Testing the Ability of JNJ-18038683 to Improve Cognition and Reduce Depressive Symptoms in Stable Bipolar Patients: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Testing the Ability of JNJ-18038683 to Improve Cognition and Reduce Depressive Symptoms in Stable Bipolar Patients (NCT02466685), a randomized phase 2 study. The primary outcome is The 8-week Evaluation of Verbal Fluency Performance After Randomization.

2. Clinical Question

3. Trial Design

This was a randomized, parallel-group, quadruple-masked phase 2 study comparing JNJ-18038683 with placebo. The registry record lists a planned enrollment of 60 participants and describes both treatments as adjunctive to standard pharmacologic treatment for bipolar disorder. The population was stable bipolar outpatients receiving treatment for depression. The recorded study start was 2015-09, primary completion was 2021-09, and status is completed.

The intervention was scheduled for eight weeks. The JNJ-18038683 arm description specifies an initial 10 mg dose for one week, then 20 mg for the remainder, with the provision for a single downward adjustment. The placebo arm is described as placebo treatment for eight weeks. The registry record does not give the detailed background-treatment regimen or explain how a dose adjustment was decided.

The primary endpoint concerned change in verbal-fluency score from baseline to week 8. Although the record lists 30 participants in each randomized group, it states that the primary analysis was restricted to people who completed all study visits. The posted analysis counts are 22 for JNJ-18038683 and 26 for placebo. The record does not describe the reasons participants were not included in that analysis.

The registry record lists two secondary outcomes: the Montgomery-Asberg Depression Rating Scale and change from baseline to week 8 on the Clinical Global Impression Severity of the Subject With Bipolar Disorder Scale. It lists three outcome measures and one analysis. More detailed information about the randomization procedure, the roles covered by quadruple masking, and the trial's original statistical design assumptions is 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.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Although the total number of participants in each group was 30, the analysis was only done for those who had completed all study visits.

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 The 8-week Evaluation of Verbal Fluency Performance After Randomization, measured over Baseline and week 8.

The primary endpoint was change in verbal-fluency performance from baseline to week 8. The task asks a participant to produce as many words as possible from a category within a given time; each correct word contributes one score. The registry describes the score as ranging from 0 to 60 and identifies it as one test in a cognitive battery addressing functions such as processing speed, attention or vigilance, working memory, and verbal and visual learning.

The registry defines a higher amount of change as a better verbal-fluency outcome. Measuring change allows the analysis to express how performance at week 8 differs from a participant's baseline score. The posted group results are least-squares means with standard errors, and the unadjusted comparison is expressed as a difference in mean change. A separate sponsor-reported analysis is described as comparing final values using ANCOVA; that is not the same stated endpoint quantity as a simple difference in change.

Verbal fluency was selected as the primary cognitive measure, but the registry record does not explain why this specific test was prioritized over other measures in the cognitive battery. Nor does it report whether the task was administered in a particular way beyond the general scoring description. The endpoint is a test score, not a direct measure of every cognitive function named in the description.

7. Results Posted in the Registry

• JNJ-18038683: mean 39.44 versus 41.12 on control; difference -1.68 (95% CI -5.34 to 1.98), standardized effect d = -0.26; z-test p = 0.368.

8. Statistical Interpretation

9. Statistical Methods Explained

The posted primary summaries are least-squares means and standard errors. A least-squares mean is an estimate from a statistical model, rather than necessarily the simple arithmetic average of all raw scores. Standard errors describe the statistical uncertainty around estimated means; they are not the same as the spread of individual participants' scores. The record does not give the full model used to obtain the posted group estimates.

What does the unadjusted comparison represent?

The unadjusted recalculation compares the posted mean changes between the two groups without applying the sponsor's ANCOVA model. It gives a mean difference of -1.68 for JNJ-18038683 relative to placebo, with an interval from -5.339 to 1.979 and p = 0.368. The negative sign indicates that the JNJ-18038683 mean change is lower than the placebo mean change under this stated comparison order. This is a calculation from posted summary statistics, not a replacement for an analysis based on individual participant records.

The recalculation uses an approximate pooled standard deviation of 6.499, derived from the posted standard errors and analysis counts. Because this dispersion is approximate, the resulting unadjusted inference should be read as an illustration of the group-summary contrast. An analysis using participant-level data could account for baseline values, covariates, or other design features in a way that this summary-statistic calculation cannot.

What does ANCOVA do?

In general, analysis of covariance (ANCOVA) compares groups while adjusting for one or more covariates. In a baseline-and-follow-up study, baseline outcome can be used as a covariate when comparing final values. Such adjustment can account for baseline differences and may improve precision, depending on the model and data. The sponsor-reported analysis is labeled ANCOVA and describes a mean difference in final values. The covariates, exact model, coding of treatment contrast, and assumptions are not reported in the registry record, so the adjustment cannot be reconstructed from the posted summary.

The sponsor's reported estimate is 5 with p < 0.05. The unadjusted recalculation instead compares change scores and produces -1.68 with p = 0.368. The estimates differ in both stated analysis and numerical result. Since the record does not explain their relationship, it would be inappropriate to treat the recalculation as a reproduction of the ANCOVA or to assume why the estimates differ.

Why does over-dispersion matter in general?

General statistical explanation: over-dispersion occurs when observed variability is greater than the model assumes. If a model understates variability, standard errors can be too small and intervals too narrow, making results appear more precise than they are. The primary score is continuous, and the posted summaries include standard errors, but the record does not report an over-dispersion assessment. This concept is not a claim that over-dispersion occurred in this trial.

What do p-values and intervals say?

General statistical explanation: a p-value summarizes how unusual the data would be under a specified null hypothesis and model. It is not the probability that the null hypothesis is true, and it does not measure the practical importance of an estimate. A confidence interval describes a range of values reasonably compatible with the data and the method used; its exact confidence level is not reported for the unadjusted interval in the registry record. The interval's width reflects uncertainty, while the range of effects it includes matters for interpretation.

The posted sponsor p-value is written as p < 0.05, not as an exact value, and its confidence limits are absent. The unadjusted recalculation reports an exact p-value of 0.368 and an interval. These values come from distinct analyses; statistical significance in one analysis does not erase the uncertainty or analytic differences in the other.

What is unknown about the plan?

The registry record does not report the original sample-size justification, the target effect used in trial planning, the detailed primary analysis plan, any interim analyses, or whether multiplicity adjustments were planned or performed. It also does not provide the ANCOVA covariates or detailed missing-data procedures. These omissions prevent a full audit of whether the posted analyses match the original protocol and how sensitive the findings are to alternative modeling choices.

10. Confidence Intervals and Precision

The unadjusted mean difference is -1.68, with an interval extending from -5.339 to 1.979. The interval spans zero, as well as negative and positive values. Under the unadjusted comparison, the observed data are therefore compatible with a range of contrasts that includes no difference and contrasts in either direction. The interval does not establish that the true difference is exactly zero.

General statistical explanation: a confidence interval conveys uncertainty around an estimate. A narrower interval indicates greater precision under the method used, while a wider interval leaves more values compatible with the data. An interval is not a guarantee that the true value lies within its endpoints, nor does it assign probabilities to individual values after the data are observed. The confidence level for the unadjusted interval is not reported in the registry record, so it should not be labeled with a particular confidence percentage.

The group estimates are 39.44 for JNJ-18038683 and 41.12 for placebo, with standard errors of 1.31 and 1.33, respectively. These are estimates for the participants included in the primary analysis, not all randomized participants. The analyzed counts were 22 and 26, while the record states that 30 participants were assigned to each group. The limited analysis population affects how precisely the findings describe the randomized population, especially because the reasons for non-completion and the missing-data method are not reported.

The sponsor-reported ANCOVA gives a mean difference of 5 and p < 0.05, but no confidence limits are posted for that result. Its precision cannot therefore be assessed from a reported interval. Also, the sponsor's final-value analysis and the unadjusted change-score calculation are not directly interchangeable. The two should be considered separately when judging what the record can establish.

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 are illustrations based on the observed effect rather than the trial's original design justification. The stated assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The assumed effect to detect is a stated fraction of the observed effect. The original sample-size assumptions are not reported in the registry record.

For planning based on 100% of the observed effect, the listed requirement is 235 participants per arm for 80% power and 315 participants per arm for 90% power. For 75% of the observed effect, the corresponding figures are 418 and 560 participants per arm. For 50% of the observed effect, they are 940 and 1,258 participants per arm. These are planning illustrations, not estimates of how many participants the completed trial actually analyzed.

In general, power is the chance that a planned analysis will identify an effect of a specified size under the assumptions used, if that effect is present. The 80% power column describes a planning scenario with an 80% chance of detecting the specified effect under those assumptions; the 90% column uses a higher target chance and consequently calls for more participants in these illustrations. The figures are not observed power and do not indicate the probability that the trial result is correct.

The smaller the effect chosen for planning, the more participants are generally needed to detect it with the same power and significance threshold. The values should be interpreted cautiously because they are derived from the observed result and rely on the stated approximation and allocation assumptions.

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 = -1.680; pooled standard deviation σ = 6.499.
2. n = 2 × (1.96 + 0.842)² × 6.499² / -1.680² = 234.9, rounded up to 235 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 <- -1.68                      # observed difference in means
sigma <- 6.49853361385742                     # 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(2466685)
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 = 235 (80% power), 315 (90% power)
 75% of observed effect: n per arm = 418 (80% power), 560 (90% power)
 50% of observed effect: n per arm = 940 (80% power), 1,258 (90% power)
power.t.test: 235.8 
Simulated power at n = 235 per arm: 0.801 (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 sponsor-reported primary analysis is an ANCOVA comparison of final values. It reports a mean difference of 5 and p < 0.05, without confidence limits. The unadjusted recalculation from the posted group summaries compares mean change and gives a difference of -1.68, with an interval from -5.339 to 1.979 and p = 0.368.

These results are not numerically concordant, and the registry record describes different analysis quantities: final values for the ANCOVA and change from baseline for the displayed group means and recalculation. It does not report enough detail to determine whether covariate adjustment, contrast coding, or another feature explains the difference. The estimates should not be combined or treated as if they answer precisely the same statistical question.

The sponsor's result is the analysis reported by the registry as the trial's ANCOVA analysis. The summary-statistic recalculation is an independent, unadjusted check based on the posted summaries and is not a substitute for the sponsor's model. Without the ANCOVA specification or participant-level data, neither its calculation nor the reason for the differing estimate can be independently reconstructed from the registry record.

14. Multiplicity, Missing Data and Analysis Populations

The record lists one primary outcome concerning verbal-fluency change and two secondary outcomes: the Montgomery-Asberg Depression Rating Scale and change on the Clinical Global Impression Severity of the Subject With Bipolar Disorder Scale. It also lists three outcome measures and one analysis. The registry record does not describe how the primary and secondary outcomes were prioritized for statistical testing beyond identifying the verbal-fluency measure as primary.

When several outcomes or comparisons are tested, the chance of at least one apparently notable result can increase. General statistical explanation: multiplicity procedures can control that risk by adjusting the testing threshold or the interpretation of intervals. Whether such adjustment is appropriate depends on the planned questions and analysis strategy. The registry record does not report whether multiplicity adjustments were planned or used, or how the sponsor handled the secondary outcomes. The p-values should therefore be read as posted, without assuming that they are adjusted for multiple testing.

The primary result is based only on participants who completed all study visits, although the record states that 30 participants were assigned to each group. The posted primary-analysis counts are 22 for JNJ-18038683 and 26 for placebo. The reasons for exclusion from the completer analysis, the amount and pattern of missing outcome information, and any imputation or other missing-data method are not reported in the registry record.

Analyzing completers alone can introduce bias if completion is related to treatment assignment or outcome. This is a general methodological concern, not evidence that such bias occurred. The record does not report whether all randomized participants were included in any other analysis, whether sensitivity analyses were done, or how missing observations were handled in the sponsor's ANCOVA. Those details are needed to judge how robust the result is to missing data and to understand the analysis population fully.

The registry record also does not report the original statistical analysis plan, interim analyses, or a detailed account of protocol deviations. As a result, the available summary does not establish whether the posted analysis choices were specified before the results were known.

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 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: serious adverse event (6.7%), serious adverse events (3.3%).

The safety summary lists 30 participants at risk for serious adverse events in each group. It records 3 participants affected in the JNJ-18038683 group and 1 in the placebo group. No deaths are listed in either group. The record also lists 6 participants affected by other adverse events in each group.

These counts describe events recorded in this study; they do not by themselves establish why an event occurred or whether groups differ reliably in safety. The registry lists serious-adverse-event terms as “Serious Adverse Event” at 6.7% and “Serious Adverse events” at 3.3%. The labels differ only in capitalization and wording, and the summary does not explain whether they are intended as distinct categories. The counts by group are clearer for reading the recorded group summaries.

General statistical explanation: uncommon events are difficult to compare precisely in a small study, and counts can be more informative than a percentage alone. The registry record does not provide event descriptions, timing, severity beyond the serious classification, or a causal assessment. These figures should be understood as a limited safety summary, not as a complete characterization of risk.

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

The posted results describe verbal-fluency score changes over eight weeks among participants who completed all study visits. In the unadjusted summary-statistic comparison, the estimated mean change was lower for JNJ-18038683 than placebo by 1.68 points, with an interval from -5.339 to 1.979 and p = 0.368. That interval includes zero, and the estimate is not precise enough to establish the direction of the unadjusted difference.

The registry also reports a different analysis: an ANCOVA of final values with a mean difference of 5 and p < 0.05. Because the ANCOVA model details and confidence limits are absent, and its stated outcome differs from the change-score calculation, the two reported results cannot be reconciled from the registry summary alone. The sponsor analysis should be understood in its own terms rather than used to overwrite or validate the unadjusted calculation.

Neither finding establishes a general conclusion about cognition in all people with bipolar disorder. The analysis is based on completers, the trial concerns a specific verbal-fluency score, and follow-up is limited to eight weeks. Other cognitive abilities, longer-term outcomes, and populations beyond the described stable outpatients are not answered by this primary result.

The study was designed to evaluate both cognitive performance and residual depressive symptoms, but the material summarized for the primary statistical comparison concerns verbal fluency. The existence of secondary outcomes does not show their results; those outcome-specific findings are not detailed in the registry record summarized for this analysis. No treatment recommendation follows from these statistics. They describe estimates and uncertainty in a particular trial population under the analyses reported.

17. Why This Trial Matters Statistically

This trial is statistically informative because it illustrates how the meaning of an estimate depends on the analysis being described. A difference in change scores and an ANCOVA comparison of final values may answer related but distinct questions. When the posted estimates differ, reporting the analysis labels and their uncertainty is essential; a single p-value cannot resolve the difference.

The gap between randomized and analyzed counts also matters. The primary result was limited to participants who completed all study visits, rather than everyone assigned to a group. Without information on incomplete visits or missing-data procedures, readers cannot assess whether the analyzed group is representative of the randomized group or how sensitive the result may be to missing outcomes.

The unadjusted interval communicates substantial uncertainty around the mean difference, while the sponsor-reported result has no posted interval. Showing both estimates and distinguishing their methods makes the limitations visible rather than implying that one number provides a complete account. Planning calculations likewise need to be separated from actual trial performance: figures based on the observed effect are prospective illustrations, not evidence of observed power.

For clinical researchers and biostatisticians, the record highlights the importance of a clear estimand, an explicit analysis population, transparent covariate and missing-data methods, and reporting uncertainty alongside point estimates. Those details are especially relevant when a trial has several outcomes and when the summary contains findings from more than one analysis.

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.

The primary analysis includes only participants who completed all study visits, with 22 analyzed in the JNJ-18038683 group and 26 in placebo, despite 30 assigned to each group. Reasons for non-completion, missing-data patterns, and procedures for handling missing scores are not reported in the registry record. This limits assessment of potential bias and generalizability to all randomized participants.

The sponsor-reported ANCOVA estimate is 5 with p < 0.05, while the unadjusted recalculation of the posted change summaries is -1.68 with an interval from -5.339 to 1.979 and p = 0.368. The registry describes the former as a final-value analysis and the latter as a comparison of change. Details needed to explain their differing results, including the ANCOVA covariates and contrast coding, are not reported.

The confidence limits for the sponsor's ANCOVA result are absent, and the confidence level for the unadjusted interval is not stated in the registry record. The original sample-size rationale and detailed statistical analysis plan are also not reported, so it is not possible to assess the planned power or confirm all analysis choices against a prespecified plan.

The study lasted eight weeks and the primary measure is a specific verbal-fluency score. The registry record does not explain why this measure was selected as primary or provide enough information to assess how the results relate to other cognitive functions. Details of the masking roles, background treatment, and adverse-event descriptions are also limited. These constraints narrow what can be concluded from the summary.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Testing the Ability of JNJ-18038683 to Improve Cognition and Reduce Depressive Symptoms in Stable Bipolar Patients — NCT02466685.
• Registry record last updated 2023-11-14; status last verified 2023-10.
• Results first posted 2023-07-03.
• 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

NCT02466685 is a completed phase 2, randomized, quadruple-masked, parallel-group study of adjunctive JNJ-18038683 versus placebo in stable bipolar outpatients receiving treatment for depression. The registry lists a planned enrollment of 60 and an eight-week treatment period.

The primary outcome was verbal-fluency score change from baseline to week 8. Although 30 participants were assigned to each group, the primary analysis was limited to participants completing all visits, with 22 analyzed for JNJ-18038683 and 26 for placebo. The unadjusted comparison reports a mean difference of -1.68, an interval from -5.339 to 1.979, and p = 0.368. The registry also reports an ANCOVA final-value mean difference of 5 with p < 0.05.

The record lists two secondary outcomes. It does not report the detailed analysis plan, missing-data procedures, or the ANCOVA specification needed to reconcile the two primary estimates.