We conducted a phase 4, open-label, randomized, 12-week study in 43 study centers in the United States, Germany, Canada, and Puerto Rico. The first patient was enrolled on September 28, 2006, and the last patient visit occurred on August 26, 2009. The institutional review board for each site approved the protocol, which was developed in accordance with the ethical guidelines of Good Clinical Practice and the Declaration of Helsinki. All patients provided written consent after the study was explained and their questions answered and before study procedures were initiated.
Outpatients with type 1 or type 2 diabetes mellitus and a glycated hemoglobin (HbA1c
) level of 12% or lower were eligible for inclusion in the study if they were aged at least 18 years, presented with DPNP (confirmed by a score of at least 3 on section B of the Michigan Neuropathy Screening Instrument),12
and had a daily pain score of 4 or higher based on a numerical rating scale (0-10 points). Also, patients had to be taking a stable dose of gabapentin (≥900 mg/d) for at least 5 weeks before randomization, had to be at least 80% adherent to gabapentin therapy, and had to agree not to change the dose of gabapentin between visits 1 and 2. In addition, the patient or physician had to have indicated a need to change from the current gabapentin therapy for pain management. Patients were excluded if they had a history or current diagnosis of mania, bipolar disorder, obsessive-compulsive disorder, or posttraumatic stress disorder or were judged before randomization to be at risk of suicide. Other exclusion criteria included historical exposure to drugs known to cause neuropathy, treatment with a monoamine oxidase inhibitor within 14 days, or treatment with fluoxetine within 30 days of randomization. Concomitant antidepressants (except monoamine oxidase inhibitors) and analgesics were allowed, provided the patient was taking a stable dose for a minimum of 3 months before study enrollment.
The screening period for eligibility occurred between visits 1 and 2 and lasted as long as 60 days (approximately 8 weeks) to accommodate the washout period for excluded medications. When no washout was required, patients were enrolled within 3 days after they completed all screening procedures. Eligible patients were randomized in a 1:1:1 ratio to open-label oral treatment with duloxetine monotherapy, pregabalin monotherapy, or a combination of duloxetine and gabapentin. Patients began taking the study drug on the next day after visit 2. Patients in the monotherapy treatment arms were cross-tapered from their current oral dose of gabapentin to 300 mg/d during a 1- to 2-week period; gabapentin therapy was then discontinued. Duloxetine therapy was initiated at 30 mg once daily for 1 week, after which patients received 60 mg once daily. In Germany, Puerto Rico, and the United States, pregabalin was dosed according to the US prescribing information13
and initiated at 150 mg/d for the first week with 50-mg doses taken 3 times daily. The following week, the dosage was increased to 300 mg/d and administered at 100 mg 3 times daily. In Canada, pregabalin was dosed according to Canadian prescribing information,14
and therapy was initiated at 150 mg/d but administered at 75 mg twice daily. The following week, the dosage was increased to 300 mg/d and administered at 150 mg twice daily. Patients randomized to combination treatment continued to take their current dose of gabapentin (≥900 mg once daily), initiated duloxetine therapy at 30 mg once daily for the first week, and then received duloxetine, 60 mg once daily, for the remainder of the study.
Patients were allowed to use concomitant medications for the management of pain or mood during the trial, provided they were taking a stable dose of the medication at study entry. Patients were not allowed to increase the dose of these medications but were allowed to decrease the dose.
The primary efficacy measure was the reduction from baseline in the weekly mean of the daily 24-hour pain (referred to hereafter as daily pain
) diary ratings at week 12. Pain was assessed using a numerical rating scale ranging from 0 for no pain to 10 for worst possible pain. Two secondary measures derived from the pain diaries that were also analyzed weekly were worst pain and night pain ratings. Other secondary measures included the Clinical Global Impression of Severity, the Brief Pain Inventory (BPI) severity and interference items,15
the Beck Depression Inventory II,16
the Patient Global Impression of Improvement, and the Sheehan Disability Scale.17
Three response criteria were analyzed on the basis of daily pain ratings: reductions of 30% or higher from baseline, reductions of 50% or higher from baseline, and a reduction of 2 or more points from baseline. Sustained response was defined as at least a 30% reduction from baseline to end point, at least a 30% reduction from baseline 2 weeks before the last week in the study period, and at least a 20% continued reduction from baseline every week in between.
Safety measures included spontaneously reported treatment-emergent adverse events (TEAEs), discontinuation due to adverse events, mean changes and categorical changes in vital signs, and mean changes and rates of abnormal laboratory analytes, including HbA1c levels.
Noninferiority in this study focuses on differences in improvement in pain reduction between duloxetine and pregabalin (primary objective) and between duloxetine and duloxetine plus gabapentin (secondary objective) at week 12. The 97.5% 1-sided confidence limit, which is equivalent to the lower limit of a 2-sided 95% confidence interval (CI), for the difference between duloxetine and pregabalin was compared with the protocol-defined MoNI. The MoNI represents the maximum disadvantage of duloxetine relative to its comparator that is not considered clinically meaningful. If the lower limit of the CI exceeds the MoNI, noninferiority is established.
The MoNI was based on results from an analysis of 3 DPNP placebo-controlled duloxetine studies18-20
using only patients previously treated with gabapentin. In these patients, the estimated mean improvement (decrease) in pain severity at week 12 was –2.74 for duloxetine, 60 mg once daily, and –1.09 for placebo, representing a mean –1.65 advantage of duloxetine over placebo. Thus, the prespecified MoNI of –0.8 represents slightly less than 50% of the previously observed advantage for duloxetine over placebo in a similar population of patients.
This study was designed to randomize 400 patients to duloxetine, pregabalin, or duloxetine plus gabapentin in a 1:1:1 ratio. With the assumption that 6% of patients would discontinue the study before providing postbaseline data, approximately 125 patients per arm were included in the primary efficacy analysis. This sample size has 92% power for the noninferiority assessment of duloxetine compared with pregabalin in pain reduction. Assumptions supporting the sample size estimate include the use of a 1-tailed 97.5% CI, a mean improvement of 2.5 for pregabalin and 2.7 for duloxetine, a common SD of 2.3 for change from baseline to end point, and a MoNI of –0.8.
In accordance with International Committee for Harmonization recommendations,21
noninferiority analyses were conducted on the intent-to-treat (ITT) population and the per-protocol population. Because results from the ITT population can produce wider CIs owing to attrition than those from the per-protocol population, there may be some bias toward finding no difference between treatments in the ITT analysis. Thus, results of the per-protocol sensitivity analysis can support the results from the ITT population if the outcomes from both analyses are consistent.22
The ITT population included all randomized patients who had weekly mean assessments of baseline and postbaseline diary-based daily pain ratings. The per-protocol population included only those patients in the ITT population who adhered to a predetermined set of key protocol criteria: adherence to study drug therapy, Michigan Neuropathy Screening Instrument score less than 3 at study entry, gabapentin taper length of not more than 14 days, no use of prohibited medications, and no HbA1c
level greater than 12% after randomization.
The primary analysis used a restricted, likelihood-based mixed-models repeated-measures (MMRM) model that included explanatory terms for the fixed, categorical effects of treatment, investigator, week, and treatment by week interaction, as well as the continuous, fixed covariates of baseline average pain severity ratings and the baseline by week interaction. An unstructured correlation matrix was used to model the within-patient errors. The noninferiority assessments were made using pairwise CIs comparing treatments at week 12.
Sensitivity analyses of noninferiority at end point were also characterized using baseline observation carried forward (BOCF) and last observation carried forward (LOCF) imputation techniques for the weekly means, with analysis based on an analysis of covariance (ANCOVA) model with terms for treatment, baseline, and investigator for both the ITT and per-protocol populations. Estimates from MMRM/ANCOVA models reported here are least-squares means.
Baseline characteristics were compared using the Fisher exact test (nominal outcomes) or analysis of variance (numeric outcomes) with terms for treatment and investigator. Analyses of secondary efficacy measures were conducted using the MMRM model as described for the primary analysis (if collected at multiple time points after baseline) and the ANCOVA model as already described (if collected at baseline or end point only) but without noninferiority assessment.
Other than for assessment of noninferiority, the term significant indicates statistical significance at a 2-sided level of .05 or less; a significance level of .10 or less is used for tests of interaction. All statistical analyses were performed using SAS statistical software, version 9.1.3 (SAS Institute, Cary, NC).