Direct and indirect costs of patients treated with extended-release oxymorphone HCl or controlled-release oxycodone HCl

Abstract

Objective:

Compare direct and indirect costs of oxymorphone extended-release (‘oxymorphone’) and oxycodone controlled-release (‘oxycodone’) users.

Methods:

Patients, aged 18+, with ≥1 claim for oxymorphone/oxycodone, Q2:2006–Q4:2009, were selected from a de-identified private payer claims database and observed from the first such claim (‘index date’) until the earliest of: use of comparator drug; end of continuous eligibility; 12 months (‘study period’). Patients with claims for any formulation of the comparator drug during the first 30 days of the study period were excluded. Direct (medical and drug) costs paid by private insurers were reported for patients aged 18–64 (n = 8354) and 65+(n = 3515), as well as sub-sets without cancer (n = 7090 and n = 2444, respectively). Indirect costs (medically-related absenteeism and disability) were reported for all employees, aged 18–64 (n = 1313), and employees without cancer (n = 1146). Multivariate models were used to estimate risk-adjusted costs controlling for patient characteristics.

Results:

Oxymorphone users, aged 18–64, had lower drug costs ($693 vs $763, p = 0.0035) and similar medical costs ($1875 vs $1976, p = 0.3570) per patient-month compared with oxycodone users (mean follow-up 236 and 280 days, respectively). Indirect costs were not different ($662 vs $670, p = 0.9370). Oxymorphone users, aged 65+, had similar Medicare supplemental drug costs ($533 vs $588, p = 0.0840) and lower medical costs ($459 vs $747, p < 0.0001). Results were comparable for subsets without cancer.

Limitations:

Patients with concomitant use of oxymorphone and oxycodone were excluded.

Conclusions:

Oxymorphone users incur lower risk-adjusted costs in several cost categories, compared with oxycodone users, and no higher costs in any of the examined categories.

Keywords::

• Oxymorphone
• Oxycodone
• Costs
• Work loss

Introduction

Opioids are commonly prescribed for the treatment of acute and chronic pain, ranging from short-term post-surgical care to long-term palliative cancer care. Over the last decade, opioid use has increased1, particularly among patients with moderate-to-severe chronic pain, such as osteoarthritis and low back pain. Many clinical and economic aspects of prescription opioid utilization have been studied in recent years, including, to name a few: costs associated with abuse and misuse of prescription opioids2–6, the burden of opioid-related gastrointestinal events7,8, and opioid-related fractures9.

A number of long acting opioids have been found to be safe and effective in treating moderate-to-severe chronic pain10, including controlled-release oxycodone HCl (‘oxycodone’) and extended-release oxymorphone HCl (‘oxymorphone’)11–13.

While many prescription opioids share common therapeutic benefits and potential adverse effects, there remain important pharmacological differences within this therapeutic class. For instance, oxycodone, like many other opioids, is metabolized through the cytochrome P450 (CYP450) enzyme system, which may place patients taking other drugs metabolized in the same fashion at increased risk of drug–drug interactions arising from drug–drug exposures14. Oxymorphone, on the other hand, is metabolized via uridine-5’-diphospho-glucuronosyltransferases (UGTs), rather than through CYP-450 mechanisms, and is not subject to drug–drug exposure and potential drug–drug interactions with CYP-450 substrate, inducer, or inhibitor agents15. Research has also suggested that patients using oxymorphone have a lower daily average consumption (DACON) than patients using oxycodone, utilizing fewer equi-analagesic daily doses while treated for chronic pain conditions16,17.

Such pharmacological differences could give rise to differences in various direct (i.e., medical and drug) and indirect (i.e., work-loss and disability) cost drivers among patients treated with opioids for moderate-to-severe pain, possibly presenting payers with cost-saving strategies. However, to the authors’ knowledge, no prior research has explicitly compared the cost profiles, both direct and indirect, of patients treated with oxymorphone and those treated with oxycodone. Such information could be of use to both practitioners and payers, as these leading long-acting opioids are widely used to treat a variety of painful conditions. With ∼$3 billion in sales in 2009, oxycodone was the most commonly prescribed treatment among on-patent (i.e., branded) long-acting opioids, accounting for ∼55% of all long-acting opioid sales18. Oxymorphone, one of the leading branded alternatives, had sales of $223 million18.

The aim of this study is evaluate and compare direct and indirect costs among patients treated with oxycodone and oxymorphone in practice (as opposed to a clinical trial setting), while accounting for observable underlying differences (e.g., demographics and comorbidity profile) between the two patient populations.

Methods

Data

The study was based on retrospective analysis of a de-identified administrative claims database for a privately-insured population. The database includes over 12 million covered lives observed over the period 1999–2009, and contains information from 55 self-insured US companies, which operate nationwide in a broad array of industries. The data include information on patient demographics, enrollment history, and medical and prescription drug claims for all beneficiaries (i.e., employees, spouses, and dependents). Medical claims contain dates of service, up to two diagnoses based on codes from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), place and type of service, procedures based on Current Procedural Terminology, Fourth Edition (CPT) codes, and amounts paid to providers. The prescription drug claims include fill date, National Drug Code (NDC), days of supply, quantity, and amounts paid.

Sample selection

To assess the costs associated with patients treated with oxymorphone and oxycodone, two mutually exclusive cohorts of patients were selected from the claims database. Patients were required to have at least one pharmacy claim for either oxymorphone or oxycodone during the period Q2:2006–Q4:2009. The beginning of the selection time period corresponds to the US Food and Drug Administration (FDA) approval date for oxymorphone, to ensure that selected patients faced similar treatment options. The date of the first observed prescription was defined as the study index date.

To facilitate comparison between cohorts, patients were required to have no claims for any formulation of the comparator drug (i.e., oxycodone or oxymorphone) during a period of at least 30 days following the index date. In addition, all patients were required to have continuous enrollment in a health plan (non-HMO) during the 6 months prior to the index date (baseline period) and at least 1 month following the index date. A variable study period was then defined, with each patient being followed until the earliest of: (i) the first use of the comparator drug (after >30 days); (ii) end of continuous eligibility (after >30 days); or (iii) 12 months from index date. Costs were then analyzed at the patient-month level, to account for differing follow-up periods across patients (see analysis below).

In addition, patients were required to have at least 30 days of supply of their index medication during the study period; have no claims associated with pregnancy (ICD-9-CM: 633, 640–646, 761, V23.2, V22, V61.6–61.7, V72.40) during the baseline or study period; and be over age 18.

As mean cost calculations can be disproportionately affected by outliers, patients whose study period direct (i.e., medical and drug) costs per patient-month fell in the top 1% of costs of the oxymorphone and oxycodone groups combined were removed from the final cohort. Note that oxycodone patients were more likely to be removed as high cost outliers.

The sample was further stratified into three dimensions: age, disease, and employment status. Patients under age 65 throughout the study period were analyzed separately from patients age 65 and older, in order to address the fact that older patients were Medicare eligible, and therefore the observed costs in the private insurer database were likely understated for this population. Costs observed for this population likely represent only supplemental coverage provided by a private insurer.

With respect to disease, patients treated for pain symptoms associated with cancer may not be representative of the overall chronic pain population. Therefore, to address any bias in estimated costs which may arise from their inclusion, sensitivity analysis was conducted excluding patients with any claims for cancer (ICD-9-CM: 140–208 excluding 173) during the baseline or study period.

Finally, to estimate indirect costs due to medically-related absenteeism or disability, a sub-sample of employed oxymorphone and oxycodone patients (aged 18–64) were examined separately.

Analysis and statistical methods

Demographic characteristics, the Charlson Comorbidity Index (CCI)19,20 and other select comorbidities, healthcare resource utilization, and monthly healthcare costs were examined over the baseline period for oxymorphone and oxycodone patients. Study drug use (e.g., morphine equivalency per day) and direct costs per patient-month were assessed over the study period for oxymorphone and oxycodone patients. Indirect costs per patient-month were estimated for the sub-sample of employees. Categorical variables were compared using chi-squared tests, and continuous variables using bias-corrected bootstrapping.

Costs were based on actual payer reimbursements to providers, and were inflated to 2009 US dollars using the Consumer Price Index for medical care services. Costs per patient-month were analyzed using patient-level weights based on length of patient follow-up. Direct costs were compared overall, and for the following sub-categories: medical services (i.e., inpatient, outpatient, emergency department), oxycodone/oxymorphone drug costs, and other prescription drugs.

Indirect costs consist of medically-related absenteeism and disability costs. Disability costs were calculated based on actual short- and long-term disability payments from employers. Medically-related absenteeism was imputed based on healthcare resource use and observed employee wage data: an inpatient or emergency department visit were taken to represent 1 day of work-loss, and an outpatient visit half a day of work-loss.

Risk-adjusted direct costs for oxymorphone and oxycodone patients and indirect costs for the sub-sample of employees were estimated using generalized estimating equations (GEE) with a log-link function and gamma distribution, and adjusted for patient demographics (age, age squared, gender, and geographic region), baseline comorbidities, baseline drug use, baseline medical resource use, and baseline costs. This regression specification addresses the skewed nature of cost data.

All analyses were conducted using SAS 9.2 statistical software package (SAS institute, Inc., Cary, NC).

Results

Selection of the study sample is presented in Table 1. After imposing all inclusion and exclusion criteria, the study sample consisted of 767 oxymorphone patients aged 18–64, and 189 patients aged 65 and up. Oxycodone was much more commonly prescribed, resulting in 7587 patients aged 18–64 and 3326 aged 65 and up. Excluding cancer patients reduced the oxymorphone sample by ∼6% under age 65, and 19% for patients aged 65+. The effect on oxycodone was more pronounced, reducing sample sizes by 16% and 31%, respectively. Disability data for active employees was available for 142 oxymorphone patients and 1171 oxycodone patients.

Table 1.  Sample selection.

Criteria	Oyxmorphone		Oxycodone
	Age 18–64	Age 65+	Age 18–64	Age 65+
1. All eligible (non-HMO) beneficiaries from 55 self-insured companies, 7/1/2006–12/31/2009	4,874,740	983,057	4,874,740	983,057
2. ≥1 pharmacy claim for oxymorphone or oxycodone^a	1777	394	13,416	5336
3. No pharmacy claims for any formulation of the comparator drug during the first 30 days of the study period^b	1123	271	13,308	5318
4. Continuous eligibility for insurance benefits during the baseline period and at least 30 days after the index date^c	876	226	10,440	4241
5. No diagnoses associated with pregnancy during the baseline or study period^d	869	226	10,354	4239
6. ≥30 days of the index medication during the study period	769	189	7702	3329
7. Exclude patients in the top 1% of study period costs	767	189	7587	3326
8. No diagnoses associated with cancer during the baseline or study period^e	719	154	6371	2290
9. Active employees with disability data	142	–	1171	–
10. Active employees with disability data and no cancer^e	137	–	1009	–

^aThe index date is the first claim for the study drug (i.e., oxymorphone or oxycodone) during 7/1/2006–12/31/2009.

^bThe study period begins on the index date and continues until the first use of the comparator drug, the end of continuous eligibility, or 12 months, whichever comes first.

^cThe baseline period is the 6-month period prior to the index date.

^dPregnancy was identified using ICD-9-CM: 633, 640–646, 761, V2.32, V22, V61.6–61.7, V72.40.

^eCancer was identified using ICD-9-CM: 140–208, excluding 173.

Table 2 reports baseline characteristics for both treatment groups, stratified by age. Oxymorphone patients were more likely to be female, and had significantly lower comorbidity burden as captured by the CCI. However, it appears the majority of that difference was driven by higher prevalence of cancer among oxycodone patients. Oxymorphone patients were more likely to have low back pain, mental disorders and osteoarthritis (aged 18–64), but appeared less likely to have suffered fractures at baseline.

Table 2.  Baseline^a characteristics.

	Age 18–64			Age 65+
	Oxymorphone (n = 767)	Oxycodone (n = 7587)	p-value^b	Oxymorphone (n = 189)	Oxycodone (n = 3326)	p-value^b
Demographics
Age, Mean (SD)	48.1 (10.0)	49.6 (9.7)	<0.0001*	74.1 (7.3)	74.6 (7.7)	0.3323
Female, %	60.0%	50.9%	<0.0001*	66.7%	54.5%	0.0011*
Comorbidities
Charlson Comorbidity Index (CCI), Mean (SD)	0.70 (1.50)	1.33 (2.44)	<0.0001*	1.74 (2.05)	2.50 (3.04)	<0.0001*
Selected Comorbidities, %
Low Back Pain	61.1%	47.4%	<0.0001*	61.4%	43.2%	<0.0001*
Mental Disorder	35.1%	29.7%	0.0021*	16.4%	17.2%	0.7697
Osteoarthritis	19.4%	16.1%	0.0185*	31.2%	26.3%	0.1347
Headache/Migraine	15.0%	11.2%	0.0018*	6.3%	5.7%	0.6875
Cancer (Excl. Non-Melanoma Skin Cancer)	3.9%	14.0%	<0.0001*	13.8%	26.4%	0.0001*
Rheumatoid Arthritis	4.6%	3.7%	0.2424	6.3%	3.8%	0.0825
Non-Opioid Substance Abuse/Dependence	5.1%	5.4%	0.6865	1.6%	1.5%	0.7573
Non-Vertebral Fracture	3.1%	4.8%	0.0340*	4.8%	6.7%	0.2889
Opioid Abuse/Dependence	2.0%	1.8%	0.7260	0.0%	0.4%	1.0000
Resource use
Unique NDCs, Mean (SD)	15.9 (9.7)	13.8 (8.6)	<0.0001*	18.8 (8.6)	16.5 (8.4)	0.0005*
Selected Drug Classes, Mean (SD)^c	2.8 (1.3)	2.6 (1.3)	0.0084*	3.6 (1.5)	3.2 (1.4)	0.0026*
Prescription Drug Use, %
Analgesics	95.7%	93.4%	0.0122*	96.8%	94.7%	0.2066
Opioids	94.5%	92.2%	0.0198*	95.8%	93.5%	0.2106
Oxycodone (any formulation)	32.2%	79.1%	<0.0001*	28.6%	76.5%	<0.0001*
Oxymorphone (any formulation)	7.0%	1.1%	<0.0001*	6.9%	0.6%	<0.0001*
Other Opioids	91.0%	59.3%	<0.0001*	91.5%	56.1%	<0.0001*
NSAIDs (Incl. Cox-2 Inhibitors)	39.8%	29.9%	<0.0001*	34.9%	27.8%	0.0338*
Other Analgesics	12.6%	8.1%	<0.0001*	4.8%	3.0%	0.1851
Antidepressants	60.4%	50.3%	<0.0001*	54.0%	46.5%	0.0448*
SSRIs	29.3%	25.8%	0.0327*	32.3%	26.5%	0.0840
Other Antidepressants	45.5%	35.0%	<0.0001*	30.7%	28.6%	0.5417
Benzodiazepines	44.9%	43.8%	0.5808	44.4%	35.3%	0.0107*
Corticosteroids	25.0%	21.3%	0.0163*	28.6%	26.2%	0.4691
Statins	19.4%	21.2%	0.2421	45.0%	40.2%	0.1961
Beta Blockers	19.2%	18.6%	0.7138	42.9%	41.0%	0.6157
Calcium Channel Blockers	9.9%	9.7%	0.8623	29.1%	25.2%	0.2303
Warfarin	2.3%	5.4%	0.0003*	16.9%	14.2%	0.3015
Medical Service Use, %
Outpatient	97.4%	95.5%	0.0133	95.8%	95.2%	0.7306
Emergency Department	34.2%	32.0%	0.2174	30.2%	37.3%	0.0493*
Inpatient	21.8%	29.6%	<0.0001*	38.1%	45.9%	0.0351*
Other	53.6%	50.2%	0.0743	56.6%	54.3%	0.5290
Monthly Costs, Mean (SD)
Total Direct Costs	$2171 ($3534)	$3469 ($6896)	<0.0001*	$1197 ($2002)	$1409 ($2766)	0.1661
Prescription Drug	$595 ($1188)	$613 ($967)	0.6724	$516 ($497)	$538 ($742)	0.5598
Opioids	$248 ($1044)	$253 ($642)	0.8745	$67 ($119)	$116 ($255)	<0.0001*
Other	$347 ($498)	$360 ($686)	0.5050	$449 ($473)	$422 ($681)	0.4641
Medical	$1576 ($3353)	$2856 ($6805)	<0.0001*	$680 ($1882)	$871 ($2606)	0.1858
Outpatient	$809 ($1531)	$1186 ($2988)	<0.0001*	$297 ($1183)	$342 ($1273)	0.6005
Inpatient	$591 ($2600)	$1442 ($5459)	<0.0001*	$242 ($977)	$367 ($1,647)	0.1109
Emergency Department	$78 ($245)	$93 ($425)	0.1289	$11 ($48)	$18 ($72)	0.0650
Other	$98 ($452)	$134 ($879)	0.0529	$130 ($700)	$144 ($753)	0.7702

SD, standard deviation; NDC, national drug code; NSAID, non-steroidal anti-inflammatory drug; SSRI, selective serotonin reuptake inhibitor.

^aThe baseline period is the 6-month period prior to the index date. The index date is the first claim for the study drug (i.e., oxymorphone or oxycodone) during 7/1/2006–12/31/2009.

^bp-values were calculated using chi-squared tests for categorical variables and t-tests with bias-corrected bootstrapping for continuous variables. *denotes p-values < 0.05.

^cSelected drug classes refer to the number of categories listed under the ‘Prescription Drug Use’ heading (e.g., analgesics, antidepressants, statins) for which a patient filled one or more prescription.

Baseline prescription drug utilization also differed across cohorts. Oxymorphone patients were considerably less likely to use other formulations of oxycodone at baseline (32.2% vs 79.1% among 18–64, p < 0.001), but much more likely to have used other opioids (91.0% vs 59.3%, p < 0.001). Baseline use of antidepressants was also more prevalent among oxymorphone patients.

Baseline prescription drug costs appeared comparable across treatment groups, though medical costs appeared higher among oxycodone patients, driven primarily by higher inpatient costs, leading to higher overall direct costs among oxycodone patients at baseline.

Descriptive (i.e., unadjusted) measures of index medication use during the study period, as well as length of follow-up, are described in Table 3. Oxymorphone patients utilized, on average, lower morphine-equivalent doses per day (120 mg vs 184 mg, p = 0.037). Length of follow-up was generally lower in the oxymorphone group (236 days vs 280 days, p < 0.001).

Table 3.  Study period drug use and length of patient follow-up^a.

	Age 18–64			Age 65+
	Oxymorphone (n = 767)	Oxycodone (n = 7587)	p-value^b	Oxymorphone (n = 189)	Oxycodone (n = 3326)	p-value^b
Oxycodone/oxymorphone use
MG of Morphine per day, Mean (SD)	120 (146)	184 (273)	0.0369*	82 (80)	86 (130)	0.5794
Median	78	84		53	51
Length of Study period follow-up
Days in Study period, Mean (SD)	236 (121)	280 (105)	<0.0001*	243 (124)	268 (109)	0.0072*
Median	244	352		274	312

SD, standard deviation.

^aThe study period begins on the index date and continues until the first use of the comparator drug, end of continuous eligibility, or 12 months, whichever comes first. The index date is the first claim for the study drug (i.e., oxymorphone or oxycodone) during 7/1/2006–12/31/2009.

^bp-values were calculated using t-tests with bias-corrected bootstrapping; *Denotes p-values < 0.05.

Table 4 reports (unadjusted) study period work-loss measures across treatment groups. While all oxymorphone patients incurred some form of work-loss (compared with 96.2% of oxycodone patients, p = 0.012), no significant differences were found in specific work-loss components in terms of number of work-loss days.

Table 4.  Study period^a work-loss measures.

	Employees (age 18–64)
	Oxymorphone (n = 142)	Oxycodone (n = 1171)	p-value^b
Any Work-loss, %	100.0%	96.2%	0.0116*
Medically-related absenteeism	86.6%	85.2%	0.6572
Disability	22.5%	24.4%	0.6199
Work-loss Days, Mean (SD)
Total Work-loss Days	40.0 (70.7)	39.9 (66.4)	0.9996
Medically-Related absenteeism	9.7 (11.9)	10.0 (12.3)	0.7604
Disability	30.4 (73.3)	29.9 (68.3)	0.9600

SD, standard deviation.

^aThe study period begins on the index date and continues until the first use of the comparator drug, end of continuous eligibility, or 12 months, whichever comes first. The index date is the first claim for the study drug (i.e., oxymorphone or oxycodone) during 7/1/2006–12/31/2009.

^bp-values were calculated using chi-squared tests for categorical variables, and t-tests with bias-corrected bootstrapping for continuous variables; *Denotes p-values < 0.05.

Analysis of risk-adjusted costs among patients aged 18–64 (Figure 1) found no significant differences in total direct or indirect costs across treatment groups. However, oxymorphone patients incurred slightly lower costs per patient-month for their index medication ($212 vs $249, p < 0.001, among patients with any diagnosis; $215 vs $257, p < 0.001, among the sub-sample with no cancer), leading to lower overall prescription drug costs per patient-month ($693 vs $763, p = 0.004 among patients with any diagnosis; $656 vs $717, p = 0.007 among the sub-sample with no cancer). Medical costs were not significantly different across treatment groups.

Figure 1.  Risk-adjusted direct and indirect costs for oxymorphone and oxycodone patients, age 18–64. *Denotes p-values < 0.05; †Indirect costs were estimated for employees only (n = 1313 for all diagnoses; n = 1146 for no cancer). Risk-adjusted direct costs for oxymorphone and oxycodone patients and indirect costs for the sub-sample of employees were estimated using GEE with a log-link function and gamma distribution, and adjusted for patient demographics, baseline comorbidities, baseline drug use, baseline medical resource use, and baseline costs. Cost components were estimated using separate equations and do not sum exactly to the risk-adjusted total direct costs.

Examination of risk-adjusted direct costs among patients aged 65 and up (Figure 2) reveals lower average costs per patient-month compared with the population aged 18–64, as would be expected given that a large share of costs is being covered by Medicare (and therefore unobservable in the database). Among the older population, no significant differences were found between oxymorphone and oxycodone patients in total prescription drug costs ($533 vs $588, p = 0.084, among patients with any diagnosis; $537 vs $538, p = 0.975, among the sub-sample with no cancer). However, total direct costs per patient-month were lower in the oxymorphone group ($966 vs $1296, p < 0.001, among patients with any diagnosis; $884 vs $1015, p = 0.037, among the sub-sample with no cancer). The difference appears to be driven primarily by lower medical costs in the oxymorphone group.

Figure 2.  Risk-adjusted direct costs for oxymorphone and oxycodone patients, age 65+. *Denotes p-values < 0.05. Risk-adjusted direct costs for oxymorphone and oxycodone patients were estimated using GEE with a log-link function and gamma distribution, and adjusted for patient demographics, baseline comorbidities, baseline drug use, baseline medical resource use, and baseline costs. Cost components were estimated using separate equations and do not sum exactly to the risk-adjusted total direct costs.

Discussion

The present study suggests that the overall healthcare costs incurred per patient-month for patients treated with oxycodone and oxymorphone are comparable for the population aged 18–64. However, oxymorphone patients incurred significantly lower costs for their index medication, as well as lower overall prescription drug costs. No differences were found in indirect costs. Among patients aged 65 and up, total healthcare costs incurred by private insurers were significantly lower among oxymorphone patients, driven by apparently lower costs associated with medical services. There were no observed cost categories across age groups or diagnosis stratification in which oxymorphone patients incurred significantly higher costs than oxycodone.

Cancer patients were more likely to be treated with oxycodone than oxymorphone. Since it is plausible that cancer patients incur higher costs, on average, than other patients treated for non-cancer pain, it was important to determine whether cancer patients played a role in driving the cost differences across cohorts. However, while costs per patient-month were reduced in both treatment groups, the comparative cost patterns were largely unchanged among the sub-set of patients without cancer. Moreover, the multivariate analysis of the broader sample (including cancer patients) explicitly controlled for baseline comorbidities, greatly reducing any such concern.

The fact that cancer patients had little effect on the comparative cost results suggests that, from a cost perspective, pharmacy policy-makers may not need to design parallel reimbursement policies for cancer and non-cancer pain patients within this therapeutic class. This issue is of increasing importance among managed care pharmacy policy-makers, as the prevalence of cancer patients and survivors with chronic pain among their long acting opioid users continues to grow21.

While both oxycodone and oxymorphone are indicated for the treatment of moderate-to-severe pain, oxycodone is more commonly prescribed by physicians. This is apparent in the sample selected for this study, in which oxycodone patients outnumber oxymorphone patients by a ratio of approximately ten-to-one, consistent with general market shares. To the extent that patients treated with these medications incur different direct or indirect costs in the course of their treatment, payers and self-insured employers could potentially develop cost-saving strategies through encouraging changes in the composition of prescribed long-acting opioids in patients with moderate-to-severe pain.

The cost findings for the over age 65 population must be interpreted with caution, as they capture only the share of costs incurred by these patients covered by their supplemental insurance (for costs not covered by Medicare). This is evident from the fact that the average costs per patient-month in this population are substantially lower than those incurred for the population aged 18–64. While private insurer payments were included in the analysis, amounts paid by Medicare were not observed in the database. However, since the estimated costs do capture the cost burden to the private payers, they represent real potential savings for third party payers. Moreover, as demographic differences across cohorts are explicitly controlled for in the multivariate modeling, Medicare eligibility does not bias the comparative nature of the results (i.e., across cohorts).

This study did not attempt to identify specific cost drivers which could explain the cost differences between treatment groups. However, prior research suggests possible explanatory channels. For instance, oxycodone is metabolized through the CYP450 enzyme system, through which many other commonly prescribed medications (e.g., statins, SSRIs) are metabolized. As such, it may put patients at greater risk of drug–drug interactions, potentially leading to increased medical resource utilization and costs22.

Oxymorphone-treated patients may face lower such risk, as oxymorphone is not similarly metabolized. The fact that medical cost differences across cohorts were more pronounced among the older (65+) patient population is consistent with the fact the older patients are more likely to be prescribed multiple additional medications metabolized through the CYP450 system. Another potential cost driver, which could contribute to the observed cost difference in index medication costs, is differences in daily average consumption (DACON) across treatments. Prior research has found that treatment with oxymorhpone is associated with lower DACON compared with oxycodone16,17. In other words, adjusting for baseline differences, patients treated with oxymorhpone required fewer morphine-equivalent tablets per day during the course of their treatment. Further research is necessary, however, to draw conclusions regarding the casual relationship between these channels and the observed cost differences in the present study.

The privately-insured claims database used in this study has a number of limitations. First, the data do not include clinical indicators or diagnostic results. Therefore, controlling for underlying differences in patient severity across treatment groups must rely on observable diagnostic and resource use measures. While the findings outlined above are consistent with previously reported findings regarding CYP450 and DACON among patients treated with opioids, the findings may be partially attributable to unobserved differences in patient severity which could not be accounted for in the statistical modeling. Second, the findings in this study represent cost profiles for a sample of privately-insured patients prescribed oxymorphone and oxycodone, and therefore may not be representative of other patient populations such as Medicaid, Medicare, or the uninsured.

The present study was designed to facilitate comparison between mutually exclusive patient cohorts treated with oxymorphone and oxycodone. This was necessary in order to independently evaluate direct and indirect costs associated with treating patients with each medication. In reality, this distinction may be artificial, as physicians may well prescribe multiple treatments to patients over time. This study did not attempt to evaluate costs associated with patients treated with both oxymorphone and oxycodone within the study period, and therefore caution is warranted in interpretation of the cost differences in a real-world setting.

In order to maintain mutually exclusive treatment groups in the real-world setting captured in the claims database it was necessary to maintain some flexibility in the length of follow-up. Differential follow-up was fully addressed in the analysis conducted at the patient-month level. However, inclusion and exclusion criteria (e.g., pregnancy, cancer) may have been applied over differing periods of time across patients. Nevertheless, given the nature and prevalence of these exclusion criteria in the study population, it appears highly unlikely that differential follow-up has introduced any bias in the comparative nature of the results.

Conclusion

Examination of risk-adjusted direct and indirect costs for several patient populations suggests that, compared with oxycodone users, oxymorphone users incur lower costs per patient-month in several cost categories (e.g., index medication for patients <65, medical and total direct costs for all patients ≥65). Oxymorphone users did not incur higher costs than oxycodone users in any of the examined costs categories.

Transparency

Declaration of funding

Research support was provided to Analysis Group, Inc. by Endo Pharmaceuticals.

Declaration of financial/other relationships

Dr Kirson, Dr White, Dr Birnbaum, Mr Schiller, Ms Waldman, and Ms Peterson are employees of Analysis Group, Inc. Dr Ben-Joseph, Dr Berner, and Dr Summers are employees of Endo Pharmaceuticals.

Acknowledgments

No assistance in the preparation of this article is to be declared.