Duloxetine for Chronic Depression: a Double-blind Study: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Duloxetine for Chronic Depression: a Double-blind Study (NCT00360724), a randomized phase 4 study. The primary outcome is Hamilton Depression Rating Scale (HDRS) - 24 Total Score.

2. Clinical Question

3. Trial Design

This was a completed Phase 4 randomized, quadruple-masked, parallel-group study sponsored by the New York State Psychiatric Institute. The registry record gives enrollment as 65. Participants were assigned to duloxetine or placebo for the Initial Phase, described as lasting 6 to 10 weeks. The primary outcome was assessed at Week 10. The record describes a subsequent 12-week Continuation Phase during which all subjects could be treated with an FDA-approved antidepressant medication. That description does not establish which participants entered the continuation phase or what treatment each received.

The parallel design compares groups followed under their assigned treatment conditions rather than crossing participants between the study arms. Random allocation can reduce systematic differences between groups at the start, while masking is intended to limit the influence of treatment knowledge on participants and study conduct. The registry record labels masking as quadruple but does not specify which roles were masked or how masking was maintained.

The arm summaries for the primary outcome include 29 participants in the duloxetine arm and 28 in the placebo arm. These are the outcome-specific counts reported for the comparison, not a statement about the number enrolled or the number who completed every phase. The primary endpoint description restricts the analysis population to patients with post-baseline data and a qualifying suicide-item score; the registry record does not provide a complete account of participant flow into that population.

Eligible subjects might also receive MRI scans to help investigators study how antidepressants work. The record does not report MRI results as part of the outcomes. It also does not report the original design assumptions, the intended sample size justification, or whether any interim analyses were planned or conducted.

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: For patients on current ineffective psychotropic medication, washout was required, with \>=7 days medication-free (\>=28 days for fluoxetine). Concurrent sleep medication (zolpidem) was allowed for a maximum of 5 days during the study.

Overall Study

Analysis population for the primary outcome, as defined in the registry: Patients for whom post-baseline data was available and scored 2 or less on the suicide item either at baseline or during the course of treatment.

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 Hamilton Depression Rating Scale (HDRS) - 24 Total Score, measured over Week 10.

The primary endpoint was the mean Hamilton Depression Rating Scale, 24-item version, total score at Week 10. HDRS-24 is a clinician-rated measure used to assess depression severity. The registry record describes a possible score range from 0 to 75, with higher scores indicating worse depression. Its stated interpretation ranges are 0-7 for no depression, 8-16 for mild depression, 17-23 for moderate depression, and 24 and up for severe depression. These ranges describe the scale; they do not by themselves define how large a difference between group means would matter clinically.

A depression rating scale provides a standardized way to compare symptom severity across participants and study arms. Reporting the mean gives a summary of the group’s scores, while the standard deviation describes how dispersed individual scores are around that mean. The Week 10 comparison therefore addresses average measured depression severity at that assessment, rather than each participant’s experience or the course of symptoms at every point in the study.

The primary outcome population was patients for whom post-baseline data were available and whose suicide-item score was 2 or less either at baseline or during treatment. This restriction is part of the reported endpoint definition and should be kept in view when interpreting the result. The registry record does not report a clinical-response threshold, a remission definition, or a minimal difference considered important for this scale.

7. Results Posted in the Registry

• Duloxetine (Cymbalta): mean 5.00 versus 10.00 on control; difference -5.00 (95% CI -7.42 to -2.58), standardized effect d = -1.08; z-test p = <0.001.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Cornell Dysthymia Rating Scale (CDRS)

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
• Result:
  • Duloxetine (Cymbalta): mean 19.10 versus 28.50 on control; difference -9.40 (95% CI -15.82 to -2.98), standardized effect d = -0.77; z-test p = 0.004.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Duloxetine (Cymbalta) than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Global Assessment of Functioning Scale (GAF)

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
• Result:
  • Duloxetine (Cymbalta): mean 69.90 versus 65.70 on control; difference 4.20 (95% CI -5.01 to 13.41), standardized effect d = 0.24; z-test p = 0.371.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Duloxetine (Cymbalta) and the control. The interval gives the range of effects that remain compatible with the data.

Beck Depression Inventory (BDI)

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
• Result:
  • Duloxetine (Cymbalta): mean 8.50 versus 10.10 on control; difference -1.60 (95% CI -5.03 to 1.83), standardized effect d = -0.24; z-test p = 0.361.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Duloxetine (Cymbalta) and the control. The interval gives the range of effects that remain compatible with the data.

The secondary outcomes broaden the assessment beyond the primary depression-rating score. They include measures of dysthymia symptoms, depressive symptoms, and functioning. Across these measures, the overall pattern is mixed in precision: some estimated differences point in the same direction as the primary comparison, while other results are less clearly distinguished from no difference by their recalculated intervals. This does not amount to a uniform secondary-outcome pattern.

These outcomes are useful as context because they address related but not identical aspects of participants’ symptoms and functioning. Agreement across different measures can strengthen the coherence of an overall interpretation, while differences among measures can reflect distinct constructs, variability, or statistical uncertainty. They should not be treated as interchangeable versions of the primary endpoint.

The registry record lists multiple secondary outcomes and reports analyses across outcomes, but it does not report a multiplicity adjustment or a formal testing hierarchy. When many outcomes are examined, the chance of obtaining an apparently notable unadjusted result increases. An unadjusted secondary p-value therefore carries less confirmatory weight than a result from a prespecified primary analysis with a clearly reported analysis plan.

The secondary findings can inform what might merit further study, but they do not independently establish an overall treatment effect or resolve uncertainties in the primary outcome population. In particular, the record does not provide enough information to determine whether these outcomes were prespecified with a particular hierarchy, how missing observations were handled for each measure, or whether all intended analyses are represented in the posted results.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Cornell Dysthymia Rating Scale (CDRS)
• Global Assessment of Functioning Scale (GAF)
• Beck Depression Inventory (BDI)
• Clinical Global Impressions Improvement(CGI-I)
• Cornell Dysthymia Rating Scale (CDRS)
• Beck Depression Inventory (BDI)
• Global Assessment of Functioning Scale (GAF)

9. Statistical Interpretation

10. Statistical Methods Explained

What does the mean difference represent?

For a continuous outcome such as the HDRS-24 total score, a mean difference expresses the gap between the group averages in the units of the original scale. The reported direction is duloxetine minus placebo. Because higher HDRS-24 scores indicate worse depression, a negative difference means the observed average score was lower in the duloxetine arm. The recalculated value is -5.0 points. It is an unadjusted comparison of the posted summary statistics, not an estimate adjusted for baseline scores or other participant characteristics.

A mean difference is useful because it preserves the scale’s units and can be compared with the range and interpretation of the measure. It does not show how much individual participants changed, whether every participant improved, or how many crossed a response threshold. Those questions require participant-level summaries or outcome definitions that are not reported in the registry record.

What does a confidence interval add?

A confidence interval describes uncertainty around an estimated difference under the method used to calculate it. The recalculated interval for the primary mean difference runs from -7.422 to -2.578. In general, an interval gives a range of estimates compatible with the observed data and the calculation’s assumptions. It communicates both the estimated direction and the degree of precision more fully than a point estimate alone. The registry record does not state the confidence level for this recalculated interval, so no confidence level should be inferred from the record.

The interval does not assign a probability to the true effect, and it does not incorporate every source of uncertainty. In particular, it does not account for uncertainty about outcome selection, missing data, analysis choices, or how well the study population represents other populations. Its interpretation is conditional on the comparison and calculation used.

How is the reported analysis different?

The sponsor-reported primary analysis is a repeated-measures ANOVA. In general, repeated-measures methods analyze measurements collected across time and can account for the relationship among observations from the same participant, depending on the model specification. The registry record reports an F statistic of 9.43 and p = .003 for the primary outcome. It does not describe the measurement occasions included in that model, its covariance structure, adjustments, or assumptions in enough detail to reproduce the analysis.

The recalculated mean difference uses the posted Week 10 group means and standard deviations in a simple unadjusted comparison. It is not a substitute for the sponsor’s repeated-measures model. The reported and recalculated analyses answer related but not necessarily identical statistical questions, because their methods and potentially their use of repeated measurements differ. The registry record does not flag a discrepancy between them. Where the sponsor’s own analysis is reported, it remains the trial’s model-based result; the recalculation offers a supplementary view of the endpoint summaries.

Why do assumptions matter?

General statistical reasoning: tests and intervals depend on assumptions about the data and the analysis. For a mean comparison, the distribution of scores, the presence of unusually influential observations, and the independence of participants can affect how well a simple calculation represents uncertainty. A repeated-measures analysis has additional modeling choices concerning the pattern of within-participant observations. The registry record does not report diagnostics or enough detail to assess those choices.

Over-dispersion is a general concern when observed variability exceeds what a model assumes. In that situation, standard errors can be too small if the extra variability is ignored, making intervals too narrow and p-values too small. The reported endpoint is continuous and its group standard deviations are available, but the registry record does not describe an over-dispersion assessment. This general issue should not be mistaken for evidence that over-dispersion occurred in this trial.

The posted p-values for the simple comparisons are unadjusted. In general, a p-value measures how incompatible the observed data are with a specified null model, conditional on the analysis assumptions. It is not the probability that the null hypothesis is true, not the size of an effect, and not a direct measure of clinical relevance. Because several outcomes and analyses are reported, interpreting isolated unadjusted p-values requires particular caution.

What do the planning illustrations mean?

The sample-size illustrations use the observed effect as a planning reference and show hypothetical participant counts per arm for stated fractions of that effect. The calculation assumptions listed in the registry material are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. General explanation: power is the probability that a specified design would detect an effect of the assumed size under the calculation’s assumptions. The illustrations do not describe the trial’s original design plan, which is not reported in the registry record, and they are not an estimate of the probability that this completed trial detected an effect.

11. Confidence Intervals and Precision

The primary recalculated mean difference is -5.0 points, with a confidence interval from -7.422 to -2.578. The interval lies on the negative side of zero, so the recalculated summary-statistic comparison is consistent with a lower Week 10 mean HDRS-24 score in the duloxetine arm. Its width indicates that the data are compatible with a range of differences, not only the point estimate. The interval does not establish the clinical importance of any value within that range.

Precision is affected by the amount of information available and by variation in participant scores. The primary arm summaries report standard deviations of 3.6 for duloxetine and 5.5 for placebo, with outcome counts of 29 and 28, respectively. These figures describe the reported endpoint summaries; they do not establish how missing outcome data were handled or whether the reported population matches every randomized participant. The registry record does not state the confidence level used for the recalculated interval.

The sponsor’s repeated-measures analysis reports an F statistic and p-value but no confidence interval in the registry record. Its uncertainty therefore cannot be read directly from a reported interval in that analysis. The recalculated interval pertains to a simpler unadjusted comparison of the posted endpoint summaries, and should not be presented as an interval from the sponsor’s model.

For secondary measures, the recalculated intervals differ in how clearly they separate from zero. This range of precision reinforces the need to consider estimates and intervals together rather than treating each p-value as a stand-alone conclusion. The registry record does not report an adjustment for multiple comparisons, and uncertainty from multiplicity is distinct from the sampling uncertainty represented by an individual interval.

12. 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 illustrations report participants per arm, not total event counts. The calculations assume a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. They use the observed effect as a reference and examine planning for 100%, 75%, and 50% of that observed effect. The original trial’s sample-size assumptions and justification are not reported in the registry record.

The 80% power column gives the planned participants per arm for a hypothetical study designed to have 80% power under those assumptions and for the stated effect size. The 90% power column gives the corresponding planning count for 90% power. In general, greater power means a higher chance of detecting the specified effect if that effect is present, while holding the other planning assumptions fixed. The figures are not descriptions of the number enrolled in this completed trial.

Smaller effects require more participants to detect under otherwise unchanged assumptions. That pattern is visible across the rows planning for progressively smaller fractions of the observed effect. The planning figures are not observed power; they are illustrations for future-study planning based on the stated assumptions and the observed effect estimate. They should not be read as evidence that the trial was adequately or inadequately designed, since its original design assumptions are not reported in the registry record.

These calculations also do not account for factors not included in the listed assumptions, such as a particular missing-data strategy, model complexity, or multiplicity adjustment. They are best read as conditional illustrations, not as a recommended sample size for a future study without further design work.

13. 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 = -5.000; pooled standard deviation σ = 4.631.
2. n = 2 × (1.96 + 0.842)² × 4.631² / -5.000² = 13.5, rounded up to 14 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 <- -5                      # observed difference in means
sigma <- 4.63117891921897                     # 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(360724)
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 = 14 (80% power), 19 (90% power)
 75% of observed effect: n per arm = 24 (80% power), 33 (90% power)
 50% of observed effect: n per arm = 54 (80% power), 73 (90% power)
power.t.test: 14.5 
Simulated power at n = 14 per arm: 0.786 (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.

14. Analyses Reported by the Sponsor

The registry record’s own primary result comes from a repeated-measures ANOVA, reported as F = 9.43 with p = .003. That result should be distinguished from the recalculated comparison of the posted Week 10 means, which gives a mean difference of -5.0 and a confidence interval from -7.422 to -2.578, with p < 0.001. The recalculation is a simple unadjusted analysis of summary statistics; it does not reproduce the sponsor’s repeated-measures model.

The sponsor’s analysis can use information from repeated observations according to its model, whereas the recalculation compares the endpoint summaries alone. Since the registry record does not give the model specification or a confidence interval for the repeated-measures result, the two analyses cannot be fully reconciled from the reported details. Both point to a difference in the direction of lower scores in the duloxetine arm, but the model-based sponsor result and the simple recalculation have different statistical foundations.

The record does not flag a discrepancy between the sponsor analysis and the recalculation. The recalculated result is therefore supplementary context rather than a correction to the trial’s analysis. For outcomes where a discrepancy is identified, the sponsor’s own analysis takes precedence; no such discrepancy is flagged for this primary result.

15. Multiplicity, Missing Data and Analysis Populations

The registry record reports one primary endpoint and several secondary outcomes, as well as multiple analyses. It does not report a multiplicity correction, a formal hierarchy for testing outcomes, or a family of hypotheses used to control the overall chance of false-positive findings. Without that information, the posted secondary p-values should be treated as unadjusted evidence. The primary outcome remains the clearest prespecified focus in the record, but the details of the statistical analysis plan are not reported in the registry record.

The primary endpoint population requires post-baseline data and a suicide-item score of 2 or less either at baseline or during treatment. This makes the analysis population definition important: it is not described simply as all enrolled participants. The registry record does not report the number excluded by this condition, the number without post-baseline data, or a full participant-flow account for the primary outcome.

The method for handling missing observations is not reported in the registry record. It is therefore not possible to determine whether the sponsor used imputation, a likelihood-based approach, complete-case analysis, or another strategy. The repeated-measures ANOVA result alone does not reveal how missing values were handled. Nor does the record describe sensitivity analyses assessing whether conclusions depend on assumptions about missing outcomes.

The registry record does not report whether analyses followed an intention-to-treat principle, whether treatment adherence affected inclusion, or whether participants were analyzed according to randomized assignment for every outcome. The primary population definition is available, but it is not enough to infer a broader analysis-population strategy.

Interim analyses, stopping rules, and details of a prespecified statistical plan are not reported in the registry record. These omissions limit assessment of outcome selection and analytic flexibility. They do not demonstrate that a particular problem occurred; they mean that the record does not provide the information needed to evaluate these aspects of the analysis.

16. 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

For serious adverse events, the registry record reports 33 participants at risk in the duloxetine group and 32 at risk in the placebo group. It reports 0 serious adverse events in each group. The figures describe the serious-event counts entered in the record for the stated groups; they do not establish that no other safety concerns occurred or that the treatment is safe.

The record does not list serious adverse event terms. Death counts are not reported in the registry record. The group counts at risk for serious events need not be interpreted as the same analysis population used for the primary efficacy endpoint, whose eligibility and data requirements are separately described.

General explanation: a zero count in a trial does not rule out an event that is rare, occurs after the observation period, or is not captured in the available records. Safety interpretation also depends on event definitions, follow-up, and completeness of reporting. The registry record does not provide enough detail to assess those dimensions or to compare the safety profiles comprehensively.

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

The posted primary outcome data show lower average Week 10 HDRS-24 scores in the duloxetine arm than in the placebo arm for the reported primary outcome population. The recalculated estimate and interval describe the size and uncertainty of that unadjusted endpoint comparison, while the sponsor’s repeated-measures analysis supplies the trial’s reported model-based result. Together, they support a statistical difference in the measured endpoint under the analyses reported.

This result does not establish that every participant had a lower score, that all symptoms changed in the same way, or that the observed mean difference is clinically important. The registry record does not report a threshold for a meaningful between-group difference, a response definition, or a remission outcome. Scale interpretation categories are not a substitute for such a threshold.

The finding is specific to Week 10 and to the primary outcome population, which requires post-baseline data and a qualifying suicide-item score. It does not by itself describe outcomes across the entire study period. The record describes a later continuation phase in which all subjects could receive an FDA-approved antidepressant medication, but does not provide enough information to interpret long-term results as a continuation of the randomized comparison.

The reported result also does not establish generalizability to every person with chronic depression. The registry record identifies the study conditions and population but does not provide the participant characteristics needed to assess how closely the enrolled group represents other clinical populations. It does not establish how results would differ under other treatment settings or for people outside the defined outcome population.

Finally, statistical evidence is not the same as a complete benefit-risk assessment. The serious adverse event record reports no serious events in either group, but the safety information is limited and does not resolve all safety questions. The result should be read as evidence about the specified rating-scale outcome, not as medical advice or a universal treatment conclusion.

18. Why This Trial Matters Statistically

This trial illustrates why an endpoint estimate, its uncertainty, and the analysis method all matter. The Week 10 mean difference is expressed in HDRS-24 scale points, so readers can see the direction and magnitude on the measure itself. The interval adds information about precision that a p-value alone cannot provide. The sponsor’s repeated-measures analysis also shows that the reported trial result is model-based, while a simple recalculation from endpoint summaries offers a related but distinct perspective.

The record highlights the importance of defining the analysis population. The primary endpoint applies to patients with post-baseline data who met a stated condition on the suicide item, not automatically to every participant enrolled. Without a complete account of missing observations and participant flow, readers should keep that population restriction visible when interpreting the estimate.

Multiple secondary outcomes can help assess whether findings align across related measures, but multiplicity affects how much weight to place on isolated unadjusted results. This is a useful statistical distinction: an outcome can be informative without providing the same level of confirmation as the primary endpoint.

Finally, planning illustrations based on observed effects are not evidence of the study’s original design adequacy. They show how hypothetical participant requirements change with the assumed effect and power, under stated calculation assumptions. The original sample-size justification is not reported in the registry record. Separating retrospective planning illustrations from actual design assumptions helps prevent a post-study calculation from being mistaken for a prospective plan.

19. 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 provides limited detail about the sponsor’s repeated-measures model. It reports the method, F statistic, and p-value, but does not report enough model specifications to evaluate its treatment of time, within-participant correlation, covariates, or missing observations. The simple recalculation offers an unadjusted comparison of endpoint summaries, but it cannot fill those gaps or reproduce the sponsor’s model.

The primary population is restricted to patients with post-baseline data and a specified suicide-item score. The record does not report how many participants were omitted from that population, why data were unavailable, or whether the analysis represents all randomized participants. This leaves uncertainty about the extent to which missingness or population selection could affect the result.

The registry record does not report the original sample-size justification, a prespecified analysis plan, multiplicity handling, interim-analysis plans, or detailed sensitivity analyses. It also does not state a minimum clinically important difference for the primary scale. These omissions limit evaluation of the design rationale and the clinical interpretation of the estimate.

The record describes a continuation phase and possible MRI scans, but does not provide those results in the outcomes. It also does not provide enough information to assess longer-term effects, treatment exposure after Week 10, or the representativeness of the study population. Serious adverse event reporting is limited to the posted counts, with deaths not reported in the registry record. These are limits on interpretation, not evidence that unreported events or analyses did or did not occur.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Duloxetine for Chronic Depression: a Double-blind Study — NCT00360724.
• Registry record last updated 2017-08-21; status last verified 2017-07.
• Results first posted 2015-11-03.
• 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.

23. Record Summary

The registry record describes a completed Phase 4 randomized, quadruple-masked, parallel-group study of duloxetine versus placebo for chronic depression, sponsored by the New York State Psychiatric Institute. The primary endpoint was the mean HDRS-24 total score at Week 10 in a defined population with post-baseline data and a qualifying suicide-item score.

The posted Week 10 mean scores were 5.0 for duloxetine and 10.0 for placebo. The sponsor reported a repeated-measures ANOVA result of F = 9.43 with p = .003. An unadjusted recalculation from the endpoint summaries gives a mean difference of -5.0, with a confidence interval from -7.422 to -2.578 and p < 0.001.

Secondary outcomes show a less uniform pattern of precision, and the registry record does not report multiplicity adjustment or missing-data methods. Serious adverse event counts were 0 in each group; death counts are not reported in the registry record. The findings describe the reported Week 10 outcome and should not be extended beyond that population and assessment without further evidence.