue for outcomes whose variability exceeds what a model assumes. A model that understates variability can yield uncertainty estimates that are too small and significance assessments that are too optimistic. The primary outcome is described as a continuous proportion and summarized by a mean and standard deviation; the registry record does not report a model or indicate that over-dispersion was assessed. This concept should not be taken as a claim that over-dispersion occurred in this study.

For proportions, the distribution of observed values and the measurement process can influence which statistical method is suitable. For example, values bounded by the limits of a proportion may not always behave like an unconstrained normal measurement. General methods can account for different distributional features, but the registry record does not specify which method was used or whether model assumptions were checked. No reanalysis can be inferred from the outcome description alone.

How should statistical significance and practical meaning be separated?

In general, statistical significance concerns compatibility with a statistical null model, while practical or scientific meaning concerns the size and relevance of an effect in context. A small estimated difference can be precise, and a substantial estimated difference can remain uncertain; the estimate and its interval are needed to distinguish these situations. The registry record does not report an estimate or interval for the eye-gaze outcome. It therefore supports neither a claim of statistical significance nor a claim about the practical importance of a condition contrast.

These methods concepts clarify how results might be interpreted, but they do not identify methods actually used in the study unless the registry record states them. The record does not report the analysis model, covariates, handling of repeated observations, missing-data method, multiplicity procedure, or interim analysis approach. Any description of those choices beyond the listed planning assumptions would be speculative.

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("ASD - Addressed", "ASD - Overheard", "TD - Addressed", "TD - Overheard")
n   <- c(80, 82, 106, 104)
m   <- c(0.42, 0.44, 0.49, 0.38)
sp  <- c(0.35, 0.33, 0.34, 0.31)
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  ASD - Addressed  80  0.420  0.350  0.342  0.498
CARM  ASD - Overheard  82  0.440  0.330  0.367  0.513
CARM  TD - Addressed  106  0.490  0.340  0.425  0.555
CARM  TD - Overheard  104  0.380  0.310  0.320  0.440

11. Multiplicity, Missing Data and Analysis Populations

The registry record lists 2 outcome measures and 6 analyses, but does not describe the full set of hypotheses, how analyses correspond to outcomes, or which comparisons were prespecified. It also does not report a multiplicity adjustment or another strategy for interpreting results across analyses. In general, evaluating many outcomes or contrasts can increase the chance of an apparently unusual result under null hypotheses. Adjustments or a clearly defined hierarchy can help manage that issue, but no particular approach should be attributed to this study.

The record says that not all enrolled participants were analyzed. It gives examples of exclusions: experimenter error and children paying insufficient attention. It does not report the number excluded, the condition or group distribution of exclusions, whether exclusions were determined before analysis, or whether reasons differed by condition. It also does not report how missing or unusable gaze measurements were handled.

Exclusions can affect interpretation if inclusion depends on participant behavior, measurement quality, or condition. In general, an analysis of only participants with usable observations can differ from an analysis that includes all enrolled participants, especially when the reasons for missingness are related to the outcome. The registry record does not describe an analysis population, an intention-to-treat approach, sensitivity analyses, or a method for handling missing data. The effect of exclusions on the reported findings cannot be assessed.

The registry record does not report the statistical analysis plan or interim analyses. It is therefore not possible to determine which analyses were planned in advance, how deviations were handled, or whether the posted analyses represent all conducted analyses. These are reporting limitations rather than evidence that any particular method was or was not used.

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 serious adverse events, other adverse events, and deaths separately for the ASD and TD groups under both addressed and overheard conditions. In each listed group-condition entry, the reported count is 0 for serious adverse events, 0 for other adverse events, and 0 for deaths. The record lists 176 at risk for each ASD condition and 166 at risk for each TD condition.

These figures describe what the registry record reports; they should not be read as proof that no adverse experiences occurred. A zero count can mean that no events were recorded under the registry's reporting definitions and collection procedures. The record does not provide further information about active safety monitoring, how events were elicited, the observation period for safety, or whether the at-risk counts represent analyzed participants.

The study's listed interventions are behavioral and involve video-based teaching interactions. The safety entries do not provide event narratives or describe how safety data were collected. The top serious-event terms list is empty. This supports a narrow reading of the posted safety record: no serious adverse events, other adverse events, or deaths are recorded in the listed groups. It does not establish that all possible adverse experiences were assessed or absent.

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

The record establishes that the study asked whether children with ASD can learn from addressed and overheard teaching via videoconferencing, with TD children serving as a control. It describes two video-based teaching conditions and identifies a primary measure of gaze toward an image of the target word immediately after a 3-minute interaction. Those details define the research question and its behavioral endpoint.

The record does not provide the primary outcome values or statistical comparisons. It therefore cannot establish whether gaze toward the target object was greater in one condition, whether a contrast differed between ASD and TD children, or how uncertain any estimated difference was. No conclusion about the direction, size, or statistical significance of an effect can be drawn from the registry description alone.

Even if gaze results were available, the endpoint would represent a specific behavior rather than a complete measure of word knowledge. Gaze directed toward the target image may be relevant to learning, but it does not by itself demonstrate later recall, use of a word, or retention. The recorded timeframe is immediate after the interaction; longer-term outcomes are not described in the primary endpoint definition.

The non-randomized allocation also matters. A condition difference in a non-randomized design may reflect factors beyond the teaching format, depending on allocation and analysis. The crossover label suggests repeated conditions, but the registry record does not explain condition order or analysis of within-child measurements. Those details are important for assessing whether a contrast isolates the effect of teaching format.

The study is marked completed, but completion status does not substitute for outcome reporting. The enrollment and safety entries provide context about the registered study, while the absence of endpoint results leaves efficacy interpretation unavailable. The registry record supports a description of what was intended and measured, not a claim that either teaching condition succeeds or fails.

14. Why This Trial Matters Statistically

This study raises a statistical question about how to compare behavioral measurements after two teaching formats when the design is described as crossover and non-randomized. The repeated-condition structure makes within-child dependence important in general, while the non-randomized allocation limits the assumptions that can be made about comparability. A clear account of condition order, allocation, and analysis would help readers understand which contrasts the estimates support.

The endpoint also illustrates the distinction between a measurable behavioral proxy and the broader concept it is intended to inform. Proportion of time looking toward a target image is concrete and quantifiable. Its relationship to word learning, retention, or use is a separate interpretive question. Statistical precision for the gaze measure would not, by itself, establish the validity of a broader inference about learning.

Reporting exclusions is important in studies where attention or measurement quality can determine whether a participant is analyzed. The registry record gives examples of reasons for exclusion but not their frequency, distribution, or analysis consequences. Without those details, readers cannot judge whether the analyzed population may differ systematically from the enrolled population.

The record also lists 2 outcome measures and 6 analyses without describing the analysis family or multiplicity approach. Transparent reporting of prespecified outcomes and comparisons helps readers distinguish planned tests from exploratory analyses. Here, the registry record does not provide the statistical plan or primary endpoint results, so the statistical evidence cannot be independently appraised.

The central statistical lesson is that a study question and design description are not enough to determine what the results show. Estimates, measures of uncertainty, analysis methods, and clear accounting of participants and exclusions are needed to assess the strength and limits of evidence.

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 registry record contains no numerical results for the primary eye-gaze outcome. Means, standard deviations, condition contrasts, confidence intervals, and p-values are not reported in the record information. As a result, the magnitude and precision of the primary finding cannot be evaluated.

Design reporting is limited. Allocation is identified as non-randomized, masking as single, and the model as crossover, but the registry record does not explain the allocation procedure, who was masked, condition sequence, order balance, or how repeated measurements were analyzed. These details are especially relevant to interpreting comparisons between teaching conditions.

The record states that some enrolled participants were not analyzed, with examples including experimenter error and insufficient attention. It does not give exclusion counts by condition or group, describe the analysis population, or report the handling of missing gaze data. The representativeness of the analyzed participants and the impact of exclusions are therefore unclear.

The registry record lists multiple outcomes and analyses but does not describe prespecification, multiplicity control, the statistical model, covariates, or interim analyses. It also does not report a sample-size justification. The planning assumptions described for an illustration should not be mistaken for the study's own design rationale.

The primary endpoint captures gaze immediately after viewing the interaction. The record does not report later measurement of word retention or use, or additional details about gaze measurement and processing. This limits interpretation of the endpoint as evidence about learning beyond the immediate behavioral response. Finally, the zero adverse-event counts should be interpreted as registry reporting, with further safety collection details not reported in the registry record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Study 2: Learning New Words From Overhearing in Children With Autism Spectrum Disorder (ASD) — NCT05192109.
• Registry record last updated 2026-04-01; status last verified 2024-11.
• Results first posted 2025-03-20.
• 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.

19. Record Summary

NCT05192109 is a completed, non-randomized crossover study sponsored by New York University. The registry record lists enrollment as 342, phase as NA, a start date of 2020-12-01, and a primary completion date of 2024-12-01.

The study concerns addressed and overheard video-based teaching in children with ASD and TD children. Its primary outcome is the proportion of time children's gaze is directed toward the target object immediately after a 3-minute interaction, summarized as a mean with standard deviation. Some enrolled participants were excluded from analysis, including for experimenter error or insufficient attention.

The registry record lists 2 outcome measures and 6 analyses but does not report primary endpoint results or the statistical analysis plan. The listed safety entries report 0 serious adverse events, 0 other adverse events, and 0 deaths for each group-condition entry. These figures describe the registry record and do not establish the completeness of safety ascertainment.