Impact of Disease Characteristics on the Efficacy of Duloxetine in Diabetic Peripheral Neuropathic Pain

The present analysis was based on pooled data from three 12-week, double-blind, randomized, placebo-controlled, multicenter trials (studies 1–3). Entry criteria and study designs were nearly identical for all three of these phase 3 clinical trials (9–11). For study 1, patients (n = 457) were randomly assigned to duloxetine, 20 mg q.d., 60 mg q.d., or 60 mg b.i.d., or placebo in a 1:1:1:1 ratio for the 12-week therapy phase, as determined by a computer-generated random sequence using an interactive voice response system. Patients were seen weekly for the first 5 weeks of treatment and biweekly thereafter. Studies 2 and 3 had an identical design and assessed the efficacy and safety of 60 mg q.d. and 60 mg b.i.d. duloxetine. In studies 2 and 3, patients were randomly assigned to duloxetine, 60 mg q.d. or 60 mg b.i.d., or placebo in a 1:1:1 ratio. The duration of treatment for studies 2 and 3 was 12 weeks of active treatment plus a 1-week taper phase.

The entry and exclusion criteria were identical in all three trials. Patients were eligible for the study if they were ≥18 years old with type 1 or type 2 diabetes and bilateral peripheral neuropathic pain of at least moderate severity. The patients’ pain had to begin in the feet and with relatively symmetrical onset. The daily pain must have been present for at least 6 months, and the diagnosis of diabetic peripheral neuropathy was to be confirmed by a score of ≥3 on the Michigan Neuropathy Screening Instrument (16). Patients with axis 1 psychiatric disorders were excluded (17,18).

The primary efficacy measure for all three studies was the weekly mean of the 24-h average pain scores (computed from the patient diary scores), as measured by an 11-point (0 = no pain, 10 = worst possible pain) Likert scale. A secondary efficacy measure was night pain (recorded daily by patients in a diary and calculated weekly), as measured by the 11-point Likert scale.

The consistency of the effect of duloxetine (60 mg q.d. and b.i.d. pooled) on both 24-h average pain and night pain severity was assessed in subgroups of patients stratified by age (<65 years, nonelderly, or ≥65 years, elderly), type of diabetes (type 1or type 2), duration of diabetes (median split <9.18 or ≥9.18 years), duration of diabetic neuropathy (<2, 2 to <6, or ≥6 years), severity of neuropathy (baseline Michigan Neuropathy Screening Instrument score median split <5 or ≥5), baseline pain severity (Brief Pain Inventory average pain, mean split <6 or ≥6), baseline A1C level (median split <7.6 or ≥7.6%), and baseline insulin use (yes/no) from pooled data of the three placebo-controlled studies.

Safety and tolerability were assessed by comparing different duloxetine doses (60 mg q.d. and b.i.d.) with placebo. Safety measures included discontinuations due to adverse events, treatment-emergent adverse events, vital signs (sitting blood pressure and heart rate), and laboratory analytes. Fasting glucose, lipids, A1C, and other clinical chemistries were determined by standard techniques at a central laboratory (Covance Laboratories, Princeton, NJ).

All analyses were conducted on an intent-to-treat basis. Data from all three studies were pooled to gain adequate statistical power to test treatment-by-subgroup interaction. Duloxetine, 20 mg q.d., in study 1 was not included in the analyses because of its lack of efficacy (9).

Baseline was defined as the last non-missing observation at or before randomization, and end point was defined as the last non-missing observation in the acute therapy phase.

For the efficacy measures, changes in 24-h average pain severity and night pain severity from baseline to end point were evaluated using the ANCOVA model with the terms of treatment, study, and baseline. For subgroup analysis, the additional terms “subgroup” and “treatment-by-subgroup interaction” were added to the above ANCOVA model and tested using the type II sum of squares. The treatment-by-subgroup interaction was evaluated at the significance level of 0.10. Using significance level 0.1 for the interaction test is considered more conventional than 0.05 for testing interaction (19). The word “interaction” is to evaluate the consistency of response to duloxetine treatment between (or across) different strata of a subgroup (example: age <65 vs. ≥65).

For safety measures, changes in continuous measures were evaluated using an ANOVA model with the terms of treatment and study. The rank-transformed changes for laboratory data were used in the model, since the distribution of raw data for most of the analytes was skewed. Categorical data were analyzed using the Cochran-Mantel-Haenszel test stratified by study.

Throughout this article, the term “significant” indicates statistical significance, and when presenting efficacy results, the “mean change” refers to “least-squares mean change.”

Baseline demographics and characteristics of patients in all three studies are summarized in Table 1. A majority of the patients in these studies were Caucasian (85.5%), with a mean age of 59.8 years. No significant differences were observed between the duloxetine and placebo groups.

Subgroup analyses from pooled data of all three studies showed that for the 24-h average pain severity and night pain scores, there were no statistically significant interactions of treatment by age, type of diabetes, duration of diabetes, duration of diabetic neuropathy, severity of neuropathy, baseline A1C level, and baseline insulin use (Tables 2 and 3). Similar results also were observed in sex and ethnicity subgroups. Treatment-by-baseline severity interactions were significant for both efficacy measures (P = 0.052 for 24-h average pain and P = 0.026 for night pain severity). Duloxetine’s effect was significantly greater than placebo for either subgroup, whereas the magnitude of treatment group difference was greater for individuals with more severe pain at baseline (Tables 2 and 3).

When the incidence of discontinuation due to adverse events was investigated in patient subgroups stratified by age, sex, type of diabetes, and baseline insulin use, no statistically significant (P > 0.10) treatment-by-subgroup interactions were observed.

The incidence of treatment-emergent adverse events was analyzed from pooled data of three DPNP studies for duloxetine, 60 mg q.d. or 60 mg b.i.d., and placebo groups, and the results are summarized in Table 4. In patients treated with duloxetine, treatment-emergent adverse events had a ≥5% incidence and twice the incidence for placebo-treated patients, and the majority of them were significantly (P < 0.05) different from the placebo group. The incidence of treatment-emergent adverse events also was analyzed in subgroups stratified by age, sex, type of diabetes, and baseline insulin use. Significant treatment-by-subgroup interactions were observed for the following treatment-emergent adverse events: constipation in age subgroups (P = 0.085), fatigue and diarrhea in sex subgroups (P = 0.028 and P = 0.053, respectively), nausea in type of diabetes subgroups (P = 0.034), and nausea in baseline insulin subgroups (P = 0.042).

When the mean changes of diastolic blood pressure and heart rate were investigated in patient subgroups stratified by age, sex, and type of diabetes, there were no statistically significant treatment-by-subgroup interactions. For the mean change of fasting glucose level, there was no significant interaction of treatment by age, sex, type of diabetes, or baseline insulin use.

The subgroup analyses for both pain severity scores showed that there were few statistically significant (P < 0.10) interactions of treatment by subgroups stratified by various baseline characteristics, suggesting that the effectiveness of duloxetine is invariant with respect to those baseline characteristics examined. The treatment-by-baseline severity assessment was not a qualitative measure.

Among the baseline variables investigated, significant treatment-by-baseline severity (Brief Pain Inventory average pain, ≥6 and <6) interactions were observed for both pain efficacy measures. These data suggest that duloxetine is more effective in patients with higher baseline pain severity or that there is a larger drug-placebo difference in patients with greater pain. We cannot exclude the statistical phenomenon of a regression toward the mean, as in both the placebo and duloxetine groups with a higher baseline 24-h average pain; night pain severity is reduced to a higher degree than in individuals with a lower baseline 24-h average pain. However, 24-h pain severity in patients with a higher versus lower baseline 24-h average pain shows a somewhat stronger reduction in the duloxetine group than in the placebo group, supporting the notion that the drug is particularly useful in patients with severe 24-h neuropathic pain.

The subgroup analyses for the efficacy measures presented here focused on weekly mean changes of the 24-h average pain severity and night pain scores. Farrar et al. (20) analyzed the pooled results of 10 placebo-controlled studies involving patients with chronic pain syndromes (diabetic neuropathy, postherpetic neuralgia, chronic low back pain, fibromyalgia, and osteoarthritis) to corroborate the association between change in pain intensity numeric rating scale and an improvement in quantifiable measures of clinical status. The analysis showed that, on average, a reduction of ∼2 points from baseline on an 11-point pain rating scale (equivalent to a 30% reduction on pain severity from baseline) corresponds to a clinically meaningful improvement (20). In the present analysis, the mean changes of the 24-h average pain severity were reduced in the range of 2.3–3.8 points across all the subgroups with duloxetine treatment and 1.4–2.7 points with placebo treatment, thus supporting the clinical relevance of duloxetine’s efficacy in the management of DPNP.

Significantly more patients treated with duloxetine were discontinued because of adverse events compared with patients treated with placebo, and no association between the baseline characteristics examined and the tolerability of duloxetine was observed. There were statistically significant treatment-by-subgroup interactions for several treatment-emergent adverse events, including constipation, fatigue, diarrhea, and nausea in age, sex, type of diabetes, and baseline insulin use subgroups, respectively. The absence of significant cardiovascular changes due to duloxetine therapy in the present analyses and other published reports (9,10) suggests that patients with diabetes undergoing duloxetine treatment do not require additional intensive assessment of their cardiovascular status.

Duloxetine-treated patients had a statistically significant decrease in A1C in these short-term studies (−0.09%; P = 0.013). It is unclear how duloxetine might affect glucose homeostasis, but some noradrenergic antidepressants (e.g., desipramine) have shown to modulate glucose homeostasis (21). It has been reported that increased noradrenergic effects may be associated with elevation of blood glucose in humans, since catecholamines can promote hyperglycemia through multiple mechanisms (e.g., inhibition of insulin secretion, stimulation of gluconeogenesis, and decrease in insulin sensitivity) (22).

The present analysis has several limitations. The results are based on post hoc analyses from pooled data of three acute treatment trials of 12–13 weeks and may not be generalized to a longer duration of treatment of DPNP, a chronic condition requiring management for >12 weeks. Because patients were selected from those with very limited or stable medical conditions and stable doses of concomitant medications, the applicability of these controlled clinical trial results to typical outpatients is limited.

In conclusion, the present analyses suggest that duloxetine was effective and well-tolerated in the management of DPNP, and its effectiveness was invariant with respect to diabetes and/or diabetic peripheral neuropathy baseline conditions, except for pain severity. Diabetes-related determinants such as duration or type of diabetes, glycemic control, and insulin treatment do not predict the effects of duloxetine in DPNP. The same holds true for the duration and severity of the neuropathic process itself. Thus, the efficacy of duloxetine is not limited to particular subgroups but rather is generalizable across a broad spectrum of diabetic patients, including those showing the highest degrees of severity of diabetes or neuropathy. In contrast, efficacy may even be enhanced in cases with relatively high pain intensity.

The authors thank the Du-Flu Product Team at Eli Lilly, the many patients for their voluntary participation in the three clinical trials, the principal investigators, and the statistical analysts. They also thank Jimmy Y. Xu, PhD, for the initial drafting of this manuscript and Elizabeth A. Agostinelli, MA, for editorial review.

All authors accept full responsibility for the conduct of this analysis, were given full access to all data from the analysis, and participated in the decision to publish the data.