Switching to Duloxetine in Patients With Depression: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Switching to Duloxetine in Patients With Depression (NCT00696774), a registered phase 4 study. The primary outcome is Change From Baseline in Brief Pain Inventory-Modified Short Form (BPI-SF) Interference Score Between Responder and Non-Responder Participants at 4 Weeks.

2. Clinical Question

This endpoint links a classification based on core depression symptoms to a measure of how much pain interferes with everyday domains. The endpoint description names mood, walking and other physical activity, work, social activity, relations with others, and sleep. The analysis therefore asks about an association between response status and change in pain interference within the study, not a randomized estimate of duloxetine's effect compared with another treatment or a control group.

3. Trial Design

The registry record identifies this as a Phase 4, parallel-model study with no masking. It lists an enrollment of 242 and a completed status. The listed condition is major depressive disorder, and the intervention is duloxetine. The study title and brief summary frame the research around switching treatment when existing treatment had not worked well.

The arm description states that participants who met screening criteria received duloxetine 60 mg once daily in an open-label manner for 4 weeks in Study Period II, identified as the acute therapy period. It then begins describing a further 4-week Study Period III for participants who did not respond, but the registry description ends before fully stating what happened in that period. The complete treatment sequence and the criteria for any later treatment decisions are not reported in the registry record.

The primary comparison is between participants categorized as responders and non-responders, with response defined by reduction in the Maier subscale from baseline. Those categories arise after treatment and do not constitute randomized study arms. The registry record does not report a randomization method or an allocation process. Its own caution is that this is not a randomized comparison and the arms are not a like-for-like control comparison. Because the status groups are based on response, differences between them cannot be attributed solely to treatment assignment.

Study start is listed as 2008-06, and primary completion as 2009-07. The record does not report a comparator treatment arm in the intervention information. It also does not report the rationale for the enrollment target, the protocol's complete sequence of periods, or whether any interim analyses were planned.

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: 270 participants began the study. 26 were screen failures (17 did not meet entry criteria, 7 decided to withdraw, and 2 withdrew based on the physician's decision). 2 participants discontinued before enrollment (1 was lost to follow-up and 1 decided to withdraw).

Period II

Reasons for not completing this period:

Period III

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Efficacy Population: Participants who had at least one post-baseline observation were included in the efficacy analysis. The group "Unclassified" was excluded from statistical testing.

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 From Baseline in Brief Pain Inventory-Modified Short Form (BPI-SF) Interference Score Between Responder and Non-Responder Participants at 4 Weeks, measured over Baseline, 4 weeks.

The primary endpoint was change from baseline in the Brief Pain Inventory-Modified Short Form interference score between participants classified as responders and non-responders at 4 weeks. The instrument asks how much pain interferes with mood, walking and other physical activity, work, social activity, relations with others, and sleep. Its score ranges from 0, meaning no interference, to 10, meaning interference completely.

Response status was defined separately, using the Maier subscale of the 17-item Hamilton Depression Rating Scale. A response was a reduction of at least 50% from baseline. The Maier subscale represents core symptoms of depression and has a total score range from 0, normal, to 24, severe. Thus, the endpoint does not use pain interference to define response; it compares pain-interference change across groups defined by depression-score response.

Measuring interference can describe how pain affects several aspects of daily functioning, rather than only recording pain intensity. Comparing the change in that score across response categories can therefore address whether the two categories had different patterns of pain-related interference over the study period. This is a descriptive and comparative question within the treated study population. It is not, by itself, an estimate of the causal effect of duloxetine, because the response groups were not randomized treatment groups.

The registry specifies least-squares means and a 95% confidence interval for this endpoint. However, it does not report the endpoint estimates, the confidence interval limits, or a test result in the available record. The direction and precision of the observed comparison therefore cannot be determined from the record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Least-squares means are model-based estimates of group means. In general, a statistical model can adjust estimates for specified factors, which may make the reported comparison differ from a simple arithmetic comparison of raw group averages. The registry identifies the primary result as a least-squares mean and says that factors were used for adjustment, but the factors themselves and the model specification are not reported in the registry record. It is therefore not possible to reconstruct the adjustment or explain how much it affected the estimate.

The primary endpoint is a continuous measure: change in the BPI-SF interference score from baseline to 4 weeks. A continuous outcome retains variation across a scale rather than reducing all observations to categories. The score ranges from 0, no interference, to 10, interference completely. The comparison is between responders and non-responders, whose response categories are based on at least a 50% reduction in the Maier subscale score from baseline. The registry identifies the primary comparison as a least-squares mean analysis with a 95% confidence interval, but the estimates and interval limits are not reported in the record.

What does a confidence interval communicate?

In general, a confidence interval represents uncertainty around an estimated quantity under the statistical model and its assumptions. It is not a guarantee that the true value lies within the particular interval, nor does it show whether an estimate is important in practical terms. The width conveys information about precision, while the position of the interval in relation to a reference value can inform interpretation of a contrast. For this endpoint, the registry states that a 95% confidence interval was used but does not report its limits. No conclusion about the precision or compatibility of the primary estimate can be drawn without those limits.

Why does responder status matter?

Responder status is a post-treatment classification rather than a baseline assignment. In this study it depends on change in a depression subscale, while the primary outcome is change in a pain-interference scale. A comparison across these categories can describe how the measures co-occurred, but it does not have the protection of randomization between the categories. In general, differences between groups formed after treatment can reflect other characteristics, changes over time, or factors associated with being classified as a responder. The registry record does not report enough detail to determine the contribution of these possibilities.

The study is described as having no masking and a parallel model. No masking means the record does not indicate blinding; it does not by itself establish how participants, assessors, or other study personnel were informed. The registry does not provide further masking details. A parallel model indicates that the study record uses parallel groups, but the primary contrast is specifically between response categories, not a randomized comparison of treatment arms. The record cautions that allocation is not stated and that the arms are not a like-for-like control comparison.

What does the efficacy population rule tell us?

The record defines the efficacy population as participants with at least one post-baseline observation and says that the unclassified group was excluded from statistical testing. This describes who could contribute under the stated efficacy rule, but it does not explain what happened when a participant had some measurements missing or did not have a measurement at a particular assessment. The missing-data method, analysis denominator for each endpoint, and any sensitivity analyses are not reported in the registry record. An inclusion rule is not the same as a complete account of missing-data handling.

The safety information classifies participants as responders, non-responders, or unclassified and gives counts at risk and counts with serious or other adverse events. These categories should be read as registry-reported counts within the stated groups. In general, event counts can be informative descriptively, but comparisons between post-treatment response categories are not randomized safety comparisons. The record does not report deaths as numerical values, and the death entries are null; that is not equivalent to a reported count of zero.

Why can multiple outcomes complicate interpretation?

When a study examines multiple outcomes or analyses, repeated testing can increase the chance of finding at least one apparently notable result even when there is no corresponding underlying effect. In general, multiplicity procedures can address this issue, but the choice depends on the analysis plan and the intended family of hypotheses. The registry lists 15 outcome measures and 17 analyses. It does not report whether the investigators defined a multiplicity family, adjusted significance levels, or used another procedure to account for multiple testing. It would be inappropriate to assume either that an adjustment was made or that it was omitted.

The record contains no numerical primary estimate, p-value, confidence interval limits, full statistical model, or complete account of covariate adjustment. It also does not report the sample-size justification or interim-analysis plan. General explanations of least-squares means and confidence intervals can clarify the intended statistical quantities, but they cannot recover results or analysis choices absent from the registry record.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists 15 outcome measures and 17 analyses, including the primary comparison and secondary outcomes involving depression and anxiety scales. Examining multiple outcomes creates a multiplicity question: in general, the more hypotheses tested, the greater the chance that at least one result may appear notable by chance alone. Statistical plans can address this using procedures such as adjusted testing or a prespecified hierarchy, but the choice depends on how outcomes and hypotheses were organized. The multiplicity strategy is not reported in the registry record. The listed number of outcomes and analyses does not establish whether adjustment was used.

The efficacy population is described as participants with at least one post-baseline observation, with the unclassified group excluded from statistical testing. This establishes an inclusion rule, not a complete missing-data strategy. The registry record does not report how missing baseline or follow-up values were handled, whether participants with incomplete measurements contributed to particular analyses, or whether sensitivity analyses were performed. It also does not report analysis-specific denominators for the primary endpoint.

The primary endpoint is specified as a least-squares mean comparison with adjustment factors listed elsewhere in the registry, but those factors are not reported in the record available for this account. The full statistical model, assumptions, and estimation method are likewise not reported. A sample-size justification, interim-analysis plan, and complete statistical analysis plan are not reported in the registry record. These omissions limit assessment of how the analysis was prespecified and how robust the reported approach would be to alternative assumptions.

11. 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 serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: depression (1.1%), dermatitis (1.1%).

The safety record separates participants into duloxetine responders, duloxetine non-responders, and an unclassified group. It reports serious adverse events in 0 participants among 115 responders, 2 among 91 non-responders, and 0 among 35 unclassified participants. It also lists other adverse events in 71 responders, 59 non-responders, and 22 unclassified participants. These are registry-reported counts, not estimates of a treatment effect.

The groups are based on response classification, not randomized assignment to different treatments. As a result, differences in event counts across these categories should not be interpreted as causal comparisons of treatment safety. The counts also do not describe severity, duration, timing, or the number of events per affected participant. The record lists Depression and Dermatitis as the top serious adverse-event terms, each with a reported value of 1.1%.

Deaths are not reported as numerical values in the registry entries. A null entry should not be read as a confirmed count of zero. The record does not provide further detail on event definitions, follow-up time for safety, or how the listed groups were handled for safety analyses. These figures are descriptive and do not, on their own, establish comparative risks.

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

The registry record defines a study question about whether change in pain-related interference differed between participants classified as duloxetine responders and non-responders. The response classification was based on reduction in the depression Maier subscale, while the primary outcome was change in the BPI-SF interference score. The record also describes open-label duloxetine treatment during a 4-week acute therapy period.

However, the registry record does not report the numerical primary endpoint estimates, their 95% confidence interval limits, or a p-value. It is therefore not possible to state whether the responder and non-responder groups had different estimated changes, in which direction any difference went, or how precise the comparison was. The endpoint definition alone does not provide the result.

Even with a reported estimate, the comparison would need to be interpreted as an association between response status and pain-interference change. Responders and non-responders were not randomized to their categories, and the record warns that this is not a like-for-like control comparison. The comparison cannot by itself establish that duloxetine caused a difference between those groups or determine how outcomes would compare with another treatment or no treatment.

The safety counts are also descriptive. They are reported across the same response-based categories, rather than treatment groups assigned by randomization. They do not establish that one category had a greater or smaller treatment-related risk. The record's null death entries do not provide numerical death counts.

The findings that can be stated from this record concern the design, outcome definitions, population rule, and listed event counts. The magnitude, direction, and statistical uncertainty of the primary result, along with the full analysis strategy, are not reported in the registry record. No conclusion about treatment benefit or lack of benefit follows from the absence of those numerical results.

13. Why This Trial Matters Statistically

This study raises a useful statistical distinction between treatment comparisons and comparisons based on outcomes observed after treatment. The primary endpoint contrasts responders with non-responders, with response defined by change in a depression subscale. Because this classification is not randomized, it answers a different question from a comparison of randomized treatment groups. Recognizing that distinction is essential when interpreting estimates as descriptive associations rather than causal effects.

The endpoint also brings together measures from different domains: response is defined using core depression symptoms, while the primary outcome concerns how pain interferes with activities and relationships. That makes the endpoint definition important to understanding what is being compared. The scale range, the direction of the response rule, and the timing of assessment help clarify the intended analysis, even though the numerical estimate is not reported.

The record is also a reminder that a stated analysis method is not the same as a complete statistical account. It names least-squares means and a 95% confidence interval, but the model, adjustment factors, interval limits, and missing-data approach are not available in the record. Multiple outcomes and analyses are listed, while the multiplicity strategy is not reported. These details matter for judging precision and the broader context of a result.

Finally, enrollment and completion information cannot replace endpoint results. Without the primary estimate and its uncertainty, the record does not permit a quantitative reading of the principal comparison. Transparent reporting of estimates, confidence intervals, analysis populations, and handling of incomplete data is necessary for readers to assess what a clinical study can support.

14. 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 most immediate limitation for statistical interpretation is that the registry record does not report the numerical primary result. The least-squares means, contrast between response groups, 95% confidence interval limits, and p-value are absent. As a result, the size, direction, and precision of the primary comparison cannot be assessed.

The comparison groups are defined by response after treatment, not by random assignment. The record states that allocation is not stated and that the arms are not a like-for-like control comparison. Consequently, an observed difference between responders and non-responders would not isolate a causal treatment effect. The record does not provide information sufficient to assess the extent of confounding or other differences between these groups.

The arm description is incomplete: it gives the open-label duloxetine regimen for Study Period II and begins, but does not finish, its description of Study Period III. The full treatment sequence and any later treatment decisions are therefore not reported in the registry record. The registry also says adjustment factors are listed in an “Other relevant information” section, but those factors are not available in this record. The statistical model and covariate adjustment cannot be evaluated.

The efficacy population rule requires at least one post-baseline observation and excludes the unclassified group from statistical testing, but the method for handling incomplete measurements is not reported. The record lists multiple outcomes and analyses without stating a multiplicity strategy. It also does not report the sample-size justification, interim-analysis plan, or complete statistical analysis plan. These limitations restrict assessment of prespecification, robustness, and the relationship between the primary comparison and the other analyses.

Safety counts are presented by response category, not randomized treatment assignment. Death entries are null rather than reported numerical values. The event counts therefore provide limited descriptive information and do not support causal comparative safety conclusions.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Switching to Duloxetine in Patients With Depression — NCT00696774.
• Registry record last updated 2010-09-08; status last verified 2010-09.
• Results first posted 2010-08-11.
• 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.

18. Record Summary

NCT00696774 is the registry record for the Phase 4 ARDENT study, “Attributes of Response in Depressed Patients Switched to Treatment With Duloxetine.” The listed condition is major depressive disorder, the intervention is duloxetine, the model is parallel, and masking is listed as none. Enrollment is 242. The record lists a study start of 2008-06, primary completion of 2009-07, and completed status.

The primary endpoint compares change from baseline in BPI-SF interference score between responders and non-responders at 4 weeks. Response is defined by at least a 50% reduction in the Maier subscale from baseline. The efficacy population includes participants with at least one post-baseline observation, while the unclassified group is excluded from statistical testing.

The registry identifies least-squares means and a 95% confidence interval but does not report numerical primary estimates or interval limits. The allocation method, full statistical plan, missing-data approach, and multiplicity strategy are not reported in the registry record. The primary comparison is not randomized and should not be read as a like-for-like treatment comparison.