Does Treatment with Duloxetine for Neuropathic Pain Impact Glycemic Control?

Clinical Care/Education/Nutrition ORIGINAL ARTICLE

Does Treatment With Duloxetine for Neuropathic Pain Impact Glycemic Control?

1 1 THOMAS HARDY, MD, PHD MARY ARMBRUSTER, MSN, CDE betic peripheral neuropathic pain (DPNP) 2 3,4 RICHARD SACHSON, MD ANDREW J.M. BOULTON, MD, FRCP vary from 3% to Ͼ20% (4). 1 SHUYI SHEN, PHD For many agents, the data supporting effectiveness is limited. In addition, nearly all treatments are associated with OBJECTIVE — We examined changes in metabolic parameters in clinical trials of duloxetine safety or tolerability issues. One category for diabetic peripheral neuropathic pain (DPNP). of side effects common to a number of the antidepressant and anticonvulsant drugs RESEARCH DESIGN AND METHODS — Data were pooled from three similarly de- is related to metabolic changes. Weight signed clinical trials. Adults with diabetes and DPNP (n ϭ 1,024) were randomized to 60 mg ϭ gain is seen with tricyclic antidepressants duloxetine q.d., 60 mg b.i.d., or placebo for 12 weeks. Subjects (n 867) were re-randomized (TCAs) (5) and anticonvulsants (e.g., val- to 60 mg duloxetine b.i.d. or routine care for an additional 52 weeks. Mean changes in plasma glucose, lipids, and weight were evaluated. Regression and subgroup analyses were used to proate, gabapentin, pregabalin) (6). identify relationships between metabolic measures and demographic, clinical, and electrophys- Changes in plasma glucose have also been iological parameters. reported with TCAs (7,8) and phenytoin (9), and dyslipidemia can be seen with RESULTS — Duloxetine treatment resulted in modest increases in fasting plasma glucose carbamazepine (10). in short- and long-term studies (0.50 and 0.67 mmol/l, respectively). A1C did not increase in Duloxetine has demonstrated efficacy placebo-controlled studies; however, a greater increase was seen relative to routine care in in several large clinical trials of DPNP long-term studies (0.52 vs. 0.19%). Short-term duloxetine treatment resulted in mean weight (11–13) and is one of only two drugs to Ϫ Ͻ loss ( 1.03 kg; P 0.001 vs. placebo), whereas slight, nonsignificant weight gain was seen in have received regulatory approval for the both duloxetine and routine care groups with longer treatment. Between-group differences were treatment of this condition. As a selective seen for some lipid parameters, but these changes were generally small. Metabolic changes did not appear to impact improvement in pain severity seen with duloxetine, and nerve conduction reuptake inhibitor of both norepineph- was also not significantly impacted by treatment. rine and serotonin, it shares some features with secondary amine TCAs such as nor- CONCLUSIONS — Duloxetine treatment was associated with modest changes in glycemia triptyline. However, it differs from TCAs in patients with DPNP. Other metabolic changes were limited and of uncertain significance. in some aspects, including the lack of in- These changes did not impact the significant improvement in pain observed with duloxetine teraction with acetylcholine receptors in treatment. vitro. In clinical trials of major depressive disorder, duloxetine was not associated Diabetes Care 30:21–26, 2007 with weight gain; however, other metabolic parameters, such as fasting glucose europathy is the most common di- risk factor for diabetic foot complications and lipids, were not tested (14). abetic microvascular complication, and nontraumatic amputations. Though The clinical trials of duloxetine for N affecting up to 50% of all individu- these late complications frequently occur DPNP provide a large database to evaluate als with diabetes (1). Substantial morbid- in the insensate foot, symptomatic DPN potential metabolic effects of duloxetine ity is associated with diabetic peripheral also significantly impacts quality of life in patients with diabetes. The current neuropathy (DPN), which is the leading (2,3). Estimates of the prevalence of dia- study examined pooled data from both ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● short- and long-term studies to assess From 1Eli Lilly, Indianapolis, Indiana; the 2Department of Internal Medicine, Southwestern Medical Center, changes in weight, glycemia, and plasma University of Texas, Dallas, Texas; the 3Division of Endocrinology, Metabolism, and Diabetes, University of lipids and to test for relationships be- Miami School of Medicine, Miami, Florida; and the 4University Department of Medicine, Manchester Royal tween metabolic changes and baseline Infirmary, Manchester, U.K. clinical factors, as well as the analgesic Address correspondence and reprint requests to Thomas Hardy, MD, PhD, Eli Lilly, Lilly Corporate Center DC 2138, Indianapolis, IN 46285. E-mail: [email protected]. response to duloxetine treatment. Received for publication 10 May 2006 and accepted in revised form 29 September 2006. R.S. has received research grants from and has been a speaker for Eli Lilly. A.J.M.B. serves on an advisory RESEARCH DESIGN AND panel for Eli Lilly. Data from this study were previously published in abstract form in 2004, specifically as “The safety of METHODS — Three randomized tri- duloxetine in the long-term treatment of diabetic neuropathic pain” at the 20th Annual Meeting of the als (each with a 12-week, double-blind, American Academy of Pain Medicine in Orlando, Florida, and as “Duloxetine at doses of 60 mg QD and 60 placebo-controlled acute phase and a 52- mg BID is effective treatment of diabetic neuropathic pain” at the 56th Annual Meeting of the American week, open-label extension phase) were Academy of Neurology in San Francisco, California. Abbreviations: BPI, brief pain inventory; DPN, diabetic peripheral neuropathy; DPNP, DPN pain; FPG, included in these analyses. Entry criteria fasting plasma glucose; TCA, tricyclic antidepressant. and study designs were nearly identical A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion for these large, phase-three clinical trials, factors for many substances. and details of these studies have been pre- DOI: 10.2337/dc06-0947. Clinical trial reg. no. NCT00058968, clinicaltrials.gov. viously published (11–13). Briefly, pa- © 2007 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby tients were adults with type 1 or type 2 marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. diabetes and with bilateral peripheral

DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 21 Metabolic parameters in duloxetine DPNP trials

Table 1—Baseline characteristics of patients in pooled DPNP studies study. Raw values were used for the analysis of fasting glucose, A1C, weight, and Acute studies Extension studies electrophysiology measures, and rank- transformed data were used for other lab- Duloxetine Placebo Duloxetine Routine care oratory analytes. The change from n 685 339 580 287 baseline to last observation carried for- Age (years) 59.7 60.1 59.4 59.2 ward in fasting glucose and A1C were also Male (%) 42.9 46.6 55.7 56.4 analyzed using subgroups based on baseline BMI (Ͻ30 and Ն30 kg/m2), A1C Caucasian (%) 85.4 85.8 86.0 83.3 Յ Ͼ Type 2 diabetes (%) 87.7 89.4 87.4 87.8 level ( 8 and 8%), and change in status Duration of DPN (years) 4.04 3.85 4.09 3.52 of peroneal F wave (same, better, and Smoking (%) 14.0 13.9 13.4 13.6 worse) using an ANCOVA model with FPG (mmol/l) 9.75 10.09 10.18 10.67 terms for treatment, subgroup, study, and A1C (%) 7.81 7.86 7.74 7.90 subgroup-by-treatment interaction, with Triglycerides (mmol/l) 2.70 2.28 2.44 2.46 baseline score as a covariate. Analysis of LDL cholesterol (mmol/l) 2.95 2.93 3.01 3.02 the primary efficacy variable (24-h aver- HDL cholesterol (mmol/l) 1.19 1.21 1.21 1.23 age pain severity) and one of the second- Total cholesterol (mmol/l) 5.24 5.14 5.26 5.34 ary efficacy variables (BPI interference) was performed using subgroups based on Weight (kg) 94.29 95.03 92.99 94.88 Յ Ͼ Pain severity (0–10 scale) 5.90 5.70 NA NA baseline A1C level ( 8 and 8%) and baseline peroneal F wave (presence or absence), as well as the same ANCOVA neuropathic pain of at least moderate se- MedRA terms. A history of hypoglycemic model terms as previously listed. Categor- verity. Peripheral neuropathy was con- events was taken from patients at each ical data (categorical change in fasting firmed with the Michigan Neuropathy visit. Patients kept a daily diary of average glucose and A1C, diabetes-related ad- Screening Instrument (15). Entry A1C pain severity during the acute-phase stud- verse events, significant hypoglycemic Ͼ12% and major depressive disorders ies (0–10 scale), which was then averaged episode, and shift of status in electrophys- were among the exclusion criteria. on a weekly basis. Interference of pain iology measures) were assessed using the Patients analyzed for this study were with daily activities was measured using Cochran-Mantel-Haenszel test to adjust randomized to placebo or 60 mg dulox- the brief pain inventory (BPI) interference for study. Pearson’s correlation coeffi- etine q.d. or 60 mg b.i.d. during the acute scale (16). Nerve conduction studies were cients (r) were calculated to evaluate cor- phase of treatment. Patients completing performed in two of three studies at base- relations between the continuous 12 weeks of treatment were then re- line and at the conclusion of both the variables of pain severity score and meta- randomized (2:1) to 60 mg duloxetine acute and extension phases using the NC- bolic parameters. Data for all subjects b.i.d. or routine care. In one of the stud- stat automated nerve conduction testing were used for correlations between two ies, some patients were also randomized system (NeuroMetrix, Waltham, MA). Pa- baseline variables, but only data from du- to 20 mg duloxetine q.d. This dose is not rameters measured included peroneal A loxetine-treated patients were used when approved for treatment of DPNP, and wave (presence or absence), deep peroneal either variable in the correlation involved those patients were not included in the F wave latency, ulnar F wave latency, ul- change during treatment. No adjustments analysis of the acute studies (but were in- nar distal sensory latency, and deep per- were made for multiple comparisons. cluded in the extension-phase data anal- oneal compound action potential Path analysis was used to test whether ysis if they entered the extension phase). amplitude. All were studied in the non- the change in 24-h average pain severity Only acetaminophen was allowed as ad- dominant extremities. Electrophysiology depends on the change in fasting glucose, juvant pain therapy in all patients during measures from extension-phase studies A1C, or individual lipid parameters— the acute phase and in duloxetine-treated were reported by study and were not versus the alternative that the reduction in patients during the extension phase. In- pooled for the current analysis. Some of 24-h average pain severity is due to a di- vestigators were not restricted in their these nerve conduction results have been rect analgesic effect of the treatment and is choice of treatments for the routine care previously reported (17). independent of the treatment effect on group, including allowance for combina- metabolic parameters. tion therapy. No specific directions were Statistical analyses included in the study protocols with re- Unless otherwise noted, the current anal- RESULTS — Pooled analysis from the gard to management of diabetes or dyslip- yses included all subjects who received at three studies included 685 duloxetine- idemia, nor were there any specific least one dose of treatment and had at treated patients from the acute studies dietary or physical exercise instructions. least one postbaseline measurement. and 580 patients treated with 60 mg du- Fasting glucose, lipids, A1C, and Mean changes from baseline to last obser- loxetine b.i.d. in extension-phase studies. other clinical chemistries were deter- vation carried forward in weight, fasting Baseline demographics are listed in Table mined by standard techniques at a central glucose, A1C, and electrophysiology mea- 1. Type 2 diabetes predominated in all laboratory (Covance Laboratories, Prince- sures were assessed using an ANCOVA treatment groups, and mean baseline A1C ton, NJ). Diabetes-related adverse events model that included baseline measure, was very similar between treatment were determined from unsolicited reports treatment, and study. Other laboratory groups in both acute and extension-phase to investigators, which were then identi- measures were examined using an ANOVA studies. Leading medications used in the fied using a search for diabetes-related model with terms for treatment and routine care group during the extension

22 DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 Hardy and Associates

Table 2—Mean changes in metabolic and pain parameters in pooled DPNP studies P ϭ 0.002, respectively). Weight de- creased in duloxetine patients during Ϫ Acute studies Extension studies acute studies, unrelated to dose ( 1.03 kg; P Ͻ 0.001 vs. placebo), and a negative Duloxetine Placebo P Duloxetine Routine care P correlation between weight change and baseline weight was observed (r ϭ n 685 339 580 287 Ϫ Ͻ FPG (mmol/l) 0.50 Ϫ0.11 0.064 0.67 Ϫ0.64 Ͻ0.001 0.27; P 0.001). No other significant A1C (%) Ϫ0.09 Ϫ0.07 0.766 0.52 0.19 Ͻ0.001 between-group changes were observed. Triglycerides (mmol/l) Ϫ0.19 0.07 0.106 0.05 0.00 0.854 LDL cholesterol (mmol/l) 0.08 Ϫ0.02 0.075 0.02 Ϫ0.09 0.068 Relationships between pain and HDL cholesterol (mmol/l) 0.03 0.00 0.008 Ϫ0.01 Ϫ0.08 0.002 metabolic parameters Total cholesterol (mmol/l) 0.07 Ϫ0.03 0.090 0.06 Ϫ0.16 0.005 We evaluated potential relationships be- Weight (kg) Ϫ1.03 0.03 Ͻ0.001 0.31 0.49 0.531 tween pain severity and demographic and Pain severity (0–10 scale) Ϫ2.70 Ϫ1.64 Ͻ0.001 NA NA metabolic parameters using correlational, subgroup, and regression (path) analyses. Mean baseline pain levels were signifi- phase were gabapentin (48%), venlax- baseline A1C Ͻ8% during acute studies cantly higher in females and in non- afine (17%), and amitriptyline (15%). On (between subgroups P Ͻ 0.001). How- Caucasians but did not differ significantly average, patients reported moderately se- ever, relative FPG change between treat- based on age (Ͻ65 or Ͼ65 years), type of vere pain before treatment (Table 1), as ment groups was similar regardless of diabetes (1 or 2), duration of neuropathy measured using the Likert scale (18). baseline glycemic control (subgroup-by- (Ͻ2orϾ2 years), or smoking status (data therapy P ϭ 0.843). The BMI categories not shown). Among the measured meta- Metabolic changes in acute and long- used were not predictive of treatment- bolic parameters, baseline BMI and tri- term studies specific, or of overall, FPG change in the glyceride levels were positively but Mean changes in glycemic parameters, acute studies (P values for subgroup and weakly correlated with baseline pain se- plasma lipids, and weight are listed in subgroup-by-therapy, both Ͼ0.79). verity (r ϭ 0.09, P Ͻ 0.001 and r ϭ 0.07, Table 2. In the short-term (12-week) Therefore, the presence of obesity was not P ϭ 0.026, respectively). Other baseline studies, duloxetine-treated patients expe- predictive of change in glycemic parame- lipid values and A1C were not correlated rienced a modest increase in mean fasting ters in these studies, but a higher baseline with pain scores. Importantly, changes in plasma glucose (FPG), whereas a small A1C was inversely related to glycemic metabolic parameters appeared to have decrease was seen in the placebo group change during duloxetine treatment. little or no impact on the improvement in (0.5 vs. Ϫ0.11 mmol/l; P ϭ 0.064 for Similar results were observed in sub- pain severity during duloxetine treat- comparison; P ϭ 0.002 for change from groups with different baseline A1C values ment. Only change in LDL and total cho- baseline in duloxetine group). The in- in the extension studies. A1C changes lesterol were significantly correlated with crease was higher in the group receiving were inversely related to pretreatment change in pain severity, but these correla- duloxetine twice daily (0.55 mmol/l) A1C levels in both duloxetine and routine tions were weak (r ϭϪ0.10, P Ͻ 0.05 compared with the once-daily group care groups, and the therapy-by- and r ϭϪ0.13, P Ͻ 0.05, respectively). (0.46 mmol/l), but the difference was not subgroup interaction was not significant Furthermore, path analysis suggests that statistically significant. These changes (P ϭ 0.765). However, when comparing Ͻ2% of the treatment effect on neuro- were not associated with a significant in- glycemic categories at baseline and end pathic pain could be attributed to changes crease in A1C in the duloxetine group. point (Ͻ7, 7–8, and Ͼ8%), more dulox- in weight, FPG, A1C, or individual lipid Instead, a small statistically significant de- etine-treated patients shifted to higher parameters. Improvements in BPI inter- crease in A1C was observed during dulox- categories, and fewer improved catego- ference scores during duloxetine treat- etine treatment (Ϫ0.09%; P ϭ 0.013). An ries, than with routine care. Treatment ment were also independent of changes in increase in FPG was also seen during du- groups did not differ in the number of these metabolic measures. loxetine treatment in extension studies diabetes-related adverse events or in the fre- (0.67 mmol/l), whereas the routine care quency of hypoglycemic events in either experienced a decline (Ϫ0.64 mmol/l). short-term or extension-phase studies. Relationships between measures of A1C increased in both the duloxetine and Duloxetine-treated patients experi- nerve function and metabolic routine care groups; however, the change enced a small decrease in triglycerides parameters was significantly larger in the duloxetine and small increases in HDL, LDL-C, and We examined potential relationships of group (0.52 vs. 0.19%; P Ͻ 0.001). These total cholesterol during acute studies. baseline electrophysiology measures with glycemic changes varied somewhat be- Only the HDL cholesterol change was sig- pain severity or metabolic parameters. tween studies, but it did not appear that nificantly different from placebo. During Only ulnar F wave latency was weakly, any single study was overly influential re- extension studies, the direction of change but significantly, correlated with baseline garding these findings. In particular, A1C in triglycerides and HDL cholesterol was pain severity (r ϭϪ0.09; P ϭ 0.048). changes in the long-term studies were reversed in the duloxetine group, but Focusing on lower-extremity nerve func- consistent. changes remained small in magnitude. tion, significant inverse correlations were Subgroup analyses of glucose change Routine care was associated with signifi- observed between baseline peroneal F were also performed as a function of base- cantly greater declines in total and HDL wave latency and baseline LDL choles- line A1C and baseline BMI. FPG increases cholesterol (Ϫ0.16 vs. ϩ0.06 mmol/l, terol (r ϭ 0.19; P Ͻ 0.001), HDL choles- were significantly greater in patients with P ϭ 0.005 and Ϫ0.08 vs. 0.01 mmol/l, terol (r ϭϪ0.26; P Ͻ 0.001), and total

DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 23 Metabolic parameters in duloxetine DPNP trials

Table 3—Mean electrophysiology measures in extension-phase studies

Duloxetine (n ϭ 73) Routine care (n ϭ 40) Baseline End point Change Baseline End point Change Study HMAVa Ulnar F wave (ms) 31.66 31.73 0.07 31.33 31.60 0.27 Ulnar distal sensory latency (ms) 3.37 3.42 0.05 3.42 3.44 0.01 Peroneal F wave (ms) 58.35 58.86 0.51 59.76 60.83 1.07 Peroneal CMAP (mV) 1.29 1.20 Ϫ0.09 1.30 1.21 Ϫ0.10

Duloxetine (n ϭ 118) Routine care (n ϭ 49) Study HMAVb Ulnar F wave (ms) 29.84 30.03 0.19 30.98 31.26 0.28 Ulnar distal sensory latency (ms) 3.32 3.29 Ϫ0.03 3.38 3.31 Ϫ0.07 Peroneal F wave (ms) 56.31 56.54 0.23 54.72 56.43 1.71* Peroneal CMAP (mV) 1.47 1.44 Ϫ0.03 1.46 1.40 Ϫ0.06 Data are mean values from the two 12-month extension-phase studies in which electrophysiology parameters were measured. Only those patients that had baseline and postbaseline measurements in the same limb are included. For some parameters, the number of patients with measurements may be less than what is listed (n) if not all measurements were completed or if different limbs were used at baseline and end point. *P ϭ 0.05 compared with the duloxetine group. No other between-group differences reached statistical significance. CMAP, compound muscle action potential. cholesterol (r ϭϪ0.15; P ϭ 0.002). No care group. There were also some be- Weight change was not associated significant correlations were seen between-group differences in lipid parame- with glycemic changes during duloxetine tween this parameter and baseline BMI, ters in both placebo-controlled and treatment. Mean weight loss was seen A1C, or triglycerides (P Ͼ 0.1 for all). No routine care–controlled studies, but these with short-term duloxetine treatment. differences were seen in acute studies be- tended to be small. Furthermore, weight change was in- tween duloxetine and placebo groups It is unclear how duloxetine might af- versely correlated with baseline weight in with respect to change in any of the mea- fect glucose homeostasis, but similar ef- these studies, such that greater degrees of sured electrophysiology parameters. Nei- fects have been reported for some TCAs. weight loss were seen in patients with ther group showed evidence of a decline For example, nortriptyline significantly higher baseline weights. In contrast, in nerve conduction velocities using mean increased glucose levels in mice (7), and a small, nonsignificant weight gain was change analyses or categorical analyses detrimental effect on glycemic control in seen in extension studies. This weight in- using a priori-defined cutoffs (data not humans has also been suggested (8). Du- crease was based on a new baseline estab- shown). In two extension-phase trials, loxetine is a selective serotonin and nor- lished after re-randomization from the nerve conduction testing was completed epinephrine reuptake inhibitor (19). acute studies, making it difficult to quan- in 191 duloxetine-treated patients and 89 While nortriptyline shares these at- tify the long-term weight effects of dulox- patients receiving routine care. Changes tributes, it also has other pharmacological etine. Nonetheless, weight change with in nerve conduction velocities and ampli- properties not seen with duloxetine (e.g., up to 15 months of treatment appears to tudes were generally small and not signif- cholinergic antagonism) (19,20). The role be small. Long-term comparative studies icantly different between treatment of serotonin in glucose homeostasis is un- with anticonvulsants used for DPNP groups (Table 3). The routine care group certain. Both inhibition and stimulation would be of interest, as weight gain is a did show a slightly greater slowing of per- of insulin secretion have been observed common effect of many agents in this oneal F wave conduction, but this barely with serotonergic agonists (21,22). Inter- class (28). reached statistical significance. Given the estingly, the selective serotonin reuptake Hyperglycemia is known to have det- small changes in the measured electro- inhibitor fluoxetine improved insulin rimental effects on nerve function and on physiology parameters during the course sensitivity in obese humans (23), and the painful symptoms in DPN (4). The mod- of these studies, we did not test for corre- serotonergic anxiolytic buspirone re- est metabolic changes observed in this lations between changes in metabolic pa- duced glucose-induced hyperglycemia in study did not impact pain relief during rameters and change in nerve conduction. mice (24). On the other hand, increased duloxetine treatment, and no decline in noradrenergic effects may be important, neurological function was detected dur- CONCLUSIONS — This analysis as catecholamines can promote hypergly- ing these studies. It is interesting that summarizes changes in metabolic param- cemia through multiple mechanisms baseline glycemic control was not corre- eters during three registration trials of du- (e.g., inhibition of insulin secretion, stim- lated with pain severity; however, limit- loxetine for DPNP. Briefly, we observed ulation of gluconeogenesis, decrease in ing the entry A1C (Յ12%) may have modest increases in FPG during both insulin sensitivity) (25,26). Finally, it has limited our ability to see an association. short- and long-term treatment with du- been reported that hypothalamic infusion Likewise, duration of diabetes or DPN loxetine. This was associated with a mod- of norepinephrine plus serotonin impairs and measures of nerve conduction were est increase in A1C only in the longer- glucose-stimulated insulin release in ro- not correlated with reported pain in pa- term studies, which was statistically dents (27). Thus, an additive effect of tients entering these studies. Baseline tri- greater than the increase seen in a routine both neurotransmitters is possible. glycerides and BMI were significantly

24 DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 Hardy and Associates correlated with baseline pain intensity, RE, Sosenko JM, Ziegler D: Diabetic neu- Mallinckrodt CH, Watkin JG, Martynov but this association was weak. Among ropathies: a statement by the American OV: Safety and tolerability of duloxetine other baseline characteristics, female pa- Diabetes Association (Position State- in the treatment of major depressive dis- tients reported a significantly higher base- ment). Diabetes Care 28:956–962, 2005 order: analysis of pooled data from eight line pain severity. Sex differences in pain 2. Vileikyte L, Peyrot M, Bundy C, Rubin RR, placebo-controlled clinical trials. Hum Leventhal H, Mora P, Shaw JE, Baker P, Psychopharmacol 20:327–341, 2005 perception have been previously reported Boulton AJ: The development and valida- 15. Feldman EL, Stevens MJ: Clinical testing (29,30), but the mechanisms underlying tion of a neuropathy- and foot ulcer–spe- in diabetic peripheral neuropathy. Can these differences remain to be defined. cific quality of life instrument. Diabetes J Neurol Sci 21:S3–S7, 1994 The current study has a number of Care 26:2549–2555, 2003 16. Cleeland CS, Ryan KM: Pain assessment: limitations. Importantly, the trials in- 3. Vinik EJ, Hayes RP, Oglesby A, Bastyr E, global use of the Brief Pain Inventory. Ann cluded in this analysis were not initially Barlow P, Ford-Molvik SL, Vinik AI: The Acad Med Singapore 23:129–138, 1994 designed to evaluate the impact of treat- development and validation of the Nor- 17. Gozani S, Megerian T, Waninger A, Xie L, ment on glycemic control or other meta- folk QOL-DN, a new measure of patients’ Wernicke J: Automated nerve conduction bolic parameters. For example, perception of the effects of diabetes and testing technology to assess peripheral investigators were not given specific in- diabetic neuropathy. Diabetes Technol nerve function in diabetic neuropathic Ther 7:497–508, 2005 pain patients during duloxetine treat- structions on glycemic targets and were 4. Boulton AJ, Malik RA, Arezzo JC, Sosenko ment. Presented at the 64th Annual Meet- allowed to adjust glucose- and lipid- JM: Diabetic somatic neuropathies. Diabe- ing of the American Diabetes Association, lowering therapies during the course of tes Care 27:1458–1486, 2004 4–8 June 2004, at the Orange County these trials. Likewise, the routine care 5. Zimmermann U, Kraus T, Himmerich H, Convention Center, Orlando, Florida group in extension studies included a Schuld A, Pollmacher T: Epidemiology, 18. Serlin RC, Mendoza TR, Nakamura Y, Ed- number of different treatments for dia- implications and mechanisms underlying wards KR, Cleeland CS: When is cancer betic neuropathic pain, which may have drug-induced weight gain in psychiatric pain mild, moderate or severe? Grading divergent effects on some of the parame- patients. J Psychiatr Res 37:193–220, 2003 pain severity by its interference with func- ters we have analyzed. For these reasons, 6. Sheth RD: Metabolic concerns associated tion. Pain 61:277–284, 1995 the current study does not allow defini- with antiepileptic medications. Neurology 19. Bymaster FP, Dreshfield-Ahmad LJ, tive conclusions about the metabolic ef- 63:S24–S29, 2004 Threlkeld PG, Shaw JL, Thompson L, 7. Erenmemisoglu A, Ozdogan UK, Saray- Nelson DL, Hemrick-Luecke SK, Wong fects of duloxetine or other medications men R, Tutus A: Effect of some antide- DT: Comparative affinity of duloxetine utilized in these studies. pressants on glycaemia and insulin levels and venlafaxine for serotonin and norepi- Obviously, the patients included in of normoglycaemic and alloxan-induced nephrine transporters in vitro and in vivo, these trials reflect a selected study popu- hyperglycaemic mice. J Pharm Pharmacol human serotonin receptor subtypes, and lation, and this may limit the generaliz- 51:741–743, 1999 other neuronal receptors. Neuropsycho- ability of our findings. For example, this 8. Lustman PJ, Griffith LS, Clouse RE, pharmacology 25:871–880, 2001 may explain some of the differences be- Freedland KE, Eisen SA, Rubin EH, Car- 20. Potter WZ, Mani HK, Rudorfer MV: Tri- tween our results and those of epidemio- ney RM, McGill JB: Effects of nortriptyline cyclics and tetracyclics. In Textbook of Psy- logical studies that have looked at on depression and glycemic control in di- chopharmacology. Schatzberg AF, Nemeroff relationships between metabolic parame- abetes: results of a double-blind, placebo- CB, Eds. Washington, DC, American Psy- controlled trial. Psychosom Med 59:241– chiatric Publishing, 1995, p. 141–160 ters and nerve function or pain symptoms 250, 1997 21. Coulie B, Tack J, Bouillon R, Peeters T, (rev. in 1). Nonetheless, these trials pro- 9. Saudek CD, Werns S, Reidenberg MM: Janssens J: 5-hydroxytryptamine-1 recep- vide a large database that is likely repre- Phenytoin in the treatment of diabetic tor activation inhibits endocrine pancre- sentative of the broader DPNP population symmetrical polyneuropathy. Clin Phar- atic secretion in humans. Am J Physiol We hope that the findings reported here macol Ther 22:196–199, 1977 274:E317–E320, 1998 will help clinicians make informed deci- 10. Verrotti A, Domizio S, Angelozzi B, Sa- 22. Adeghate E, Ponery AS, Pallot D, Parvez sions regarding the risks and benefits of batino G, Morgese G, Chiarelli F: Changes SH, Singh J: Distribution of serotonin and duloxetine in patients seeking relief from in serum lipids and lipoproteins in epilep- its effect on insulin and glucagon secre- painful diabetic neuropathy. tic children treated with anticonvulsants. tion in normal and diabetic pancreatic tis- J Paediatr Child Health 33:242–245, 1997 sues in rat. Neuro Endocrinol Lett 20:315– 11. Goldstein DJ, Lu Y, Detke MJ, Lee TC, 322, 1999 Iyengar S: Duloxetine vs. placebo in pa- 23. Maheux P, Ducros F, Bourque J, Garon J, Acknowledgments— This study was sup- tients with painful diabetic neuropathy. Chiasson JL: Fluoxetine improves insulin ported by Eli Lilly. Pain 116:109–118, 2005 sensitivity in obese patients with non-in- The authors thank Drs. Edward Bastyr, 12. Raskin J, Smith TR, Wong K, Pritchett YL, sulin-dependent diabetes mellitus inde- Amy Chappell, Louis Vignati, and Joachim D’Souza DN, Iyengar S, Wernicke JF: Du- pendently of weight loss. Int J Obes Relat Vernicke for their contributions to the inter- loxetine versus routine care in the long- Metab Disord 21:97–102, 1997 pretation of data for this manuscript; Beth term management of diabetic peripheral 24. Sugimoto Y, Takashima N, Noma T, Pangallo for her technical assistance in pre- neuropathic pain. J Palliat Med 9:29–40, Yamada J: Effects of the serotonergic an- paring reports on which this manuscript was 2006 xiolytic buspirone on plasma glucose and based; and Fujun Wang for his contribu- 13. Raskin J, Pritchett YL, Wang F, D’Souza glucose-induced hyperglycemia in mice. tions to the statistical analysis of data for this DN, Waninger AL, Iyengar S, Wernicke J Pharmacol Sci 93:446–450, 2003 manuscript. JF: A double-blind, randomized multi- 25. Henequin JC: Cell biology of insulin se- center trial comparing duloxetine with cretion. In Joslin’s Diabetes Mellitus. 13th placebo in the management of diabetic ed. Kahn C, Weir G, Eds. Baltimore, References peripheral neuropathic pain. Pain Med MD, Williams and Wilkins, 1994, p. 1. Boulton AJ, Vinik AI, Arezzo JC, Bril V, 6:346–356, 2005 56–80 Feldman EL, Freeman R, Malik RA, Maser 14. Hudson JI, Wohlreich MM, Kajdasz DK, 26. Lembo G, Capaldo B, Rendina V, Iac-

DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 25 Metabolic parameters in duloxetine DPNP trials

carino G, Napoli R, Guida R, Trimarco B, infusion of norepinephrine plus serotonin Drug Saf 24:969–978, 2001 Sacca L: Acute noradrenergic activation into the ventromedial hypothalamus im- 29. Sun LS: Gender differences in pain sensi- induces insulin resistance in human skel- pairs pancreatic islet function. Metabolism tivity and responses to analgesia. J Gend etal muscle. Am J Physiol 266:E242–E247, 48:1287–1289, 1999 Specif Med 1:28–30, 1998 1994 28. Jallon P, Picard F: Bodyweight gain and 30. Berkley KJ: Sex differences in pain. Behav 27. Liang Y, Luo S, Cincotta AH: Long-term anticonvulsants: a comparative review. Brain Sci 20:371–380, 1997

26 DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007