- Email: [email protected]
Incidence and antecedents of nonmedical prescription opioid use in four US communities
Drug and Alcohol Dependence 85 (2006) 171–176
Incidence and antecedents of nonmedical prescription opioid use in four US communities The Coronary Artery Risk Development in Young Adults (CARDIA) prospective cohort study Mark J. Pletcher a,b,∗ , Stefan G. Kertesz c,e , Stephen Sidney d , Catarina I. Kiefe c,e , Stephen B. Hulley a a
Department of Epidemiology and Biostatistics, University of California, San Francisco, 185 Berry Street, Suite 5700, San Francisco, CA 94107, United States b Division of General Internal Medicine, University of California, San Francisco, CA 94143, United States c Division of Preventive Medicine, University of Alabama at Birmingham, AL 35205, United States d Division of Research, Kaiser Permanente, Oakland, CA 94612, United States e Deep South Center on Effectiveness, Veteran’s Affairs Medical Center, Birmingham, AL 35205, United States Received 18 January 2006; received in revised form 11 April 2006; accepted 14 April 2006
Abstract Background: Nonmedical use of prescription opioids has emerged as a major public health problem during the last decade, but direct measures of incidence and predisposing factors are lacking. Methods: We prospectively measured incidence and antecedents of nonmedical prescription opioid use in The Coronary Artery Risk Development in Young Adults study among 28–40-year-old African- and European-American men and women with no prior history of nonmedical opioid use. Results: Among 3163 participants, 23 reported new nonmedical prescription opioid use in 2000–2001 (5-year incidence 0.7%; 95%CI: 0.4–1.0%). All 23 had previously reported marijuana use (p < 0.001). Five-year incidence was significantly higher among European-American men (OR = 3.3; 95%CI: 1.3–8.3), and among participants reporting a history of amphetamine use (OR = 24; 95%CI: 6.9–83) or medical opioid use for treatment of pain (OR = 8.6; 95%CI: 2.5–30). These associations remained strong when examined among marijuana users and after adjusting for demographics, social factors, and other antecedent substance use. Amphetamine use was the best single predictor of future nonmedical use (sensitivity 87%, specificity 79%). Conclusions: Initiation of nonmedical prescription opioid use is generally rare in 28–40-year-old adults, but is observed to be more common with a previous history of substance abuse and legal access to opioids through prescription by a physician. © 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Opioid-related disorders; Pain; Depression; Substance-related disorders
1. Introduction Nonmedical use of prescription opioid analgesics has emerged as a major public health problem during the last decade (Zacny et al., 2003). Cross-sectional studies in the United States (US) conducted by the Substance Abuse and Mental Health Services Administration (SAMHSA) show a tripling in the rate of initiation on nonmedical prescription
∗
Corresponding author. Tel.: +1 415 514 8008; fax: +1 415 514 8150. E-mail address: [email protected] (M.J. Pletcher).
0376-8716/$ – see front matter © 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.drugalcdep.2006.04.011
opioid use among young adults since 1990 (US, 2005a, 2005b, 2005c, 2005d) and concomitant increases in attributable emergency department visits (US, 2004a, 2004b, 2004c) and substance abuse treatment admissions (US, 2003), both of which have doubled during the last decade. More than 2 million US adults start using prescription opioids for nonmedical purposes every year (US, 2004a, 2004b, 2004c; US, 2005a, 2005b, 2005c, 2005d), and prescription opioid dependence is now more than four times as prevalent as heroin dependence among the non-institutionalized US population sampled in the National Survey on Drug Use and Health (US, 2005a, 2005b, 2005c, 2005d). Nonmedical use of prescription opioids appears
172
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
to be an increasing problem outside the US as well (Hall et al., 2000; Hao et al., 2002; Mattoo et al., 1997; Shah et al., 2001). While these national surveys have helped to identify and describe this emerging phenomenon, cross-sectional data are less useful in estimating incidence or in identifying antecedents of developing disease. For example, nonmedical prescription opioid use has been associated with mental illness (SimoniWastila et al., 2004) and use of illegal substances (Dowling et al., 2004), but published data do not show which problem typically comes first. Similarly, it has been difficult to determine the degree to which prescribing opioids for pain increases the risk for nonmedical prescription opioid use. Previous research (mostly uncontrolled case series (Medina and Diamond, 1977; Perry and Heidrich, 1982; Portenoy and Foley, 1986; Porter and Jick, 1980)) suggests that patients treated for pain are at low risk for developing opioid addiction (Ballantyne and Mao, 2003; Passik and Kirsh, 2004), but surveys of opioid-addicted patients (Potter et al., 2004; Rayport, 1954) suggest that opioid treatment for pain often predates addiction to heroin. Longitudinal data are needed to estimate directly how often patients start misusing prescription opioids, and to identify which putative risk factors precede, rather than result from, the development of nonmedical prescription opioid use (Passik and Kirsh, 2003) in order to inform the efforts of both public health officials and individual physicians trying to reduce the potential harm from addiction caused by these otherwise useful pain medications. The Coronary Artery Risk Development in Young Adults (CARDIA) study measured substance abuse, mental health, and use of prescribed medications in a large community-based cohort of young black and white men and women in 1985–1986 and at five subsequent examinations. We identified a cohort of young adults aged 28–40 who in 1995–1996 denied misusing opioids of any kind, and characterized incidence and antecedents of new, self-reported nonmedical prescription opioid use 5 years later. 2. Methods 2.1. Study design and sample CARDIA is a longitudinal study of risk factors for coronary artery disease in a community-based cohort of European- and African-American women and men (n = 5115) aged 18–30 years and healthy at the time of enrollment in 1985–1986. With the informed consent of participants and the approval of Institutional Review Boards at each site (Oakland, Chicago, Minneapolis, and Birmingham), participants underwent a baseline examination and follow-up examinations at years 2, 5, 7, 10, and 15, with 74% retention at year 15 (2000–2001). Details of the study design, recruitment and procedures have been published (Friedman et al., 1988; Hughes et al., 1987). For this investigation, we identified CARDIA participants who reported no nonmedical opioid use (heroin or prescription) on a substance use questionnaire at year 10 (1995–1996), and measured associations between medical use of opioids, illegal substance use, and depression at year 10 with new self-reported nonmedical prescription opioid use at year 15.
2.2. Primary predictors Participants filled out a substance use questionnaire at each examination asking if they had ever used “marijuana”, “cocaine or crack”, “amphetamines (‘Speed’ or ‘Uppers’)”, and “opiates (Heroin, Dilaudid, Morphine, Demerol)” for nonmedical reasons. In order to capture only new nonmedical opioid users
at year 15, participants who reported nonmedical opioid use of any sort on this questionnaire at year 10 were excluded from the analysis. Among the remaining participants, we tested whether measuring past use of marijuana, cocaine and/or amphetamines might help predict new nonmedical prescription opioid use 5 years later. Note that all references to “amphetamine use” refer to nonmedical amphetamine use. We identified participants who were receiving opioids for medical purposes at year 10 through manual review of prescription medications reported on a nonsubstance abuse-oriented questionnaire concerning medications prescribed by a physician. We found participants using propoxyphene (n = 7), codeine (with and without acetaminophen, n = 19), hydrocodone (n = 27), oxycodone (n = 2), and methadone (n = 1). Participants reporting any of these medications on their prescription medication list were deemed to be “medical users” of prescription opioids at year 10. Depressive symptoms were measured at the year 10 examination using the Center for Epidemiologic Studies-Depression (CES-D) scale, a 0–60 point scale for which scores of 16 or greater are considered to indicate significant depressive symptomatology (Naughton and Wiklund, 1993; Radloff, 1977). We categorized the CES-D score into four groups – 0–7, 8–15, 16–23, and ≥24 – for purposes of adjustment, but used the standard cutoff score of ≥16 as our primary depressive symptoms predictor.
2.3. Outcome: nonmedical prescription opioid use at year 15 The substance use questionnaire (described above) was also administered in year 15, 5 years after the year 10 exam that serves as the baseline for this analysis. Participants in our sample who newly answered yes to, “Have you ever used opiates (Heroin, Dilaudid, Morphine, Demerol) for nonmedical reasons?”, and indicated use of either Dilaudid, morphine, Demerol, or an “Other (specify)” opioid that we identified as a prescription opioid (not heroin, opium, etc.) were considered to be incident cases of nonmedical prescription opioid use.
2.4. Other covariates Sex, ethnicity and birthday were self-reported at baseline. Educational grade attained (no college, some college, college graduate), family income (<$25,000, $25,000–49,000, ≥$50,000/year), and health insurance were self-reported at the year 10 examination. Smoking at year 10 was categorized as never, past or current, and alcohol consumption was categorized as 0, 1–6, 7–13, or ≥14 drinks/week. Participants with missing data on income, education, health insurance status, alcohol or smoking (n = 29) were excluded from multivariable analyses. These participants were not statistically different from included participants in terms of demographics and substance abuse, and none developed the outcome at year 15 (exact p = 1.0).
2.5. Statistical analysis We described the study sample at baseline in reference to use or non-use of opioid medications for treatment of pain at baseline, testing for bivariate associations using Fisher’s exact tests. We then measured the 5-year cumulative incidence of initiating nonmedical prescription opioid use, and calculated exact 95% confidence intervals using the binomial distribution. We compared incidence in subgroups of participants defined by our primary predictors and tested differences with Fisher’s exact tests. We used logistic regression to measure the association between each primary predictor and year 15 nonmedical prescription opioid use. Results are offered as follows: unadjusted in all participants; unadjusted among marijuana users only; adjusted for age, sex and ethnicity (model 1); additionally adjusted for income, education and health insurance status (model 2); then additionally adjusted for the other primary predictors and alcohol and smoking status (model 3). Adjusted models are presented for marijuana users only because no non-users of marijuana reported year 15 nonmedical prescription opioid use. All models adjust for study site. Because the number of outcomes was small (n = 23), we were concerned that calculated standard errors for each coefficient would not accurately represent the true level of confidence we had in our results. We, therefore, estimated coefficients, 95% confidence intervals and p-values using bootstrap methods with 1000 repetitions.
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
173
To identify subgroups of participants at especially high or low risk for initiating nonmedical prescription opioid use, we used unguided recursive partitioning, a procedure that produces a “classification tree” by splitting a dataset into increasingly homogeneous subsets (Breiman et al., 1984). Because the outcome was rare, we set the cost of misclassifying one of the rare outcomes much higher than the cost for misclassifying one of the non-outcomes, testing the range of misclassification cost ratios between 2:1 and 1000:1. We also set the threshold complexity parameter at zero in order to generate a deep tree, and then determined the final size of the tree using cost-complexity pruning and cross-validation (Breiman et al., 1984). All analyses were performed using Stata version 8 (Stata Corporation, College Station, Texas) except the recursive partitioning, which was performed using R (The Comprehensive R Archive Network, http://cran.r-project.org).
among participants with severe depressive symptoms, but this difference was not statistically robust (p = 0.38). Of the 36 participants who reported medical use of prescription opioid analgesics at year 10, two subsequently reported initiation of nonmedical prescription opioid use during the next 5 years, yielding an incidence of 5.6% (exact 95%CI: 0.7–19%), eight-fold greater than the 0.7% observed among all other participants (exact p = 0.027). Both participants were previous medical users of hydrocodone (Table 2).
3. Results
Marijuana use universally preceded nonmedical prescription opioid use, so non-users of marijuana (n = 949) were excluded from all adjusted analyses. Among marijuana users, past amphetamine and cocaine use were both strongly associated with subsequent nonmedical prescription opioid use. Though the association with cocaine mostly disappeared with adjustment for other substance use, it is difficult to determine if cocaine or amphetamine use is truly more strongly associated with the outcome given the high incidence of co-abuse of these substances (79% of cocaine users also used amphetamines at baseline, p < 0.001) and the small number of outcomes. Amphetamine use and medical use of prescription opioids, which were not significantly correlated with each other, both remained strong predictors of future nonmedical prescription opioid use in the fully adjusted model (Table 3). Depressive symptoms were not
3.1. Study sample Of 3950 year 10 CARDIA examination participants, 3894 (99%) answered questions about medical use of prescribed medications and illegal drug use, 237 (6%) baseline opioid users were excluded (118 for heroin use and 119 for nonmedical prescription opioid use), and 3163 (86%) of the remaining 3657 eligible participants attended the year 15 exam and were included in our study sample. Eligible participants who did not follow up at year 15 (n = 494) were slightly younger (34 versus 35 years), more likely to be an African-American man (30% versus 20%), and slightly less likely to use illegal substances at year 10 (66% versus 70% overall, similar pattern for each substance) than average eligible participants. Participants included in our study sample were 28–40 years old (mean age 35.1 ± 3.6 years) at the year 10 examination, about half female (56%), and half African-American (47%). Only 36 participants were current users of opioids for treatment of pain. Though these participants may be systematically different than those who were not medical opioid users, the differences were not statistically significant given the small numbers. Past cocaine, amphetamine and especially marijuana use were common, as were depressive symptoms (Table 1). 3.2. Incidence of new nonmedical prescription opioid use Overall, 23 participants (0.7%; exact 95%CI: 0.4–1.0%) reported new nonmedical use of prescription opioid analgesics at the year 15 examination. These included participants misusing Dilaudid (n = 5), morphine (n = 14), Demerol (n = 9), hydrocodone (n = 1), oxycodone (n = 3), and methadone (n = 1). Some participants reported nonmedical use of more than one of these opioids (n = 9), including combinations of morphine/Demerol (n = 6), morphine/Dilaudid (n = 1), morphine/Dilaudid/Demerol (n = 1) and hydrocodone/oxycodone (n = 1). Nonmedical use appeared to be more common in EuropeanAmerican men (1.5%, exact 95%CI: 0.8–2.7%), and among participants who had previously reported use of illegal drugs at year 10 (1.0%, exact 95%CI: 0.7–1.5%). Of note, nonmedical prescription opioid use did not occur among any of the participants who had never used illegal drugs at year 10 (0%, exact 95%CI: 0–0.4%). Nonmedical use was slightly more common
3.3. Multivariable analyses
Table 1 Baseline characteristics of study sample, 1995–1996 Characteristicsa
Baseline medical use of opioids No (n = 3127)
Age, mean years ±S.D. 35 ± 3.6 Sex, n (%) female 1757 (56) Ethnicity, n (%) 1466 (47) African-American Education, n (%) college 1417 (45) graduates Income, n (%) ≥$50,000/year 1268 (41) Health insurance, n (%) 2562 (85) insured Tobacco smoking, n (%) 690 (22) current Alcohol use, n (%) ≥1 1642 (53) drinks/week Marijuana use, n (%) ever 2180 (60) Cocaine use, n (%) ever 979 (31) Amphetamine use, n (%) ever 675 (22) Depressive symptoms, n (%) 643 (21) CES-D > 15
p-Value†
Yes (n = 36) 36 ± 4.0 25 (69) 19 (53)
0.43 0.13 0.51
13 (36)
0.31
11 (31) 34 (94)
0.24 0.16
12 (33)
0.11
15 (42)
0.24
27 (75) 9 (25) 9 (25) 10 (28)
0.59 0.47 0.68 0.30
CES-D: Center for Epidemiologic Studies-Depression Scale. Data from participants in the Coronary Artery Risk Development in Young Adults (CARDIA) study located in Birmingham (AL), Chicago (IL), Minneapolis (MN), and Oakland (CA) 1995–1996. a Note that 29 persons are missing information about one or more of the following pieces of information: education (n = 5), family income (n = 19), health insurance status (n = 7), smoking status (n = 10), or alcohol consumption (n = 5). Also, education, income, smoking and tobacco use have been dichotomized for this table only. † p-Value refers to a Fisher’s exact test, or a t-test (for age).
174
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
independently associated with the outcome, and neither were alcohol use or tobacco smoking. European-American men were more likely than other sex/ethnic groups to initiate nonmedical prescription opioid use overall (OR 3.3; 95%CI: 1.3–8.3%, p = 0.010), in marijuana users (OR 2.9; 95%CI: 1.2–7.0%, p = 0.017), and after full adjustment (OR 3.0; 95%CI: 1.2–7.9%, p = 0.022).
Table 2 Incidence of new nonmedical prescription opioid use Characteristics at year 10, 1995–1996
5-Year incidence of new nonmedical prescription opioid use p-Value‡
n/Na
%
Exact 95%CIb
23/3163
0.7
0.4–1.0%
Medical use of prescription opioids None 21/3127 Anyc 2/36 Hydrocodone 2/18 Codeine 0/13 Propoxyphene 0/5 Oxycodone 0/1
0.7 5.6 11 0.0 0.0 0.0
0.4–1.0% 0.7–19% 1.4–35% 0–25% 0–52% –
0.027 0.007 1 1 1
0.0 1.0 1.0 2.9 2.1 2.7
0–0.4% 0.7–1.5% 0.7–1.6% 1.8–4.5% 1.3–3.3% 1.1–5.4%
<0.001 <0.001 <0.001 <0.001 <0.001
0.5 0.9 0.5 1.3
0.2–1.1% 0.4–1.6% 0.06–1.7% 0.3–3.7%
Overall
Illegal drug use None Anyd Marijuana Amphetamines Powdered cocaine Crack cocaine
0/940 23/2223 23/2207 20/684 19/892 7/262
Depressive symptoms, CES-D score 0–7 7/1361 8–15 10/1119 16–23 2/417 24+ 3/236
3.4. Exploratory subgroup analysis and recursive partitioning
0.38
CES-D: Center for Epidemiologic Studies-Depression Scale. Data from participants in the Coronary Artery Risk Development in Young Adults (CARDIA) study located in Birmingham (AL), Chicago (IL), Minneapolis (MN), and Oakland (CA) 1995–2000. a The denominator is the number of participants in the subgroup defined by the row heading; the numerator is the number of these participants who reported new nonmedical prescription opioid use at year 15. b Exact 95%CI’s calculated using a binomial distribution. c Note that two persons used more than one of the listed medical opioids. d Note that many persons used more than one illegal substance, and that persons reporting heroin use at year 10 were excluded from this and subsequent analyses. ‡ p-Values refer to Fisher’s exact test (e.g. marijuana users vs. marijuana nonusers).
Using the strongest predictors from the logistic regression model, we identified four subgroups with 5-year risk estimates for incident nonmedical prescription opioid use that differed by orders of magnitude: 0% (exact 95%CI: 0–0.4%) for never users of marijuana; 0.2% (exact 95%CI: 0.04–0.6%) for past marijuana but never amphetamine users; 2.7% (exact 95%CI: 1.6–4.2%) for past marijuana and amphetamine users but without an opioid prescription; 25% (exact 95%CI: 3.2–65%) for past marijuana and amphetamine users who were prescribed an opioid. Unsupervised recursive partitioning consistently identified year 10 amphetamine use as the single best predictor of future nonmedical prescription opioid use. If clinicians were to use past amphetamine use as a test for identifying participants who will go on to use prescription opioids for nonmedical purposes in the next 5 years, we estimate such a test would have a sensitivity of 87% (exact 95%CI: 66–97%) and a specificity of 79% (exact 95%CI: 77–80%). Our sample includes too few participants to directly estimate the excess risk associated with prescribing opioids to different subgroups of patients. By assuming independence of associations and applying Bayesian analysis, however, we calculated these risks indirectly and estimate that the risk of future nonmedical prescription opioid use among amphetamine users prescribed opioids is approximately 20%, and among non-amphetamine users prescribed opioids is less than 1%.
Table 3 Multivariable predictors of new nonmedical prescription opioid use Characteristics at year 10, 1995–1996
Odds ratio for new nonmedical prescription opioid use at year 15, 2000–2001 (95%CI) Unadjusted, all participantsa
Medical use of prescription opioids Marijuana use Amphetamines use Cocaine use Depressive symptoms, CES-D score (16+)
8.6 (2.5–30) Cannot calculatef 24 (6.9–83) 10 (3.0–35) 1.1 (0.3–3.4)
Among marijuana users only (n = 2185)b Unadjusted
Model 1c
Model 2d
Model 3e
8.1 (2.3–29) – 15 (4.3–51) 6.0 (1.9–19) 1.1 (0.3–3.6)
10 (2.7–40) – 16 (4.0–65) 5.5 (1.7–19) 1.3 (0.4–4.2)
11 (2.2–59) – 15 (3.8–64) 5.2 (1.6–17) 0.9 (0.2–3.1)
15 (1.5–153) – 12 (1.9–71) 1.7 (0.4–6.7) 1.5 (0.3–7.6)
CES-D: Center for Epidemiologic Studies-Depression Scale. Data from participants in the Coronary Artery Risk Development in Young Adults (CARDIA) study located in Birmingham (AL), Chicago (IL), Minneapolis (MN), and Oakland (CA) 1995-2000. CI’s are based on the Wald test for regression coefficients (not exact). a Analysis restricted to participants with complete data on all covariates (n = 3134; n = 29 excluded). b Because no events occurred among participants not using marijuana, non-users (n = 949) were excluded from regression models. c Model 1 estimates are adjusted for age, sex and race. d Model 2 estimates are additionally adjusted for social factors including income, education and health insurance status. e Model 3 is fully adjusted, and includes the other primary predictors in left column as well as alcohol and smoking status. f No events occurred among non-marijuana users, so the odds ratio cannot be calculated (denominator would be 0).
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
4. Discussion In this analysis, we directly measured the 5-year incidence of initiating nonmedical prescription opioid use among young adults aged 28–40 participating in The CARDIA study from 1995 to 2000. Overall, the incidence was very low (0.7%), did not occur among participants without a history of marijuana use, and was more common among participants who had been prescribed an opioid analgesic by their physician. Past use of amphetamines, reported by 20 of the 23 participants who went on to initiate nonmedical prescription opioid use, was the single best indicator of future nonmedical prescription opioid use in our sample. Depressive symptoms were not associated with future nonmedical prescription opioid use. Our study was limited in several ways. Nonmedical use of prescription opioids is not as common in this age group as in younger age groups (<25 years). Given the rarity of the outcome, we observed incident nonmedical prescription opioid use only 23 times over 5 years in our sample of 3163 eligible participants. We were able to demonstrate some important, large associations with statistical confidence, but could not confidently exclude more subtle associations, or directly estimate interactions between medical opioid use and other factors. Our attempt to identify independent predictors through multivariable analyses must also be viewed with caution given the small number of outcomes. Also, our data are based on self-report, which is known to under-estimate actual substance use behavior (Harrison and Hughes, 1997). Underreporting may be due to social stigmatization and privacy concerns, or confusion about the jargon or conceptual information in a questionnaire (Morral et al., 2003), either of which may have occurred. On the other hand, CARDIA participants used self-administered non-verbal questionnaires (better than verbal reporting (Harrison and Hughes, 1997)), and were comfortable enough to report use of marijuana, and also cocaine and amphetamine (two highly stigmatized drugs) with high frequency. Misunderstanding of the term “opiates” by some participants, however, is more likely, and presumably led to some underreporting and downward bias in our incidence estimates. The prospective nature of our study is an important strength. Previous estimates of incidence have been derived indirectly from self-reported age of initiation among nonmedical users from cross-sectional studies such as in the 2004 National Survey on Drug Use and Health (NSDUH, formerly the National Household Survey on Drug Abuse (NHSDA)), which reports an annual nonmedical prescription opioid use incidence of 0.3%/year (1.5%/5 years) among adults over age 25 (US, 2005a, 2005b, 2005c, 2005d). Our estimate is a bit lower (0.7%/5 years), but does not include former heroin users and reflects a direct, prospective measure of incidence among young- to middle-aged adults (age 28–40) that may be useful to practitioners or policymakers. Our prospective data are also useful for identifying antecedents of nonmedical prescription opioid use. Our results confirm that illegal drug use often (in our study invariably) precedes nonmedical prescription opioid use, as suggested by cross-sectional studies of college students (McCabe et al., 2005a,
175
2005b, 2005c), high-school seniors (McCabe et al., 2005a, 2005b, 2005c), and the US population (2001 NHSDA) (US, 2002; Dowling et al., 2004). They also begin to provide clarity to the issue of whether or not treatment of pain predisposes to prescription opioid addiction. Previous studies of pain and subsequent opioid addiction have mostly been uncontrolled case series studies of patients treated for pain, showing low incidence of subsequent nonmedical use (Medina and Diamond, 1977; Perry and Heidrich, 1982; Portenoy and Foley, 1986; Porter and Jick, 1980), or opioid addiction, showing a high prevalence of previous opioid treatment for pain (Potter et al., 2004; Rayport, 1954). Our results help resolve this apparent contradiction by finding modest evidence of an association between opioid treatment for pain and subsequent nonmedical use in a prospective study. Our prospective study confirms results from a recent cross-sectional study of college students (McCabe et al., 2005a, 2005b, 2005c) with a study design less prone to prevalence-incidence, effect-cause, and length-time bias, and helps to fill a gap in the literature that has been widely noted (Klein, 2004; Passik and Kirsh, 2004; Von Korff and Deyo, 2004). We did not find an association with previous depressive symptoms, but this result must be viewed as less definitive given the limited power we had to detect small associations with our sample size. In summary, we found the incidence of nonmedical prescription opioid use in this age group (28–40 years) to be quite low overall, less than 1% in 5 years. The incidence is higher among patients prescribed opioid medications by their physicians (about 6% in 5 years). While pain may sometimes result from opioid use (Ballantyne and Mao, 2003), our findings suggest that pain, and access to opioids provided by a physician, at least sometimes precedes nonmedical opioid use, supporting the logical presupposition that providing access to an addictive drug may increase the frequency of misuse. On the other hand, only a small minority of patients treated for pain in our study went on to use prescription opioids for nonmedical purposes, and this was always preceded by use of other illegal substances (particularly amphetamines). Physicians considering prescribing opioid medications for pain should be generally reassured by our results, and even when considering a prescription for a higher-risk patient, the risk of nonmedical use and addiction should be balanced against the harm caused by inadequate pain control. Future studies should address specific prescribing practices, alternate opioid formulations (Grudzinskas et al., 2006), and other ways of optimizing pain management that minimize the harm caused by the addictive potential of this otherwise useful class of medications. Acknowledgements The CARDIA study is supported by contracts N01-HC48047, N01-HC-48048, N01-HC-48049, N01-HC-48050, and N01-HC-95095 from the National Heart, Lung, and Blood Institute, which provides scientific review of all CARDIA manuscripts. Dr. Kertesz received general support from the National Institute on Drug Abuse (K23-DA-15487). Neither funder played any role in design or execution of this analysis.
176
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
References Ballantyne, J.C., Mao, J., 2003. Opioid therapy for chronic pain. N. Engl. J. Med. 349, 1943–1953. Breiman, L., Friedman, J.H., Olshen, R.A., Stone, C.J., 1984. Classification and Regression Trees. Chapman and Hall/CRC Press, Boca Raton. Dowling K., Storr C., Chilcoat H., 2004. Who’s misusing analgesics in the general population of the US? [Abstract]. Presented at the 2004 Annual Scientific Meeting of the College on Problems of Drug Dependence. Friedman G.D., Cutter G.R., Donahue R., Hughes G.H., Hulley S., Jacobs D.R.J., Liu K., P.J. S., 1988. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol. 41, 1105– 1116. Grudzinskas C., Balster R.L., Gorodetzky C.W., Griffiths R.R., Henningfield J.E., Johanson C.E., Mansbach R.S., McCormick C.G., Schnoll S.H., Strain E.C., Wright C., 2006. Impact of formulation on the abuse liability, safety and regulation of medications: the expert panel report. Drug Alcohol Depend. Epub ahead of print, March 29. Hall, W., Lynskey, M., Degenhardt, L., 2000. Trends in opiate-related deaths in the United Kingdom and Australia, 1985–1995. Drug Alcohol Depend. 57, 247–254. Hao, W., Xiao, S., Liu, T., Young, D., Chen, S., Zhang, D., Li, C., Shi, J., Chen, G., Yang, K., 2002. The second National Epidemiological Survey on illicit drug use at six high-prevalence areas in China: prevalence rates and use patterns. Addiction 97, 1305–1315. Harrison, L., Hughes, A., 1997. The Validity of Self-Reported Drug Use: Improving the Accuracy of Survey Estimates. National Institute on Drug Abuse. Hughes, G.H., Cutter, G.R., Donahue, R., Friedman, G.D., Hulley, S., Hunkeler, E., Jacobs, D.R.J., Liu, K., Orden, S., 1987. Recruitment in the Coronary Artery Disease Risk Development in Young Adults (CARDIA) study. Control Clin. Trials. 8, 68S–73S. Klein, M.J., 2004. Opioid therapy for chronic pain. N. Engl. J. Med. 350, 840–842, author reply 840–842. Mattoo, S.K., Basu, D., Sharma, A., Balaji, M., Malhotra, A., 1997. Abuse of codeine-containing cough syrups: a report from India. Addiction 92, 1783–1787. McCabe, S.E., Boyd, C.J., Teter, C.J., 2005a. Illicit use of opioid analgesics by high school seniors. J. Subst. Abuse Treat. 28, 225–230. McCabe, S.E., Teter, C.J., Boyd, C.J., 2005b. Illicit use of prescription pain medication among college students. Drug Alcohol Depend. 77, 37–47. McCabe, S.E., Teter, C.J., Boyd, C.J., Knight, J.R., Wechsler, H., 2005c. Nonmedical use of prescription opioids among U.S. college students: prevalence and correlates from a national survey. Addict. Behav. 30, 789– 805. Medina, J.L., Diamond, S., 1977. Drug dependency in patients with chronic headaches. Headache 17, 12–14. Morral, A.R., McCaffrey, D.F., Chien, S., 2003. Measurement of adolescent drug use. J. Psychoact. Drugs 35, 301–309. Naughton, M.J., Wiklund, I., 1993. A critical review of dimension-specific measures of health-related quality of life in cross-cultural research. Qual. Life Res. 2, 397–432. Passik, S.D., Kirsh, K.L., 2003. The need to identify predictors of aberrant drugrelated behavior and addiction in patients being treated with opioids for pain. Pain Med. 4, 186–189.
Passik, S.D., Kirsh, K.L., 2004. Opioid therapy in patients with a history of substance abuse. CNS Drugs 18, 13–25. Perry, S., Heidrich, G., 1982. Management of pain during debridement: a survey of U.S. burn units. Pain 13, 267–280. Portenoy, R.K., Foley, K.M., 1986. Chronic use of opioid analgesics in nonmalignant pain: report of 38 cases. Pain 25, 171–186. Porter, J., Jick, H., 1980. Addiction rare in patients treated with narcotics. N. Engl. J. Med. 302, 123. Potter, J.S., Hennessy, G., Borrow, J.A., Greenfield, S.F., Weiss, R.D., 2004. Substance use histories in patients seeking treatment for controlled-release oxycodone dependence. Drug Alcohol Depend. 76, 213–215. Radloff, L., 1977. The CES-D scale: a self-report depression scale for research in the general population. Appl. Psychol. Measure. 1, 385–401. Rayport, M., 1954. Experience in the management of patients medically addicted to narcotics. J. Am. Med. Assoc. 156, 684–691. Shah, R., Uren, Z., Baker, A., Majeed, A., 2001. Trends in deaths from drug overdose and poisoning in England and Wales 1993–1998. J. Public Health Med. 23, 242–246. Simoni-Wastila, L., Ritter, G., Strickler, G., 2004. Gender and other factors associated with the nonmedical use of abusable prescription drugs. Subst. Use Misuse. 39, 1–23. United States. Substance Abuse and Mental Health Services Administration. The NHSDA report: nonmedical use of prescription-type drugs among youths and young adults. http://www.oas.samhsa.gov/2k3/ prescription/prescription.htm, (accessed on January 11, 2006). United States. Substance Abuse and Mental Health Services Administration. The DASIS report: treatment admissions involving narcotic painkillers. http://www.oas.samhsa.gov/2k3/painTX/painTX.htm, (accessed on January 11, 2006). United States. Substance Abuse and Mental Health Services Administration. The DAWN report: narcotic analgesics, 2002 update. http://oas.samhsa.gov/2k4analgesics.pdf, (accessed on January 11, 2006). United States. Substance Abuse and Mental Health Services Administration. The NSDUH report: nonmedical use of prescription pain relievers. http:// www.oas.samhsa.gov/2k4/pain/pain.pdf, (accessed on January 11, 2006). United States Substance Abuse and Mental Health Services Administration, 2004. National survey on drug use and health: dependence, abuse and treatment tables. http://oas.samhsa.gov/NSDUH/2k4nsduh/ 2k4tabs/LOTSect5pe.htm#TopOfPage, (accessed on January 11, 2006). United States Substance Abuse and Mental Health Services Administration, 2004. National survey on drug use and health: incidence tables. http://oas. samhsa.gov/NSDUH/2k4nsduh/2k4tabs/LOTSect4pe.htm#TopOfPage, (accessed on January 11, 2006). United States Substance Abuse and Mental Health Services Administration, 2004. National survey on drug use and health: prevalence tables. http://oas.samhsa.gov/NSDUH/2k4nsduh/2k4tabs/LOTSect1pe.htm# TopOfPage, (accessed on January 11, 2006). Von Korff, M., Deyo, R.A., 2004. Potent opioids for chronic musculoskeletal pain: flying blind? Pain 109, 207–209. Zacny, J., Bigelow, G., Compton, P., Foley, K., Iguchi, M., Sannerud, C., 2003. College on Problems of Drug Dependence taskforce on prescription opioid non-medical use and abuse: position statement. Drug Alcohol Depend. 69, 215–232.
Incidence and antecedents of nonmedical prescription opioid use in four US communities The Coronary Artery Risk Development in Young Adults (CARDIA) prospective cohort study Mark J. Pletcher a,b,∗ , Stefan G. Kertesz c,e , Stephen Sidney d , Catarina I. Kiefe c,e , Stephen B. Hulley a a
Department of Epidemiology and Biostatistics, University of California, San Francisco, 185 Berry Street, Suite 5700, San Francisco, CA 94107, United States b Division of General Internal Medicine, University of California, San Francisco, CA 94143, United States c Division of Preventive Medicine, University of Alabama at Birmingham, AL 35205, United States d Division of Research, Kaiser Permanente, Oakland, CA 94612, United States e Deep South Center on Effectiveness, Veteran’s Affairs Medical Center, Birmingham, AL 35205, United States Received 18 January 2006; received in revised form 11 April 2006; accepted 14 April 2006
Abstract Background: Nonmedical use of prescription opioids has emerged as a major public health problem during the last decade, but direct measures of incidence and predisposing factors are lacking. Methods: We prospectively measured incidence and antecedents of nonmedical prescription opioid use in The Coronary Artery Risk Development in Young Adults study among 28–40-year-old African- and European-American men and women with no prior history of nonmedical opioid use. Results: Among 3163 participants, 23 reported new nonmedical prescription opioid use in 2000–2001 (5-year incidence 0.7%; 95%CI: 0.4–1.0%). All 23 had previously reported marijuana use (p < 0.001). Five-year incidence was significantly higher among European-American men (OR = 3.3; 95%CI: 1.3–8.3), and among participants reporting a history of amphetamine use (OR = 24; 95%CI: 6.9–83) or medical opioid use for treatment of pain (OR = 8.6; 95%CI: 2.5–30). These associations remained strong when examined among marijuana users and after adjusting for demographics, social factors, and other antecedent substance use. Amphetamine use was the best single predictor of future nonmedical use (sensitivity 87%, specificity 79%). Conclusions: Initiation of nonmedical prescription opioid use is generally rare in 28–40-year-old adults, but is observed to be more common with a previous history of substance abuse and legal access to opioids through prescription by a physician. © 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Opioid-related disorders; Pain; Depression; Substance-related disorders
1. Introduction Nonmedical use of prescription opioid analgesics has emerged as a major public health problem during the last decade (Zacny et al., 2003). Cross-sectional studies in the United States (US) conducted by the Substance Abuse and Mental Health Services Administration (SAMHSA) show a tripling in the rate of initiation on nonmedical prescription
∗
Corresponding author. Tel.: +1 415 514 8008; fax: +1 415 514 8150. E-mail address: [email protected] (M.J. Pletcher).
0376-8716/$ – see front matter © 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.drugalcdep.2006.04.011
opioid use among young adults since 1990 (US, 2005a, 2005b, 2005c, 2005d) and concomitant increases in attributable emergency department visits (US, 2004a, 2004b, 2004c) and substance abuse treatment admissions (US, 2003), both of which have doubled during the last decade. More than 2 million US adults start using prescription opioids for nonmedical purposes every year (US, 2004a, 2004b, 2004c; US, 2005a, 2005b, 2005c, 2005d), and prescription opioid dependence is now more than four times as prevalent as heroin dependence among the non-institutionalized US population sampled in the National Survey on Drug Use and Health (US, 2005a, 2005b, 2005c, 2005d). Nonmedical use of prescription opioids appears
172
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
to be an increasing problem outside the US as well (Hall et al., 2000; Hao et al., 2002; Mattoo et al., 1997; Shah et al., 2001). While these national surveys have helped to identify and describe this emerging phenomenon, cross-sectional data are less useful in estimating incidence or in identifying antecedents of developing disease. For example, nonmedical prescription opioid use has been associated with mental illness (SimoniWastila et al., 2004) and use of illegal substances (Dowling et al., 2004), but published data do not show which problem typically comes first. Similarly, it has been difficult to determine the degree to which prescribing opioids for pain increases the risk for nonmedical prescription opioid use. Previous research (mostly uncontrolled case series (Medina and Diamond, 1977; Perry and Heidrich, 1982; Portenoy and Foley, 1986; Porter and Jick, 1980)) suggests that patients treated for pain are at low risk for developing opioid addiction (Ballantyne and Mao, 2003; Passik and Kirsh, 2004), but surveys of opioid-addicted patients (Potter et al., 2004; Rayport, 1954) suggest that opioid treatment for pain often predates addiction to heroin. Longitudinal data are needed to estimate directly how often patients start misusing prescription opioids, and to identify which putative risk factors precede, rather than result from, the development of nonmedical prescription opioid use (Passik and Kirsh, 2003) in order to inform the efforts of both public health officials and individual physicians trying to reduce the potential harm from addiction caused by these otherwise useful pain medications. The Coronary Artery Risk Development in Young Adults (CARDIA) study measured substance abuse, mental health, and use of prescribed medications in a large community-based cohort of young black and white men and women in 1985–1986 and at five subsequent examinations. We identified a cohort of young adults aged 28–40 who in 1995–1996 denied misusing opioids of any kind, and characterized incidence and antecedents of new, self-reported nonmedical prescription opioid use 5 years later. 2. Methods 2.1. Study design and sample CARDIA is a longitudinal study of risk factors for coronary artery disease in a community-based cohort of European- and African-American women and men (n = 5115) aged 18–30 years and healthy at the time of enrollment in 1985–1986. With the informed consent of participants and the approval of Institutional Review Boards at each site (Oakland, Chicago, Minneapolis, and Birmingham), participants underwent a baseline examination and follow-up examinations at years 2, 5, 7, 10, and 15, with 74% retention at year 15 (2000–2001). Details of the study design, recruitment and procedures have been published (Friedman et al., 1988; Hughes et al., 1987). For this investigation, we identified CARDIA participants who reported no nonmedical opioid use (heroin or prescription) on a substance use questionnaire at year 10 (1995–1996), and measured associations between medical use of opioids, illegal substance use, and depression at year 10 with new self-reported nonmedical prescription opioid use at year 15.
2.2. Primary predictors Participants filled out a substance use questionnaire at each examination asking if they had ever used “marijuana”, “cocaine or crack”, “amphetamines (‘Speed’ or ‘Uppers’)”, and “opiates (Heroin, Dilaudid, Morphine, Demerol)” for nonmedical reasons. In order to capture only new nonmedical opioid users
at year 15, participants who reported nonmedical opioid use of any sort on this questionnaire at year 10 were excluded from the analysis. Among the remaining participants, we tested whether measuring past use of marijuana, cocaine and/or amphetamines might help predict new nonmedical prescription opioid use 5 years later. Note that all references to “amphetamine use” refer to nonmedical amphetamine use. We identified participants who were receiving opioids for medical purposes at year 10 through manual review of prescription medications reported on a nonsubstance abuse-oriented questionnaire concerning medications prescribed by a physician. We found participants using propoxyphene (n = 7), codeine (with and without acetaminophen, n = 19), hydrocodone (n = 27), oxycodone (n = 2), and methadone (n = 1). Participants reporting any of these medications on their prescription medication list were deemed to be “medical users” of prescription opioids at year 10. Depressive symptoms were measured at the year 10 examination using the Center for Epidemiologic Studies-Depression (CES-D) scale, a 0–60 point scale for which scores of 16 or greater are considered to indicate significant depressive symptomatology (Naughton and Wiklund, 1993; Radloff, 1977). We categorized the CES-D score into four groups – 0–7, 8–15, 16–23, and ≥24 – for purposes of adjustment, but used the standard cutoff score of ≥16 as our primary depressive symptoms predictor.
2.3. Outcome: nonmedical prescription opioid use at year 15 The substance use questionnaire (described above) was also administered in year 15, 5 years after the year 10 exam that serves as the baseline for this analysis. Participants in our sample who newly answered yes to, “Have you ever used opiates (Heroin, Dilaudid, Morphine, Demerol) for nonmedical reasons?”, and indicated use of either Dilaudid, morphine, Demerol, or an “Other (specify)” opioid that we identified as a prescription opioid (not heroin, opium, etc.) were considered to be incident cases of nonmedical prescription opioid use.
2.4. Other covariates Sex, ethnicity and birthday were self-reported at baseline. Educational grade attained (no college, some college, college graduate), family income (<$25,000, $25,000–49,000, ≥$50,000/year), and health insurance were self-reported at the year 10 examination. Smoking at year 10 was categorized as never, past or current, and alcohol consumption was categorized as 0, 1–6, 7–13, or ≥14 drinks/week. Participants with missing data on income, education, health insurance status, alcohol or smoking (n = 29) were excluded from multivariable analyses. These participants were not statistically different from included participants in terms of demographics and substance abuse, and none developed the outcome at year 15 (exact p = 1.0).
2.5. Statistical analysis We described the study sample at baseline in reference to use or non-use of opioid medications for treatment of pain at baseline, testing for bivariate associations using Fisher’s exact tests. We then measured the 5-year cumulative incidence of initiating nonmedical prescription opioid use, and calculated exact 95% confidence intervals using the binomial distribution. We compared incidence in subgroups of participants defined by our primary predictors and tested differences with Fisher’s exact tests. We used logistic regression to measure the association between each primary predictor and year 15 nonmedical prescription opioid use. Results are offered as follows: unadjusted in all participants; unadjusted among marijuana users only; adjusted for age, sex and ethnicity (model 1); additionally adjusted for income, education and health insurance status (model 2); then additionally adjusted for the other primary predictors and alcohol and smoking status (model 3). Adjusted models are presented for marijuana users only because no non-users of marijuana reported year 15 nonmedical prescription opioid use. All models adjust for study site. Because the number of outcomes was small (n = 23), we were concerned that calculated standard errors for each coefficient would not accurately represent the true level of confidence we had in our results. We, therefore, estimated coefficients, 95% confidence intervals and p-values using bootstrap methods with 1000 repetitions.
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
173
To identify subgroups of participants at especially high or low risk for initiating nonmedical prescription opioid use, we used unguided recursive partitioning, a procedure that produces a “classification tree” by splitting a dataset into increasingly homogeneous subsets (Breiman et al., 1984). Because the outcome was rare, we set the cost of misclassifying one of the rare outcomes much higher than the cost for misclassifying one of the non-outcomes, testing the range of misclassification cost ratios between 2:1 and 1000:1. We also set the threshold complexity parameter at zero in order to generate a deep tree, and then determined the final size of the tree using cost-complexity pruning and cross-validation (Breiman et al., 1984). All analyses were performed using Stata version 8 (Stata Corporation, College Station, Texas) except the recursive partitioning, which was performed using R (The Comprehensive R Archive Network, http://cran.r-project.org).
among participants with severe depressive symptoms, but this difference was not statistically robust (p = 0.38). Of the 36 participants who reported medical use of prescription opioid analgesics at year 10, two subsequently reported initiation of nonmedical prescription opioid use during the next 5 years, yielding an incidence of 5.6% (exact 95%CI: 0.7–19%), eight-fold greater than the 0.7% observed among all other participants (exact p = 0.027). Both participants were previous medical users of hydrocodone (Table 2).
3. Results
Marijuana use universally preceded nonmedical prescription opioid use, so non-users of marijuana (n = 949) were excluded from all adjusted analyses. Among marijuana users, past amphetamine and cocaine use were both strongly associated with subsequent nonmedical prescription opioid use. Though the association with cocaine mostly disappeared with adjustment for other substance use, it is difficult to determine if cocaine or amphetamine use is truly more strongly associated with the outcome given the high incidence of co-abuse of these substances (79% of cocaine users also used amphetamines at baseline, p < 0.001) and the small number of outcomes. Amphetamine use and medical use of prescription opioids, which were not significantly correlated with each other, both remained strong predictors of future nonmedical prescription opioid use in the fully adjusted model (Table 3). Depressive symptoms were not
3.1. Study sample Of 3950 year 10 CARDIA examination participants, 3894 (99%) answered questions about medical use of prescribed medications and illegal drug use, 237 (6%) baseline opioid users were excluded (118 for heroin use and 119 for nonmedical prescription opioid use), and 3163 (86%) of the remaining 3657 eligible participants attended the year 15 exam and were included in our study sample. Eligible participants who did not follow up at year 15 (n = 494) were slightly younger (34 versus 35 years), more likely to be an African-American man (30% versus 20%), and slightly less likely to use illegal substances at year 10 (66% versus 70% overall, similar pattern for each substance) than average eligible participants. Participants included in our study sample were 28–40 years old (mean age 35.1 ± 3.6 years) at the year 10 examination, about half female (56%), and half African-American (47%). Only 36 participants were current users of opioids for treatment of pain. Though these participants may be systematically different than those who were not medical opioid users, the differences were not statistically significant given the small numbers. Past cocaine, amphetamine and especially marijuana use were common, as were depressive symptoms (Table 1). 3.2. Incidence of new nonmedical prescription opioid use Overall, 23 participants (0.7%; exact 95%CI: 0.4–1.0%) reported new nonmedical use of prescription opioid analgesics at the year 15 examination. These included participants misusing Dilaudid (n = 5), morphine (n = 14), Demerol (n = 9), hydrocodone (n = 1), oxycodone (n = 3), and methadone (n = 1). Some participants reported nonmedical use of more than one of these opioids (n = 9), including combinations of morphine/Demerol (n = 6), morphine/Dilaudid (n = 1), morphine/Dilaudid/Demerol (n = 1) and hydrocodone/oxycodone (n = 1). Nonmedical use appeared to be more common in EuropeanAmerican men (1.5%, exact 95%CI: 0.8–2.7%), and among participants who had previously reported use of illegal drugs at year 10 (1.0%, exact 95%CI: 0.7–1.5%). Of note, nonmedical prescription opioid use did not occur among any of the participants who had never used illegal drugs at year 10 (0%, exact 95%CI: 0–0.4%). Nonmedical use was slightly more common
3.3. Multivariable analyses
Table 1 Baseline characteristics of study sample, 1995–1996 Characteristicsa
Baseline medical use of opioids No (n = 3127)
Age, mean years ±S.D. 35 ± 3.6 Sex, n (%) female 1757 (56) Ethnicity, n (%) 1466 (47) African-American Education, n (%) college 1417 (45) graduates Income, n (%) ≥$50,000/year 1268 (41) Health insurance, n (%) 2562 (85) insured Tobacco smoking, n (%) 690 (22) current Alcohol use, n (%) ≥1 1642 (53) drinks/week Marijuana use, n (%) ever 2180 (60) Cocaine use, n (%) ever 979 (31) Amphetamine use, n (%) ever 675 (22) Depressive symptoms, n (%) 643 (21) CES-D > 15
p-Value†
Yes (n = 36) 36 ± 4.0 25 (69) 19 (53)
0.43 0.13 0.51
13 (36)
0.31
11 (31) 34 (94)
0.24 0.16
12 (33)
0.11
15 (42)
0.24
27 (75) 9 (25) 9 (25) 10 (28)
0.59 0.47 0.68 0.30
CES-D: Center for Epidemiologic Studies-Depression Scale. Data from participants in the Coronary Artery Risk Development in Young Adults (CARDIA) study located in Birmingham (AL), Chicago (IL), Minneapolis (MN), and Oakland (CA) 1995–1996. a Note that 29 persons are missing information about one or more of the following pieces of information: education (n = 5), family income (n = 19), health insurance status (n = 7), smoking status (n = 10), or alcohol consumption (n = 5). Also, education, income, smoking and tobacco use have been dichotomized for this table only. † p-Value refers to a Fisher’s exact test, or a t-test (for age).
174
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
independently associated with the outcome, and neither were alcohol use or tobacco smoking. European-American men were more likely than other sex/ethnic groups to initiate nonmedical prescription opioid use overall (OR 3.3; 95%CI: 1.3–8.3%, p = 0.010), in marijuana users (OR 2.9; 95%CI: 1.2–7.0%, p = 0.017), and after full adjustment (OR 3.0; 95%CI: 1.2–7.9%, p = 0.022).
Table 2 Incidence of new nonmedical prescription opioid use Characteristics at year 10, 1995–1996
5-Year incidence of new nonmedical prescription opioid use p-Value‡
n/Na
%
Exact 95%CIb
23/3163
0.7
0.4–1.0%
Medical use of prescription opioids None 21/3127 Anyc 2/36 Hydrocodone 2/18 Codeine 0/13 Propoxyphene 0/5 Oxycodone 0/1
0.7 5.6 11 0.0 0.0 0.0
0.4–1.0% 0.7–19% 1.4–35% 0–25% 0–52% –
0.027 0.007 1 1 1
0.0 1.0 1.0 2.9 2.1 2.7
0–0.4% 0.7–1.5% 0.7–1.6% 1.8–4.5% 1.3–3.3% 1.1–5.4%
<0.001 <0.001 <0.001 <0.001 <0.001
0.5 0.9 0.5 1.3
0.2–1.1% 0.4–1.6% 0.06–1.7% 0.3–3.7%
Overall
Illegal drug use None Anyd Marijuana Amphetamines Powdered cocaine Crack cocaine
0/940 23/2223 23/2207 20/684 19/892 7/262
Depressive symptoms, CES-D score 0–7 7/1361 8–15 10/1119 16–23 2/417 24+ 3/236
3.4. Exploratory subgroup analysis and recursive partitioning
0.38
CES-D: Center for Epidemiologic Studies-Depression Scale. Data from participants in the Coronary Artery Risk Development in Young Adults (CARDIA) study located in Birmingham (AL), Chicago (IL), Minneapolis (MN), and Oakland (CA) 1995–2000. a The denominator is the number of participants in the subgroup defined by the row heading; the numerator is the number of these participants who reported new nonmedical prescription opioid use at year 15. b Exact 95%CI’s calculated using a binomial distribution. c Note that two persons used more than one of the listed medical opioids. d Note that many persons used more than one illegal substance, and that persons reporting heroin use at year 10 were excluded from this and subsequent analyses. ‡ p-Values refer to Fisher’s exact test (e.g. marijuana users vs. marijuana nonusers).
Using the strongest predictors from the logistic regression model, we identified four subgroups with 5-year risk estimates for incident nonmedical prescription opioid use that differed by orders of magnitude: 0% (exact 95%CI: 0–0.4%) for never users of marijuana; 0.2% (exact 95%CI: 0.04–0.6%) for past marijuana but never amphetamine users; 2.7% (exact 95%CI: 1.6–4.2%) for past marijuana and amphetamine users but without an opioid prescription; 25% (exact 95%CI: 3.2–65%) for past marijuana and amphetamine users who were prescribed an opioid. Unsupervised recursive partitioning consistently identified year 10 amphetamine use as the single best predictor of future nonmedical prescription opioid use. If clinicians were to use past amphetamine use as a test for identifying participants who will go on to use prescription opioids for nonmedical purposes in the next 5 years, we estimate such a test would have a sensitivity of 87% (exact 95%CI: 66–97%) and a specificity of 79% (exact 95%CI: 77–80%). Our sample includes too few participants to directly estimate the excess risk associated with prescribing opioids to different subgroups of patients. By assuming independence of associations and applying Bayesian analysis, however, we calculated these risks indirectly and estimate that the risk of future nonmedical prescription opioid use among amphetamine users prescribed opioids is approximately 20%, and among non-amphetamine users prescribed opioids is less than 1%.
Table 3 Multivariable predictors of new nonmedical prescription opioid use Characteristics at year 10, 1995–1996
Odds ratio for new nonmedical prescription opioid use at year 15, 2000–2001 (95%CI) Unadjusted, all participantsa
Medical use of prescription opioids Marijuana use Amphetamines use Cocaine use Depressive symptoms, CES-D score (16+)
8.6 (2.5–30) Cannot calculatef 24 (6.9–83) 10 (3.0–35) 1.1 (0.3–3.4)
Among marijuana users only (n = 2185)b Unadjusted
Model 1c
Model 2d
Model 3e
8.1 (2.3–29) – 15 (4.3–51) 6.0 (1.9–19) 1.1 (0.3–3.6)
10 (2.7–40) – 16 (4.0–65) 5.5 (1.7–19) 1.3 (0.4–4.2)
11 (2.2–59) – 15 (3.8–64) 5.2 (1.6–17) 0.9 (0.2–3.1)
15 (1.5–153) – 12 (1.9–71) 1.7 (0.4–6.7) 1.5 (0.3–7.6)
CES-D: Center for Epidemiologic Studies-Depression Scale. Data from participants in the Coronary Artery Risk Development in Young Adults (CARDIA) study located in Birmingham (AL), Chicago (IL), Minneapolis (MN), and Oakland (CA) 1995-2000. CI’s are based on the Wald test for regression coefficients (not exact). a Analysis restricted to participants with complete data on all covariates (n = 3134; n = 29 excluded). b Because no events occurred among participants not using marijuana, non-users (n = 949) were excluded from regression models. c Model 1 estimates are adjusted for age, sex and race. d Model 2 estimates are additionally adjusted for social factors including income, education and health insurance status. e Model 3 is fully adjusted, and includes the other primary predictors in left column as well as alcohol and smoking status. f No events occurred among non-marijuana users, so the odds ratio cannot be calculated (denominator would be 0).
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
4. Discussion In this analysis, we directly measured the 5-year incidence of initiating nonmedical prescription opioid use among young adults aged 28–40 participating in The CARDIA study from 1995 to 2000. Overall, the incidence was very low (0.7%), did not occur among participants without a history of marijuana use, and was more common among participants who had been prescribed an opioid analgesic by their physician. Past use of amphetamines, reported by 20 of the 23 participants who went on to initiate nonmedical prescription opioid use, was the single best indicator of future nonmedical prescription opioid use in our sample. Depressive symptoms were not associated with future nonmedical prescription opioid use. Our study was limited in several ways. Nonmedical use of prescription opioids is not as common in this age group as in younger age groups (<25 years). Given the rarity of the outcome, we observed incident nonmedical prescription opioid use only 23 times over 5 years in our sample of 3163 eligible participants. We were able to demonstrate some important, large associations with statistical confidence, but could not confidently exclude more subtle associations, or directly estimate interactions between medical opioid use and other factors. Our attempt to identify independent predictors through multivariable analyses must also be viewed with caution given the small number of outcomes. Also, our data are based on self-report, which is known to under-estimate actual substance use behavior (Harrison and Hughes, 1997). Underreporting may be due to social stigmatization and privacy concerns, or confusion about the jargon or conceptual information in a questionnaire (Morral et al., 2003), either of which may have occurred. On the other hand, CARDIA participants used self-administered non-verbal questionnaires (better than verbal reporting (Harrison and Hughes, 1997)), and were comfortable enough to report use of marijuana, and also cocaine and amphetamine (two highly stigmatized drugs) with high frequency. Misunderstanding of the term “opiates” by some participants, however, is more likely, and presumably led to some underreporting and downward bias in our incidence estimates. The prospective nature of our study is an important strength. Previous estimates of incidence have been derived indirectly from self-reported age of initiation among nonmedical users from cross-sectional studies such as in the 2004 National Survey on Drug Use and Health (NSDUH, formerly the National Household Survey on Drug Abuse (NHSDA)), which reports an annual nonmedical prescription opioid use incidence of 0.3%/year (1.5%/5 years) among adults over age 25 (US, 2005a, 2005b, 2005c, 2005d). Our estimate is a bit lower (0.7%/5 years), but does not include former heroin users and reflects a direct, prospective measure of incidence among young- to middle-aged adults (age 28–40) that may be useful to practitioners or policymakers. Our prospective data are also useful for identifying antecedents of nonmedical prescription opioid use. Our results confirm that illegal drug use often (in our study invariably) precedes nonmedical prescription opioid use, as suggested by cross-sectional studies of college students (McCabe et al., 2005a,
175
2005b, 2005c), high-school seniors (McCabe et al., 2005a, 2005b, 2005c), and the US population (2001 NHSDA) (US, 2002; Dowling et al., 2004). They also begin to provide clarity to the issue of whether or not treatment of pain predisposes to prescription opioid addiction. Previous studies of pain and subsequent opioid addiction have mostly been uncontrolled case series studies of patients treated for pain, showing low incidence of subsequent nonmedical use (Medina and Diamond, 1977; Perry and Heidrich, 1982; Portenoy and Foley, 1986; Porter and Jick, 1980), or opioid addiction, showing a high prevalence of previous opioid treatment for pain (Potter et al., 2004; Rayport, 1954). Our results help resolve this apparent contradiction by finding modest evidence of an association between opioid treatment for pain and subsequent nonmedical use in a prospective study. Our prospective study confirms results from a recent cross-sectional study of college students (McCabe et al., 2005a, 2005b, 2005c) with a study design less prone to prevalence-incidence, effect-cause, and length-time bias, and helps to fill a gap in the literature that has been widely noted (Klein, 2004; Passik and Kirsh, 2004; Von Korff and Deyo, 2004). We did not find an association with previous depressive symptoms, but this result must be viewed as less definitive given the limited power we had to detect small associations with our sample size. In summary, we found the incidence of nonmedical prescription opioid use in this age group (28–40 years) to be quite low overall, less than 1% in 5 years. The incidence is higher among patients prescribed opioid medications by their physicians (about 6% in 5 years). While pain may sometimes result from opioid use (Ballantyne and Mao, 2003), our findings suggest that pain, and access to opioids provided by a physician, at least sometimes precedes nonmedical opioid use, supporting the logical presupposition that providing access to an addictive drug may increase the frequency of misuse. On the other hand, only a small minority of patients treated for pain in our study went on to use prescription opioids for nonmedical purposes, and this was always preceded by use of other illegal substances (particularly amphetamines). Physicians considering prescribing opioid medications for pain should be generally reassured by our results, and even when considering a prescription for a higher-risk patient, the risk of nonmedical use and addiction should be balanced against the harm caused by inadequate pain control. Future studies should address specific prescribing practices, alternate opioid formulations (Grudzinskas et al., 2006), and other ways of optimizing pain management that minimize the harm caused by the addictive potential of this otherwise useful class of medications. Acknowledgements The CARDIA study is supported by contracts N01-HC48047, N01-HC-48048, N01-HC-48049, N01-HC-48050, and N01-HC-95095 from the National Heart, Lung, and Blood Institute, which provides scientific review of all CARDIA manuscripts. Dr. Kertesz received general support from the National Institute on Drug Abuse (K23-DA-15487). Neither funder played any role in design or execution of this analysis.
176
M.J. Pletcher et al. / Drug and Alcohol Dependence 85 (2006) 171–176
References Ballantyne, J.C., Mao, J., 2003. Opioid therapy for chronic pain. N. Engl. J. Med. 349, 1943–1953. Breiman, L., Friedman, J.H., Olshen, R.A., Stone, C.J., 1984. Classification and Regression Trees. Chapman and Hall/CRC Press, Boca Raton. Dowling K., Storr C., Chilcoat H., 2004. Who’s misusing analgesics in the general population of the US? [Abstract]. Presented at the 2004 Annual Scientific Meeting of the College on Problems of Drug Dependence. Friedman G.D., Cutter G.R., Donahue R., Hughes G.H., Hulley S., Jacobs D.R.J., Liu K., P.J. S., 1988. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol. 41, 1105– 1116. Grudzinskas C., Balster R.L., Gorodetzky C.W., Griffiths R.R., Henningfield J.E., Johanson C.E., Mansbach R.S., McCormick C.G., Schnoll S.H., Strain E.C., Wright C., 2006. Impact of formulation on the abuse liability, safety and regulation of medications: the expert panel report. Drug Alcohol Depend. Epub ahead of print, March 29. Hall, W., Lynskey, M., Degenhardt, L., 2000. Trends in opiate-related deaths in the United Kingdom and Australia, 1985–1995. Drug Alcohol Depend. 57, 247–254. Hao, W., Xiao, S., Liu, T., Young, D., Chen, S., Zhang, D., Li, C., Shi, J., Chen, G., Yang, K., 2002. The second National Epidemiological Survey on illicit drug use at six high-prevalence areas in China: prevalence rates and use patterns. Addiction 97, 1305–1315. Harrison, L., Hughes, A., 1997. The Validity of Self-Reported Drug Use: Improving the Accuracy of Survey Estimates. National Institute on Drug Abuse. Hughes, G.H., Cutter, G.R., Donahue, R., Friedman, G.D., Hulley, S., Hunkeler, E., Jacobs, D.R.J., Liu, K., Orden, S., 1987. Recruitment in the Coronary Artery Disease Risk Development in Young Adults (CARDIA) study. Control Clin. Trials. 8, 68S–73S. Klein, M.J., 2004. Opioid therapy for chronic pain. N. Engl. J. Med. 350, 840–842, author reply 840–842. Mattoo, S.K., Basu, D., Sharma, A., Balaji, M., Malhotra, A., 1997. Abuse of codeine-containing cough syrups: a report from India. Addiction 92, 1783–1787. McCabe, S.E., Boyd, C.J., Teter, C.J., 2005a. Illicit use of opioid analgesics by high school seniors. J. Subst. Abuse Treat. 28, 225–230. McCabe, S.E., Teter, C.J., Boyd, C.J., 2005b. Illicit use of prescription pain medication among college students. Drug Alcohol Depend. 77, 37–47. McCabe, S.E., Teter, C.J., Boyd, C.J., Knight, J.R., Wechsler, H., 2005c. Nonmedical use of prescription opioids among U.S. college students: prevalence and correlates from a national survey. Addict. Behav. 30, 789– 805. Medina, J.L., Diamond, S., 1977. Drug dependency in patients with chronic headaches. Headache 17, 12–14. Morral, A.R., McCaffrey, D.F., Chien, S., 2003. Measurement of adolescent drug use. J. Psychoact. Drugs 35, 301–309. Naughton, M.J., Wiklund, I., 1993. A critical review of dimension-specific measures of health-related quality of life in cross-cultural research. Qual. Life Res. 2, 397–432. Passik, S.D., Kirsh, K.L., 2003. The need to identify predictors of aberrant drugrelated behavior and addiction in patients being treated with opioids for pain. Pain Med. 4, 186–189.
Passik, S.D., Kirsh, K.L., 2004. Opioid therapy in patients with a history of substance abuse. CNS Drugs 18, 13–25. Perry, S., Heidrich, G., 1982. Management of pain during debridement: a survey of U.S. burn units. Pain 13, 267–280. Portenoy, R.K., Foley, K.M., 1986. Chronic use of opioid analgesics in nonmalignant pain: report of 38 cases. Pain 25, 171–186. Porter, J., Jick, H., 1980. Addiction rare in patients treated with narcotics. N. Engl. J. Med. 302, 123. Potter, J.S., Hennessy, G., Borrow, J.A., Greenfield, S.F., Weiss, R.D., 2004. Substance use histories in patients seeking treatment for controlled-release oxycodone dependence. Drug Alcohol Depend. 76, 213–215. Radloff, L., 1977. The CES-D scale: a self-report depression scale for research in the general population. Appl. Psychol. Measure. 1, 385–401. Rayport, M., 1954. Experience in the management of patients medically addicted to narcotics. J. Am. Med. Assoc. 156, 684–691. Shah, R., Uren, Z., Baker, A., Majeed, A., 2001. Trends in deaths from drug overdose and poisoning in England and Wales 1993–1998. J. Public Health Med. 23, 242–246. Simoni-Wastila, L., Ritter, G., Strickler, G., 2004. Gender and other factors associated with the nonmedical use of abusable prescription drugs. Subst. Use Misuse. 39, 1–23. United States. Substance Abuse and Mental Health Services Administration. The NHSDA report: nonmedical use of prescription-type drugs among youths and young adults. http://www.oas.samhsa.gov/2k3/ prescription/prescription.htm, (accessed on January 11, 2006). United States. Substance Abuse and Mental Health Services Administration. The DASIS report: treatment admissions involving narcotic painkillers. http://www.oas.samhsa.gov/2k3/painTX/painTX.htm, (accessed on January 11, 2006). United States. Substance Abuse and Mental Health Services Administration. The DAWN report: narcotic analgesics, 2002 update. http://oas.samhsa.gov/2k4analgesics.pdf, (accessed on January 11, 2006). United States. Substance Abuse and Mental Health Services Administration. The NSDUH report: nonmedical use of prescription pain relievers. http:// www.oas.samhsa.gov/2k4/pain/pain.pdf, (accessed on January 11, 2006). United States Substance Abuse and Mental Health Services Administration, 2004. National survey on drug use and health: dependence, abuse and treatment tables. http://oas.samhsa.gov/NSDUH/2k4nsduh/ 2k4tabs/LOTSect5pe.htm#TopOfPage, (accessed on January 11, 2006). United States Substance Abuse and Mental Health Services Administration, 2004. National survey on drug use and health: incidence tables. http://oas. samhsa.gov/NSDUH/2k4nsduh/2k4tabs/LOTSect4pe.htm#TopOfPage, (accessed on January 11, 2006). United States Substance Abuse and Mental Health Services Administration, 2004. National survey on drug use and health: prevalence tables. http://oas.samhsa.gov/NSDUH/2k4nsduh/2k4tabs/LOTSect1pe.htm# TopOfPage, (accessed on January 11, 2006). Von Korff, M., Deyo, R.A., 2004. Potent opioids for chronic musculoskeletal pain: flying blind? Pain 109, 207–209. Zacny, J., Bigelow, G., Compton, P., Foley, K., Iguchi, M., Sannerud, C., 2003. College on Problems of Drug Dependence taskforce on prescription opioid non-medical use and abuse: position statement. Drug Alcohol Depend. 69, 215–232.