Pioglitazone for the Treatment of Bipolar Depression: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Pioglitazone for the Treatment of Bipolar Depression (NCT01717040), a randomized phase 4 study. The primary outcome is Change in Inventory of Depressive Symptomatology-Clinician Rated (IDS-C30) Total Score.

2. Clinical Question

3. Trial Design

This was a Phase 4, randomized, parallel-group trial comparing pioglitazone with placebo. The registry describes the masking as quadruple. It does not identify which groups of people were masked, so the precise masking arrangements are not reported in the registry record. The title describes the trial as double-blind, while the structured registry entry gives the masking category as quadruple.

The registry records enrollment of 37 participants: 17 in the pioglitazone arm and 20 in the placebo arm for the primary outcome. The trial status is terminated. The record lists a start date of 2012-09 and primary completion of 2016-03. The reason for termination, the dates of recruitment and the number of participants who completed follow-up are not reported in the registry record.

The primary outcome is described as change in IDS-C30 total score, with a timeframe of baseline and Week 8. The posted arm estimates are least-squares means with standard errors. The registry does not explain the model used to obtain the least-squares means, the covariates included, or the handling of data from participants with incomplete follow-up. These details are not reported in the registry record.

The brief summary describes secondary objectives concerning potential moderators and mediators. However, the registry record has no posted secondary outcome results and no posted analysis details for those questions. It also does not give the original recruitment target or the rationale for the planned sample size. The sample-size illustrations shown elsewhere on the page are based on the observed effect and should not be read as the trial’s original design calculations.

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:

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 Change in Inventory of Depressive Symptomatology-Clinician Rated (IDS-C30) Total Score, measured over Baseline and Week 8.

The primary endpoint is change in the Inventory of Depressive Symptomatology-Clinician Rated total score, abbreviated IDS-C30. The registry describes this clinician-rated measure as assessing the severity of depressive symptoms. Total scores range from 0 to 84, with higher scores indicating greater symptom severity. The outcome is assessed from baseline to Week 8.

A change score can summarize how a participant’s measured symptoms differ between the start and a later assessment. Comparing change between randomized groups is intended to address whether the groups’ measured outcomes followed different courses over that period. A clinician-rated measure is one way to make symptom severity observable and comparable, rather than relying only on a broad statement that symptoms improved or worsened.

The registry reports arm-level least-squares means and standard errors for this outcome, but does not specify the model, adjustment variables or exact convention used to code the change. Since higher total scores indicate greater severity, interpretation of the sign of a change estimate depends on how the change was defined. That coding convention is not reported in the registry record. The posted values can therefore be described as recorded, but their direction should not be translated into symptom improvement or worsening without that information.

7. Results Posted in the Registry

• Pioglitazone: mean 31.40 versus 24.30 on control; difference 7.10 (95% CI -0.81 to 15.01), standardized effect d = 0.58; z-test p = 0.079.

8. Statistical Interpretation

9. Statistical Methods Explained

What does a least-squares mean represent?

The registry labels each arm estimate a least-squares mean. In general, a least-squares mean is an estimate from a statistical model rather than necessarily the simple arithmetic mean of the observed values. A model may adjust for specified variables or account for features of the design. Here, the exact model and any adjustment variables are not reported in the registry record. The reported arm values should therefore be presented as least-squares means, without assuming what adjustments were made.

The arm estimates are 31.4 for pioglitazone and 24.3 for placebo. The registry gives standard errors of 3.0 and 2.7, respectively. A standard error describes the estimated sampling uncertainty of an estimate: in general, smaller standard errors indicate more precise estimates, while larger standard errors indicate more uncertainty. It is not the same thing as the spread of individual participants’ scores. The record does not provide the raw participant-level values or a standard deviation for the primary outcome.

How should the mean difference be read?

The registry reports a mean difference of 7.1 for pioglitazone relative to placebo. This is the contrast between the posted arm estimates, expressed in the units of the IDS-C30 change score. It is not a ratio, a percentage change or a count of participants. Because the score’s direction is defined but the change-score coding convention is not, the sign alone cannot be translated confidently into which arm had greater symptom improvement.

The registry also reports an unadjusted p-value of 0.079. In general, a p-value measures how inconsistent the observed data would be with a specified null hypothesis under the statistical model and its assumptions. It is not the probability that the null hypothesis is true, nor does it measure the size or clinical importance of an effect. The record calls this value unadjusted; details of the hypothesis test and model are not reported.

What does the interval add?

A confidence interval expresses uncertainty around an estimate by giving a range of values compatible with the data under the interval’s statistical assumptions. The posted interval for the mean difference includes zero and extends to 15.01 at its upper end. The confidence level and the method used to construct the interval are not reported in the registry record. In general, an interval that includes zero for a difference does not distinguish clearly between a zero difference and differences on either side of zero.

Interval width also matters. A wide interval leaves more plausible values open than a narrow one. Here, the group sizes and the interval indicate that considerable uncertainty remains about the magnitude and direction of the difference. The interval should be considered alongside the point estimate, rather than relying on the p-value alone.

What are the planning illustrations?

The sample-size illustrations use the observed effect as their starting point. They show planning values for detecting the full observed effect, 75% of the observed effect, or 50% of the observed effect. In general, planning around a smaller effect requires more participants to achieve the same target power, because a smaller difference is harder to distinguish from random variation. The illustrations assume a two-sided significance level of 0.05, equal allocation between the two arms and a normal approximation. They are not the trial’s original design calculations; the registry does not report the original sample-size assumptions.

The standardized effect information in the comparison record is based on a pooled standard deviation that is marked as approximate. General standardization expresses a difference relative to a measure of variability, which can help compare effects measured on different scales. Since the pooled spread is approximate and the underlying analysis details are not reported, the standardized value should not be treated as a more definitive result than the posted mean difference and its interval.

What analysis details are unavailable?

The registry does not report the statistical analysis plan, whether the primary analysis adjusted for baseline score, the approach to missing outcome data, or the population included in the analysis. It also does not report whether an interim analysis was planned or conducted. The secondary objectives concern moderators and mediators, but no secondary analysis results are posted. These omissions limit the ability to reconstruct the analysis or judge how closely it followed a prespecified plan.

In general, a statistical method depends on assumptions about the outcome, the design and the data available. With a continuous outcome, analysts may need to consider the distribution of scores, model fit and the effects of incomplete observations. The registry record does not provide diagnostics or enough model information to assess those issues. This is a limitation of the available reporting, not evidence that any particular analysis choice was made.

10. Confidence Intervals and Precision

The posted mean difference is 7.1, and the interval runs from a negative lower bound to 15.01. Because the interval includes zero, the posted data do not pin down a single direction or magnitude of difference. The upper end allows for a difference as large as 15.01 in the reported scale units, while the negative lower end allows for a value on the other side of zero. The confidence level and interval construction method are not reported in the registry record.

In general, a confidence interval is useful because it shows a range of estimates consistent with the data under a statistical procedure. A point estimate alone can give a misleading sense of certainty. Here, the interval communicates substantial uncertainty around the point estimate. It does not mean that all values inside it are equally likely, and it does not guarantee that the true value lies within the interval.

The standard errors for the arm estimates are 3.0 for pioglitazone and 2.7 for placebo. They likewise describe uncertainty in the estimated arm means, rather than the variability among individual participants. The registry does not provide the model details needed to independently evaluate how those standard errors and the interval were obtained.

Precision is also affected by the limited number of participants contributing to each arm estimate: 17 for pioglitazone and 20 for placebo. The registry does not report how many enrolled participants had missing Week 8 measurements or how missing values were handled, so the extent to which incomplete data affected precision cannot be assessed from this record. The interval is therefore central to interpreting the result, and the reported point estimate should not be read as a highly precise estimate of the difference.

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 gives planning illustrations in participants per arm, not total events. It assumes a two-sided significance level of 0.05, equal allocation between the two arms and a normal approximation. The calculation uses the observed effect as a reference and considers planning for 100%, 75% or 50% of that observed effect. These are planning assumptions for the illustration; the original trial design assumptions are not reported in the registry record.

For planning around 100% of the observed effect, The table shows 47 participants per arm for 80% power and 63 participants per arm for 90% power. For 75% of the observed effect, it lists 83 participants per arm for 80% power and 111 participants per arm for 90% power. For 50% of the observed effect, it lists 186 participants per arm for 80% power and 249 participants per arm for 90% power.

In plain terms, the 80% power column describes the number per arm used in the illustration for a design intended to have an 80% chance of detecting the specified effect under its assumptions. The 90% power column uses a higher target chance and therefore lists more participants per arm for each effect assumption. Smaller effects are harder to distinguish from random variation, so the table requires more participants per arm when planning for a smaller fraction of the observed effect.

These figures are planning illustrations based on the observed result, not observed power for the completed trial. They are not evidence about how likely the trial was to detect an effect after it had been run, and they do not establish what sample size the investigators originally intended.

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 = 7.100; pooled standard deviation σ = 12.210.
2. n = 2 × (1.96 + 0.842)² × 12.210² / 7.100² = 46.4, rounded up to 47 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 <- 7.1                      # observed difference in means
sigma <- 12.2103001016121                     # 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(1717040)
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 = 47 (80% power), 63 (90% power)
 75% of observed effect: n per arm = 83 (80% power), 111 (90% power)
 50% of observed effect: n per arm = 186 (80% power), 249 (90% power)
power.t.test: 47.4 
Simulated power at n = 47 per arm: 0.796 (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 brief summary identifies secondary objectives concerning potential moderators and mediators of antidepressant efficacy. However, the record posts no secondary outcome results. It does not describe which moderator or mediator analyses were planned, how many comparisons were intended, or how such analyses would be interpreted. Those details are not reported in the registry record.

Multiplicity arises when several outcomes, comparisons or subgroup questions are examined. In general, testing more questions can increase the chance of finding an apparently notable result by chance if the analysis does not account for the number of tests. The registry does not report a multiplicity strategy, such as a hierarchy of outcomes or an adjustment across tests. No particular strategy should be inferred from the presence of secondary objectives.

The primary comparison is reported with an unadjusted p-value of 0.079. The record does not explain whether this was the only primary test or whether other analyses of the primary endpoint were conducted. The statistical analysis plan is not reported, so it is not possible to determine whether the posted test matches a prespecified procedure.

Missing-data handling is also not reported. The endpoint timeframe is baseline and Week 8, but the registry does not state how many participants contributed measurements at each time, why any observations were absent, or what method was used for incomplete outcome data. Without those details, readers cannot assess whether the estimates represent all randomized participants, only participants with observed outcome data, or another analysis population.

The record does not define an intention-to-treat, per-protocol or safety analysis population. It also does not report protocol deviations or participant flow. These omissions matter because the participants included in an analysis and the treatment of missing observations can affect both the estimated result and its uncertainty. The available registry information is therefore insufficient to evaluate the analysis populations or the robustness of the results to missing data.

14. Safety Results

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

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent non-serious adverse event terms

The registry lists serious adverse events for 0 of 17 participants in the pioglitazone group and 0 of 20 participants in the placebo group. These figures describe the posted counts of participants affected by serious adverse events. The record lists no serious adverse event terms. It does not give additional detail about how serious events were defined or collected.

For other adverse events, the registry lists 6 of 17 participants in the pioglitazone group and 8 of 20 participants in the placebo group. The record does not provide event terms or further breakdowns in the information. These counts should be read as the registry’s reported participant counts, not as a complete description of the type, timing or severity of every event.

Deaths are not reported in the registry record. The safety information also does not state the period during which events were monitored, whether participants could experience more than one event, or whether events were judged related to treatment. The counts alone therefore do not permit a broad comparison of safety or establish that an event did or did not occur outside the information captured in the record.

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

The result is a registry-posted comparison of change in clinician-rated depressive symptom scores between a pioglitazone arm and a placebo arm. The reported mean difference is 7.1, with an interval that includes zero and reaches 15.01 at its upper end. The unadjusted p-value is 0.079. Taken together, these results leave substantial uncertainty about the size and direction of the difference.

The scale description says higher IDS-C30 totals indicate greater depressive symptom severity. However, the registry does not specify how change was coded. It is therefore not appropriate to translate the positive difference into a claim that one group improved more or worsened more. The values can be reported as recorded, while acknowledging this missing interpretive detail.

The posted estimates do not establish that the groups were identical. An interval that includes zero can reflect uncertainty that leaves both a zero difference and nonzero differences plausible. Nor does the result establish a definitive difference. The registry does not provide the model details, missing-data methods, analysis population or statistical analysis plan that would help evaluate the strength and robustness of the comparison.

The trial enrolled 37 participants and is marked terminated. Its termination reason is not reported in the registry record. The record does not post secondary results for the stated moderator and mediator objectives, and it does not report the original sample-size rationale. These limits restrict what can be concluded from the registry results alone.

In general, a clinical trial result is one contribution to an evidence base, and its interpretation depends on the design, conduct, analysis and completeness of reporting. This page describes the posted data and their precision; it does not offer medical advice or determine what treatment is appropriate for any individual.

16. Why This Trial Matters Statistically

This trial raises several statistical questions that are useful beyond this particular comparison. The outcome is a continuous clinician-rated symptom score, and the registry reports least-squares means rather than describing a simple difference in raw participant averages. That makes it important to distinguish the reported estimate from the model details needed to interpret it fully.

The result also shows why an effect estimate should be considered alongside its uncertainty. The reported mean difference is 7.1, but the interval includes zero and extends to 15.01 at its upper end. The unadjusted p-value is 0.079. A single threshold-based reading would omit the range of differences compatible with the data and the limits imposed by the small enrolled population.

The record’s incomplete reporting is statistically consequential. It does not specify the change-score coding convention, analysis population, missing-data procedure, model adjustments, or multiplicity strategy. Those are not minor presentation details: they shape how readers understand what was estimated and how dependable the estimate may be. The registry also gives no reason for the terminated status, so the status cannot be linked to a particular analytical or operational explanation.

Finally, planning calculations based on the observed effect should be kept separate from an original sample-size justification. The illustrations show how the required participants per arm change when planning for smaller effects, but they do not reconstruct the trial’s initial design. This distinction helps prevent a post hoc planning exercise from being mistaken for evidence about the adequacy of the trial’s original plan.

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 registry record leaves important aspects of the trial and its analysis unspecified. Although the title calls the study double-blind and the structured entry lists quadruple masking, the record does not identify who was masked or how masking was maintained. Arm descriptions are blank, and details such as the pioglitazone dose, treatment schedule and permitted background care are not reported in the registry record.

The study is marked terminated, but the reason for termination is not reported in the registry record. The record gives enrollment of 37 and arm counts for the posted endpoint, but it does not provide a participant flow account, completion counts or the number with missing Week 8 data. The original recruitment target and sample-size justification are also not reported.

The outcome is described as change from baseline to Week 8, but the coding convention for the change score is not stated. This limits interpretation of the positive mean difference, especially because higher IDS-C30 scores indicate greater symptom severity. The model used to obtain the least-squares means, the variables included in it, the confidence level and construction method for the interval, and the missing-data approach are not reported in the registry record.

The record provides no statistical analysis plan, analysis-population definitions, multiplicity strategy or interim-analysis information. It also posts no secondary analyses despite the summary’s stated moderator and mediator objectives. Safety information is limited to event counts; deaths are not reported in the registry record, and details on event types, monitoring period and attribution are not included. These limitations mean the results should be understood as a limited registry account, not a complete report of trial conduct and analysis.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Pioglitazone for the Treatment of Bipolar Depression — NCT01717040.
• Registry record last updated 2017-01-27; status last verified 2016-11.
• Results first posted 2017-01-27.
• 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

This Phase 4 randomized, parallel-group trial compared adjunctive pioglitazone with placebo for bipolar depression. The registry lists quadruple masking, enrollment of 37 participants and a status of terminated. The primary endpoint was change in IDS-C30 clinician-rated total score from baseline to Week 8.

The posted least-squares means were 31.4 for pioglitazone and 24.3 for placebo, with standard errors of 3.0 and 2.7. The reported mean difference was 7.1, with an interval including zero and an unadjusted p-value of 0.079. The change-score coding convention and key analysis details are not reported in the registry record.

The registry lists no serious adverse events in either arm, reports other adverse events for 6 of 17 pioglitazone participants and 8 of 20 placebo participants, and does not report deaths. It posts no secondary outcome analyses or reason for trial termination.