Abstract
Objective:
Patients managing chronic non-cancer pain with cytochrome P450 (CYP450)-metabolized opioid analgesics who concurrently take another CYP450-metabolized medication experience a drug–drug exposure (DDE), which puts them at risk for a pharmacokinetic drug–drug interaction (PK DDI). This study examined the economic impact of incident DDEs with the potential to cause PK DDIs compared to similar patients without such exposure.
Study Design:
This retrospective analysis used paid claims from a large, commercially insured population during January 1, 2004 through December 31, 2008.
Methods:
Propensity matching was used to control for baseline differences in comparisons between 85,043 exposed and 85,043 non-exposed patients.
Results:
Comparisons yielded mean total costs 6 months after the DDEs that were significantly higher in subjects with DDE versus matched subjects without DDE [$8165 (SD $11,357) vs. $7498 (SD $11,668), respectively, p < 0.01] resulting in a difference of $667. This was driven by medical costs [$5520 (SD $10,505) vs. $5222 (SD $10,689), respectively, p < 0.01] a $298 difference, and total prescription costs [$2646 (SD $3262) vs. $2276 (SD $3907), respectively, p < 0.01] a $369 difference.
Limitations:
The study design demonstrates associations only and cannot establish causal relationships. Further, relevant DDEs were not included if concurrent consumption occurred outside the index period and when CYP450 substances were consumed that are not reflected in pharmacy claims (herbals, over-the-counter medications).
Conclusion:
Since concurrent exposure to DDEs with the potential to cause PK DDIs may be relatively common, policy decisions-makers should consider the use of long-acting opioids that are not metabolized through the CYP450 pathway.
Introduction
Opioid analgesics are recommended for patients with moderate or severe chronic non-cancer pain (CNCP) that adversely affects their function and when there is a favorable risk-benefit ratio, according to evidence-based guidelines issued jointly by the American Pain Society and American Academy of Pain MedicineCitation1. While support exists for use of opioids to manage CNCP, questions persist regarding their appropriate useCitation2,Citation3.
When multiple drugs are metabolized simultaneously through the cytochrome P450 (CYP450) system, patients experience a drug–drug exposure (DDE) with potential risk of a pharmacokinetic drug–drug interaction (PK DDI)Citation4. Among the opioid analgesics commonly prescribed to CNCP patients, codeine, fentanyl, hydrocodone, methadone, oxycodone and tramadol are metabolized by the CYP450 enzymesCitation5. A Black Box warning appears in the prescribing information for OxyContin (oxycodone hydrochloride controlled-release) tablets regarding concomitant use of OxyContin with CYP450 3A4 inhibitors and inducersCitation6. Small clinical studies of oxycodone DDEs have demonstrated PK changes and associated clinical manifestationsCitation7,Citation8. The clinical impact of PK DDIs vary in severity and can be difficult to predictCitation9,Citation10.
A DDE may result in a PK DDI and manifest as an adverse drug experience (ADE), resulting in longer length of hospital stay, higher operating expenses and decreased profitsCitation11–15. While the global incidence of PK DDIs is not known, a sharp increase in such DDIs was noted among seniors from 1992 to 2005, with a prevalence of 19.2% in patients ≥70 years of ageCitation16. Previous research found a 27% prevalence of DDEs with the potential to cause PK DDIs among ambulatory chronic low back pain patients using CYP450 opioid analgesicsCitation17 and 26% in osteoarthritisCitation18.
While well-controlled, randomized trials would provide the best source of information regarding the clinical consequences of specific DDEs, their availability is limited and the significance and recognition of their occurrence in common practice is likely variable. Their economic impact is relatively unknown. Our objective was to assess among ambulatory CNCP patients using CYP450-metabolized opioid analgesics, the associated clinical events and economic outcomes comparing those with DDEs of interest with similar subjects whose concurrent medications do not have the potential to cause PK DDIs (no-DDE).
Methods
This retrospective analysis used medical and pharmacy claims from a commercially insured population in the MarketScan Commercial Claims and Encounters Database, containing comprehensive claims on more than 40 million beneficiaries, representing a large proportion of commercially insured beneficiaries in the US. All patient-unique identifiers are removed to ensure patient privacy.
The database has been evaluated and certified by an independent third party to be in compliance with the HIPAA statistical de-identification standards and satisfies conditions set forth in Sections 164.514 (a)--(b)1ii of the HIPAA privacy rule regarding the documentation of de-identified data. The study design was submitted for review by the Independent Investigational Review Board, Inc., which determined that it was exempt from human subject research review.
All 18–65-year-old subjects were identified for the period July 1, 2006 through September 30, 2009. Included were subjects who received CYP450 opioid analgesics for ≥30 days to indicate chronic use. Excluded were those with pregnancy-related claims (ICD-9-CM: 633, 640–646, 761, V23.2, V22, V61.6–61.7, V72.40), health maintenance organization enrollment or claims indicative of cancer (ICD-9 codes 140–208), but including non-melanoma skin cancer (ICD-9: 173.xx) during the baseline and observation periods. Subjects had continuous enrollment (medical and pharmacy benefits) during the 6 months prior to and after the index date.
The baseline period was 6 months prior to the index date, the first dispense date for a CYP450 opioid, including codeine, fentanyl, hydrocodone, methadone, oxycodone or tramadolCitation5. Hydromorphone, morphine and oxymorphone were excluded as their chief metabolic pathway is through the UGT2B7 system and they offer minimal potential for CYP450-related PK DDIs since this system is not an important substrate, inhibitor or inducer of the CYP450 system. The index period was 30-days immediately following the index date during which a second prescription was dispensed (concurrent medication date). A DDE occurred when the concurrent medication was metabolized through the CYP450 system, with at least 1 day overlap. For no-DDE, the concurrent medication was a drug not metabolized through CYP450. The observation period occurred 6 months following the concurrent medication date. ()
Figure 1. Study timeline. The index date, defined as the first day of a prescription of at least a 30-day supply of a CYP450-metabolized opioid analgesic, commences the 30-day index period. During this period, each patient received a concurrent prescription for another agent with at least 1 day’s overlap with the opioid prescription. The dispense date of the second prescription is the concurrent medication date, which may fall anywhere in the index period. Subjects whose concurrent medications are CYP450-metabolized drugs are then assigned to the DDE group; subjects whose concurrent medications are not CYP450-metabolized drugs belong in the no-DDE group. The observation period is the 6-month period following the concurrent medication date.
Associated clinical events were claims for services that represent potential ADEs that may arise from PK DDIs, including: cardiovascular events, respiratory symptoms, ileus, drug-induced symptoms, and other potential ADEs temporally associated with the incidence of DDEsCitation19–28. Drug-induced symptoms include mental disorders, delirium (confusion), hypersomnia (sedation), anxiety disorders, dermatitis (pruritus), and headache. These were calculated as the percentage of subjects experiencing a relevant event within 30 days of the concurrent medication date, based on ICD-9 codes. As a sensitivity analysis, data were assessed again at 90 and 180 days.
Healthcare services utilization and cost data occurred during the observation period, regardless of diagnosis. This included claims related to physician office, outpatient and emergency department (ED) visits as well as inpatient hospitalizations. Multiple claims could arise from a single office or outpatient visit. Average number of hospital days was derived from inpatient visits for any cause and calculated as the sum of all days from inpatient visits divided by the number of subjects. Medical costs were payments for office, outpatient and ED visits as well inpatient hospitalizations. Prescription costs were for opioid and non-opioid drugs. All costs were adjusted to 2008 values using the Consumer Price Index for Medical Care and expressed in US dollars. All analyses were from the payer’s perspective, based on total plan payments (excluding patient payments).
Propensity score (PS) matchingCitation29,Citation30 was used to control for potential selection bias between the two groups: DDE and no-DDE. Estimates of the predicted likelihood of DDE occurrence were derived from age, gender, geographical location, number of unique concurrent medications, healthcare services utilization and total costs during the baseline period. Univariate analysis was performed to compare mean differences after applying PS matching. The bootstrap t-tests and chi-square test were used for analysis of continuous and categorical variables, respectively. Since cost data can be sensitive to extreme outliers, subjects were excluded when total costs during the study period exceeded the 99th percentileCitation31. The statistical package used for all analyses was SAS 9.2 for WindowsCitation32.
Several sensitivity analyses were conducted, comparing costs between the groups:
Among DDE subjects, those with a ≥15-day versus <15-day overlap
DDE subjects with a ≥15-day overlap versus no-DDE subjects
DDE subjects with a <15-day overlap versus no-DDE subjects
Results
Of 10,279,623 subjects with ≥1 prescription for an opioid analgesic, 1,346,165 received a CYP450 opioid (or opioids) with ≥30-day supply. After exclusions, 322,343 subjects remained. The 1:1 matching yielded a sample of 170,086 – 85,043 per group (). Prior to PS matching, significant differences (all p < 0.01) in all baseline characteristics occurred comparing those with and without a DDE. After PS matching, differences remained for age, number of unique prescriptions, depression and anxiety rates and percentage utilizing office, inpatient and ED visits, as well as inpatient admissions ().
Table 1. Sample selection, Inclusion and exclusion criteria defining patient cohorts.
Table 2. Patient characteristics at baseline (before and after PS matching).
Associated clinical events were evaluated 30 days after the DDE concurrent medication date. Compared to similar subjects with no-DDEs, the DDE group had a lower percentage with ≥1 claims consistent with potential ADRs related to cardiovascular events [13.7% (1636/85,043) vs. 15.1% (12821/85,043), p < 0.01] within 30 days of a DDE, but a higher percentage with drug-induced symptoms [0.4% (340/85,043) vs. 0.3% (277/85,043), p < 0.01]. As a sensitivity analysis, these events were evaluated at 90 and 180 days. At both points, the direction and magnitude of these differences remained.
During the 6-month observation period, subjects with DDEs had significantly higher average number of claims related to office visits (19.1 vs.18.3, p < 0.01) and outpatient visits (6.7 vs. 6.4, p < 0.01) than similar subjects without DDEs. Most subjects had 2–3 claims associated with one service date for office and outpatient visits. In addition, DDE subjects had significantly higher average number of ED visits (0.46 vs. 0.43, p < 0.01), inpatient hospitalizations (0.13 vs. 0.12, p < 0.01) and longer average number of hospital days (0.54 vs. 0.47, p < 0.01) than similar subjects without DDEs ().
Table 3. Medical services utilization.
Subjects with DDEs had significantly greater 6-month total costs than similar subjects with no DDE [$8165 (SD $11,357) vs. $7498 (SD $11,668), respectively, p < 0.01], a $667 difference. This was driven by medical costs [$5520 (SD $10,505) vs. $5222 (SD $10,689), respectively, p < 0.01] a $298 difference, and total prescription costs [$2646 (SD $3262) vs. $2276 (SD $3907), respectively, p < 0.01] a $369 difference. Compared to those without a DDE, those with a DDE incurred $294 less cost for opioid prescriptions but $663 more for non-opioid prescriptions. (, )
Figure 2. Mean costs for DDE and no-DDE subjects, post matching. *p < 0.01. DDE, drug–drug exposure. Bias-corrected bootstrapping t-test used to calculate p-values (1000 replications).
Table 4. Medical and prescription payments ($) for patients with and without drug–drug exposure (DDE).
In this study, an incident DDE was based on at least 1 day’s overlap between a CYP450 opioid and a concurrent medication metabolized via the CYP450 system. However, it is possible that subjects who concurrently possessed two potentially interacting prescriptions for a short period of time did not take them at the same time. For DDE subjects, mean total costs for those with an overlap of ≥15 days (n = 26,480) was $496 more (p < 0.01) than those with <15 days overlap (n = 58,563). In addition, mean total costs for DDE subjects with a <15-day overlap was a $513 more (p < 0.01) than those with no-DDE. Further, mean total costs for DDE subjects with a ≥15-day overlap was $1008 more (p < 0.01) than those with no-DDE. Finally, we compared costs of all subjects, including those with cancer diagnoses. The mean total costs for the DDE group including cancer was $726 more (p < 0.01) than for those with no-DDE (n = 89,862).
Discussion
Our study evaluated DDEs which may result in potentially serious PK DDIs. Subjects with DDEs with the potential to cause PK DDIs utilized more healthcare resources and had significantly greater associated costs compared to matched subjects whose DDEs did not pose this risk. In numerous sensitivity analyses, the direction, magnitude and statistical significance of results support the findings in the primary analysis, supporting the robustness of the findings.
The cost differential of over $667 in a 6-month period is not trivial and represents a substantial burden to health plans, which is a potentially avoidable cost. To reduce the risk of a PK DDIs, with potentially important economic consequences, decision-makers should consider opioid analgesics that do not rely on the CYP450 enzyme system as their primary means of metabolism, such as morphine, hydromorphone and oxymorphone when making utilization management policies for patients at riskCitation33.
To isolate the independent economic impact of incident DDEs with the potential to cause PK DDIs, we excluded subjects with DDEs in the baseline period. Further, in the no-DDE group, we excluded those with a DDE during the observation period. These criteria combined to exclude a large portion of potential patient observations. The size of this excluded population is not entirely surprising, given that previous research has indicated that the prevalence rate of DDEs in this population is about 26%Citation17,Citation18. Thus, by focusing narrowly on incident DDEs, this study may provide a conservative estimate of the economic impact of prevalent DDEs.
Moreover, this study was retrospective and identified prescription drug use from claims records which provide, at best, an imperfect record of actual medication consumptionCitation34. Data sources did not report the concentrations of drugs used or allow us to determine if PK DDIs resulted in adjustments to dosages or changes in medication, including discontinuationsCitation35. However, this potential limitation was evaluated with sensitivity analyses which supported the robustness of our findings.
While standard PS matching was used to mitigate confounding due to potential sample selection bias, it requires the assumption that any unobserved factors affecting the tendency to experience a DDE can be ignored. If this assumption is incorrect, our study results remain subject to the potential for sample selection bias. Further, the 1:1 PS matching used in this study provided an imperfect direct comparison, with some differences remaining.
We could not observe use of over-the-counter (OTC) medications, supplements and herbal products, which must be taken into account when evaluating the results from our study in that many subjects use such productsCitation36. As a result, our results are likely subject to false negatives (misclassification of subjects experiencing a DDE with an OTC drug or supplement as not having a DDE). Thus, our study must be viewed as a conservative estimate of the economic impact of DDEs.
This analysis was foundational in nature, intended to gain a better understanding of the potential scope of the economic consequences of DDEs associated with certain opioid analgesics. It did not examine results based on the nature of the DDE, such as type of CYP450 enzyme (CYP2D6, CYP3A4) or whether the DDE involved exposure to inhibitors, inducers or substrates.
Conclusions
When payers evaluate policies related to opioid analgesic utilization, they may want to consider the metabolism of product alternatives. Subjects with DDEs had $667 greater total costs in the 6 months following a DDE, compared to similar subjects with no DDE. Thus, DDEs place a potentially avoidable economic burden on the healthcare system in terms of both resource utilization and costs. Since concurrent exposure to multiple drugs metabolized through the CYP450 enzyme system may be relatively common, policy decisions for patients at risk should include a consideration of long-acting opioids not metabolized through the CYP450 pathway. While policies that encourage use of non-CYP450 opioid analgesics would not eliminate DDEs and PK DDIs, they would likely mitigate their economic impact.
Transparency
Declaration of funding
This research was sponsored by Endo Pharmaceuticals.
Declaration of financial/other relationships
Four of the authors are employees of Endo Pharmaceuticals (K.H.S., R.A.P., R.B-J.), and one author is a contract employee (N.R.). One author is a consultant to Endo Pharmaceuticals (R.R.O.).
Acknowledgments
The authors wish to acknowledge an Endo employee and contractors: Steve Cooper for helping develop the research concept; Seongjung Joo, PhD for programming and Chunmay Fu, MS, for quality control.
References
- Chou R, Fanciullo GJ, Fine PG, et al. Clinical guidelines for the use of chronic opioid therapy in chronic noncancer pain. J Pain 2009;10:113-30
- Altman R. Practical considerations for the pharmacologic management of osteoarthris. Am J Manag Care 2009;15:S236-S243
- Nicholas MK, Molloy AR, Brooker C. Using opioids with persisting noncancer pain: a biopsychosocial perspective. Clin J Pain 2006;22:137-46
- Hartshorn EA, Tatro DS. Principles of drug interactions. In: Tatro DS, ed. Drug Interaction Facts, 5th edn. St. Louis: Facts and Comparisons, 1996
- Holmquist G. Opioid metabolism and effects of cytochrome P450. Pain Med 2009;10:S20-9
- Prescribing Information, OxyContin® (oxycodone hydrochloride controlled-release) Tablets Purdue Pharma L.P. Stamford, CT 301734-0C May 10, 2010
- Kummer O, Hammann R, Moser C, et al. Effect of the inhibition of CYP3A4 or CYP2D6 on the pharmacokinetics and pharmacodynamics of oxycodone. Eur J Clin Pharmacol 2010. [Epub ahead of print]
- Nieminen TH, Hagelberg NM, Saari TI, et al. Oxycodone concentrations are greatly increased by the concomitant use of ritonavir or lopinavir/ritonavir. Eur J Clin Pharmacol 2010;66:977-85
- Chien JY, Mohutsky MA, Wrighton SA. Physiological approaches to the prediction of drug-drug interactions in study populations. Curr Drug Metab 2003;4:347-56
- Shou M LY, Lu P, Tang C, et al. Enzyme kinetics of cytochrome P450-mediated reactions. Curr Drug Metab 2000;2:17-36
- Reducing and preventing adverse drug events to decrease hospital costs. Research in Action, Issue 1. AHRQ, Agency for Healthcare Research and Quality http://www.ahrq.gov/qual/aderia/aderia.htm. 2001. Accessed July 26, 2010
- Bates D, Spell N, Cullen D, et al. The costs of adverse drug events in hospitalized patients. JAMA 1997;277:307-11
- Classen D, Pestotnik S, Eans R, et al. Adverse drug events in hospitalized patients: excess length of stay, extra costs, and attributable mortality. JAMA 1997;227:301-6
- Saurat J-H, Guerin A, Yu A, et al. High prevalence of potential drug–drug interactions for psoriasis patients prescribed methotrexate or cyclosporine for psoriasis: associated clinical and economic outcomes in real-world practice. Dermatology 2010;220:128-37
- Cruciol-Souza JM, Thomson JC. A pharmacoepidemiologic study of drug interactions in a Brazilian teaching hospital. Clinics (Sao Paulo) 2006;61:515-20
- Becker ML, Hofman A, Stricker BHC. Increasing exposure to drug–drug interactions between 1992 and 2005 in people aged =>55 years. Drugs Aging 2008;25:145-52
- Pergolizzi J, Labhsetwar S, Puenpatom R, et al. Exposure to potential CYP450 prevalence of exposure to potential CYP450 pharmacokinetic drug–drug interactions among patients with chronic low back pain taking opioids. Pain Pract 2010. [Epub ahead of print]
- Pergolizzi J, Labhsetwar S, Puenpatom R, et al. Exposure to potential CYP450 pharmacokinetic drug–drug interactions among osteoarthritis patients: incremental risk of multiple prescriptions. Pain Pract 2010 doi: 10.1111/j, 1533-2500
- Haijar E, Hanlon J, Artz M, et al. Adverse drug reaction risk factors in older outpatients. Am J Geriatr Pharmacother 2003;1:82-9
- McCaffery M, Pasero M. Pain: Clinical Manual. Philadelphia: Mosby, 1999
- Slone Epidemiology Center. Patterns of Medication Use in the United States 2006. http://www.bu.edu/slone/SloneSurvey/AnnualRpt/SloneSurveyWebReport2006.pdf. Accessed October 5, 2010
- Nguyen J, Fouts M, Kotabe S, et al. Polypharmacy as a risk factor for adverse drug reactions in geriatric nursing home residents. Am J Geriatr Pharmacother 2006;4:36-41
- Gurwitz J, Field T, Harrold L, et al. Incidence and preventability of adverse drug events among older persons in the ambulatory setting. JAMA 2003;289:1107-16
- Lötsch J. Pharmacokinetic-pharmacodynamic modeling of opioids. J Pain Symptom Manage 2005;29(5 Suppl):S90-103
- Kurz A, Sessler D. Opioid-induced bowel dysfunction: pathophysiology and potential new therapies. Drugs Aging 2003;63:649-71
- Pappagallo M. Incidence, prevalence, and management of opioid bowel dysfunction. Am J Surg 2001;182(5A Suppl):11-18S
- Cadet P. Mu opiate receptor subtypes. Med Sci Monit 2004;10:MS28-32
- Hale M, Ahdieh H, Ma T, et al. Efficacy and safety of OPANA ER (oxymorphone extended release) for relief of moderate to severe chronic low back pain in opioid-experienced patients: a 12-week, randomized, double-blind, placebo-controlled study. J Pain Symptom Manage 2007;8:175-84
- Robins J, Mark S, Newey W. Estimating exposure effects by modeling the expectation of exposure conditional on confounders. Biometrics 1992;48:479-95
- Rosenbaum P, Rubin D. Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc 1984;79:516-24
- Barnett V, Lewis T. Outliers in Statistical Data, 3rd edn. Chichester, UK: Wiley, 1994
- SAS 9 on Microsoft Windows, 2008, SAS Institute Inc., Cary, NC, USA
- Armstrong S, Cozza K. Pharmacokinetic drug interactions of morphine, codeine, and their derivatives: theory and clinical reality, part I. Psychosomatics 2003;44:167-71
- Solberg L, Hurley J, Roberts M, et al. Measuring patient safety in ambulatory care: potential for identifying group drug–drug interaction rates using claims data. Am J Manag Care 2005;10:753-9
- Dickson M, Bramley T, Kozma C, et al. Potential drug–drug interactions with antiepileptic drugs in Medicaid recipients. Am J Health-System Pharm 2008;65:1983-91
- Qato D, Alexander G, Conti R, et al. Use of prescription and over-the-counter medications and dietatry supplements among older adults in the United States. JAMA 2008;300:2867-78