Sex Differences, Cognitive Training & Emotion Processing: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Sex Differences, Cognitive Training & Emotion Processing (NCT03137654), a randomized na study. The primary outcome is Emotional Stroop: Change From Baseline in Reaction Time.

2. Clinical Question

3. Trial Design

This was a randomized, parallel-group pilot project with no masking. The registry lists three groups: Affective Training, Neutral Training, and Control (Non-active). The registered enrollment is 82. The primary outcome results are reported for 15 participants in Affective Training, 13 in Neutral Training, and 14 in Control (Non-active). These outcome counts do not explain the relationship between enrolled participants and those included in each outcome analysis; participant flow and reasons for any differences are not reported in the registry record.

The active groups completed baseline assessment, 3 wks of training, and post-training assessment. The training tasks differed in their stimuli: emotionally laden material for Affective Training and neutral material for Neutral Training. The non-active control completed baseline assessment and another assessment approximately three weeks later, without active intervention. Each arm description also says participants would be contacted monthly for 3 months after discharge for follow-up interviews.

The primary outcome timeframe is Baseline and Post-Training (3 weeks). The record lists Timeline Follow-back, Moos Health and Daily Living Scale, Change in Profile of Mood States, and Mini Alcohol Craving Experience Questionnaire as secondary outcomes. It also lists 13 outcome measures and 9 analyses. The protocol’s allocation procedure, masking rationale, recruitment setting, participant characteristics, and original sample-size justification are not reported in the registry record. The registered trial status is completed.

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

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 Emotional Stroop: Change From Baseline in Reaction Time, measured over Baseline and Post-Training (3 weeks).

The primary endpoint is change from baseline in reaction time on incongruent trials of an Emotional Stroop task, measured in milliseconds. In the task, target words are presented over non-target visual stimuli; an example is a negative word presented over a happy face. The stated dependent measure focuses on reaction time when the word and visual stimulus are incongruent.

Change is defined as baseline reaction time minus post-training reaction time. Under that definition, a positive change score indicates faster processing at post-training than at baseline. The endpoint therefore expresses the direction and magnitude of within-person change over the training interval, rather than a post-training reaction time alone.

Reaction time is used as a continuous measure of performance on the interference task. It can capture variation in processing speed across participants and changes across assessments. The endpoint is specific to performance on this task; by itself, it does not establish change in alcohol use, treatment outcomes, or everyday functioning. The registry lists several secondary measures related to alcohol use, health and daily living, mood, and craving, but their numerical results are not included in the registry record described for this analysis.

7. Results Posted in the Registry

• Affective Training: mean 336.00 versus 78.30 on control; difference 257.70 (95% CI 42.07 to 473.33), standardized effect d = 0.85; z-test p = 0.019.
• Neutral Training: mean 162.00 versus 78.30 on control; difference 83.70 (95% CI -316.46 to 483.86), standardized effect d = 0.16; z-test p = 0.682.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary measure is continuous: it is a reaction-time change score measured in milliseconds. A mean summarizes the group’s average change, while a standard deviation describes how spread out individual scores are around that mean. A larger standard deviation means the observed scores are more variable; it does not describe uncertainty in the mean by itself. In these results, the reported standard deviations differ noticeably across groups, so the group averages should be read alongside the within-group variability.

What does a mean difference describe?

A mean difference expresses the distance between two group means on the outcome’s original scale. The recalculated comparisons use the control mean of 78.3 milliseconds as the reference. The Affective Training comparison has a mean difference of 257.7 milliseconds, and the Neutral Training comparison has a mean difference of 83.7 milliseconds. Because the endpoint is defined so positive change represents faster processing, a positive difference indicates a higher average change score in the training group than in control. It does not, on its own, show that the difference is precise or that it is a direct comparison between the two training strategies.

These estimates are recalculated from posted summary statistics and are unadjusted. An unadjusted comparison does not incorporate additional covariates or the structure of a model unless those are explicitly part of the calculation. The registry’s reported Linear Mixed Model is a distinct analysis. A mixed model can account for relationships among repeated observations or other modeled data structure, depending on its specification. The model details are not reported in the registry record, so the exact role of the repeated measurements and the model’s contrast definitions cannot be established from the reported result alone.

How should the p-values be read?

A p-value describes how compatible the data are with a specified statistical null model, given the assumptions of the test. It is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect. The recalculated unadjusted p-values are 0.019 for Affective Training versus control and 0.682 for Neutral Training versus control. The registry’s Linear Mixed Model result reports p = .0078 for a superiority analysis involving all three groups. Because the contrast represented by that model result is not described, it should not be read as if it were necessarily the same test as either recalculated pairwise comparison.

When multiple groups, outcomes, or comparisons are considered, the chance of obtaining at least one apparently unusual result can change. General statistical practice may address this through a prespecified hierarchy or adjustment, but the registry record does not report a multiplicity procedure. The reported p-values therefore need to be interpreted in the context of the analysis that generated each one; the available information does not establish whether they were adjusted for other outcomes or comparisons.

Why does variability matter?

Uncertainty in a group comparison depends on both the size of the observed difference and the variability and amount of information in the groups. High variation makes it harder to estimate a mean precisely. Neutral Training has a standard deviation of 713.0 milliseconds, which is large relative to its mean change of 162.0 milliseconds and contributes to the broad uncertainty in its recalculated comparison with control. A broad interval can be compatible with a range of effects, including effects on either side of zero.

Over-dispersion is a general concept used when observed variability exceeds what a model’s assumed distribution would allow. It can make standard errors too small if the model fails to account for that extra variability. The summary statistics alone do not establish that a particular model had over-dispersion; the registry does not report diagnostics or sufficient model detail to assess this issue. The large and differing standard deviations are reasons to attend carefully to the model assumptions and uncertainty, not proof that a specific assumption was violated.

What role does the randomized design play?

Random allocation is intended to support comparisons by reducing systematic differences between groups at assignment, on average. The registry identifies the study as randomized, but the method used to generate or conceal allocation is not reported. No masking was used. The parallel-group structure means the listed groups received their assigned study conditions during the same trial rather than crossing through each condition, as described in the record.

Randomization does not ensure that every group is identical on all characteristics, especially in a pilot study, and it does not remove uncertainty from a small or variable outcome dataset. The participant counts attached to the primary outcome are 15, 13, and 14 in the three respective arms. The registry does not explain the analysis population or missing-data handling, so the effect of exclusions or incomplete assessments cannot be evaluated from these results.

What is the distinction between the recalculation and sponsor analysis?

The pairwise estimates and intervals are unadjusted recalculations from the posted group summaries. The reported sponsor analysis is a Linear Mixed Model with a superiority result across the three groups. These methods can differ in their use of repeated observations, assumptions, and contrast definitions. The model’s estimate and confidence interval are absent from the registry record, so its effect size and precision cannot be compared directly with the recalculated pairwise estimates. Because the record flags a discrepancy, the sponsor’s own analysis takes precedence. The recalculations are useful as transparent descriptive checks, not as replacements for the sponsor’s analysis.

10. Confidence Intervals and Precision

A confidence interval describes a range of effect estimates that are reasonably compatible with the data and the method used to calculate the interval. Its width conveys precision: narrower intervals indicate less uncertainty in the estimated difference, while wider intervals indicate more. An interval is not a guarantee that the true effect lies within its bounds, and its interpretation depends on the analysis assumptions.

For Affective Training compared with Control (Non-active), the recalculated mean difference is 257.7 milliseconds, with a 95% confidence interval from 42.07 to 473.3. This interval is entirely above zero, but its breadth still permits a range of possible magnitudes. The result is therefore more informative about the direction of the unadjusted difference than about a single precise size of that difference.

For Neutral Training compared with Control (Non-active), the recalculated mean difference is 83.7 milliseconds, with a 95% confidence interval from -316.5 to 483.9. This interval includes zero and spans negative and positive values. The data summarized by this comparison are compatible with a range of possible differences, so the estimate is imprecise.

Both intervals are recalculated and unadjusted. The registry’s reported Linear Mixed Model result has no associated estimate or confidence interval, so its precision cannot be assessed from the reported p-value alone. The standard deviations, including 713.0 milliseconds in Neutral Training, provide further context for the variability in the underlying outcome, but they are not confidence intervals for the group means or differences.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are participants per arm, not total events. They apply to the continuous primary outcome comparisons between each training group and the non-active control. The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, a normal approximation, and an effect to detect set to the stated fraction of the observed effect. The registry’s original design assumptions and sample-size justification are not reported in the registry record.

The 80% power column gives the planned participants per arm under those assumptions for an 80% chance of detecting the specified effect if that effect is present. The 90% power column gives the corresponding participants per arm for a 90% chance. These are design-planning quantities, not a statement about the probability that the trial’s findings are correct.

For Affective Training versus control, the planning figures for 100% of the observed effect are 22 participants per arm for 80% power and 29 participants per arm for 90% power. For 75% of the observed effect, they are 39 and 52; for 50%, they are 87 and 116. For Neutral Training versus control, the corresponding figures are 589 and 789 for 100% of the observed effect, 1,047 and 1,402 for 75%, and 2,356 and 3,154 for 50%. Smaller effects require more participants under the same assumptions. These figures are planning illustrations based on observed estimates, not observed power.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = 257.700; pooled standard deviation σ = 302.439.
2. n = 2 × (1.96 + 0.842)² × 302.439² / 257.700² = 21.6, rounded up to 22 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
delta <- 257.7                      # observed difference in means
sigma <- 302.439219482992                     # pooled standard deviation
n_per_arm <- function(power, frac = 1) {
  ceiling(round(2 * (z_a + qnorm(power))^2 * sigma^2 / (delta * frac)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (t-based, so very slightly larger)
cat("power.t.test:", round(power.t.test(delta = delta, sd = sigma, power = 0.80)$n, 1), "\n")

# Simulation: exact two-sample t-test using simulated means and variances
set.seed(3137654)
n    <- n_per_arm(0.80)
reps <- 20000
m_diff <- rnorm(reps, delta, sigma * sqrt(2 / n))             # difference in sample means
s2     <- sigma^2 * rchisq(reps, 2 * (n - 1)) / (2 * (n - 1))   # pooled sample variance
tstat  <- m_diff / sqrt(s2 * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(tstat) > qt(0.975, 2 * (n - 1))), reps))

Output from running the code

100% of observed effect: n per arm = 22 (80% power), 29 (90% power)
 75% of observed effect: n per arm = 39 (80% power), 52 (90% power)
 50% of observed effect: n per arm = 87 (80% power), 116 (90% power)
power.t.test: 22.6 
Simulated power at n = 22 per arm: 0.789 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Analyses Reported by the Sponsor

The registry reports a sponsor analysis using a Linear Mixed Model, classified as a superiority analysis across the three groups, with p = .0078. The result does not include a model-based effect estimate, confidence interval, or enough detail to identify the precise contrasts being tested. It should therefore be kept distinct from the recalculated pairwise comparisons with control.

The recalculated Affective Training comparison with control has a mean difference of 257.7 milliseconds, a 95% confidence interval from 42.07 to 473.3, and an unadjusted p-value of 0.019. The recalculated Neutral Training comparison with control has a mean difference of 83.7 milliseconds, a 95% confidence interval from -316.5 to 483.9, and an unadjusted p-value of 0.682. Those estimates use posted summary statistics and do not reproduce the reported mixed-model analysis.

The record flags a discrepancy between the recalculation and the sponsor’s reported analysis. The sponsor’s own analysis takes precedence when interpreting the trial’s reported inferential result. The pairwise calculations can serve as transparent checks of the posted group summaries, but they should not be used to override, combine with, or claim to explain the mixed-model result. The registry does not report enough methodological detail to resolve the discrepancy.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists 13 outcome measures and 9 analyses. It identifies four secondary outcomes: Timeline Follow-back, Moos Health and Daily Living Scale, Change in Profile of Mood States, and Mini Alcohol Craving Experience Questionnaire. The primary endpoint is the Emotional Stroop change in reaction time. The record does not describe a prespecified hierarchy among outcomes, an adjustment for multiple comparisons, or how the reported analyses were selected from the listed measures. Whether multiplicity was addressed is therefore not reported in the registry record.

Multiplicity matters because evaluating several outcomes or contrasts can increase the chance that at least one result appears noteworthy under a null model. This general concern does not establish that any particular reported result is false. It does mean that isolated p-values should be interpreted with awareness of the broader analysis set, which is not fully described in the reported results.

Missing-data methods are not reported in the registry record. The record does not state whether incomplete outcome measurements were imputed, excluded, or handled by a model-based method. It also does not give participant flow by assessment, reasons for missing observations, or the number contributing to the reported mixed model. The outcome arm counts are 15 for Affective Training, 13 for Neutral Training, and 14 for Control (Non-active); the record does not explain how these counts relate to all randomized participants.

The analysis population is not reported in the registry record. It is not stated whether the sponsor’s analysis followed intention-to-treat principles, used a per-protocol population, or applied another inclusion rule. The planned follow-up interviews do not establish which follow-up data were collected or analyzed. These omissions limit assessment of possible bias from missing outcomes and of how closely the analyses followed the randomized assignment.

15. 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 lists serious adverse events and other adverse events separately for each arm. For Affective Training, the record lists 0 serious events among 15 at risk and 0 other events among 15 at risk, with 0 deaths. For Neutral Training, it lists 0 serious events among 13 at risk and 0 other events among 13 at risk, with 0 deaths. For Control (Non-active), it lists 0 serious events among 14 at risk and 0 other events among 14 at risk, with 0 deaths.

These figures describe what is recorded for the listed categories and groups. They do not establish that no symptoms or other experiences occurred, nor do they describe event severity beyond the serious-versus-other classification. The registry record does not provide event terms, observation details, or additional information needed to assess the completeness of safety reporting. The figures should be read as the reported adverse-event counts, not as proof that an intervention has no risks.

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

The posted primary outcome results describe changes in reaction time on incongruent Emotional Stroop trials from baseline to post-training. Average change scores are positive in all three groups under the stated scoring convention, with the largest group mean in Affective Training. In unadjusted recalculations against control, the Affective Training difference is estimated as 257.7 milliseconds, while the Neutral Training difference is estimated as 83.7 milliseconds. The uncertainty differs: the Affective Training interval lies above zero, whereas the Neutral Training interval spans zero and a wide range on either side.

The sponsor’s reported Linear Mixed Model analysis gives a superiority result of p = .0078 across the three groups, but the corresponding estimate, interval, and contrast details are not reported in the registry record. Because the record identifies a discrepancy, the sponsor’s analysis takes precedence over the simple recalculations. Neither the reported p-value nor the recalculated comparisons alone fully describe the size and precision of the model-based result.

The findings are limited to the reported task and post-training timeframe. They do not establish effects on alcohol use or the secondary outcomes, whose numerical results are not included in the registry record. They also do not establish whether men and women responded differently. A direct affective-versus-neutral estimate is not reported, so separate comparisons against control cannot determine whether the two training approaches differ from one another.

General interpretation of a pilot study should account for uncertainty, variability, and the study’s limited outcome information. The registry record does not report the original sample-size justification, participant flow, missing-data methods, analysis population, or detailed model specification. These gaps constrain conclusions about robustness and generalizability. The results are best understood as estimates of task performance in this trial, not as a definitive account of clinical benefit or harm.

17. Why This Trial Matters Statistically

This trial poses several statistical questions that commonly arise in clinical research. It includes two active strategies that differ in stimulus content and a non-active control, allowing the study to examine both training conditions against control within a randomized, parallel design. The primary outcome is a continuous change score, which makes interpretation dependent on the scoring direction, the variability of individual measurements, and the precision of group contrasts.

The results also illustrate why a group mean alone is not enough. Mean change scores differ across the arms, but their standard deviations vary substantially, and the recalculated confidence intervals convey very different levels of uncertainty. In particular, the Neutral Training comparison has a wide interval spanning zero. Reading estimates together with intervals helps distinguish the direction of an observed difference from how precisely its magnitude has been estimated.

A further statistical issue is the coexistence of a sponsor-reported mixed-model result and unadjusted pairwise recalculations. These analyses are not interchangeable. The registry does not provide the sponsor model’s estimate or contrast details, while it flags a discrepancy; careful interpretation therefore requires giving precedence to the sponsor’s analysis and treating the recalculations as descriptive checks rather than substitutes.

Finally, the project’s questions about sex differences and the relative value of affective versus neutral stimuli are not answered by the reported aggregate comparisons. Subgroup results and a direct active-arm contrast are not reported in the registry record. The trial therefore highlights the importance of reporting the estimates, uncertainty, analysis choices, and contrasts needed to connect a study’s stated questions with its numerical results.

18. 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 reported for this trial. The primary outcome results are available as group summaries, but the registry does not report participant flow or explain why the outcome counts differ from the registered enrollment. The counts contributing to each analysis, the analysis population, and the handling of incomplete assessments are not described. Without those details, the potential influence of missing observations or exclusions cannot be evaluated.

The sponsor’s Linear Mixed Model is reported with p = .0078, but the model-based estimate, confidence interval, covariates, repeated-measure structure, and precise contrasts are absent. The record flags a discrepancy with the simple unadjusted recalculations. This limits direct comparison between the sponsor result and the recalculated pairwise estimates, and prevents identifying the source of the discrepancy from the registry information alone.

The study description raises questions about sex differences and whether affective training differs from neutral training, but sex-specific estimates and a direct comparison of the two active arms are not reported. The displayed recalculations each compare a training arm with control; they do not test the difference between the active strategies. The registry also lists secondary outcomes without their numerical results in the registry record for this analysis.

Multiplicity procedures, missing-data methods, and the original sample-size justification are not reported in the registry record. No masking was used, and the allocation method is not described. The outcome is a task-based reaction-time measure at post-training, so these results alone do not establish effects on alcohol use or longer-term clinical outcomes. The planned follow-up schedule does not indicate which follow-up observations were collected or analyzed.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Sex Differences, Cognitive Training & Emotion Processing — NCT03137654.
• Registry record last updated 2025-05-31; status last verified 2025-05.
• Results first posted 2025-05-31.
• 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.

22. Record Summary

The registry record describes a completed randomized, unmasked, parallel-group pilot trial of Affective Training, Neutral Training, and a non-active control for Alcohol Use Disorder. The primary endpoint is change in reaction time on incongruent Emotional Stroop trials from baseline to post-training, with positive change indicating faster processing. Reported group means are 336.0 milliseconds, 162.0 milliseconds, and 78.3 milliseconds, respectively.

Unadjusted recalculations compare each training group with control, while the registry separately reports a Linear Mixed Model superiority analysis across the three groups with p = .0078. Because a discrepancy is flagged, the sponsor’s own analysis takes precedence. Model details, sex-specific findings, a direct active-arm comparison, missing-data methods, and the original sample-size justification are not reported in the registry record.