Evaluation of healthcare resource utilization and costs in employees with pain associated with diabetic peripheral neuropathy treated with pregabalin or duloxetine

Abstract

Objective:

To evaluate changes in healthcare resource use and costs after initiating pregabalin or duloxetine in employees with pain associated with diabetic peripheral neuropathy (pDPN).

Methods:

Employees (18–64 years old) with a DPN diagnosis and at least one pDPN-related pain medication claim were identified using the MarketScan Commercial Database (2005–2008). Propensity scored matched pregabalin and duloxetine new starts were evaluated in the 6-month pre- and 6-month post-initiation periods. Study outcomes including imputed medically-related work loss, prescription and healthcare utilization, and associated expenditures were analyzed using univariate statistics and multivariate models in a difference-in-difference approach.

Results:

A total of 473 employees in each treatment group were identified. Mean age was 53.6 (SD 7.0) years for pregabalin and 53.5 (SD 7.4) years for duloxetine. There were no pre-index differences between groups. Adjusted marginal effects were not statistically significant for pre-to-post changes in opioid utilization (p = 0.328), number of pDPN-related analgesic medications (p = 0.506), all-cause healthcare costs (p = 0.895), indirect costs (p = 0.324), or pDPN-attributable expenditures (p = 0.359).

Limitations:

Claims analysis is limited in accounting for all patient and plan differences, and by the reliability of medical claims for diagnosis coding. The sample size of the matched cohorts may have limited the power of the analysis to detect differences.

Conclusions:

There were no significant pre-to-post differences between pregabalin and duloxetine treatment groups in pDPN-related analgesic medication use, or pDPN-attributable, all-cause, and indirect expenditures.

Key words::

• Diabetic peripheral neuropathy
• Duloxetine
• Neuropathic pain
• Pregabalin

Introduction

Diabetic peripheral neuropathy (DPN) is the most common type of neuropathy in patients with diabetes mellitus (DM), affecting approximately half of the over 23 million patients with DM in the United States (US)1,2. The prevalence of DPN increases with poor glycemic control, age, and duration of DM, with the highest rates among patients who have had DM for 25 years or longer3. An estimated 10–20% of DPN patients have chronic painful symptoms that are severe enough to interfere with activities of normal daily living and warrant treatment4. The symptoms of painful DPN, which include burning, aching pain, or shooting pain in the lower limbs, usually last in excess of several months. As no therapies are yet available that will restore or improve nerve function, the current treatment goals are to optimize glycemic control to minimize progression of DPN and provide pain management sufficient for improving the patient's function and quality of life5,6.

Several evidence-based guidelines and algorithms have been published over the last several years in the US and Europe for the diagnosis and management of neuropathic pain and specifically for pain associated with DPN (pDPN). These guidelines are generally consistent in recommending anticonvulsants, tricyclic antidepressants (TCAs), and opioids as first-line treatment for neuropathic pain, depending on the patient's clinical presentation, condition, and comorbidities1,3,7–10. Only pregabalin and duloxetine have US Food and Drug Administration (FDA) approval for treatment of pDPN11–13. A study by Quilici et al. indirectly compared duloxetine and pregabalin in the treatment of pDPN and found them of comparable effectiveness and tolerability. It was also reported that patient response and tolerability to the various types of pharmacological treatment may be highly individual14.

pDPN is commonly associated with comorbidities of chronic pain, such as sleep disruption, depression, and interference with activities of daily living, in addition to other comorbidities attributed to DM. DPN has been shown to be associated with increased healthcare costs, reduced productivity in the workplace, and reduced quality of life for affected individuals15–17. Gordois et al. employed a simulated cost-of-illness model estimating the 2001 US annual costs of managing symptomatic DPN and its complications to be between $4.6 and $13.7 billion18. Ritzwoller et al., using retrospective data from 2003, found that the likelihood of a hospital admission for pDPN patients was more than 2.5-fold higher relative to diabetic patients without pDPN, and estimated the marginal cost per patient associated with pDPN to be $5,907 (SD $278.58) (p < 0.001) for the calendar year19. However, a paucity of evidence on the impact of pDPN treatments on healthcare resource use and costs in clinical practice has been reported. This study was undertaken to evaluate healthcare resource utilization and costs incurred by patients treated with either of the two pharmacologic agents having specific indications from the US FDA for management of pain associated with DPN.

The objective of this study was to evaluate the impact of initiating pregabalin or duloxetine on healthcare resource utilization, including disease-related medication utilization and associated costs in employees treated for pDPN, comparing changes in these outcomes of interest between pregabalin and duloxetine patients from the self-insured employer perspective.

Patients and methods

Data source

Administrative medical and pharmacy claims from October 2005 through March 2009 from the Thomson Reuters MarketScan Commercial Database were used for this study. This database includes complete longitudinal records of inpatient services, outpatient services, long-term care, and prescription drug claims covered under a variety of fee-for-service and capitated health plans, including exclusive provider organizations, PPOs, POS plans, indemnity plans, and health maintenance organizations (HMOs). All database records are de-identified and fully compliant with US patient confidentiality requirements (HIPAA).

Subject selection

Employees were selected who had at least one non-diagnostic claim (i.e., no laboratory or diagnostic radiology claims) with an ICD-9-CM diagnosis code for DPN (250.6x, 357.2x) anytime during the intake period of April 1, 2006, through June 30, 2008. Evidence of treatment for symptomatic relief of DPN pain, consisting of at least one claim for an oral or transdermal medication used in treating painful DPN (Table 1) within 60 days after a DPN diagnosis, was required as a proxy for pDPN. To enable assessment of indirect costs, the study was limited to employees aged 18 years or older. All subjects having continuous eligibility in a health plan, including both medical and pharmacy benefits, for a minimum of 6 months pre-index and 6 months post-index were selected. Patients were excluded who had claims for either pregabalin or duloxetine during the pre-index period; any claims with an ICD-9-CM diagnosis code for epilepsy (345.xx, 780.39), postherpetic neuralgia (053.1x), fibromyalgia (729.1x), cancer (140.xx – 172.xx, 174.xx – 208.xx, 235.xx – 239.xx, except for basal cell and squamous cell skin cancers and benign neoplasms); or a diagnosis or procedure indicative of transplant surgery or pregnancy during the study period (codes available upon request). Patients in a long-term care facility residence for 90 days or longer during the study period were also excluded due to unavailability of prescription claims during such stays. The index event was defined as the first treatment with either pregabalin or duloxetine within 60 days of a DPN diagnosis. The pre-index period was the 6-month period prior to the index event, and the follow-up period was the 6-month period from the date of the index event. Table 2 depicts the incremental attrition associated with each of the above criteria and the resulting cohort sample sizes.

Table 1.  Medications used for managing painful DPN*.

Medication class	Generic name
Non-opioid analgesics (‘NSAIDS’ – Rx only)	Celecoxib, diclofenac, etodolac, fenoprofen, flurbiprofen, ibuprofen, indomethacin, ketoprofen, ketorolac tromethamine, mefenamic acid, meloxicam, nabumetone, naproxen, naproxen sodium, oxaprozin, piroxicam, sulindac, tolmetin sodium, salsalate
Opioid analgesics	Strong opioids: Alfentanil, buprenorphine, fentanyl, hydromorphone, levorphanol tartrate, meperidine, methadone, morphine sulphate, oxycodone, oxymorphone, remifentanil Weak opioids: Butorphanol, codeine, dihydrocodeine, hydrocodone, nalbuphine, pentazocine, propoxyphene, tramadol, and combinations (with acetaminophen, aspirin, etc.)
Anticonvulsants	1st generation: Carbamazepine, divalproex, ethosuximide, phenytoin, primidone, valproate sodium, valproic acid 2nd generation: Gabapentin, lamotrigine, levetiracetam, oxcarbazepine, pregabalin, tiagabine, topiramate, zonisamide
Antidepressants	Tricyclics: Amitriptyline and its combinations, amoxapine, clomipramine, desipramine, doxepin, imipramine, nortriptyline, protriptyline, trimipramine Selective serotonin reuptake inhibitors (SSRIs): Citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline Serotonin norepinephrine reuptake inhibitors (SNRIs): Duloxetine, desvenlafaxine, nefazodone, venlafaxine Other: Bupropion, isocarboxazid, maprotiline, mirtazapine, phenelzine, selegiline, tranylcypromine, trazodone
Anesthetics	Lidocaine patch (Lidoderm)

*Only oral and transdermal (patch) medications were included.

Table 2.  Sample size attrition table.

Attrition criterion	Sample
Diabetic peripheral neuropathy diagnosis 4/1/2006–6/  30/2008	160,679
Treated for pain within 60 days on or after DPN diagnosis	53,120
Six months of continuous eligibility both pre- and post-index	37,718
No epilepsy, postherpetic neuralgia, transplant surgery, cancer, or fibromyalgia	32,705
Less than 90 total days in a long-term care facility pre- or post-index	32,633
No pregnancy pre- or post-index	32,496
Age 18–64 years at index	32,468
No pregabalin or duloxetine pre-index	28,060
Employees only (no spouses or dependents)	17,727
Pregabalin index medication before matching*	1,576
Duloxetine index medication before matching*	479
Pregabalin patients after matching	473
Duloxetine patients after matching	473

*Patients in the matching pool met all inclusion and exclusion criteria (index dates were reset to the first pregabalin or duloxetine prescription claim in the index period).

Patient matching

Propensity score matching was used20, and included the following covariates: diabetes type; age; gender; geographic region; plan type; use of opioid analgesics (strong and/or weak) in 60 days pre-index; use of other pDPN-related analgesics in 60 days pre-index; pre-index use of other medications for pDPN comorbidities (anxiolytics, hypnotics, antihypertensives, statins, corticosteroids, and muscle relaxants); number of unique medications overall in 60 days pre-index; presence of specific pre-index comorbidities including diabetes complications (ischemic heart disease, atherosclerosis, peripheral vascular disease, aortic aneurysm, heart failure, myocardial infarction, coronary artery bypass grafts, angioplasty, stroke/TIA, diabetic nephropathy, diabetic retinopathy), hypertension, dyslipidemia, musculoskeletal pain, back pain, neck pain, arthritis (osteoarthritis and rheumatoid), causalgia/other painful neuropathies, anxiety, depression and sleep disorders; pre-index healthcare utilization and pre-index total healthcare expenditures. Variables that were not significant and those that did not result in a high standardized difference after matching were eliminated, resulting in final matching criteria that included: diabetes type; non-opioid and antidepressant use in the 60 days pre-index period; pre-index comorbidities of back pain, dyslipidemia, anxiety, depression, and foot conditions (foot ulcers, amputations, and gangrene); and pre-index healthcare utilization. Patients receiving pregabalin were matched from the pool of duloxetine patients who had a similar predicted probability using nearest-neighbor matching21 without replacement with a 25% caliper at a matching ratio of 1:1.

Key outcome measures

Diagnoses and pharmacologic therapy were derived from medical and pharmacy claims received according to service dates in the study period. Diagnoses and comorbidities were determined from non-diagnostic medical claims that had the associated ICD-9-CM diagnosis codes in any position on the claim. Analgesic medications commonly used for treating pDPN were analyzed by drug class as shown in Table 1. Anxiolytic, hypnotic, and other commonly used medications by pDPN patients, such as insulin, were also evaluated. In addition to the mean number of prescriptions, persistence was measured for pregabalin and duloxetine as the days from the first pharmacy claim post-index to the date of the last prescription claim post-index with no breaks in therapy exceeding 30 days.

All-cause healthcare utilization data included hospital admissions, length of stay, emergency department visits, physician office visits, other outpatient services, and outpatient prescriptions. All-cause and pDPN-attributable expenditures (i.e., costs associated with pDPN-related analgesic and DPN-attributable healthcare resource use) were evaluated. Expenditures, including copayments, were adjusted to 2008 constant dollars, measured as the gross covered payments for all healthcare services or products (i.e., the amount eligible for payment after applying pricing guidelines such as fee schedules and discounts, and before applying deductibles, copayments, and coordination of benefits). Utilization and expenditures were measured for all medical and pharmacy claims.

Indirect costs were imputed from the number of hours associated with work loss due to all-cause hospitalization and outpatient visits. Each hospitalization day counted for a full day of work loss (8 hours) and an outpatient visit accounted for a half day of work loss (4 hours)22. The cost of absenteeism was estimated based on proxies of a patient's income according to their zip code using the most recent (1999) income data by zip code obtained from the US Census Bureau and escalated to 2008 levels using the US Bureau of Labor Statistics Employment Cost Index (constant dollar). The 2008 median household income was estimated by calculating the annual salary increase from December 2001 (earliest data available) to December 2008 and applying it to the US Census Bureau Data. The estimated 2008 median household income was divided by the yearly hours (2,080 hours per year) to get the median household hourly income.

Analysis

Baseline patient demographics, clinical characteristics, and pre- and post-index pDPN medication utilization, including anxiolytics, hypnotics, and other commonly used medications in pDPN patients, were compared between the pregabalin and duloxetine cohorts using chi-square tests for categorical variables and t-tests for continuous variables. Difference-in-difference analyses (‘difference-in-differences’ = (duloxetine cohort post-index period – duloxetine cohort pre-index period) – (pregabalin cohort post-index period – pregabalin cohort pre-index period)) were conducted to examine the incremental pDPN-related analgesic medication use, healthcare resource utilization, imputed absences, and expenditures across comparison groups. Generalized estimating equation (GEE) models were implemented23. Logit-link and binomial variance function were used for analyzing dichotomous outcomes. Log-link and gamma distribution were used for expenditures. Log-link and negative binomial distribution were used for count outcomes. The analysis of indirect costs, which had a large proportion of zero values, used a two-part model in which the first part was a GEE regression of any all-cause absence, and the second part was a GEE regression of indirect cost using the subsample of patients with absence. Logit-link and binomial variance were used in the first-part regression, while log-link and gamma variance functions were used in the second. A model building approach with backward selection was used to identify the potential confounding factors such as demographics and clinical comorbidities. The threshold for statistical significance was set at 0.05. Analyses were conducted using Stata release 11 MP.

Results

A total of 2,055 pDPN patients met all subject selection criteria (Table 2) and comprised the matching pool from which 473 pregabalin patients and 473 duloxetine patients were subsequently matched. The baseline demographics are shown in Table 3. Mean age was 53.6 (SD 7.0) years for pregabalin and 53.5 (SD 7.4) years for duloxetine (p = 0.864). Members were located primarily in the southern region of the US (52.2% and 51.6% of pregabalin and duloxetine patients, respectively), with another two-fifths split between north central and western regions, and the remainder in northeastern US. The distribution of insurance coverage was similar between the two cohorts (p = 0.941), with most patients having PPO coverage (59.2% of pregabalin and 57.9% of duloxetine patients), followed by HMO and POS coverage. Gender distribution showed slightly more males, but was not significantly different between the two cohorts. No significant differences in the Deyo–Charlson Comorbidity index or prevalence of selected comorbidities were seen in pre-index or post-index periods, with the exception of a higher prevalence of depression in the post-index period (9.5 vs. 5.9%, p = 0.038) for the duloxetine cohort. There were no differences in pre-index and post-index use of anxiolytic, hypnotic, or other commonly used medications in pDPN patients with the exception of a greater percentage of insulin use in the duloxetine cohort post-index (43.3% of pregabalin patients vs. 51.2% of duloxetine patients, p = 0.016). Total pre-index all-cause costs were similar between treatment groups.

Table 3.  Baseline patient demographic and clinical characteristics.

Characteristic	Pregabalin		Duloxetine		p-value
	n/mean	%/SD	n/mean	%/SD
Number of patients	473	100.0%	473	100.0%
Age, in years	53.6	7.0	53.5	7.4	0.864
Gender					0.268
Male	242	51.2%	259	54.8%
Female	231	48.8%	214	45.2%
Geographic region					0.649
Northeast	42	8.9%	38	8.0%
North Central	103	21.8%	110	23.3%
South	247	52.2%	244	51.6%
West	81	17.1%	79	16.7%
Other	0	0.0%	2	0.4%
Insurance plan type					0.941
Comprehensive	32	6.8%	33	7.0%
Health maintenance org.	79	16.7%	86	18.2%
Point-of-service	50	10.6%	51	10.8%
Preferred provider org.	280	59.2%	274	57.9%
Other	32	6.8%	29	6.1%
Deyo–Charlson Comorbidity Index	2.8	1.5	2.8	1.5	0.601
Select pre-index comorbidities
Diabetes type 1	64	13.5%	71	15.0%	0.515
Diabetes type 2	373	78.9%	367	77.6%	0.636
Other diabetes complications	149	31.5%	141	29.8%	0.573
Hypertension	160	33.8%	155	32.8%	0.730
Dyslipidemia	87	18.4%	92	19.5%	0.678
Musculoskeletal pain	236	49.9%	224	47.4%	0.435
Back pain with neuropathic involvement	44	9.3%	39	8.2%	0.566
Other back pain	76	16.1%	79	16.7%	0.792
Neck pain with neuropathic involvement	11	2.3%	17	3.6%	0.250
Other neck pain (spinal)	41	8.7%	36	7.6%	0.552
Arthritis	62	13.1%	65	13.7%	0.775
Causalgia and other painful neuropathies	62	13.1%	69	14.6%	0.510
Anxiety	103	21.8%	104	22.0%	0.937
Depression	37	7.8%	41	8.7%	0.636
Sleep disorders	45	9.5%	51	10.8%	0.518
Pre-index medication use
Insulin only (without other antihyperglycemics)	85	18.0%	92	19.5%	0.560
Non-insulin antihyperglycemics only	200	42.3%	175	37.0%	0.097
Insulin and other antihyperglycemics	118	24.9%	133	28.1%	0.269
Anxiolytics	81	17.1%	78	16.5%	0.794
Hypnotics	62	13.1%	71	15.0%	0.400
Antihypertensives	347	73.4%	362	76.5%	0.260
Statins	254	53.7%	238	50.3%	0.298
Systemic corticosteroids	45	9.5%	33	7.0%	0.156
Muscle relaxants	55	11.6%	50	10.6%	0.605
Total all-cause costs pre-index, $	16,635	36,560	17,024	45,797	0.885

In the post-index period, the mean number of prescriptions for pregabalin and duloxetine were 2.9 (SD 1.9) and 3.1 (SD 2.1), respectively. Persistence on pregabalin was 87.1 (SD 62.7) days and for duloxetine was 102.7 (SD 66.0) days. There were 6.3% of patients initiating duloxetine who received pregabalin in the post-index period, while 3.6% of pregabalin new starts received duloxetine. As shown in Table 4, the percentage of patients with at least one pDPN-related analgesic prescription in the post-index period significantly increased relative to the pre-index period by 40.8 percentage points in patients initiating pregabalin and 36.2 percentage points in patients starting duloxetine. The mean number of unique pDPN-related analgesic medications significantly increased pre-post by 1.2 and 1.1 for pregabalin and duloxetine, respectively. The average number of pDPN-related analgesic medication prescriptions increased pre-post by 3.1 and 2.9 for pregabalin and duloxetine, respectively. There were no significant differences between treatment groups in these pre-post pDPN-related analgesic utilization measures. There was a significant pre-post increase in the percentage of patients with any opioid use driven by the use of ‘strong opioids only’ in the duloxetine group, but between-group differences were not significant. Tricyclic antidepressant use significantly reduced pre-post in the duloxetine cohort, but was not significantly different from changes in the pregabalin cohort. A significant within- and between-group pre-post reduction in use of SSRI antidepressants was observed in the duloxetine cohort. SNRI use other than duloxetine decreased pre-post in the duloxetine and pregabalin cohorts with a significantly greater decrease in the duloxetine cohort. A significant pre-post decrease in the use of second generation anticonvulsants other than pregabalin was observed in the pregabalin group that was significantly different from pre-post changes in the duloxetine cohort. There were no significant pre-post changes within or between groups in use of other antidepressants, first generation anticonvulsants, lidocaine patch, and non-opioid analgesics. Bonferroni correction was made to adjust the p-values given that multiple significance tests were done. The difference-in-difference in the use of pregabalin, second generation anticonvulsants, duloxetine and SSRIs remained statistically significant (p < 0.05) even after Bonferroni correction.

Table 4.  Pre-index to post-index pDPN medication utilization.

	Pregabalin		Duloxetine		DID†	DID p-value
	Pre-index	Differences pre- to post-index*	Pre-index	Differences pre- to post-index*
Percentage with any pDPN-related analgesic	59.2%	40.8%	63.8%	36.2%	−4.7%	0.142
No. of unique pDPN-related analgesics	1.4 ± 1.8	1.2‡	1.5 ± 1.7	1.1‡	−0.1	0.302
Mean no. of pDPN-related prescriptions	2.9 ± 4.4	3.1‡	3.0 ± 3.9	2.9‡	−0.1	0.621
Non-opioid analgesics	22.4%	0.0%	22.4%	−3.2%	−3.2%	0.285
Opioid analgesics	14.6%	2.1%	13.5%	3.8%‡	1.7%	0.417
Strong only	8.0%	0.8%	7.8%	2.3%‡	1.5%	0.326
Weak only	6.1%	0.8%	5.3%	1.5%	0.6%	0.696
Both	0.4%	0.4%	0.4%	0.0%	−0.4%	0.480
Anticonvulsants
Pregabalin	0.0%	100.0%	0.0%	6.3%	−93.7%	<0.001
1st generation	1.1%	0.0%	1.1%	0.4%	0.4%	0.414
2nd generation other than pregabalin	23.0%	−9.5%‡	23.3%	0.2%	9.3%	0.001
Antidepressants
Duloxetine	0.0%	3.6%	0.0%	100.0%	96.4%	<0.001
Tricyclic antidepressants	10.8%	−2.1%	9.9%	−3.4%‡	−1.3%	0.471
Selective serotonin reuptake inhibitors	18.2%	1.5%	21.4%	−11.4%‡	−12.9%	<0.001
Serotonin norepinephrine reuptake  inhibitors other than duloxetine	4.7%	−0.2%	5.9%	−2.7%‡	−2.5%	0.028
Other antidepressants	8.5%	0.2%	8.5%	0.8%	0.6%	0.680
Lidocaine patch	2.1%	0.2%	3.0%	1.1%	0.8%	0.493
Anxiolytics	17.1%	0.4%	16.5%	1.5%	1.1%	0.612
Hypnotics	13.1%	2.5%	15.0%	0.8%	−1.7%	0.371

*Differences pre- to post-index are post-index minus pre-index.

†Difference-in-differences is duloxetine pre- to post-index differences minus pregabalin pre- to post-index differences.

‡Statistically significant difference pre- to post-index, at p < 0.050 significance level.

There were significant pre-post reductions in both cohorts for the percentage of patients with a physician office visit or other outpatient service visit, although differences between treatment groups in these measures were not significant (Table 5). Duloxetine was associated with a significant reduction in the mean number of all-cause physician office visits and other outpatient services, although only the pre-post change in physician office visits was significantly different from the pregabalin cohort, and that difference became non-significant after Bonferroni correction. There were no significant within- or between-group pre-post differences in the percentage of patients using or the mean number of emergency department visits for any reason. The percentage of patients with and the mean number of all-cause inpatient visits was significantly reduced pre-post in the duloxetine group, although this was not significantly different from changes in the pregabalin cohort. No within- or between-group differences were observed for changes in length of stay. The mean number of unique medications and prescriptions for any medication significantly increased pre-post in both treatment cohorts with no significant differences in changes between pregabalin and duloxetine.

Table 5.  Unadjusted all-cause utilization differences pre-index to post-index and differences between cohorts.

All-cause utilization	Differences Pre- to post-index*		DID†	DID p-value
	Pregabalin	Duloxetine
Inpatient admissions
Patients with an admission	−3.6%	−7.0%‡	−3.4%	0.269
Admissions per patient (all patients)	−0.06	−0.11‡	−0.05	0.349
Length of stay	−2.4	−2.6	−0.3	0.891
Emergency department visits
Patients with a visit	−4.4%	0.8%	5.3%	0.117
Visits per patient (all patients)	−0.10	0.04	0.06	0.486
Physician office visits
Patients with a visit	−2.5%‡	−1.7%‡	0.8%	0.446
Visits per patient (all patients)	0.17	−0.52‡	−0.69	0.021
Other outpatient services
Patients with other services	−2.5%‡	−2.1%‡	0.4%	0.762
Other service visits per patient (all patients)	−0.46	−2.43‡	−1.98	0.256
Medication utilization
Prescriptions per patient	4.6‡	4.2‡	−0.4	0.626
Unique medications per patient	1.30‡	1.26‡	−0.04	0.916
Imputed absence hours	−5.0	−8.3‡	−3.3	0.545

*Differences pre- to post-index are post-index minus pre-index.

†Difference-in-differences (DID) is duloxetine pre- to post-index differences minus pregabalin pre- to post-index differences.

‡Statistically significant difference pre- to post-index, at p < 0.050 significance level.

Expenditures per employee with pDPN (Table 6) did not change significantly pre- to post-index in either cohort for all-cause inpatient admissions, emergency department visits, other outpatient services, or total all-cause and pDPN-attributable healthcare utilization. Further, there were no significant differences-in-differences between cohorts for these expenditure categories. All-cause physician office expenditures for duloxetine patients decreased pre- to post-index by $90 (p < 0.001) and were $88 lower pre- to post-index compared with pregabalin patients (p = 0.004). Expenditures for pDPN-related analgesic medications increased significantly pre- to post-index in both cohorts ($476 for pregabalin patients, p < 0.001; $530 for duloxetine patients, p < 0.001); however, the difference-in-differences between cohorts of $53 was not statistically significant (p = 0.221). Similarly, the total all-cause medication costs pre- to post-index also significantly increased ($645 for pregabalin patients, p < 0.001; $695 for duloxetine patients, p < 0.001), but again the difference-in-differences between cohorts of $51 was not statistically significant (p = 0.616). The significant expenditure differences mentioned above remained statistically significant with p < 0.05 after Bonferroni correction.

Table 6.  Unadjusted all-cause and pDPN-attributable expenditures post-index, differences pre-index to post-index and differences between cohorts.

	Pregabalin		Duloxetine		DID†	DID p-value
	Post-index	Difference pre- to post-index*	Post-index	Difference pre- to post-index*
All-cause expenditures
Inpatient admissions	$4,743 ± 23,315	−$2,239	$4,063 ± 20,057	−$3,337	−$1,098	0.642
Emergency dept. visits	$215 ± 720	−$25	$282 ± 933	−$41	−$15	0.856
Physician office visits	$610 ± 501	−$2	$591 ± 483	−$90‡	−$88	0.004
Other outpatient services	$5,743 ± 22,343	−$116	$5,054 ± 16,644	−$518	−$402	0.572
Total medication	$3,164 ± 4,532	$645‡	$3,268 ± 2,647	$695‡	$51	0.616
Total all‐cause healthcare	$14,885 ± 38,320	−$1,750	$13,680 ± 31,208	−$3,344	−$1,594	0.550
Imputed all-cause absences	$663 ± 2,224	−$113	$514 ± 1,939	−$235	−122	0.465
pDPN-attributable Expenditures
pDPN-related analgesics	$824 ± 1,669	$476‡	$845 ± 804	$530‡	$53	0.221
Total pDPN	$1,996 ± 9,128	$267	$1,591 ± 3,982	−$1,186	−$1,454	0.088

*Expenditure differences are measured as the difference in the mean per member during the 6-month pre-index to 6-month post-index periods (post-index minus pre-index).

†Difference-in-differences (DID) is duloxetine pre- to post-index expenditure differences minus pregabalin pre- to post-index expenditure differences.

‡Statistically significant difference pre- to post-index, at p < 0.050 significance level.

All-cause imputed absence hours (Table 5) decreased pre- to post-index for both cohorts, with a significant decrease of 8.3 hours (p = 0.040) for duloxetine patients and a non-significant decrease of 5.0 hours (p = 0.183) for pregabalin patients. The difference-in-differences between cohorts of 3.3 hours was not statistically significant (p = 0.545). The attributed cost of absence similarly decreased for both cohorts, but was not statistically significant pre- to post-index, as shown in Table 6 ($113 decrease for pregabalin patients, p = 0.295; $235 decrease for duloxetine patients, p = 0.064). The difference-in-differences of $122 was not statistically significant (p = 0.465).

The marginal effects of the difference-in-difference estimator (Table 7) show that, controlling for baseline patient characteristics, there were no statistically significant differences between cohorts for changes pre-index to post-index in the probability of opioid use, the number of pDPN-related analgesic medications used, pDPN-attributable expenditures, and total all-cause direct or indirect costs.

Table 7.  Marginal effects comparing pregabalin and duloxetine for pDPN-specific medications, expenditures, and indirect costs.

Parameter	Marginal effect*	Standard error†	p-value	95% Confidence limits
Reference: duloxetine				Lower	Upper
Probability of opioid use	−0.013	0.013	0.328	−0.040	0.013
Number of pDPN-related analgesics used	0.108	0.162	0.506	−0.210	0.426
Total pDPN-attributable expenditure (2008 USD)	$145	$158	0.359	−$165	$454
Total all-cause expenditure (2008 USD)	$154	$1,174	0.895	−$2,147	$2,456
Probability of all-cause absence	0.034	0.028	0.215	−0.020	0.088
Indirect costs among individuals with any absence	−$458	$464	0.324	−$1,367	$452

*Marginal effects are the adjusted effects for the 6-month follow-up period.

†Standard errors of the marginal effects were computed using the delta method.

Discussion

Several pharmacological options for symptomatic treatment of pDPN have been recognized on the basis of clinical evidence of efficacy. The 2006 EFNS guidelines9 for painful polyneuropathy found treatments with established efficacy on the basis of class I trials to include pregabalin, duloxetine, tricyclic antidepressants, venlafaxine, gabapentin, opioids and tramadol (level A). Consensus guidelines developed in the US by the Diabetic Peripheral Neuropathic Pain Consensus Treatment Guidelines Advisory Board12 (2006) for treatment of pDPN identified first-tier agents based on the level of evidence available from clinical trials and the committee's clinical experience to include duloxetine, pregabalin, oxycodone CR, and the TCAs. They further recognized that therapy selection must take into account patient factors such as comorbidities, concomitant medication, adverse events, treatment goals, and cost. This study compares the healthcare utilization and expenditures for pDPN patients treated with the two agents having specific indications from the US FDA for management of pain associated with DPN, providing pharmacoeconomic evidence that clinicians and payers may consider when making treatment decisions. Utilization and cost comparisons of these two indicated treatments with the other recommended treatments should be considered for additional future research.

Changes in pDPN-related analgesic medication usage were overall not significantly different between cohorts, with two notable exceptions. There was a decrease of 11.5 percentage points post-index in the percentage of duloxetine patients who were taking SSRI antidepressants. This decrease may reasonably be attributed to patients taking an SSRI pre-index for either depression or pDPN switching to duloxetine for both conditions. There were a significantly greater percentage of insulin users in the duloxetine cohort post-index, which increased by four percentage points from pre-index – an increase not seen in the pregabalin group. The prescribing information for duloxetine includes a statement in the Warnings and Precautions section regarding blood glucose control in diabetes, namely that small increases in fasting blood glucose and HbA_1c have been observed in pDPN patients11. The finding in this study of increased numbers of insulin-using diabetics post-index in the duloxetine cohort may warrant further research for any potential clinical significance.

The prevalence of comorbidities during the post-index period was only significantly different between cohorts in the percentage of patients with a depression diagnosis, which was lower for pregabalin. Although the reasons for this difference are not known, there is a possibility that providers may be specifically prescribing duloxetine for pDPN patients with comorbid depression, and there may also be more depression coding for duloxetine patients to facilitate insurance coverage for the drug. Otherwise, there were no further significant differences in observed comorbidities.

No significant differences in utilization were found, except there were 0.7 fewer office visits pre- to post-index per patient for duloxetine. This difference carried over into the office expenditures, which decreased significantly more ($88 less in office visit expenditures) for duloxetine patients; however, it did not significantly impact the overall pDPN-attributable or all-cause healthcare expenditure differences-in-differences, none of which were significant between cohorts. Medication utilization increased post-index for pDPN-related pain medications overall, opioids specifically, and in general for all medications, but this occurred commensurately with no significant differences between the cohorts. Medication expenditure likewise increased post-index but commensurately between cohorts for both pDPN-related medication expense as well as total medication expense with no significant differences between cohorts. Further, there were no significant differences pre- to post-index for indirect costs.

The relationship between treatment group membership and key outcomes was assessed for the key dependent variables of opioid use, number of pDPN-related analgesic medications, pDPN-attributable healthcare expenditures, total all-cause healthcare expenditures, and indirect costs using multivariable analytic techniques. The use of multivariate regression for results of propensity-score-matched cohort outcomes provides an additional level of robustness to the results. Marginal effects of the difference-in-difference estimator confirmed the unadjusted results that showed no statistically significant differences between the cohorts for changes pre-index to post-index in any of these key dependent variables.

Limitations

Claims analysis is limited in its ability to account for all possible differences in patients and providers residing in different plans. In the absence of patient charts or provider attestations, misclassification error is possible when relying on diagnosis coding from administrative claims data, where the extent of under- or over-coding for DPN is unknown. This analysis focused on pregabalin and duloxetine use for pDPN, and the results should not be interpreted outside of these labeled indications. With no specific diagnosis code for pDPN, this study relied on pharmaceutical claims for pain treatment subsequent to DPN diagnosis for identifying patients with painful manifestations of their condition. Further, it assumes the use of these medications was at least in part for their effect on the alleviation of DPN pain, and in the absence of diagnosis coding on pharmacy claims, that drugs shown to be used for treating pDPN (Table 1) were actually and properly used as such. The size of the matched cohorts (473 patients per cohort) limited the power of the analysis to detect differences between the cohorts. It is also acknowledged that the difference-in-difference approach can only account for observed changes in covariates that were measurable across treatment groups.

Conclusion

pDPN is a common and often debilitating complication of DM for which pharmacologic therapy can provide symptomatic relief. This study evaluated the effects on direct and indirect costs of initiating pregabalin or duloxetine in employed individuals diagnosed with pDPN. Medication use, healthcare resource use, and healthcare expenditures were compared pre-index to post-index within and between the two propensity-score-matched treatment groups, including the use of a difference-in-difference modeling approach with calculation of marginal effects. There were no significant pre-to-post differences between pregabalin and duloxetine treatment groups in opioid use, pDPN-related analgesic medication use, pDPN-attributable expenditures, all-cause direct, or indirect expenditures.

Transparency

Declaration of funding

This study was sponsored by Pfizer Inc, New York, NY, USA.

Declaration of financial/other relationships

J.M., Z.C., and R.F are employees of Thomson Reuters, who was paid by Pfizer Inc in connection with the development of this manuscript. S.L.S. was a paid consultant to Pfizer Inc in connection with the development of this manuscript. J.H., R.J.S. and J.M. are employees of Pfizer Inc, manufacturer of pregabalin.

Acknowledgments

The authors wish to acknowledge the key contributions of Bevan Kirley, MS, whose tireless work in defining, extracting, assembling, and analyzing the data helped make this research possible.

Selected results from this study were previously presented in a poster at the International Society for Pharmacoeconomics and Outcomes Research 15th Annual International Meeting, May 15–19, 2010, in Atlanta, GA, USA.