Influence of Stress on Encoding and Prediction: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Influence of Stress on Encoding and Prediction (NCT05920161), a randomized na study. The primary outcome is Changes in fMRI Signal.

2. Clinical Question

The clinical question is therefore best understood as a laboratory question about stress exposure, learning, and neural measures, rather than as a comparison of clinical treatments. The registry describes a learning period lasting 1-2 hours and says fMRI signal would be assessed using connectivity, multivariate, and univariate approaches. It does not report the exact operational definitions of pair familiarity or the full set of analyses used for each fMRI approach. Nor does it describe how the broader outcomes of statistical learning and episodic encoding were combined into a single conclusion.

Random assignment makes the two groups relevant for studying effects of assignment to the stress-induction condition, but the primary reported quantity is a correlation within each group. A correlation describes how two measured quantities vary together; it is not itself a randomized between-group contrast. The record lists separate correlation values and one-sided tests for the two groups, but does not report a formal test comparing their correlations. That distinction limits what can be concluded about whether stress altered the relationship between connectivity change and familiarity.

3. Trial Design

NCT05920161, titled “Influence of Stress on Encoding and Prediction,” is listed as completed. The sponsor is Yale University. The registry lists the study start as 2023-03-02, primary completion as 2024-10-04, and enrollment as 106. The phase is recorded as NA. The study used randomized allocation, no masking, and a parallel model. These design labels describe assignment and organization, but do not establish how randomization was generated, whether allocation was concealed, or how outcome assessment was handled; those details are not reported in the registry record.

The two arms were Acute Stress and No Stress. The former involved the socially evaluated cold pressor test, described as an ice-bath exposure, while the latter involved a matched warm-water condition without stress-related exposure. Both interventions are categorized as behavioral. The record describes the purpose as investigating neural mechanisms by which acute stress influences statistical learning and episodic encoding. The fMRI learning period is described as lasting 1-2 hours.

The posted primary outcome is “Changes in fMRI Signal,” measured using Pearson’s rho. The arm-level outcome entries include 45 participants for Acute Stress and 43 for No Stress. The registry also lists two secondary outcomes: Cortisol Reactivity and Alpha-amylase Reactivity. Their reported outcome counts are 49 for Acute Stress and 47 for No Stress in each case. The reasons that the outcome denominators differ from one another and from the primary outcome denominators are not reported in the registry record.

The record lists 6 outcome measures and 15 analyses, while only two secondary outcomes are posted in the results. It does not describe the full analysis population, the handling of missing measurements, or whether the reported analyses follow an intention-to-treat principle. It also does not report a sample-size justification or a prespecified statistical analysis plan. These omissions matter when interpreting how the planned study relates to the posted estimates and tests.

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: 106 participants were randomized, 9 participants withdrew prior to receiving any intervention therefore 97 started

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 Changes in fMRI Signal, measured over 1-2 hours.

The primary outcome measures Pearson’s rho, a correlation coefficient, for the relationship between change in CA1-EC background connectivity and pair familiarity. The registry says the fMRI signal during learning was assessed with connectivity, multivariate, and univariate approaches over a 1-2 hour learning period. It defines rho as the population correlation, with values from -1 to +1: values nearer +1 describe a stronger positive linear association, values nearer -1 a stronger negative linear association, and values nearer zero little or no linear association.

This outcome is used to describe whether variation in a neural measure accompanies variation in familiarity across learning observations. It is not a measure of the size of a randomized group difference, and it does not by itself show that a change in connectivity causes a change in familiarity. The posted correlations are 0.261 for Acute Stress and -0.264 for No Stress. These point in different directions, but the registry does not report confidence intervals for either correlation or a direct inferential comparison between them. The separate one-sided tests listed by arm do not, on their own, establish that the two correlations differ.

Although the primary outcome title refers broadly to changes in fMRI signal, the specific description focuses on the correlation between connectivity change and pair familiarity. The record does not give a single endpoint definition covering each connectivity, multivariate, and univariate approach, nor does it explain how these approaches relate to the posted rho values. Interpretations should therefore remain tied to the stated correlation rather than treating it as a comprehensive measure of all neural activity or learning.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The statistical quantities in this study answer different questions, so their roles should be kept distinct. The primary outcome is Pearson’s rho, a measure of the direction and strength of a linear association between change in CA1-EC background connectivity and pair familiarity. A positive rho indicates that higher values of one measure tend to accompany higher values of the other; a negative rho indicates that higher values of one tend to accompany lower values of the other. A value nearer zero indicates a weaker linear association. Correlation does not establish cause and effect, and it can fail to capture non-linear patterns.

What does an arm-specific correlation show?

The reported primary values are 0.261 for Acute Stress and -0.264 for No Stress. Each describes an association among measurements in its own group. A question about whether stress changes that association requires a direct comparison of the group correlations, or another model testing the relevant interaction. The registry does not report such a comparison. Separate tests within each arm do not substitute for it, even when the signs of the estimates differ.

The registry lists one-sided t-tests with p-values for each arm. A p-value describes how compatible the observed test statistic is with a specified null hypothesis under the test’s assumptions. A one-sided test evaluates evidence in a specified direction, but the record does not state the exact null, direction, or detailed test construction. The p-values therefore cannot be interpreted as a fully documented test of a stress-versus-control difference. A p-value is also not the probability that the null hypothesis is true, nor is it a measure of the size or practical importance of an association.

How should the secondary comparisons be read?

Cortisol Reactivity and Alpha-amylase Reactivity are reported as means with standard deviations in each arm. A mean summarizes the central value of the measured outcome, while a standard deviation describes the spread of individual measurements around that mean. The recalculated comparisons use the posted summary statistics to estimate a simple, unadjusted difference in means, expressed as No Stress relative to Acute Stress. They are not adjusted model estimates, and they should not be mistaken for the sponsor’s separately reported arm-specific tests.

A confidence interval gives a range of values compatible with the data and the method used to construct the interval. A comparatively narrow interval indicates greater precision than a wider interval under the same method. If an interval includes zero for a difference in means, the data do not clearly distinguish the estimate from no mean difference at the interval’s confidence level. If it excludes zero, that is evidence against a zero difference under the method’s assumptions; it does not establish that the difference is important or that the analysis was prespecified. The registry does not state an adjustment for the number of outcomes or tests, so the recalculated secondary intervals and p-values are unadjusted.

What do the planning illustrations mean?

The sample-size illustrations are based on the observed secondary effect, not on a reported design calculation. They use a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation, with the target effect set to the stated fraction of the observed effect. The illustrations report participants per arm for 80% or 90% power. They are planning scenarios, not observed power and not evidence that the trial was designed with those sample sizes. The registry does not report its own sample-size justification.

Observed-effect planning illustrations can be highly sensitive to the estimated effect. When an observed difference is close to zero relative to its variability, the illustration can imply a very large sample requirement. That is a property of the calculation’s assumptions and the observed estimate, not a finding that a future study must enroll a particular number. Planning with a smaller fraction of the observed effect changes the target effect and therefore changes the illustrative sample size. These calculations do not account for unreported design features such as covariate adjustment, clustering, missing measurements, or multiplicity.

Why do analysis denominators matter?

The primary outcome lists 45 participants for Acute Stress and 43 for No Stress. The two secondary outcomes each list 49 and 47, respectively. Different denominators mean that the outcomes are not necessarily based on the same set of observed measurements. Without a reported analysis population, missing-data method, or account of exclusions, it is not possible to determine how those differences arose or how they may affect comparability. Randomization describes assignment; it does not by itself guarantee that every randomized participant contributes data to every outcome.

Finally, a statistically significant result is not a substitute for an estimate and its precision. The primary correlation estimates have no posted confidence intervals, and the primary arm-specific p-values do not provide a direct group comparison. The secondary recalculations offer mean differences and intervals, but their unadjusted status and secondary role should remain explicit. Statistical methods are informative only to the extent that their target questions, assumptions, and analysis populations are clear.

10. Analyses Reported by the Sponsor

The registry’s own primary analyses list one-sided t-tests separately for Acute Stress and No Stress, with p-values of 0.023 and 0.068. Those reported tests accompany the arm-specific correlations, but the record does not give a direct test comparing the correlations between randomized groups. The sponsor’s entries therefore do not answer the same question as a between-group comparison would, and their exact null hypotheses are not described.

For the secondary outcomes, the registry likewise lists one-sided tests separately by arm. The recalculated comparisons instead use the posted means and standard deviations to estimate unadjusted differences between arms. For Cortisol Reactivity, the registry entries are p < 0.000 for Acute Stress and p < 0.001 for No Stress; the recalculated between-arm mean comparison has an unadjusted p-value of <0.001. For Alpha-amylase Reactivity, the registry entries are p = 0.083 and p = 0.088, whereas the recalculated between-arm comparison has an unadjusted p-value of 0.984. These values should not be treated as competing estimates of one identical test: they concern different reported analyses.

No discrepancy is flagged between the posted summary data and the recalculated secondary comparisons. Recalculation is limited to simple unadjusted comparisons from those summaries and cannot reproduce an analysis that depends on participant-level data or unreported modeling choices. The sponsor’s own results remain the record’s reported analyses; the recalculations offer a transparent, limited comparison rather than a replacement for a complete statistical analysis.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 6 outcome measures and 15 analyses, but it does not report how multiplicity was handled. Multiplicity arises when several outcomes, contrasts, or tests are examined: without a plan, the overall chance of at least one apparently unusual result can be greater than the error rate associated with any one test. A prespecified hierarchy, adjustment procedure, or clear distinction between confirmatory and exploratory analyses can help readers understand how a set of results should be interpreted. None is described in the registry record.

This omission is relevant to both the primary and secondary results. The primary outcome has separate one-sided tests listed for each group, but no reported test directly comparing the group correlations. The secondary recalculated comparisons are explicitly unadjusted. The registry’s arm-specific secondary tests are also listed without an explanation of their role among the 15 analyses. It is therefore not possible to determine from the record which results, if any, were intended as the principal confirmatory tests or whether the reported p-values were adjusted.

Missing-data handling and analysis populations are also not reported in the registry record. The primary outcome counts are 45 for Acute Stress and 43 for No Stress; the secondary outcome counts are 49 and 47 for each group. These denominators establish that the posted outcome summaries use different numbers of participants, but do not explain why. The record does not state whether data were missing, whether participants were excluded from particular analyses, how incomplete measurements were handled, or whether analysis followed assignment regardless of available outcome data.

Randomized allocation does not answer those questions. If availability of measurements is related to group assignment or outcome values, analyses restricted to participants with observed data may not represent the full randomized groups in the same way. That is a general methodological concern, not a claim that such bias occurred in this study. Without an account of missingness and analysis populations, readers cannot assess its relevance. A statistical analysis plan, participant flow, and outcome-specific denominators with reasons for differences would clarify these issues; they are not reported in 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.

The registry reports zero serious adverse events in both groups: the Acute Stress group lists 50 participants at risk, and the No Stress group lists 47. It also reports zero other adverse events in each group and zero deaths in each group. No serious adverse-event terms are listed.

These counts should be read as the adverse events recorded in the registry for the stated at-risk groups. They are not the same denominators as the posted primary or secondary outcome analyses, and the record does not explain the difference. Zero recorded events does not demonstrate that no event could occur, nor does it establish that the conditions are risk-free. The counts also do not convey the completeness or duration of adverse-event monitoring, which are not reported in the registry record.

The intervention was a laboratory stress-induction procedure, and the comparator was a warm-water condition. The adverse-event figures describe the reported safety record for those study arms; they do not support broader conclusions about other settings, exposure patterns, or populations. No medical advice or clinical safety conclusion follows from these counts alone.

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

The results describe a randomized laboratory study of acute stress, learning, and fMRI measures. For the primary outcome, the observed relationship between CA1-EC background connectivity change and pair familiarity is positive in the Acute Stress group and negative in the No Stress group. The posted arm-specific tests give different p-values, but the registry does not report a direct test of the difference between those correlations or confidence intervals for either estimate. The primary results therefore describe the observed associations without establishing how precisely the correlations differ across groups.

The secondary results address cortisol and alpha-amylase reactivity through group means and, in the recalculations, unadjusted between-arm differences. Those findings do not measure the same quantity as the primary correlation. They also come with analysis denominators that differ from the primary outcome, and the reason is not reported. Secondary outcomes and multiple analyses warrant cautious interpretation because no multiplicity strategy is reported.

The study’s randomized design supports a structured comparison of assignment to the two laboratory conditions. However, design labels alone do not answer whether the primary neural association was altered by assignment. That requires an appropriate direct contrast, which is not reported in the registry record. Nor do these outcomes establish mechanisms beyond the particular measures described, show that one measure causes another, or establish effects outside the study setting.

The registry record also does not report participant characteristics, detailed eligibility criteria, full analysis methods, missing-data handling, or the study’s own sample-size rationale. Those omissions limit assessments of generalizability, precision, and susceptibility to analytic choices. The planning illustrations are based on observed effects and stated assumptions; they are not observed power and do not reveal the original design rationale. Overall, the posted results are best read as outcome-specific estimates and tests whose meaning depends on the distinction between within-group association and between-group comparison, and on statistical details that are not reported.

14. Why This Trial Matters Statistically

This study illustrates a central issue in interpreting randomized studies with biological and neural outcomes: the type of estimand matters. The primary outcome is a correlation within each arm, while the randomized comparison of interest concerns whether assignment to Acute Stress versus No Stress changes that relationship. Two arm-specific estimates can point in different directions, but a direct statistical contrast is needed to quantify evidence that they differ. The registry record does not report that contrast.

The study also highlights why p-values must be matched to their hypotheses. The reported one-sided tests are listed separately by arm, while the recalculated secondary results compare means between arms. Their p-values are not interchangeable because the tests address different questions. Likewise, an unadjusted secondary result should not automatically be interpreted as confirmatory when the record lists multiple outcomes and analyses without describing a multiplicity plan.

Differences in outcome denominators matter for the same reason: randomization happens at assignment, but each outcome analysis may include a different set of observed measurements. Without explanations for those denominators, readers cannot assess whether analysis populations are comparable across outcomes or whether incomplete data could affect interpretation. These are not reasons to discard the results; they identify what information is needed to evaluate them fully.

The registry’s sample-size illustrations further distinguish planning from inference. They use observed effects under stated assumptions and describe hypothetical participants per arm for target power levels. They are not observed power and do not document the original study’s design rationale. Separating these concepts helps prevent a post-result calculation from being mistaken for evidence that the study was adequately powered for its primary question.

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 trial-specific limits follow from what is and is not reported. The primary outcome has arm-level correlations and one-sided p-values, but no confidence intervals and no direct comparison between the randomized groups. As a result, the precision of each correlation and the evidence for a between-group difference cannot be assessed from the registry’s primary results. The exact null hypotheses and construction of the arm-specific tests are also not described.

The registry gives a broad description of fMRI assessment using connectivity, multivariate, and univariate approaches, but does not specify how these approaches relate to the posted primary rho values. Details of pair familiarity, the connectivity-change measure, and any selection among analytic approaches are not reported in the registry record. The record also does not report whether outcomes or analyses were prespecified in a protocol or statistical analysis plan.

Outcome denominators vary: the primary outcome lists 45 and 43 participants, while each secondary outcome lists 49 and 47. Reasons for these differences, participant flow, exclusions, missing-data methods, and analysis-population definitions are not reported. Participant characteristics and eligibility details are also absent from the record, limiting judgments about how well the study population represents other groups or settings.

Finally, multiplicity handling is not reported despite the listed 6 outcome measures and 15 analyses. The secondary recalculations are unadjusted, and the planning illustrations use the observed effect rather than the trial’s own reported design assumptions. These limitations do not establish that a particular bias or analytic problem occurred; they identify uncertainties that constrain interpretation of the posted estimates.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Influence of Stress on Encoding and Prediction — NCT05920161.
• Registry record last updated 2025-10-15; status last verified 2025-09.
• Results first posted 2025-10-15.
• 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

NCT05920161 is a completed, randomized, parallel, unmasked study sponsored by Yale University. The registry lists enrollment of 106, with Acute Stress and No Stress arms. The stress condition used a socially evaluated cold pressor test involving an ice bath; the comparator used a matched warm-water condition without stress-related exposure. The study’s stated purpose was to investigate how acute stress influences statistical learning and episodic encoding.

The primary outcome concerns Pearson’s rho for the relationship between change in CA1-EC background connectivity and pair familiarity during a 1-2 hour learning period. Posted correlations are 0.261 for Acute Stress and -0.264 for No Stress. Separate one-sided tests are listed, but a direct comparison between the correlations is not reported. Secondary outcomes are Cortisol Reactivity and Alpha-amylase Reactivity. Their recalculated unadjusted comparisons address differences in means, not the primary correlation question.

The registry does not report multiplicity handling, missing-data methods, analysis-population definitions, or a sample-size justification. It records zero serious adverse events, zero other adverse events, and zero deaths in both groups. These findings describe the posted registry results and should be interpreted with the stated analysis and reporting limitations.