Duloxetine compliance and its association with healthcare costs among patients with diabetic peripheral neuropathic pain

Abstract

Objective: Duloxetine is approved to treat diabetic peripheral neuropathic pain (DPNP) in the US. The study objective was to examine the predictors of duloxetine compliance, and its association with healthcare costs among DPNP patients.

Methods: The study used administrative claims databases to identify non-depressed DPNP patients with a duloxetine prescription dispensed between October 1, 2004 and December 31, 2006. Two cohorts of patients were constructed based on compliance to duloxetine therapy over 1-year follow-up with high compliance defined as a medication possession ratio (MPR) ≥0.80. All-cause, diabetes-, and DPNP-related healthcare costs during 1-year follow-up were estimated. Logistic regressions were performed to examine how average daily dose (ADD) of duloxetine and other factors may influence compliance. Multivariate regressions were estimated to examine the association between compliance and healthcare costs.

Results: The study included 1,380 commercially insured (mean age 55 years) and 974 patients with employer-sponsored Medicare supplemental insurance (mean age 75 years). In both populations, patients with an ADD >30 mg were more likely to be compliant with the therapy compared with those with an ADD of ≤30 mg (odds ratio ranged 1.79–3.38, all p<0.05). Controlling for differences in demographics, clinical and economic characteristics, commercially insured low duloxetine compliance patients had greater all-cause ($5,334, p<0.05) and diabetes-related healthcare costs ($3,414, p<0.05) than high-compliance patients, with the biggest difference from inpatient costs (all-cause: $7,508; diabetes-related: $3,785, all p<0.05). Similar trends were found in the Medicare supplemental insured population; however, differences in all-cause healthcare costs were not significant.

Conclusions: DPNP patients with a higher ADD of duloxetine over a 1-year follow-up period were more compliant with the therapy. Duloxetine patients with high compliance were also associated with lower healthcare costs. Due to the use of a retrospective cohort design on administrative claims database, limitations of this analysis include a lack of formal diagnostic testing of patients, and inability to infer causality or measure factors such as DPNP severity that are not captured in such database.

Key words::

• average daily dose
• compliance
• diabetic peripheral neuropathic pain
• duloxetine
• healthcare costs

Introduction

Diabetic peripheral neuropathic pain (DPNP) is a common complication of diabetes mellitus that manifests itself as a painful electric, burning, or shooting sensation. Up to half of all diabetes patients have some degree of DPNP, with 10–20% experiencing moderate or severe pain^1–6. Most DPNP patients experience pain on a daily basis and are at increased risk of mood, anxiety, and sleep disorders^7–9. DPNP also severely compromises patient health-related quality of life and functioning, causing significant disability and distress^9–14.

DPNP is associated with substantial direct and indirect costs¹⁵. Patients who suffer from DPNP have high levels of healthcare resource utilisation, including physician visits, medication use, and hospital stays^12,16. DPNP also leads to significant indirect costs resulting from work loss (manifested by absenteeism), reduced productivity in the workplace, and reduced ability to perform domestic chores^12,16,17. As diabetes becomes more prevalent in an aging and overweight population, the burden of DPNP is expected to rise^13,15,18,19.

While DPNP is incurable, its symptoms can be alleviated with pharmacological therapies, including antidepressants and anticonvulsants²⁰. Among these, duloxetine and pregabalin have been approved by the US Food and Drug Administration to treat DPNP. Duloxetine was officially approved to treat DPNP in late 2004, and clinical guidelines rank it as a first-tier agent^1,21. A recent study found that duloxetine was cost-effective for treating DPNP compared with routine treatment with other pain therapies including antidepressants, anticonvulsants, analgesics, and non-pharmacological treatments²². Duloxetine is available in 20, 30, and 60 mg, and the recommended initial dose to treat DPNP is 60 mg²³. Although duloxetine has been available for treating DPNP since its approval in 2004, no prior studies have examined how duloxetine is prescribed among DPNP patients. Furthermore, no known studies have assessed the distribution of the average daily dose (ADD) of duloxetine for DPNP and its association with compliance or the correlation between compliance and subsequent direct healthcare costs.

The study objectives were to describe patterns of duloxetine treatment for DPNP, including ADD and compliance, and to examine the associations between duloxetine ADD, compliance, and healthcare costs among the commercially insured and Medicare supplemental insured populations, respectively. To address these objectives, the study employed a retrospective cohort design using a large US administrative claims database. Patients who were covered by commercial insurance (18–64 years of age) and patients over age 65 with employer-sponsored Medicare supplemental insurance were analysed separately, and trends were compared across the two populations. Multivariate regression analysis was undertaken to examine the factors associated with duloxetine compliance and its impact on healthcare costs after controlling for differences in demographic and clinical characteristics. By evaluating the relationship between ADD and compliance, the study could address the importance of maintaining an appropriate dose of duloxetine in treating DPNP based on empirical evidence. Furthermore, the findings of association between compliance and healthcare costs will inform the economic implications of continuous treatment of DPNP with duloxetine.

Methods

Data source

The study used the Thomson's Medstat MarketScan Commercial Claims and Medicare Supplemental databases from 2003 to 2007 with a retrospective cohort design. These databases include files for enrolment records as well as medical and pharmacy claims. Information across files can be linked by an encrypted patient identification number. The enrolment file contains information on age, gender, US geographic region, health-plan type, and employment status, as well as monthly enrolment status. Medical claims files contain detailed information about inpatient and outpatient care including date and place of service, provider type, payments and up to 15 International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes. Pharmacy claims contain information for each prescription filled including National Drug Code (NDC), dispense date, quantity, days supplied, and plan and patient paid amounts.

Study sample selection

The commercially insured patients were required to be 18–64 years of age and the Medicare supplemental insured patients at least 65 years. All included patients were required to have a pharmacy claim for duloxetine between October 1, 2004 and December 31, 2006. The corresponding dispense date for the initial claim was denoted as the ‘index date’. Furthermore, patients were required to have continuous enrolment during the 12 months before and after the index date. All patients included had to have a diabetes-related (ICD-9-CM: 250.xx) and a DPNP-related (ICD-9-CM: 250.6x [diabetes with neurological manifestations] or 357.2x [polyneuropathy in diabetes]) diagnosis recorded in any of the diagnosis fields in the medical claims over the 12 months pre-index period. Selected patients also had to have a prescription filled for pain medication (including antidepressants, anticonvulsants, non-narcotics, narcotics, and non-steroidal anti-inflammatory drugs [NSAIDs]) during the 12-month pre-index period. All patients were required to have at least 30 days of duloxetine supply during the 12-month post-index period. Finally, individuals were excluded if they had any pharmacy claim for duloxetine or any diagnosis of depression (ICD-9-CM 296.2, 296.3, 300.4, 309.1, 311) recoded anywhere in the medical claims in the 12 months prior to the index date.

Analysis

Compliance with duloxetine therapy was estimated based on medication possession ratio (MPR), calculated as the total days of supply for all duloxetine prescriptions (across all doses) over the 12-month post-index period divided by 365²⁴. Two patient cohorts were constructed based on MPR: (1) patients with high compliance (MPR ≥80%), and (2) patients with low compliance (MPR <80%)²⁵. For both compliance cohorts, the differences of demographics, healthcare costs and resource utilisation by place of service (e.g., inpatient and outpatient) over the pre-index period were examined. All-cause, diabetes-, and DPNP-related costs were further assessed. Diabetes- and DPNP-related healthcare costs were extracted based on the medical service claims with such diagnoses coded. Diabetes-related medication costs were estimated based on prescriptions of insulin and oral anti-diabetic drugs (OADs). DPNP-related medication costs were estimated based on prescriptions of pharmacological therapies recommended for DPNP treatment¹: (1) tricyclic antidepressants, (2) venlafaxine, (3) duloxetine, (4) pregabalin, (5) gabapentin, and (6) opioids (e.g., tramadol, oxycodone, morphine, hydrocodone, methadone, levorphanol). All costs were adjusted to 2006 US dollars using the medical component of the consumer price index²⁶. Descriptive statistics were summarised separately for both commercial and Medicare supplemental insured populations. Percentages were reported for categorical variables (e.g., gender, health insurance plan type, geographic region, and healthcare utilisation) and statistical differences between cohorts were analysed using chi-square tests. For continuous variables (e.g., age and healthcare costs), mean and standard deviation were reported and S* atudent t-tests were employed to detect statistical differences.

To measure comorbid medical conditions over the 12-month pre-index period, diabetes- and pain-related conditions (as listed in Table 6) were examined. For both populations, the prevalence (percentage of patients) of each medical condition between cohorts was compared via chi-square tests. The prevalence differences between populations were further examined. Charlson comorbidity index (CCI) was estimated based on the 12-month pre-index period to assess the health status²⁷.

The distribution of ADD for duloxetine (calculated based on the prescribed number of days) over the 12-month post-index period was compared between the high- and low-compliance cohorts. Copayment of index duloxetine prescription standardised to 30-day of supply was assessed across study cohorts. Use of pain medications (e.g., antidepressants, anticonvulsants, narcotics, non-narcotics, and NSAIDs), diabetes-related (insulin and/or OADs), DPNP-related (e.g., tricyclic antidepressants, venlafaxine, duloxetine, pregabalin, gabapentin, and opioids)¹, and total prescription medications were also examined.

Logistic regressions were performed to assess the factors associated with high compliance to duloxetine therapy for the commercial and Medicare supplemental insured populations separately. The explanatory variables included age, gender, prescribed by specialist (yes/no indicator) for the initial duloxetine prescription dispensed (physician specialty from the most recent physician visit over the 14 days prior to the duloxetine index claim), ADD (≤30 mg, 31–59 mg, 60 mg, and >60 mg), index duloxetine prescription copayment, pre-index use of insulin or OADs, CCI, and select diabetes- and pain-related comorbidities. Adjusted odds ratios (OR) and 95% confidence intervals (CI) were estimated.

Finally, all-cause, diabetes- and DPNP-related healthcare costs by type of service (inpatient, outpatient, and pharmacy) were compared between compliance cohorts for each population. Healthcare costs were analysed using multivariate generalised linear regression with a log link and gamma distribution²⁸, controlling for age, gender, region, prescribed by specialist, index duloxetine prescription copayment, pre-index use of insulin or OADs, and CCI. Cost measures (e.g., inpatient costs) with a relatively large proportion of zero values (>10% ) were analysed using two-part models²⁹, where the probability of non-zero values was first modelled using a logistic regression and the conditional means for non-zero values were estimated using multivariate generalised linear regression with a log link and gamma distribution in the second stage. Adjusted differences were calculated as the mean differences of all individuals when their compliance group was alternatively coded as ‘0’ (low compliance) and ‘1’ (high compliance)²⁹. Confidence intervals (CIs) for adjusted differences were obtained by bootstrapping with 1000 iterations, and the 95% CIs were based on the 2.5 and 97.5 percentiles³⁰. All analyses were conducted using SAS version 9.1 (SAS Institute, Inc., Cary, NC, USA), and findings of p-values of <0.05 were considered statistically significant.

Results

Demographics and prior healthcare utilisation and costs

The study identified a total of 1,380 commercially insured and 974 Medicare supplemental insured DPNP patients of approximate mean ages of 55 and 75, respectively (Table 1). Most patients had low compliance to duloxetine therapy (66.7% of the commercial patients and 62.0% of the Medicare supplemental insured patients). Patients in the high-compliance cohort were slightly older than those in the low-compliance cohort among the commercially insured patients (55.6 vs. 54.3 years, p<0.05), but there was no significant difference in age between compliance cohorts in the Medicare supplemental insured population. Approximately 50% of the patients in the commercial population were enrolled in a preferred provider organisation, while the majority of Medicare supplemental insured patients had comprehensive insurance (73.2%). For both populations, approximately 80% of patients lived in the north central or southern regions of the US.

Approximately 30% of the patients had duloxetine prescribed by a specialist. In the commercially insured population, there were no significant differences in pre-index utilisation of inpatient and outpatient services between the low- and high-compliance cohorts. However, a significantly larger proportion of patients in the Medicare supplemental insured high-compliance cohort utilised DPNP-related outpatient services than the low-compliance cohort (94 vs. 89%, p<0.05). There were no statistically significant differences in pre-index costs between compliance cohorts except for DPNP-related costs among the commercially insured, where patients in the low-compliance cohort incurred significantly higher DPNP-related costs than those in the high-compliance cohort ($4,091 vs.$3,001, p<0.05).

Pre-index comorbidities

The most common diabetes-related comorbidities were cardiovascular disease and cerebrovascular and peripheral vascular disease, and the most common pain-related conditions were low back pain and osteoarthritis (Table 2). Compared with the commercially insured patients, those with Medicare supplemental insurance had a significantly higher rate of cardiovascular diseases, cerebrovascular and peripheral vascular diseases, low back pain, and osteoarthritis. However, the prevalence of obesity, fibromyalgia syndrome, and migraine was significantly higher among the commercial population. Medicare population also have higher mean CCI than commercial population (1.36 vs. 0.77, p<0.05).

When comparing the comorbid conditions and CCI across study cohorts, the distribution was similar between compliance groups with few exceptions. Retinopathy was more prevalent among low-compliance patients with Medicare supplemental insurance (13.3 vs. 8.9%, p<0.05), and migraine was more common among low-compliance patients in the commercially insured population (4.4 vs. 1.7%, p<0.05). Other metabolic diseases were also more common among low-compliance patients with commercial insurance (12.0 vs. 8.5%, <0.05).

Medication use in post-index period

Overall, approximately two-thirds of the commercially insured patients and half of the Medicare supplemental insured patients had an ADD equal to or greater than 60 mg (Table 3). For both insurance populations, a higher proportion of patients in the low-compliance cohort had an ADD of duloxetine ≤30 mg compared with those in the high-compliance cohort (commercial: 16.4 vs. 7.4%; Medicare supplemental insured: 26.2 vs. 13.5%, both p<0.05). Compared with low-compliance patients, a significantly higher proportion of high-compliance patients with either insurance type were found to have an ADD of 31–59 mg and >60 mg (all p<0.05); however, no significant differences were found for ADD of 60 mg between compliance cohorts. For both populations, patients in the high-compliance group had a higher mean ADD than those in the low-compliance group (commercial: 61 mg vs. 57 mg; Medicare supplemental insured: 56 mg vs. 51 mg; both p<0.05). Patients in the Medicare population had similar copayment for a standard 30-day duloxetine supply (among index claims), however those in the commercially insured low-compliance cohort had significantly higher duloxetine copayment ($29 vs. $24, p<0.05).

Roughly one-third of patients used any antidepressants, with SSRIs being the most commonly dispensed in the low-compliance groups and tricyclic antidepressants being the most frequently prescribed in the high-compliance groups among both populations. Close to 60% of patients used anticonvulsants, and the only significant difference between compliance cohorts was the use of pregabalin in the Medicare supplemental insured population (24.8% in low-compliance cohort vs. 15.1% in high-compliance cohort, p<0.05). Furthermore, more than two-thirds of patients used narcotic medications; hydrocodone and oxycodone were the most frequently used among all commercially insured and the Medicare supplemental insured high-compliance patients, whereas hydrocodone and proproxyphene were the most common opioid agents used among the low-compliance Medicare supplemental insured cohort. Approximately one-third of all patients used NSAIDs; among the commercially insured patients, those in the high-compliance cohort were more likely to use COX-2 inhibitors than those in the low-compliance group (11.7 vs. 7.2%, p<0.05). On average, low-compliance patients with either commercial or Medicare supplemental insurance had a greater number of DPNP-related medications (commercial: 2.3 vs. 2.2; Medicare: 2.2 vs. 2.0, both p<0.05). Moreover, high-compliance patients with Medicare supplemental insurance were dispensed more diabetes-related medications than patients with low duloxetine compliance (1.8 vs. 1.6, p<0.05).

Factors associated with compliance to duloxetine therapy

Table 4 presents the results from the logistic regression models that examined factors associated with duloxetine compliance. Among the commercially insured, patients older than 55 years of age were more likely to be compliant with duloxetine therapy than those aged 18–44 years (OR=2.12, 95% CI=1.30–3.46). Patients paying >$30 copayment were less likely to be compliant compared with those paying $0–10 copayment (OR=0.63, 95% CI=0.47–0.86). Among the Medicare supplemental insured patients, those aged between 75 and 84 were less likely to adhere to duloxetine therapy than those aged 65–74 years (OR=0.75, 95% CI=0.56–0.99), and patients taking OADs were more likely to be compliant (OR=1.38, 95% CI=1.01–1.89). Controlling for demographics and pre-index comorbidities, patients in both insurance populations with an ADD of duloxetine >30 mg were significantly more likely to adhere to the therapy compared to those with an ADD of ≤30 mg (odds ratio ranged 1.79–3.38, all p<0.05).

Compliance and healthcare costs

Table 5 presents the average healthcare costs during the 12-month post-index period as well as the adjusted differences in costs between compliance cohorts. Among commercially insured patients, those with low compliance incurred higher diabetes-related ($10,024 vs.$6,093, p<0.05) and all-cause healthcare costs ($32,407 vs.$25,747, p<0.05) than those with high compliance, primarily due to differences in inpatient costs ($6,584 vs.$2,232 for diabetes-related, $12,851 vs.$4,833 for all-cause, all p<0.05). However, low-compliance patients spent less on DPNP-related pharmacy costs ($1,464 vs.$2,474, p<0.05) than high-compliance patients. Controlling for demographics, pre-index anti-diabetic agent use, index copayment, prescribed by specialist, and CCI, low-compliance patients with commercial insurance had significantly greater all-cause ($5,334, p<0.05) and diabetes-related costs ($3,414, p<0.05) than high-compliance patients, where the difference was primarily attributable to inpatient costs. Patients in the low-compliance cohort incurred significantly lower pharmacy costs (diabetes-related: –$489; DPNP-related: –$1,037: all-cause: –$2,376, all p<0.05), and DPNP-related costs (–$802, p<0.05). Similar trends were found among the Medicare supplemental insured patients for diabetes-related and DPNP-related costs; however the difference in all-cause costs was not statistically significant.

Discussion

This study investigated factors associated with compliance with duloxetine and examined the relationship between compliance and healthcare costs among commercially and Medicare supplemental insured DPNP patients. For both populations, patients with high compliance had, on average, higher ADD and lower healthcare costs. Controlling for differences in demographics and comorbidities between cohorts, patients who maintained a higher ADD (greater than 30 mg) were significantly more likely to be compliant with duloxetine therapy. Also, patients with better compliance with duloxetine therapy had significantly lower healthcare costs than those with low compliance.

For both the commercial and Medicare supplemental insured populations, most patients did not have high compliance with duloxetine therapy, as less than 40% had an MPR of ≥80%. Although the recommended dose for DPNP is 60 mg of duloxetine, approximately one-third of patients had an ADD of <60 mg, and many patients never received an ADD >30 mg. Although the study was unable to investigate the clinical justification for prescribing a lower ADD of duloxetine, patients with ADD over 30 mg were significantly associated with high compliance. This might be due to the fact that duloxetine 30 mg was not the efficacious dose for treating DPNP. Without proper relief of the pain symptoms, patients might stop the treatment. Hence, it is important for clinicians to prescribe the recommended DPNP duloxetine dosage of 60 mg in order to ensure its efficacy and reinforce compliance with the therapy. Similar to a previous study³¹, older commercially-insured patients in the current analysis were found to be more compliant on duloxetine. This might be because age is highly correlated with duration and severity of DPNP³² which were not captured in the study dataset. As DPNP progresses, the associated pain symptoms may worsen over time which may require increased pain medication control. Patients with higher copayment for duloxetine were also less likely to be compliant, which indicates that the cost-sharing burden may influence medication compliance.

Patients in both compliance cohorts had similar comorbidities and healthcare costs in the 12-month pre-index period; however, concomitant drug use and post-index costs were different.

Compared with low-compliance patients, high-compliance patients incurred substantially lower diabetes-related and all-cause healthcare costs, largely due to lower costs of inpatient care. It is not surprising that high-compliance patients had higher pharmacy costs, as by definition they have filled more prescriptions (at least for duloxetine). However, these greater medication costs for high-compliance patients were offset by lower costs in medical services. Inpatient admissions, in particular, have the potential to be reduced among patients with better duloxetine compliance, who may also have good overall compliance on all medications.

DPNP patients took many prescriptions. On average, these patients with either commercial or Medicare supplemental insurance took 19 different medications (including two different diabetes- and two DPNP-related medications) over a 1-year follow-up period. The complexity of existing medication regimens due to chronic medical conditions such as diabetes may impose a significant burden for patients to be compliant with any additional new therapies. Hence, it is important to initiate DPNP patients with treatment at a desirable dosage to avoid unnecessary future adjustment and to ensure the best clinical response. Although NSAIDs are not recommended for DPNP use³³, over one-third of the DPNP patients used this treatment. This number might be under-estimated if patients were obtaining NSAIDs over the counter, which were not captured in the data. Since there is little evidence to support the efficacy of NSAIDs in DPNP and most DPNP patients have many comorbid pain-related medical conditions, the use of NSAIDs observed in this analysis may be driven by pain comorbidities rather than DPNP. Further study is needed to examine the use of NSAIDs in clinical practice for treating DPNP, specifically as their use may result in renal impairment³⁴.

Although Medicare supplemental insured patients in both compliance cohorts had high healthcare utilisation during the pre-index period, they had much lower costs than those with commercial insurance. This is primarily due to the fact that the reimbursed amount by Medicare is generally lower than the commercial insurance companies. Hence, the analyses for commercially-insured and Medicare supplemental insured populations were conducted separately. Furthermore, the distribution of ADD across insurance populations was slightly different as a higher proportion of patients with Medicare supplemental insurance had a lower ADD of duloxetine.

Several limitations need to be noted when interpreting the results for this study. First, confounding may have occurred as a result of unobserved heterogeneity that correlated with the variables of interest (e.g., ADD, compliance) and study outcomes, which prevents causality of the associations identified in the study being inferred. Second, medical conditions were identified based on administrative claims using ICD-9-CM diagnosis codes and were not validated with medical chart records. This information is subject to billing issues, clerical errors, or other erroneous reporting, leading to potentially inaccurate identification of medical conditions. Specifically, medical conditions not recorded in administrative claims were not recognised in this study. Similarly, the data used here did not directly measure the severity of DPNP, which has been found to significantly predict healthcare utilisation and costs³⁵ and associate with diminished quality of life¹⁰. Although the models controlled for diabetes-related comorbidities as a proxy of DPNP severity, the direct measure of severity such as pain scores were not available. The study calculated direct treatment costs based on health insurance reimbursed amount and patient out-of-pocket payment for each service and did not capture indirect costs incurred by patients, including lost productivity and opportunity costs associated with time lost seeking treatment. The prescription claims analysed in the current study could not specify for what medical condition the medications were used, and several of the medications included here were also used for medical problems other than DPNP. Therefore, the DPNP-related pharmacy costs might be over-estimated. On the other hand, the study may under-estimate the overall pharmacy utilisation and associated costs, as over-the-counter medications were not captured in the administrative claims databases. Thus, the overall impact on the DPNP-related pharmacy costs and utilisation was unclear. Moreover, the data provided no information on why some patients remained on a lower ADD of duloxetine; however, physicians' prior experience with duloxetine and certain clinical considerations, such as intolerance of duloxetine or renal impairment among some patients, could have contributed to this observation. Also, the study was not able to assess the clinical intermediate endpoint of duloxetine use (i.e., serum level) or the clinical outcomes (i.e., efficacy and tolerability). Finally, compliance measures were based on medication refill records, which may not represent actual consumption.

Conclusions

Among DPNP patients treated with duloxetine, those with either low or high compliance had similar demographics, comorbidities and pre-index healthcare utilisation and costs. However, patients who remained on a higher ADD of duloxetine over a 1-year follow-up period were more compliant with duloxetine treatment. Patients with higher compliance also incurred lower healthcare costs.

Table 1. Demographics and prior healthcare utilisation and costs.

Demographic	Commercial			Medicare supplemental insured
	Low compliance	High compliance	p-value*	Low compliance	High compliance	p-value*
Number of patients	920	460		604	370
Proportion of patients (%)	66.7	33.3		62.0	38.0
Age (years): mean (SD)	54.3 (7.3)	55.6 (6.2)	<0.05	74.9 (5.6)	74.2 (6.1)	0.07
Female (%)	57.4	61.3	0.16	57.3	55.1	0.51
Plan type (%)			0.09			0.95
Comprehensive	17.1	19.4		74.0	72.4
Point of service	9.7	9.8		0.2	0.3
Preferred provider organisation	50.1	51.1		20.5	20.8
Health maintenance organisation	15.2	14.6		4.3	4.9
Missing	0.6	0.9		0.8	1.4
Region (%)			0.21			0.94
Northeast	4.6	6.5		9.4	8.1
North central	31.1	35.0		46.2	47.8
South	51.9	46.7		33.3	33.8
West	12.1	11.1		10.6	9.7
Missing	0.4	0.7		0.5	0.5
Prescribed by specialist (%)	30.0	30.7	0.80	29.9	30.5	0.60
Prior healthcare utilisation (%)
Inpatient
Diabetes-related	23.0	19.0	0.09	19.0	16.0	0.29
DPNP-related	11.0	10.0	0.54	10.0	7.0	0.22
All-cause	34.0	29.0	0.06	39.0	35.0	0.25
Outpatient
Diabetes-related	97.0	98.0	0.46	96.0	97.0	0.41
DPNP-related	91.0	92.0	0.42	89.0	94.0	<0.05
All-cause	100.0	100.0	—	100.0	100.0	—
Prior healthcare costs ($): mean (SD)
Diabetes-related	10,681 (26,146)	8,319 (19,308)	0.06	5,236 (11,506)	4,905 (10,636)	0.65
DPNP-related	4,091 (11,637)	3,001 (5,716)	<0.05	1,894 (3,765)	2,647 (8,939)	0.13
All-cause	29,235 (42,211)	26,307 (32,985)	0.16	22,264 (23,609)	21,902 (19,147)	0.79

*Chi-square test was used for categorical variables (gender, health-plan region, prescribed by specialist, prior healthcare utilisation) and Student t-test was used for continuous variables (age, prior healthcare costs).

SD, standard deviation. DPNP, diabetic peripheral neuropathic pain-related.

Table 2. Prevalence of comorbid medical conditions over 12-month pre-index period.

	Commercial			Medicare supplemental insured
	Low compliance	High compliance	p-value^†	Low compliance	High compliance	p-value⁁
Diabetes-related conditions (%)
Cardiovascular disease*	63.2	65.7	0.36	77.7	75.4	0.42
Cerebrovascular and peripheral vascular disease*	26.2	28.7	0.32	45.5	44.1	0.65
Retinopathy	12.6	13.3	0.73	13.3	8.9	<0.05
Other metabolic diseases	12.0	8.5	<0.05	10.6	11.1	0.81
Hypoglycaemic events	11.4	10.9	0.76	11.3	13.0	0.42
Skin problems	11.3	10.2	0.54	11.4	13.0	0.47
Nephropathy	6.7	7.6	0.55	8.8	6.5	0.20
Obesity*	5.4	4.8	0.61	1.3	1.1	0.74
Anxiety	4.2	3.3	0.38	2.8	3.2	0.70
Diabetes related infections	2.0	0.9	0.13	2.3	2.4	0.91
Pain-related conditions (%)
Low back pain*	32.4	30.4	0.46	39.2	34.3	0.12
Osteoarthritis*	13.5	13.5	1.00	24.0	22.7	0.64
Fibromyalgia syndrome*	8.7	6.3	0.12	3.8	4.3	0.69
Migraine*	4.4	1.7	<0.05	1.7	0.5	0.13
Rheumatoid arthritis	2.9	2.8	0.91	3.0	3.2	0.82
Psoriatic arthropathy	1.6	1.3	0.64	1.8	3.0	0.24
Charlson comorbidity index: mean (SD)*	0.77 (1.25)	0.76 (1.39)	0.92	1.40 (1.56)	1.31 (1.47)	0.37

*Difference between commercially insured and Medicare-insured population reached statistical significance at p<0.05 via chi-square test for each of the prevalence comparison and t-test for Charlson comorbidity index.

^†Chi-square test was used for categorical variable (prevalence of each medication condition) and Student t-test was used for continuous variables (Charlson comorbidity index).

SD, standard deviation.

Table 3. Medication use in 12-month post-index period.

	Commercial			Medicare supplemental insured
	Low compliance	High compliance	p-value	Low compliance	High compliance	p-value
Average daily dose* (%)
≤30 mg	16.4	7.4	<0.05	26.2	13.5	<0.05
31–59 mg	14.7	21.7	<0.05	20.2	30.0	<0.05
60 mg	55.7	52.6	0.28	45.7	44.1	0.62
>60 mg	13.3	18.3	<0.05	8.0	12.4	<0.05
Average daily dose: mean (SD)	57.0 (19.5)	61.4 (20.3)	<0.05	51.2 (18.1)	55.7 (17.7)	<0.05
Duloxetine index claim copayment^†, mean (SD)	28.8 (20.8)	23.8 (20.5)	<0.05	18.4 (23.5)	16.8 (22.9)	0.28
Any antidepressant agents (%)	46.4	35.2	<0.05	38.4	29.7	<0.05
SSRI	24.2	12.2	<0.05	18.7	11.9	<0.05
Tricyclic	18.0	14.8	0.13	13.9	12.2	0.44
Venlafaxine	5.1	2.0	<0.05	4.0	3.0	0.42
Bupropion	7.2	6.3	0.55	2.5	1.4	0.23
Other antidepressants	7.2	6.7	0.77	6.0	6.5	0.74
Any anticonvulsant agents (%)	59.7	57.6	0.46	61.8	55.7	0.06
Gabapentin	36.6	37.2	0.84	40.7	35.1	0.08
Pregabalin	23.9	20.9	0.21	24.8	15.1	<0.05
Other anticonvulsants	19.2	16.5	0.22	13.9	17.3	0.15
Any non-narcotic agents (%)	2.2	2.6	0.61	1.0	0.8	0.77
Any narcotic agents (%)	77.6	74.6	0.21	73.2	67.3	0.05
Hydrocodone	52.8	52.0	0.76	42.9	44.1	0.72
Oxycodone	24.0	23.7	0.89	15.4	18.4	0.22
Proproxyphene	19.1	17.0	0.33	25.3	16.0	<0.05
Tramadol	18.2	15.0	0.14	18.1	14.9	0.20
Codeine	8.6	8.7	0.95	9.8	7.6	0.24
Other narcotics	19.4	17.2	0.33	12.1	12.7	0.78
Any NSAID agents (%)	34.6	37.8	0.23	32.3	33.0	0.82
COX-2 inhibitors	7.2	11.7	<0.05	11.4	10.3	0.58
Other NSAIDs	29.5	30.2	0.77	23.7	24.9	0.67
Number of medications, mean (SD)
Diabetes-related^‡	2.0 (1.3)	2.0 (1.2)	0.08	1.6 (1.1)	1.8 (1.1)	<0.05
DPNP-related^§	2.3 (1.2)	2.2 (1.1)	<0.05	2.2 (1.0)	2.0 (1.0)	<0.05
Total medications	18.9 (8.5)	19.3 (7.6)	0.45	19.5 (7.4)	19.1 (7.1)	0.45

*Average daily dose based on days of duloxetine treatment.

^‡Copayment of duloxetine adjusted to a standard 30-day supply.

^‡Diabetes-related medications include: insulin and oral anti-diabetic drugs (OADs).

^§DPNP-related medications include: tricyclic antidepressants, venlafaxine, duloxetine, pregabalin, gabapentin, and opioids (i.e., tramadol, oxycodone, morphine, hydrocodone, methadone, levorphanol).

SD, standard deviation; DPNP, diabetic peripheral neuropathic pain-related.

Table 4. Factors associated with compliance to duloxetine therapy.

Variable*	Commercial		Medicare supplemental insured
	Odds ratio	95% confidence interval	Odds ratio	95% confidence interval
Age (years)
18–44	reference
45–54	1.63	(0.99–2.71)
55–64	2.12	(1.30–3.46)
Age (years)
65–74			reference
75–84			0.75	(0.56–0.99)
85+			0.84	(0.48–1.47)
Female	1.22	(0.96–1.55)	0.94	(0.72–1.24)
Prescribed by specialist	0.97	(0.75–1.25)	1.10	(0.82–1.48)
Average daily dose
≤30 mg	reference		reference
31–59 mg	3.37	(2.12–5.36)	2.90	(1.91–4.40)
60 mg	2.23	(1.48–3.37)	1.79	(1.22–2.62)
>60 mg	3.38	(2.10–5.44)	3.13	(1.84–5.31)
Duloxetine index claim copayment^†
$0–10	reference		reference
$11–20	1.39	(0.95–2.05)	0.93	(0.59–1.46)
$21–30	1.10	(0.78–1.56)	0.91	(0.60–1.38)
>$30	0.63	(0.47–0.86)	0.80	(0.57–1.12)
Pre–index use of insulin	1.11	(0.86–1.42)	0.98	(0.72–1.32)
Pre-index use of OADs	1.14	(0.87–1.49)	1.38	(1.01–1.89)

*Additional independent variables included Charlson comorbidity index and select diabetes- and pain-related conditions over the 12-month pre-index period.

^†Copayment of duloxetine adjusted to a standard 30-day supply.

Table 5. Healthcare costs in post-index period.

Healthcare costs ($)	Commercial			Medicare supplemental insured
	Low compliance Mean (SD)	High compliance Mean (SD)	Adjusted difference^†	Low compliance Mean (SD)	High compliance Mean (SD)	Adjusted difference^†
All-cause	32,407 (55,910)	25,747 (24,944)*	5,334*	24,622 (29,769)	23,281 (23,856)	63
Inpatient	12,851 (43,998)	4,833 (13,716)*	7,508*	6,754 (24,217)	4,595 (11,736)	1,055
Outpatient	11,888 (18,996)	10,718 (14,408)	1,105	10,598 (12,829)	9,284 (16,699)	726
Pharmacy	7,667 (9,324)	10,195 (6,496)*	–2,376*	7,270 (5,488)	9,402 (5,222)	–2,221*
Diabetes-related	10,024 (29,778)	6,093 (10,456)*	3,414*	5,015 (13,653)	3,739 (6,967)	1,080*
Inpatient	6,584 (28,891)	2,232 (8,662)*	3,785*	2,606 (13,309)	1,362 (6,387)	805
Outpatient	1,989 (3,974)	1,822 (3,905)	192	1,231 (2,447)	984 (1,631)	259*
Pharmacy	1,451 (1,430)	2,039 (1,737)*	–489*	1,179 (1,196)	1,393 (1,326)*	–131
DPNP-related	3,565 (10,847)	3,732 (6,792)	–802*	2,384 (12,138)	2,638 (3,716)	–805*
Inpatient	1,739 (10,362)	876 (6,094)	639	1,048 (12,102)	274 (3,341)	1,283
Outpatient	362 (1,416)	382 (1,660)	–26	181 (612)	241 (643)	–61
Pharmacy	1,464 (2,159)	2,474 (1,792)*	–1,037*	1,155 (1,254)	2,123 (1,226)*	–986*

*Significant at p<0.05.

^†Adjusted differences (low compliance–high compliance) estimated from generalised linear regression with log link and gamma distribution or two-part model (inpatient costs and DPNP-related outpatient costs) controlling for differences in demographics, prescribed by specialist, index prescription copayment, pre-index use of insulin and/or oral anti-diabetic agents, and Charlson comorbidity index.

SD, standard deviation.

Table 6. Diabetes and pain-related comorbidities.

Comorbidities	ICD-9-CM codes
Diabetes-related
Anxiety	300.0x, 300.2x, 300.3x, 309.81
Cardiovascular disease	390-398.xx, 401.x–403.xx, 404.1, 404.9, 405.xx, 410.xx–417.x, 420.xx–429.xx
Cerebrovascular and peripheral vascular disease	430.xx–437.xx, 440.xx–444.xx, 447.xx–454.xx, 457.xx–459.xx, 456.3x, 785.4x
Diabetes related infections	038.xx, 790.7x
Hypoglycaemic events	250.8x, 251.0x–251.2x
Nephropathy	580.9x, 581.81, 581.9x, 582.9x, 583.xx, 588.8x, 593.9x
Obesity	278.xx
Other metabolic diseases	251.3x, 270.3x, 276.xx
Retinopathy	362.0x–362.2x, 362.41, 363.31, 365.44, 366.41
Skin problems	680.xx–686.xx, 707.xx
Pain-related
Fibromyalgia syndrome	729.1x
Low back pain	720.1x, 721.3x, 721.9x, 722.1x, 724.02, 724.2x, 724.5x, 724.8x, 733.82
Migraine	346.0x–346.2x, 346.8x, 346.9x
Osteoarthritis	715.xx
Psoriatic arthropathy	696.xx
Rheumatoid arthritis	714.xx

Acknowledgment

Declaration of interest: The funding of this study was provided by Eli Lilly and Company. The statements contained in this paper are solely those of the authors and no endorsement by Eli Lilly and Company should be inferred or implied.

N.W., S.C., L.B. and K.F. have disclosed they are employees of Abt Bio-Pharma Solutions, Inc., which received funding from Eli Lilly. S.L.B. and Y.Z. Have disclosed they are employees of Eli Lilly.