Medical cost savings associated with an extended-release opioid with abuse-deterrent technology in the US

Abstract

Objectives:

In the US, prescription opioids with technology designed to deter abuse have been introduced to deter drug abuse without hindering appropriate access for pain patients. The objective of this study was to estimate changes in medical costs following the introduction of a new formulation of extended-release (ER) oxycodone HCl (oxycodone) with abuse-deterrent technology in the US.

Methods:

Health insurance claims data were used to estimate changes in rates of diagnosed opioid abuse among continuous users of extended-release opioids (EROs) following the introduction of reformulated ER oxycodone in August 2010. This study also calculated the excess medical costs of diagnosed opioid abuse using a propensity score matching approach. These findings were integrated with published government reports and literature to extrapolate the findings to the national level. All costs were inflated to 2011 US dollars.

Results:

The introduction of reformulated ER oxycodone was associated with relative reductions in rates of diagnosed opioid abuse of 22.7% (p < 0.001) and 18.0% (p = 0.034) among commercially-insured and Medicaid patients, respectively. There was no significant change among Medicare-eligible patients. The excess annual per-patient medical costs associated with diagnosed opioid abuse were $9456 (p < 0.001), $10,046 (p < 0.001), and $11,501 (p < 0.001) for commercially-insured, Medicare-eligible, and Medicaid patients, respectively. Overall, reformulated ER oxycodone was associated with annual medical cost savings of ∼$430 million in the US.

Limitations:

Because of the observational research design of this study, caution is warranted in any causal interpretation of the findings. Portions of the study relied on prior literature, government reports, and assumptions to extrapolate the national medical cost savings.

Conclusions:

This study provides evidence that reformulated ER oxycodone has been associated with reductions in abuse rates and substantial medical cost savings. Payers and policy-makers should consider these benefits as they devise and implement new guidelines and policies in this therapeutic area.

Keywords::

• Prescription opioids
• Abuse
• Abuse-deterrent formulation
• Medical costs

Introduction

Pain is a common cause of disability in the US1. The Food and Drug Administration (FDA) has noted that prescription opioids (RxOs) are an important component of modern pain management2. At the same time, the agency has acknowledged that RxO abuse is a serious public health problem. In 2011, an estimated 1.8 million Americans aged 12 or older (or 0.7%) had pain reliever dependence or abuse3.

Opioids are prescribed for the treatment and management of pain. Extended-release opioids (EROs) provide a longer period of drug release and can, thus, be taken less frequently than immediate-release opioids (IROs)4. However, EROs can be manipulated and tampered with in an attempt to overcome their extended-release properties. Because of their higher drug dosages, EROs may be more attractive to some abusers than immediate-release formulations5.

Some EROs have been developed with properties designed to deter abuse in an attempt to make the manipulation of EROs more difficult and, thereby, reduce the abuse of EROs without hindering appropriate access to care for pain patients. These new products include physical or chemical properties to hinder extraction of the active ingredient or reduce the reward or likability of tampering with a drug6,7. Douglas Throckmorton, MD, deputy director for regulatory programs in the FDA’s Center for Drug Evaluation and Research, has said, ‘An important step towards the goal of creating safer opioids is the development of products that are specifically formulated to deter abuse’8.

RxO abuse has been estimated to account for ∼$56 billion in annual societal costs in the US9. The introduction of these new EROs using technology designed to deter abuse has the potential to reduce societal costs, in part by reducing RxO abuse in a sub-set of abusers (e.g., those who tamper with the drug delivery system) and reducing the economic costs RxO abuse imposes on society. To date, the economic impact of the introduction of such products has not yet been quantified. Two prior studies estimated the potential societal impact of abuse-deterrent formulations using assumed changes in rates of opioid abuse and market share rather than actual data following the introduction of abuse-deterrent formulations. One of the studies estimated potential annual healthcare savings from a hypothetical abuse-deterrent formulation of US$0.6–US$1.6 billion in the US in 200610. The other study estimated potential annual direct and indirect cost savings of C$0.5 billion in Canada in 201111.

This study is the first to report actual changes in rates of opioid abuse and medical cost savings associated with a new formulation of extended-release (ER) oxycodone HCl (oxycodone) with abuse-deterrent technology in the US by analyzing patient-level medical and pharmacy claims data before and after the introduction of reformulated ER oxycodone in August 2010. This study is particularly timely given the FDA’s consideration of the development of abuse-deterrent formulations as ‘a high public health priority’2 and the agency’s approval of abuse-deterrent labeling for the new formulation of ER oxycodone12.

Methods

Our analytic approach consisted of three components. First, we measured reductions in rates of diagnosed opioid abuse following the introduction of reformulated ER oxycodone. Second, we calculated the excess medical costs associated with diagnosed opioid abuse. The first two components of our study were conducted using actual medical and pharmacy claims data for continuous ERO users (defined below). Third, we integrated these findings with additional data published in government reports and prior literature in order to extrapolate our findings to estimate total medical cost savings associated with the launch of reformulated ER oxycodone at the US national level. Since a significant share of abuse is not formally diagnosed10, we extrapolated our findings to account for potential savings due to reductions in undiagnosed abuse. Our calculations accounted for a mix of insurance programs, including commercial plans, Medicaid, Medicare, and the uninsured. Medicaid is a federal and state government program that provides health insurance to low-income individuals and families or those with certain disabilities. Medicare is a federal government program that provides health insurance to individuals aged 65 and older, regardless of income or medical history.

This study used 2009–2011 Truven Health Analytics (Truven) de-identified medical and pharmacy claims data for commercially-insured, Medicaid, and Medicare-eligible patients who had at least one prescription drug claim for an RxO during this period. RxOs were identified using the Medi-Span Generic Product Identifier for drug group 65 for Analgesics–Narcotic, which included both extended-release and immediate-release opioids. Truven claims data include information on patient demographics, enrollment/eligibility, and medical and prescription drug claims. Truven’s Commercial MarketScan database reflects the healthcare utilization of individuals covered by a variety of commercial health plans, including those eligible for Medicare (aged 65 or above) but covered by supplemental insurance. Our sample of Medicare-eligible patients consisted of patients aged 65 or above who were enrolled in employer-sponsored Medicare Supplemental plans. Truven’s Medicaid database reflects the healthcare utilization of ∼7 million Medicaid enrollees from multiple states.

The claims data analyses focused on the time period before and after the introduction of reformulated ER oxycodone and on a population of continuous ERO users. While there are other EROs on the market that use technologies designed to deter abuse, our analyses focused on the launch of reformulated ER oxycodone as it is the only ERO that has obtained FDA approval for abuse-deterrent labeling13, and it has garnered a substantial market share. Continuous ERO users were defined as patients with at least 120 days supply of EROs over a 6-month period14. Non-continuous ERO users may have used EROs only briefly or may have switched between multiple EROs, making it difficult to attribute changes in abuse rates to the launch of reformulated ER oxycodone. We extended our findings to non-continuous ERO users assuming a similar proportional effect of reformulated ER oxycodone.

In addition, patients were only included in the claims data analyses if they were aged 18 or above. Patients were also required to be continuously enrolled in non-capitated/non-HMO healthcare plans during the study period, as the healthcare costs and utilization of patients in capitated or HMO healthcare plans may not be fully captured in the claims data. Additional inclusion criteria are described below for each component of the study.

All claims data analyses were stratified by payer type. Medical costs included the amount paid to providers by payers and by patients out of pocket for emergency department (ED), inpatient, and outpatient services and were inflated to 2011 dollars using the Consumer Price Index for Medical Care from the Bureau of Labor Statistics, US Department of Labor. For Medicare-eligible patients, we included payments from all payers, including commercial insurance plans and Medicare, in order to capture all of their medical costs. We assumed that the medical costs of Medicare-eligible patients reflected the medical costs of individuals whose primary insurance coverage was Medicare.

Changes in rates of diagnosed opioid abuse associated with the introduction of reformulated ER oxycodone

For this component of the study, we analyzed changes in rates of diagnosed opioid abuse after the introduction of reformulated ER oxycodone. The relevant study period was February 2010 to May 2011. The analysis was divided into a 6-month pre-reformulation period (February 2010 to August 2010) and a 6-month post-reformulation period (November 2010 to May 2011). Note that the pre-reformulation and post-reformulation periods were separated by a 3-month transition period in order to allow treatment regimens to stabilize after potential switching, the new formulation to reach pharmacies, and patients to use up remaining supplies from prescriptions obtained pre-reformulation. A study on changes in abuse post-reformulation also excluded a transition period to allow for pharmacies’ supplies of the original formulation to be depleted15.

Since patients may use several different EROs concomitantly or rotate between EROs, each patient was assigned a ‘primary’ ERO which was defined as the ERO that accounted for 70% or more of the patient’s days supply of EROs during the relevant 6-month pre-reformulation or post-reformulation period. The 70% threshold was based on clinical input, and alternative threshold values (50%, 60%, 80%) were considered in sensitivity analyses but did not meaningfully affect the findings.

Patients were classified as diagnosed abusers if they had any medical claims with International Classification of Diseases, Ninth Revision (ICD-9-CM) diagnosis codes for opioid abuse or dependence (304.0, 304.7, 305.5, and 965.0, excluding 965.01 for heroin), during the relevant 6-month pre-reformulation or post-reformulation period. This is consistent with the Substance Abuse and Mental Health Services Administration’s National Survey on Drug Use and Health (NSDUH) which combines prescription pain reliever abuse and dependence3. In addition, clinical input suggests that opioid abuse and dependence diagnosis codes are often used interchangeably. A prior study on the excess costs of opioid abuse found that the magnitude of the excess costs was similar, regardless of whether abuse-only or dependence-only codes were used16. Therefore, we use both abuse and dependence diagnosis codes and hereinafter use only the term abuse for simplicity.

The rate of diagnosed abuse was calculated as the percentage of all eligible continuous ERO users with an opioid abuse diagnosis during the relevant 6-month period. As the codes only capture users being treated for abuse, the diagnosed abuse rates are a conservative estimate of RxO abuse. We address this issue in our national-level estimates below.

We examined whether rates of diagnosed abuse differed between patients whose primary ERO was reformulated ER oxycodone (in the post-reformulation period) and patients with any primary ERO (in the pre-reformulation period). Chi-square tests were used to test differences in rates of abuse between the two groups.

Excess medical costs associated with diagnosed opioid abuse

We estimated the excess medical costs associated with diagnosed abuse among continuous ERO users. Diagnosed abusers were matched to non-abusers (i.e., controls with no diagnoses of opioid abuse). Each patient was assigned an index date corresponding to the date of the first abuse diagnosis for abusers and the date of a random medical claim for non-abusers. As the medical costs associated with diagnosed abuse may accrue before a formal diagnosis, medical costs were calculated over a 12-month period centered on the index date. We imposed a wash-out period of 12 months prior to the first abuse diagnosis to ensure that we were capturing the first abuse diagnosis and that there were no abuse diagnoses in the baseline period, so the index date had to occur after January 1, 2010. The 12-month period over which medical costs were calculated was preceded by a 6-month baseline period used to calculate baseline healthcare resource utilization for matching purposes. Pre-matching, baseline differences between abusers and non-abusers were assessed using chi-squared tests for categorical variables and Wilcoxon rank-sum tests for continuous variables.

Abusers were matched to non-abusers on their primary ERO (e.g., an ER oxycodone patient could only be matched to another ER oxycodone patient) and propensity score. Propensity scores were estimated for the entire sample of eligible patients based on demographics (e.g., age, gender, type of healthcare plan, and year of index date), comorbidity profile (Charlson Comorbidity Index, arthritis, fibromyalgia, hepatitis, lower back pain, other back or neck pain, neuropathic pain, psychiatric disorders, and non-opioid substance abuse diagnoses), drug use (total days supply of EROs during the baseline period, use of other abusable prescription drugs (i.e., IROs, stimulants, and depressants)), baseline healthcare resource utilization (any inpatient or ED visit), and baseline healthcare costs (medical costs and prescription drug costs). Within these primary ERO cohorts, abusers and non-abusers were then matched using an iterative greedy match with calipers for the propensity score. Post-matching, baseline differences between abusers and non-abusers were assessed using McNemar tests for categorical variables and Wilcoxon signed-rank tests for continuous variables.

Following matching, the excess medical costs associated with diagnosed abuse were calculated as the difference in medical costs in the 12-month period centered on the index date between abusers and non-abusers. Differences in prescription drug costs between abusers and non-abusers were also examined. Paired t-tests were used to test differences in costs between abusers and non-abusers.

Medical cost savings at the US level

The next component of the study involved calculating the total medical cost savings associated with reformulated ER oxycodone in the US. Cost savings were estimated for adults aged 18 and up. All of the following calculations were done separately for each payer type (i.e., commercially-insured, Medicaid, Medicare-eligible, and uninsured) and then combined based on the distribution of payer populations among US adults. Since we lacked medical and pharmacy claims data for uninsured patients, rates of abuse and medical costs of the uninsured were assumed to mirror those of Medicaid enrollees.

To estimate the total number of opioid abusers in the US population, we applied the rate of prescription painreliever abuse and dependence from the NSDUH3 to the total US population aged 18 and up from the US Centers for Disease Control and Prevention. To determine how many of these abusers were diagnosed abusers vs undiagnosed abusers, we used a 1:5 ratio of diagnosed to undiagnosed abusers that builds on NSDUH data and previous prevalence estimates of diagnosed abuse10, but with adjustments to account for the considerably lower rate of abuse among the over-65 population, which reduces the overall rate and number of diagnosed abusers estimated in the US population.

Next, we estimated the actual reductions in abuse associated with the introduction of reformulated ER oxycodone for both diagnosed and undiagnosed abusers. As mentioned above, for our core analysis we assumed that the impact of reformulated ER oxycodone on rates of abuse among continuous ERO users was equivalent to that among non-continuous ERO users. Observed reductions in rates of diagnosed abuse were estimated in the claims data analysis described above, but these observational estimates may overstate the effect of the abuse-deterrent formulation, as recent studies have documented that some abusers may switch away from reformulated ER oxycodone to other abusable substances such as other EROs without abuse-deterrent technology17 or heroin18. To account for this. we assumed that reductions in rates of abuse were only 75% of the reductions observed in our claims data analysis and tested the assumption in our sensitivity analyses.

We applied the adjusted reductions in rates of diagnosed abuse associated with the introduction of reformulated ER oxycodone to the estimated number of diagnosed abusers in the US. This calculation provided an estimate of the reduction in the number of diagnosed abusers associated with the introduction of reformulated ER oxycodone. The analysis was extended proportionately to undiagnosed abusers.

The reduction in the number of abusers associated with the introduction of reformulated ER oxycodone was then multiplied by the average excess medical costs of patients diagnosed with abuse, estimated in the claims data analysis described above, to calculate the total medical cost savings associated with reductions in abuse post-reformulation. The excess annual per-patient medical costs associated with undiagnosed abuse were assumed to be 80% of the costs of diagnosed abuse based on prior literature. Sensitivity analyses were conducted to examine the effect of this assumption on our results.

Results

Changes in rates of diagnosed opioid abuse associated with the introduction of reformulated ER oxycodone

Among all continuous ERO users, 3.6% of commercially-insured patients, 1.2% of Medicare-eligible patients, and 6.2% of Medicaid patients were diagnosed with abuse in the 6-month pre-reformulation period (Table 1). In the 6-month post-reformulation period, rates of diagnosed abuse among users of reformulated ER oxycodone were 2.8% for the commercially-insured, 1.3% for Medicare-eligible, and 5.1% for Medicaid patients. These post-reformulation rates represent statistically significant declines in the rates of diagnosed abuse of 22.7% (p < 0.001) among commercially-insured patients and 18.0% (p = 0.034) among Medicaid patients post-reformulation. The change in rates of diagnosed abuse was not statistically significant among Medicare-eligible patients, which is an age group in which abuse is less common19.

Table 1. Changes in rates of abuse among continuous ERO users following the introduction of reformulated ER oxycodone.

Population	Continuous users of EROs (pre-reformulation)		Continuous users of reformulated ER oxycodone (post-reformulation)		Relative change in rates of abuse
	n	Rate of abuse [A]	n	Rate of abuse [B]	([B] − [A])/[A]	95% CI	p Value
Commercial	69,315	3.6%	14,480	2.8%	−22.7%	(−31.2%, −14.3%)	<0.001
Medicare-eligible	17,899	1.2%	3680	1.3%	6.1%	(−27.8%, 40.0%)	0.707
Medicaid	11,883	6.2%	2490	5.1%	−18.0%	(−34.0%, −2.0%)	0.034

Due to rounding, the differences may not exactly equal the estimates implied by the rounded numbers presented in the table. Chi-squared tests were used to test differences in rates of abuse between continuous users of EROs (pre-reformulation) and continuous users of reformulated ER oxycodone (post-reformulation).

Excess medical costs associated with opioid abuse

Before matching, there were significant differences between abusers and non-abusers along various baseline characteristics including demographics, comorbidities, and utilization of EROs (Table 2). Using the matching methods described above, nearly all eligible abusers in our sample were matched: 99% of commercially-insured abusers, 99% of Medicare-eligible abusers, and 98% of Medicaid abusers were successfully matched to non-abusers. After matching, differences between abusers and non-abusers were substantially diminished and, in most cases, no longer statistically significant (Table 2).

Table 2. Baseline characteristics of abusers and non-abusers.

Characteristics	Unmatched			Matched
	Abusers	Non-abusers	p Value	Abusers	Non-abusers	p Value
I. Commercially-insured patients
n	2532	61,421		2510	2510
Age on index date, mean (SD)	47.9 (10.2)	51.2 (9.1)	<0.001	48.0 (10.1)	48.0 (9.9)	0.937
Male, %	47.2%	43.9%	0.001	47.1%	48.5%	0.287
Charlson Comorbidity Index (CCI), mean (SD)	0.7 (1.3)	0.9 (1.5)	<0.001	0.69 (01.3)	0.72 (01.3)	0.307
Select co-morbidities
Lower back pain, %	72.9%	67.5%	<0.001	72.9%	73.4%	0.722
Arthritis, %	59.8%	55.3%	<0.001	59.6%	59.8%	0.931
Psychiatric disorders, %	44.7%	29.3%	<0.001	44.5%	35.3%	<0.001
Trauma or accidental injury, %	25.7%	19.0%	<0.001	25.4%	24.3%	0.358
Neuropathic pain, %	19.2%	17.1%	0.005	19.2%	18.0%	0.295
Fibromyalgia, %	17.5%	15.0%	<0.001	17.5%	17.9%	0.741
ERO days supply, mean (SD)	177.5 (117.6)	151.2 (93.0)	<0.001	174.6 (96.8)	173.7 (102.5)	0.256
II. Medicare-eligible patients
n	272	19,564		268	268
Age on index date, mean (SD)	74.1 (6.7)	76.9 (7.9)	<0.001	74.2 (6.7)	74.3 (7.3)	0.850
Male, %	30.5%	33.3%	0.333	30.6%	29.1%	0.705
Charlson Comorbidity Index (CCI), mean (SD)	1.4 (1.5)	1.6 (1.9)	0.249	1.41 (01.5)	1.28 (01.4)	0.494
Select comorbidities
Lower back pain, %	73.9%	57.6%	<0.001	73.5%	73.9%	0.916
Arthritis, %	62.1%	58.8%	0.267	61.6%	60.4%	0.785
Psychiatric disorders, %	38.2%	20.3%	<0.001	38.1%	25.4%	0.003
Trauma or accidental injury, %	30.9%	22.4%	<0.001	31.3%	24.3%	0.069
Neuropathic pain, %	22.1%	16.1%	0.008	21.6%	21.6%	1.000
Fibromyalgia, %	11.4%	6.9%	0.004	10.8%	10.1%	0.785
ERO days supply, mean (SD)	173.3 (83.6)	138.6 (84.1)	<0.001	172.7 (83.1)	167.8 (86.7)	0.185
III. Medicaid patients
n	548	7770		536	536
Age on index date, mean (SD)	46.9 (9.6)	48.9 (9.6)	<0.001	47.1 (9.5)	47.5 (9.8)	0.386
Male, %	37.8%	40.2%	0.259	38.2%	36.0%	0.446
Charlson Comorbidity Index (CCI), mean (SD)	1.5 (2.1)	1.7 (2.0)	0.004	1.49 (02.1)	1.42 (1.8)	0.755
Select comorbidities
Lower back pain, %	75.9%	66.9%	<0.001	75.7%	76.7%	0.692
Arthritis, %	69.0%	64.5%	0.033	68.7%	69.4%	0.786
Psychiatric disorders, %	65.0%	52.5%	<0.001	64.7%	57.1%	0.008
Trauma or accidental injury, %	30.5%	25.4%	0.008	30.4%	28.0%	0.366
Neuropathic pain, %	24.8%	22.0%	0.119	24.3%	23.7%	0.824
Fibromyalgia, %	17.2%	15.3%	0.253	17.0%	17.2%	0.936
ERO days supply, mean (SD)	175.0 (94.2)	148.4 (100.4)	<0.001	173.3 (93.2)	179.6 (104.8)	0.597

This table shows a selection of baseline characteristics. The full set of baseline characteristics is available upon request. For categorical variables, differences between abusers and non-abusers were assessed using chi-squared tests (pre-match) and McNemar tests (post-match). For continuous variables, differences between abusers and non-abusers were assessed using Wilcoxon rank-sum tests (pre-match) and Wilcoxon signed-rank tests (post-match).

Across all payer groups, diagnosed abusers had higher mean annual medical costs than non-abusers (Table 3, Panel I). Among commercially-insured patients, diagnosed abusers had mean annual per-patient medical costs of $37,846, compared with $28,390 for matched non-abusers. All costs reported in this section reflect mean annual per-patient medical costs. Among Medicare-eligible patients, diagnosed abusers had costs of $34,965, compared with $24,920 for matched non-abusers. Among Medicaid patients, diagnosed abusers had costs of $34,607, compared with $23,106 for matched non-abusers. The annual per-patient excess medical costs associated with diagnosed abuse were $9456 (p < 0.001) for commercially-insured patients, $10,046 (p < 0.001) for Medicare-eligible patients, and $11,501 (p < 0.001) for Medicaid patients. Differences in prescription drug costs between abusers and non-abusers were not statistically significant (Table 3, Panel II). Continuous ERO users have high prescription drug costs to begin with, and the incremental prescription drug costs associated with the diagnosis and treatment of abuse were not significant. Therefore, prescription drug costs were not included in the calculation of total cost savings associated with reformulated ER oxycodone at the US level.

Table 3. Excess medical and prescription drug costs associated with diagnosed opioid abuse among continuous ERO users.

Population (n for abusers)	Abusers [A]	Non-abusers [B]	Excess costs (difference)		p Value
			[A] − [B]	95% CI
I. Annual medical costs, mean (SD)
Commercial (n = 2510)	$37,846 ($53,282)	$28,390 ($50,575)	$9456	($6648, $12,264)	<0.001
Medicare-eligible (n = 268)	$34,965 ($44,885)	$24,920 ($44,423)	$10,046	($2565, $17,527)	0.009
Medicaid (n = 536)	$34,607 ($44,852)	$23,106 ($39,096)	$11,501	($6632, $16,369)	<0.001
II. Annual prescription drug costs, mean (SD)
Commercial	$11,230 ($13,784)	$11,529 ($14,155)	−$299	(−$1060, $461)	0.440
Medicare-eligible	$10,649 ($9387)	$12,152 ($19,769)	−$1504	(−$4113, $1106)	0.258
Medicaid	$12,851 ($24,320)	$12,165 ($11,994)	$686	(−$1580, $2953)	0.552

All costs are reported in 2011 US dollars. The sample sizes provided in the first column refers to the sample sizes for abusers by payer; there are identical sample sizes for the matched non-abusers by payer (e.g., the commercially-insured population includes 2510 abusers and 2510 non-abusers). Due to rounding, the differences may not exactly equal the estimates implied by the rounded numbers presented in the table. Paired t-tests were used to test differences in costs between abusers and non-abusers.

Medical cost savings at the US level

By applying the rate of prescription pain reliever abuse and dependence (0.7%) from the 2011 NSDUH to the total US population aged 18 and up from the US Centers for Disease Control and Prevention, we estimated 1.53 million Americans ages 18 and up with opioid abuse. Using the 1:5 ratio of diagnosed to undiagnosed abusers described above, we estimated that there were 254,781 diagnosed abusers and 1,273,907 undiagnosed abusers.

The reduction in the number of abusers yielded medical cost savings by averting the excess medical costs associated with abuse. We estimated annual cost savings from reductions in diagnosed abuse of $35 million, $16 million, and $35 million for commercially-insured, Medicaid, and uninsured patients, for a total of $86 million across all payers (Table 4, Panel I). In addition, we estimated annual cost savings from reductions in undiagnosed abuse of $139 million, $63 million, and $142 million for commercially-insured, Medicaid, and uninsured patients, for a total of $344 million across all payers (Table 4, Panel II). No significant reduction in rates of diagnosed abuse was found among Medicare-eligible patients, so we did not include any medical cost savings associated with reformulated ER oxycodone for this population.

Table 4. Calculations for medical cost savings at the US level.

Population	Reduction in number of abusers	Annual medical cost savings (2011 USD)
		Per patient	Total
I. Diagnosed abusers
Commercially-insured	3673	$9456	$34,730,641
Medicare-eligible	0	$10,046	$0
Medicaid	1371	$11,501	$15,769,878
Uninsured	3079	$11,501	$35,408,151
Sub-total for diagnosed abusers			$85,908,669
II. Undiagnosed abusers
Commercially-insured	18,364	$7565	$138,922,564
Medicare-eligible	0	$8036	$0
Medicaid	6856	$9200	$63,079,510
Uninsured	15,394	$9200	$141,632,603
Sub-total for undiagnosed abusers			$343,634,677
Total			$429,543,346

All costs are reported in 2011 US dollars.

Combining these estimates, we estimate that at the US population level, the total annual medical cost savings associated with reformulated ER oxycodone were $430 million. This includes $86 million in annual medical cost savings from reductions in diagnosed abuse and $344 million in annual medical cost savings from reductions in undiagnosed abuse.

Sensitivity analyses

We conducted sensitivity analyses to examine the impact of several assumptions on our results. We assumed that the impact of reformulated ER oxycodone on reducing rates of diagnosed abuse may be diminished by abusers switching away to other abusable substances such as EROs without abuse-deterrent technology17 or heroin18, so we included an adjustment factor that the actual impact of reformulated ER oxycodone would be 75% of what was observed in the claims data analysis (Table 1). To the extent that there is even greater switching to other abusable substances by abusers, changing this assumption to 50% would yield total annual medical cost savings of $286 million. Conversely, if no abusers switched away to other abusable substances (contrary to some of the published studies), then changing this assumption to 100% would yield total annual medical cost savings of $573 million.

Our claims data analyses were conducted among continuous ERO users, and we assumed that the medical costs of continuous ERO users and non-continuous ERO users were equivalent (i.e., 100%). Medical costs for non-continuous ERO users may actually be lower than those for continuous ERO users, as the latter are more likely to be chronic pain patients with higher medical costs. Changing this assumption to 75% yields total annual medical cost savings of $355 million.

The medical costs of undiagnosed abuse were assumed to be 80% of the costs of diagnosed abuse, as undiagnosed abuse may not have resulted in a crisis event such as an ED visit (or other interaction with a medical professional resulting in a medical claim with an abuse diagnosis), suggesting possibly lower costs. Varying this assumption by 20 percentage points yields total annual medical cost savings of $344 million (60%) to $515 million (100%).

Our sensitivity analyses ranged from $286 million to $573 million in annual medical cost savings at the US level and demonstrated that varying different assumptions still resulted in substantial medical cost savings associated with reformulated ER oxycodone.

Discussion

Our estimates suggest that the introduction of reformulated ER oxycodone has yielded substantial economic benefits. We observed relative reductions in rates of diagnosed abuse of 22.7% and 18.0% among commercially-insured and Medicaid continuous ERO users, respectively, following the introduction of reformulated ER oxycodone. Changes in rates of diagnosed abuse among Medicare-eligible continuous ERO users were small in magnitude and not statistically significant. This finding is consistent with the lower rates of abuse observed among older adults20. While reformulated ER oxycodone may have a lesser impact among older adults themselves, there could be societal spillover effects, as it has been reported that some elderly individuals divert RxOs for economic reasons21. Consistent with prior literature, we observed significant excess medical costs associated with diagnosed abuse20. The annual per-patient excess medical costs associated with diagnosed abuse ranged from $9456–$11,501, depending on the payer.

Overall, reformulated ER oxycodone was associated with total annual medical cost savings of $430 million in the US. Prescription drug cost savings were not included in this calculation since the excess prescription drug costs associated with the diagnosis and treatment of opioid abuse were not statistically significant, which is not surprising given that continuous ERO users have high prescription drug costs to begin with.

We did not attempt to account for potential reductions in abuse-related indirect costs22 (e.g., criminal justice and work loss costs), abuse-related mortality, or emotional and financial burden on family members. For example, RxOs now account for more overdose deaths than heroin and cocaine combined23. RxO abuse accounted for more than 15,500 deaths in 2009, nearly a 4-fold increase compared to 199924. Further research is necessary to establish the effect of EROs with abuse-deterrent technology on abuse-related mortality. Recent studies using an adverse event reporting database found reductions in the number of fatal adverse events (including abuse and overdose) occurring after reformulated ER oxycodone was introduced, as compared to the original formulation of ER oxycodone without abuse-deterrent technology22,25. To the extent that reformulated ER oxycodone contributes to reduced mortality, the societal benefits associated with reformulated ER oxycodone could be substantially higher than those presented above.

Results from our claims data analyses were consistent with previous research. Our rate of diagnosed abuse among commercially-insured continuous ERO users in the pre-reformulation period (3.6%) was comparable to the rate (3.8%) found in a study of adults receiving daily opioid therapy26 and the rate (3.3%) found in a study of continuous non-cancer pain patients on continuous opioid therapy27. Another study found that the percentage of abusers at substance abuse treatment centers who reported abusing reformulated ER oxycodone was significantly lower than the percentage of abusers at substance abuse treatment centers who reported abusing the original formulation of ER oxycodone before the introduction of the reformulation with abuse-deterrent technology17. This study measured the share of abusers who used a given drug in the last 30 days, which is different from the measure of the rate of abuse used in this study, but the results were directionally consistent with our finding of reductions in rates of abuse post-reformulation.

A key strength of this study is that it analyzed actual changes in rates of diagnosed opioid abuse following the introduction of reformulated ER oxycodone, and estimated the excess medical costs of opioid abuse using observed medical and prescription drug claims. Previous studies on the impact of abuse-deterrent formulations relied on projections of what would happen in the advent of an abuse-deterrent formulation. As new EROs with characteristics designed to deter abuse are introduced, additional research should be conducted to examine their impacts on abuse and medical costs.

This study has several limitations. First, due to the observational research design used in the claims data analyses, caution is warranted in any causal interpretation of the findings. While the introduction of reformulated ER oxycodone allows for a quasi-experimental study design, it is not a full substitute for a true randomized experiment. Therefore, while we observed changes in rates of diagnosed opioid abuse associated with the introduction of reformulated ER oxycodone, it is still possible that unobserved factors may have contributed to these changes. Moreover, we were only able to identify abusers who were diagnosed for their abuse.

Second, due to data limitations, portions of the study had to rely on prior literature, government reports, and assumptions, rather than primary data analyses. For example, we were only able to study Medicare-eligible patients enrolled in employer-sponsored Medicare Supplemental plans, as we did not have access to actual Medicare claims data. While these patients are eligible for Medicare due to their age, Medicare Secondary Payer Rules imply that their employers are the primary payers. For these Medicare-eligible patients, both their commercial and Medicare payments were included in our medical cost calculations in order to capture all of their medical costs. We assumed that these Medicare-eligible patients (i.e., Medicare Secondary Payer patients) were representative of patients for whom Medicare was the primary payer, but the patient populations may differ. We also lacked data on the uninsured, so we assumed that their abuse rates and medical costs mirrored those of Medicaid enrollees. Nevertheless, our estimates are based on the best evidence currently available from government reports and published literature.

Third, our claims data analyses were conducted among continuous ERO users so we were unable to examine changes in abuse and medical costs among individuals who used EROs sporadically or among individuals primarily using IROs or illicit drugs. We assumed that continuous ERO users and non-continuous ERO users would be impacted by reformulated ER oxycodone in the same way (i.e., that reformulated ER oxycodone would impact the rates of abuse and medical costs of both groups in the same way). Future research should examine the impact of reformulated ER oxycodone in the broader population of patients, part of which may not use EROs continuously. Recent studies have found that, while reformulated ER oxycodone is aimed at deterring abuse, serious abusers may find other substances to abuse, such as EROs or IROs without abuse-deterrent properties17 or illicit drugs such as heroin18. Future research should also examine the potential unintended consequences of reformulated ER oxycodone such as its impact on the abuse of drugs without abuse-deterrent characteristics, as determined abusers may switch to other more abusable drugs.

Conclusions

EROs with abuse-deterrent properties are part of a multi-faceted approach to RxO abuse reduction, including education about appropriate RxO use, policy, and co-ordination among healthcare providers, pharmaceutical companies, government, and researchers. Use of abuse-deterrent EROs can help achieve a balance between ensuring appropriate pain management and deterring abuse. This study provides evidence that reformulated ER oxycodone with abuse-deterrent technology has yielded substantial health and economic benefits in the US, with annual savings in direct medical costs of ∼$430 million. Payers and policymakers should consider these benefits as they devise and implement new guidelines and policies in this therapeutic area.

Transparency

Declaration of funding

This study was funded by Purdue Pharma, L.P.

Declaration of financial/other relationships

RBJ is an employee of Purdue Pharma L.P.; NYK, AS, AGW, and HGB are employees of Analysis Group, Inc., a consulting company that received funding for this research from Purdue Pharma L.P. LFR and EM are consultants on this project and received funding from Analysis Group. JME Peer Reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

The authors gratefully acknowledge the contributions of Katharine Bodnar, Alice Kate Cummings, Michael Kaminsky, and Sarah King.