Exploring the Effects of Corticosteroids on the Human Hippocampus: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Exploring the Effects of Corticosteroids on the Human Hippocampus (NCT03896659), a randomized phase 4 study. The primary outcome is Hippocampal Subfield Activation (Left Hemisphere).

2. Clinical Question

The study description places the imaging question in a broader investigation of declarative memory and hippocampal function. It describes prior concerns about corticosteroid exposure and memory, and proposes examining both neurocognitive performance and multimodal neuroimaging. The registry lists four outcome measures, but only the primary outcome is described in detail.

The central statistical issue is not simply whether any one mean is above or below zero. The outcome is a standardized activation measure, and the comparison estimates must be interpreted in light of the groups and reference condition actually shown in the results. In particular, the posted contrast estimates use the depressed placebo group as their reference, including the comparisons labelled for healthy controls. They therefore should not be read as direct hydrocortisone-versus-placebo estimates within each participant group.

The page presents what the registry record reports and how uncertain the estimates are. It cannot establish from these summaries whether treatment sequences, within-participant correlations, or other crossover features were incorporated into a formal analysis.

3. Trial Design

This was a Phase 4, randomized, quadruple-masked crossover study sponsored by the University of Texas Southwestern Medical Center. The registry lists enrollment of 66 and status as completed. The study began on 2019-10-01, with primary completion on 2024-09-16.

The record describes two oral conditions: hydrocortisone and matching placebo. Each exposure lasted 3 days. The arm descriptions specify a washout period of 25 days between periods and include both treatment sequences: hydrocortisone followed by placebo, and placebo followed by hydrocortisone. These sequence descriptions appear for both the depressed group and healthy controls.

The study was intended to examine changes in declarative memory and use high-resolution structural and functional neuroimaging. The brief description identifies healthy controls and a depressed group, with the depressed group described as an exploratory aim. The primary outcome is left-hemisphere hippocampal subfield activation measured after 3 days.

The crossover structure means participants can contribute observations under more than one condition, but the posted summary is organized by group and condition. It does not show participant-level paired observations, sequence-specific results, or a model that accounts for repeated measurements. The record also does not explain how many people completed each period or whether all randomized participants contributed to every outcome summary.

For the primary outcome, the reported group counts are 13 for Depressed: Hydrocortisone, 11 for Depressed: Placebo, 15 for Healthy Controls: Hydrocortisone, and 16 for Healthy Controls: Placebo. These are the counts shown alongside the outcome summaries, not a reconstruction of the trial flow. The registry record does not report a participant-flow account that reconciles these outcome counts with enrollment.

Randomization and masking are stated in the registry, but details of sequence generation, concealment, masking procedures, and the original statistical design are not reported in the registry record. The posted results contain no registered analyses, so the methods underlying the sponsor’s primary analysis cannot be evaluated from this 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 Intervention (3 Days)

Washout (25 Days)

Second Intervention (3 Days)

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 Hippocampal Subfield Activation (Left Hemisphere), measured over 3 days.

The primary endpoint is hippocampal subfield activation in the left hemisphere, expressed as a Z-score, at 3 days. The score summarizes the standardized difference in activation between correct and incorrect conditions on a mnemonic discrimination task. The record explains that voxel-wise beta weights were estimated separately for correct and incorrect trials and represent blood oxygen level-dependent signal associated with each condition.

A score of 0 represents the mean difference across all voxels. A positive score indicates greater activation for correct than incorrect trials; a negative score indicates greater activation for incorrect than correct trials. Larger absolute scores indicate stronger condition-related effects. The registry explicitly notes that there are no established clinical thresholds for these experimental activation differences.

This measure is used to quantify a task-related brain signal, rather than a clinical diagnosis or a direct measure of everyday memory functioning. It gives researchers a standardized way to describe the difference between task conditions in the hippocampal subregions. The study’s broader rationale connects hippocampal function with declarative memory, but the posted primary outcome is the imaging activation score itself.

Because this is an experimental measure without established clinical thresholds, the direction and size of a score need to be interpreted as features of the task-related signal. They do not, on their own, establish the clinical importance of a difference. The registry provides group means and standard deviations for the outcome, but does not report confidence intervals for those individual group means.

7. Results Posted in the Registry

• Depressed: Hydrocortisone: mean -0.35 versus 0.03 on control; difference -0.38 (95% CI -0.82 to 0.06), standardized effect d = -0.65; z-test p = 0.090.
• Healthy Controls: Hydrocortisone: mean -0.08 versus 0.03 on control; difference -0.11 (95% CI -0.43 to 0.21), standardized effect d = -0.24; z-test p = 0.506.
• Healthy Controls: Placebo: mean -0.07 versus 0.03 on control; difference -0.10 (95% CI -0.40 to 0.20), standardized effect d = -0.23; z-test p = 0.512.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: it is a Z-score summarizing task-related activation. The posted group summaries therefore use means and standard deviations. A mean describes the center of the observed scores, while a standard deviation describes their spread around that mean. The mean and standard deviation are descriptive summaries; by themselves, they do not account for a crossover design or show how precisely a mean is estimated.

Reading the contrasts

The comparison estimates are presented as mean differences. A mean difference expresses the distance between two group means on the outcome’s scale. Here, the reference mean shown for all three comparisons is 0.03, from Depressed: Placebo, with 11 participants. The comparison means for the other groups are -0.35, -0.08, and -0.07. The corresponding reported differences are -0.38, -0.11, and -0.1.

This shared reference matters. The comparison labelled Depressed: Hydrocortisone compares that group with Depressed: Placebo. The comparisons labelled for healthy controls also use Depressed: Placebo as the reference. They are not within-healthy-control treatment contrasts. A crossover analysis would ordinarily need to reflect that observations from the same participant are related; general statistical practice recognizes that paired data differ from independent groups. The registry record does not report whether the posted comparisons incorporate this relationship, treatment sequence, or period.

Uncertainty and p-values

A confidence interval gives a range of values compatible with an estimate under the assumptions used to calculate it. Its width reflects both variability and the amount of information available. The reported 95% intervals for all three contrasts include zero. This means the posted intervals do not rule out a zero mean difference, while also including nonzero differences. The intervals should be considered together with their estimates and the study’s design, rather than as proof of no effect.

The reported p-values are unadjusted: 0.09, 0.506, and 0.512. A p-value measures how unusual the observed data, or more extreme data, would be under a specified null model and its assumptions. It is not the probability that the null hypothesis is true, nor does it measure the magnitude or clinical relevance of an estimate. Because the values are unadjusted, they do not account for testing multiple outcomes or contrasts. The registry does not report a multiplicity strategy or analysis plan.

Design and model considerations

General statistical practice for crossover trials distinguishes between within-participant comparisons and independent-group comparisons. Repeated observations from a person tend to be related, and the study sequence or period can also be relevant. The arm descriptions show treatment order and a washout period, but the posted results do not contain participant-level data or a statistical model description. It is therefore not possible from the registry record to establish whether these specific considerations were incorporated into an inferential analysis.

The record identifies the primary outcome’s dispersion as standard deviation and gives means and standard deviations for each arm. It does not report outcome distributions, diagnostic checks, transformations, or a method for handling outliers. General explanation: if a continuous outcome has markedly different variability across groups or is strongly skewed, model assumptions and the reliability of standard errors may be affected. There is no basis in this record to say whether those features occurred.

There are no registered analysis records, despite the primary outcome summaries and calculated comparisons appearing in the posted results. The registry record does not report the sponsor’s original analysis model, its covariates, its treatment of sequence and period, or its approach to incomplete measurements. Consequently, the comparison values can be described, but they cannot be verified as the intended or definitive analysis of the crossover trial.

Planning calculations

The accompanying sample-size figures are planning illustrations based on observed effects, not the trial’s original design justification. The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set to the stated fraction of the observed effect. The registry record does not report the original sample-size justification or original planning assumptions.

The table gives participant counts per arm for these continuous outcomes. General explanation: a power calculation estimates how many participants a hypothetical study would need under specified assumptions to have a chosen chance of detecting a specified effect, if that effect and model assumptions are correct. Power is not a property that can be inferred after the fact simply by looking at the observed p-value. These planning values should not be taken as evidence about the adequacy of the original trial design or as a substitute for a design based on a prespecified clinically meaningful effect.

General explanation: over-dispersion, when data vary more than a model assumes, can make standard errors too small and confidence intervals too narrow if not appropriately handled. That concept is relevant to model checking generally, but the registry does not report evidence of over-dispersion or identify a model to which it applies. Similarly, no interim analyses or stopping rules are reported in the registry record.

10. Confidence Intervals and Precision

The comparison estimates are accompanied by 95% confidence intervals. For Depressed: Hydrocortisone versus the Depressed: Placebo reference, the reported mean difference is -0.38, with an interval from -0.819 to 0.059. For Healthy Controls: Hydrocortisone versus that same reference, the estimate is -0.11, with an interval from -0.434 to 0.214. For Healthy Controls: Placebo versus the reference, the estimate is -0.1, with an interval from -0.399 to 0.199.

All three intervals include zero, and each also spans values in both directions. The range of compatible values is therefore broad enough that the posted results do not determine a single direction for each contrast with high precision. The depressed hydrocortisone interval is especially broad relative to its point estimate. These intervals describe uncertainty for the comparisons as calculated; they do not resolve whether the calculations appropriately account for the repeated-measures crossover design.

The per-group sample counts for this outcome are 13, 11, 15, and 16. The group standard deviations are 0.76, 0.25, 0.57, and 0.53 in the order shown in the endpoint summaries. Differences in spread and the number of observations contribute to precision, but the registry does not provide confidence intervals around each group mean. It also does not provide a full analysis report, participant-level measurements, or details needed to reproduce the intervals.

Precision is distinct from whether a result passes a threshold. An unadjusted p-value does not convey the full range of plausible values, and a confidence interval that includes zero does not establish equivalence. There is no equivalence margin or clinically meaningful activation threshold reported. The record explicitly states that established clinical thresholds do not exist for these experimental activation 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 sample-size table is a planning illustration for continuous-outcome comparisons. Its figures are participants per arm, not total events. For each comparison, the rows consider planning for 100%, 75%, or 50% of the observed effect. The columns show the participants per arm associated with 80% power and 90% power under the listed assumptions.

The stated assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is the stated fraction of the observed effect. The planning values are derived from the observed effect, not from the trial’s original design plan. The registry record does not report the study’s own sample-size justification or original design assumptions.

In plain terms, the 80% power column describes the participant count per arm for a hypothetical study designed to have an 80% chance of detecting the specified effect under those assumptions. The 90% power column gives the corresponding count per arm for a 90% chance. These are prospective planning quantities, not a measure of what the completed study achieved. This is not observed power.

The table shows that planning for a smaller effect requires more participants per arm. For example, planning for 50% of the observed effect requires more participants than planning for the full observed effect. This is expected because smaller differences are harder to distinguish from random variation. The accompanying curve should be read as the visual counterpart of those planning scenarios, not as evidence that the trial observed each hypothesized effect or had the listed power.

These calculations are especially illustrative because the study is a crossover trial while the stated planning assumptions describe equal allocation between two arms. The registry does not report whether the hypothetical calculations account for within-person correlation or other crossover features. They should be read as planning references rather than as a reconstruction of the actual study design.

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 = -0.380; pooled standard deviation σ = 0.586.
2. n = 2 × (1.96 + 0.842)² × 0.586² / -0.380² = 37.3, rounded up to 38 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 <- -0.38                      # observed difference in means
sigma <- 0.586057707366464                     # 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(3896659)
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 = 38 (80% power), 50 (90% power)
 75% of observed effect: n per arm = 67 (80% power), 89 (90% power)
 50% of observed effect: n per arm = 150 (80% power), 200 (90% power)
power.t.test: 38.3 
Simulated power at n = 38 per arm: 0.797 (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. Multiplicity, Missing Data and Analysis Populations

The registry lists four outcome measures, but the record provides the primary endpoint and does not provide secondary outcome results. It reports three comparison estimates for the primary outcome, each with an unadjusted p-value. No multiplicity adjustment is reported, and the analysis record contains no registered analysis entries.

Multiplicity arises when researchers conduct more than one statistical test. General statistical explanation: as the number of tests increases, the chance of at least one apparently unusual result can increase if each test is interpreted independently. Researchers may prespecify an adjustment, a testing hierarchy, or a designated primary comparison. The registry record does not report whether any such strategy was planned or used. The p-values should therefore be read as unadjusted values for the contrasts shown, not as family-wise or otherwise multiplicity-controlled results.

The record does not report how missing outcome measurements were handled, whether an intention-to-treat or per-protocol population was used, or how participants who did not complete both crossover periods were treated. The endpoint counts differ across the four condition summaries, but the registry does not provide enough information to attribute those differences to missingness, eligibility, completion, or another reason. No imputation method, missing-data sensitivity analysis, or participant-flow account is reported.

The trial is described as randomized and crossover, but the posted results do not specify the analysis population or the statistical approach for repeated observations. Sequence, period, and within-participant dependence are not described in the analysis record. It is therefore not possible to assess whether the reported comparisons correspond to a prespecified primary estimand or whether they account for the design.

General explanation: analysis-population definitions clarify which randomized or treated participants contribute to an estimate, and missing-data methods make assumptions about unobserved outcomes. Those choices can affect interpretation. Because they are not reported in the registry record, this page cannot determine their impact on these results.

14. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: suicidal ideation (4.8%).

The adverse-event summaries report serious adverse events and deaths by condition group. The serious-event figures are 0 among 19 at risk in Depressed: Hydrocortisone, 1 among 21 in Depressed: Placebo, 0 among 42 in Healthy Controls: Hydrocortisone, and 0 among 37 in Healthy Controls: Placebo. The death count is 0 in each group.

The registry also lists suicidal ideation as a top serious-event term at 4.8%. The record does not provide a group-specific breakdown for that term in the information, so it should not be assigned to a particular condition group based on these figures alone. The number of other adverse events reported as affected participants is 3, 3, 6, and 5 in the same group order.

These are descriptive registry counts. They do not establish whether an event was caused by treatment, and they do not show how event risk compares after accounting for the crossover design, timing, or participant characteristics. The at-risk counts vary across groups, and the registry record does not explain the reason. Rare or serious outcomes can be difficult to compare using small counts; the record does not provide inferential safety analyses.

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

The posted data describe mean activation scores in four condition groups and provide unadjusted mean-difference estimates with confidence intervals. All three reported comparisons use Depressed: Placebo as their reference. Each confidence interval includes zero, and the estimates are imprecise enough to include values on either side of zero. These results do not establish a clear difference, but they also do not demonstrate that the groups are equivalent.

The outcome is an experimental Z-score of task-related hippocampal activation, not a clinical outcome with an established threshold. A numerical difference on this measure cannot, from the registry record alone, be translated into a meaningful change in memory or health. The study description links the imaging measure to declarative memory, but the primary results do not provide a posted analysis of memory outcomes.

The comparison structure also limits what can be concluded. The healthy-control estimates are contrasted with the depressed placebo group, rather than comparing hydrocortisone with placebo within healthy controls. The record does not show an analysis that uses each participant’s measurements across the two crossover periods or accounts for treatment sequence and period. Thus, these summaries do not establish a treatment effect within either participant group.

The findings should be read as a description of the posted summaries, not as a definitive assessment of the study hypothesis. No analysis records, secondary outcome results, multiplicity plan, missing-data methods, or original statistical plan are reported in the registry record. The posted results therefore leave important questions about estimand, analysis population, and model specification unresolved.

Nothing in these summaries determines whether hydrocortisone should be used or avoided in any person. The statistical results concern the reported study outcome and its uncertainty. They do not constitute medical guidance or establish the clinical significance of the experimental activation differences.

16. Why This Trial Matters Statistically

This trial is statistically interesting because its design and its posted summaries raise distinct interpretation issues. The registry describes a randomized, masked crossover study, while the results are presented as condition-group means and contrasts against a single reference group. In a crossover study, the same participant may contribute data under more than one condition; the analysis must reflect the relationship between those observations if it is to estimate within-participant contrasts appropriately. The posted record does not show whether that occurred.

The comparison reference is also consequential. The healthy-control hydrocortisone and healthy-control placebo estimates are each reported against Depressed: Placebo. They are not the within-group treatment contrasts a reader might initially expect from the condition labels. Recognizing the reference group prevents interpreting these numbers as direct evidence about hydrocortisone versus placebo among healthy controls.

The confidence intervals draw attention to uncertainty, not merely a yes-or-no reading of the p-values. Each includes zero, and the experimental outcome has no established clinical threshold. Precision and clinical meaning are therefore separate questions: the data provide uncertain estimates on a research measure, without a reported threshold that defines a clinically important difference.

Finally, the sample-size section is useful as a planning illustration but is not a reconstruction of the trial’s rationale. It highlights how the participant requirements change when a smaller effect is targeted, while the original design assumptions remain unreported. Taken together, these points show why registry results need to be read alongside the design, comparison definition, and analysis details—not just the point estimates.

17. 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 outcome summaries report condition-group means and standard deviations, but do not show participant-level data, paired differences, or the number of participants who completed both crossover periods. The registry record does not clarify why outcome counts vary by condition, nor does it provide a participant-flow account that connects the endpoint counts with enrollment of 66.

The comparison estimates use the depressed placebo group as the reference for all three contrasts. This limits direct interpretation of the healthy-control comparisons as treatment contrasts within that group. The registry does not report a separate hydrocortisone-versus-placebo estimate within healthy controls or within the depressed group using a crossover-aware model.

Although the study is described as randomized, quadruple-masked, and crossover, important design and analysis details are not reported in the registry record. These include the randomization procedure, treatment-sequence balance, analysis of period or sequence, handling of within-participant correlation, analysis population, missing-data approach, and model diagnostics. The record also does not report the original sample-size justification, interim analyses, or an analysis plan.

The outcome is an experimental activation measure without established clinical thresholds. The registry’s broader rationale includes declarative memory and multimodal imaging, but secondary results are not posted. Thus, interpretation is limited to the primary outcome summaries and cannot establish how the imaging results relate to other measures or to clinically meaningful outcomes.

The reported p-values are unadjusted, and no multiplicity strategy is reported. The record does not establish how these comparisons relate to any prespecified testing hierarchy. In addition, the serious adverse-event counts are descriptive and do not provide a causal or crossover-adjusted safety comparison.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Exploring the Effects of Corticosteroids on the Human Hippocampus — NCT03896659.
• Registry record last updated 2025-12-16; status last verified 2025-12.
• Results first posted 2025-12-16.
• 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.

21. Record Summary

The registry describes NCT03896659, “Exploring the Effects of Corticosteroids on the Human Hippocampus,” as a completed Phase 4 randomized, quadruple-masked crossover study sponsored by the University of Texas Southwestern Medical Center. Enrollment is listed as 66.

The study compared oral hydrocortisone at 160 mg per day for 3 days with matching placebo, with a 25-day washout period in the sequence descriptions. It included healthy controls and a depressed group, identified as exploratory in the study description.

The primary outcome was left-hemisphere hippocampal subfield activation, measured as a Z-score at 3 days. The posted summaries include group means, standard deviations, and three comparisons against Depressed: Placebo. All reported confidence intervals include zero. The registry record does not report registered analysis entries, secondary outcome results, the original analysis plan, or key details of missing-data handling and crossover modeling.