The Journal of Pain, Vol 16, No 2 (February), 2015: pp 126-134 Available online at www.jpain.org and www.sciencedirect.com
Original Reports Nonopioid Substance Use Disorders and Opioid Dose Predict Therapeutic Opioid Addiction Kelly L. Huffman,*,y Elizabeth R. Shella,* Giries Sweis,*,y Sandra D. Griffith,y,z Judith Scheman,*,y and Edward C. Covington* *Cleveland Clinic Neurological Center for Pain, yCleveland Clinic Lerner College of Medicine of Case Western Reserve University, and zCleveland Clinic Department of Quantitative Health Sciences, Cleveland, Ohio.
Abstract: Limited research examines the risk of therapeutic opioid addiction (TOA) in patients with chronic noncancer pain. This study examined TOA among 199 patients undergoing long-term opioid therapy at the time of admission to a pain rehabilitation program. It was hypothesized that nonopioid substance use disorders and opioid dosage would predict TOA. Daily mean opioid dose was 132.85 mg ± 175.39. Patients with nonopioid substance use disorders had 28 times the odds (odds ratio [OR] = 28.58; 95% confidence interval [CI] = 10.86, 75.27) of having TOA. Each 50-mg increase in opioid dose nearly doubled the odds of TOA (OR = 1.73; 95% CI = 1.29, 2.32). A 100-mg increase was associated with a 3-fold increase in odds (OR = 3.00; 95% CI = 1.67, 5.41). Receiver operating characteristic analysis revealed that opioid dose was a moderately accurate predictor (area under the curve = .75; 95% CI = .68, .82) of TOA. The sensitivity (.70) and specificity (.68) of opioid dose in predicting TOA was maximized at 76.10 mg; in addition, 46.00 mg yielded 80% sensitivity in identifying TOA. These results underscore the importance of obtaining a substance use history prior to prescribing and suggest a low screening threshold for TOA in patients who use opioids in the absence of improvement in pain or functional impairment. Perspective: This article examines TOA in patients with chronic noncancer pain undergoing longterm opioid therapy. Results suggest that patients should be screened for nonopioid substance use disorders prior to prescribing. In the absence of improvement in pain or function, there is a low threshold (50 mg daily opioid dose) for addiction screening. ª 2015 by the American Pain Society Key words: Long-term opioid therapy, chronic pain, therapeutic opioid addiction, opioid dosage, substance abuse.
S
ince the late 1980s, the prescribing of opioid analgesics to patients with chronic noncancer pain has risen dramatically.7,32,33,57 There is now a widely held belief that this practice has contributed to an epidemic of prescription opioid abuse and addiction.23,34,43 Prescription opioids are now the leading cause of unintentional drug overdose deaths.44 Thus, clinicians find themselves faced with the dilemma of balancing Received June 27, 2014; Revised October 14, 2014; Accepted October 24, 2014. Conflict of Interest/Disclosures: No funding was received for this study. All authors report no conflict of interest and have nothing to disclose. Address reprint requests to Kelly L. Huffman, PhD, Cleveland Clinic, 9500 Euclid Avenue/C21, Cleveland, OH 44195. E-mail: [email protected] 1526-5900/$36.00 ª 2015 by the American Pain Society http://dx.doi.org/10.1016/j.jpain.2014.10.011
126
the potential benefits of long-term opioid therapy with the risk of misuse or addiction. The interplay of chronic pain and addiction is a significant challenge faced by primary care physicians, pain specialists, and addictionologists alike. The decision on whether to initiate long-term opioid therapy is complicated by a dearth of reliable information on the prevalence of misuse in those who receive it. Recent estimates of the prevalence of misuse and addiction in patients with chronic pain vary widely, ranging anywhere from 0% to 50%.5,12 Our own data from the Cleveland Clinic Chronic Pain Rehabilitation Program, an interdisciplinary approach that incorporates opioid weaning, suggests that approximately one-third of patients present with a co-occurring therapeutic opioid addiction (TOA).27 This is a highly select group, given
Huffman et al that many patients were referred precisely because they were experiencing difficulties with medication use. Therefore, these data may not be reflective of the prevalence of TOA in other settings. The factors that increase the likelihood that a patient will develop TOA are not completely understood. It is generally believed that past misuse of other substances increases the likelihood that long-term opioid therapy will result in aberrant behavior12,42,50; however, there is little empirical data to support this.12,42 The majority of studies (66%) in a Cochrane review of randomized clinical trials of long-term opioid therapy not only excluded participants with substance use disorders but also did not assess whether addiction occurred.42 There is some evidence, mostly from small clinical studies, that other substance use disorders increase the risk of prescription opioid misuse or abuse.36,50,53 In a large study, Edlund et al reviewed the medical records of $15,000 veterans and found that a personal substance abuse history was the single strongest predictor of therapeutic opioid abuse or addiction.17 Further, as the authors note, this relationship was likely underestimated because substance use disorders often go undetected.13,37 Despite the known risks, patients with histories of substance abuse are more likely to be prescribed opioids, more likely to be prescribed Schedule II drugs, and more likely to be prescribed higher dosages.57 For example, a large-scale study of adult health plan enrollees demonstrated that 51.5% of patients with a prior history of an opioid use disorder were on long-term opioid therapy. In addition, 17.0% of patients with substance use disorder histories were on long-term opioid therapy versus only 3.9% of those without.57 Similarly, Morasco et al examined the prevalence of high-dose opioid use among veterans with chronic pain in a Veterans Affairs regional health network and found that patients receiving high doses had the highest rates of comorbid substance use disorders.38 There is a clear need for physicians to balance the benefits of long-term opioid therapy with the risk of addiction. The aim of the present study was to examine the associations of nonopioid substance use disorders, opioid dosage, and TOA in a population of patients thought to
The Journal of Pain
127
be representative of those at highest risk. It was hypothesized that nonopioid substance use disorders and increasing opioid dosages would be associated with increased odds of TOA. An additional aim was to examine whether opioid dosage can be used as a diagnostic indicator of opioid addiction and to determine dosage thresholds for prompting screening for TOA. It was hypothesized that opioid dosage is predictive of TOA.
Methods Participants This retrospective study examined clinical outcomes, tracked in an institutional review board–approved data registry, among adult outpatients with chronic noncancer pain treated in an interdisciplinary chronic pain rehabilitation program. The majority of patients were admitted after having failed to respond favorably to extensive treatment, such as interventional procedures, surgery, long-term opioid therapy, other medication management, and physical therapy. Many participants had additional concerns such as multiple medical conditions, TOA, a personal history of a substance use disorder, or psychological comorbidities. Thus, participants in this study were representative of those believed to be at the highest risk of developing TOA. Data were available for 352 patients admitted between January 2010 and March 2011. Cases were included in the study if patients 1) were undergoing long-term opioid therapy at the time of admission or 2) had been weaned from opioids as part of chemical dependency treatment (or by virtue of running out of opioids) directly before admission. Patients were excluded if 1) they had a history of nonmedicinal opioid use that preceded the introduction of therapeutic opioids; 2) they had abused therapeutic opioids but did not meet full criteria for TOA; or 3) the diagnosis of TOA could not be excluded or made conclusively (however, in most such cases, there were clear indications of misuse or abuse). Patients were not excluded on the basis of gender, race/ethnicity, disability status, or comorbid conditions such as chronic disease, psychiatric illness, and
Figure 1. Inclusion flow sheet. *Includes 13 participants weaned in chemical dependency treatment prior to admission, 5 participants with TOA in remission, and 3 participants who ran out of opioids prior to admission. yIncludes 12 participants on tramadol and no other opioids and 8 participants on buprenorphine. zIncludes 2 participants with known and 6 with probable nonmedicinal use of opioids preceding the introduction of therapeutic opioids. Abbreviation: COT, chronic opioid therapy.
128
Predictors of Therapeutic Opioid Addiction
The Journal of Pain
substance use disorders. Fig 1 details the inclusion/exclusion process. There were 199 participants remaining for inclusion in the study. Participants were predominantly married (64.82%, n = 129), female (60.80%, n = 121), and between ages 19 and 81 years (mean = 46.48, standard deviation = 13.62). Participants did not differ from those excluded in age (F(1, 350) = .01, P = .91), gender (c2(1, n = 352) = 2.71, P = .10), marital status (c2(5, n = 352) = 4.00, P = .55), likelihood of having a lifetime history of a nonopioid substance use disorder (c2(1, n = 309) = .00, P = .99), or rates of program completion (c2(1, n = 352) = .26, P = .61).
Measures Long-Term Opioid Therapy Long-term opioid therapy was defined as the use of legitimately obtained prescribed opioids at least 5 days per week for 90 days.24 This criterion is consistent with previous literature24 and drug adherence studies.4
TOA The term ‘‘therapeutic opioid’’ is used to distinguish prescribed opioid analgesia from recreational use, and the term ‘‘addiction’’ rather than ‘‘dependence’’ is used to avoid the ambiguity inherent in the latter term. Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revised, criteria,3 supplemented by the consensus definitions developed by the American Academy of Pain Medicine, the American Pain Society, and the American Society of Addiction Medicine,2 were used to diagnose TOA. Per consensus definitions, Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revised, diagnostic criteria of tolerance and withdrawal3 were not used as diagnostic criteria, as physical dependence is a normal response to appropriate long-term use. Diagnosis instead utilized the remaining criteria, which focus on such characteristics of addiction as loss of control, compulsive use, and continued use despite consequences. A positive diagnosis required the presence of 3 of these remaining criteria. It is important to note that the diagnosis of TOA can be difficult. For example, the diagnostic criteria refer to reduced function due to the substance; however, it can be difficult to ascertain whether impairment is due to pain, substance use, deconditioning, or some combination of the former. To be conservative in our diagnoses, failure to fulfill role expectations (employment, family, etc) was not considered adequate evidence of addiction; rather, this was attributed to pain. Examples of substance-related impairment included behaviors such as somnolence during conversations or at meals, cognitive impairment that waxed and waned with medication use, or personality changes clearly associated with medication use. Readers are referred to the consensus definitions for more information about the subtleties of diagnosing TOA in this population.2 The diagnostic process was multifaceted and included clinical interview, review of medical records, urine toxi-
cology, and collateral interviews with family. Diagnostic interviews were conducted by either a master’s-level chemical dependency counselor (E.R.S.) or a licensed psychologist (K.C.H. or G.S.) with experience in treating patients with chronic pain and co-occurring substance disorders. The final diagnosis was confirmed by a physician board certified in both pain management and psychiatry (E.C.C. or another). TOA was diagnosed only if the participant had at least initially lawfully obtained prescription opiates for a legitimate medical purpose and had no known history of recreational opioid abuse or dependence.
Nonopioid Substance Use Disorders We use the terms ‘‘substance use disorder,’’ ‘‘substance abuse,’’ and ‘‘substance dependence’’ specifically in reference to nonopioid substance use disorders, diagnosed using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revised3 criteria. All nonopioid substance use disorders, including prescription sedative-hypnotic abuse or dependence, were diagnosed using the same process used to diagnose TOA, as described above. It was not possible from our data set to determine whether the onset of a substance use disorder predated the introduction of therapeutic opioids.
Opioid Dosage Opioid dosages at the time of admission were converted to daily oral morphine equivalence dosages based on ratios derived from empirical studies,11,35 package inserts,8,9,26,30,47,49 and clinical guidelines.1,15,16,31,39,54
Treatment Participants underwent treatment in a wellestablished (>30 years) interdisciplinary 3- to 4-week chronic pain rehabilitation program. The program is a time-intensive (5 days/wk, 7:30 AM–5 PM) adult outpatient program that includes individual and group psychotherapy, physical therapy, occupational therapy, medication management, weaning of opioids and other habituating medications, substance use education, and monthly aftercare.
Data Analysis Analyses were conducted using SPSS 20 (IBM Corp, Armonk, NY)29 and R version 3.0.1 (R Foundation for Statistical Computing, Vienna, Austria).48,51 Descriptive statistics, including means, standard deviations, counts, and percentages, as applicable, were examined for patients with and without TOA. Group differences were analyzed using 1-way analysis of variance, Pearson chi-square analysis without a continuity correction, or nonparametric statistics, as appropriate. A binary logistic regression, including gender and age as controls, was used to determine whether a lifetime history of a nonopioid substance use disorder or morphine equivalence dosage at time of admission predicted increased odds of having TOA. Data were examined for outliers (63 standard deviation). Missing data were excluded on a listwise basis. Post hoc power
Huffman et al
The Journal of Pain
analyses were conducted using G*Power software (Heinrich-Heine University, Dusseldorf, Germany).19,20 Descriptive statistics were used to examine the rates of TOA as a function of the specific type of substance use disorder (ie, nonprescription, sedative hypnotic, or nonopioid substance use disorder in remission). To test whether opioid dose discriminated between patients with and without TOA, 3 receiver operating characteristic (ROC) curves were built plotting the truepositive rate (sensitivity) against the false-positive rate (1-specificity); in addition to empirical curves, smoothed ROC curves were constructed using the binormal distribution method.51 The first curve examined the relationship between opioid dose and TOA for all patients, the second for patients with a lifetime history of a substance use disorder, and the third for patients without a lifetime history of a substance use disorder. In each case, the area under the curve (AUC) and 95% confidence intervals were calculated. AUC scores were evaluated using current conventions.21 Optimal thresholds (Youden’s index) for maximizing both sensitivity and specificity were calculated,58 as well as corresponding positive predictive values (PPV) and negative predictive values (NPV).
Results Eighty-seven patients were diagnosed with TOA (43.72%). In patients with no known history of a substance use disorder, 25.00% had TOA. In patients with a known history of a substance use disorder, 82.54% also had TOA. Patients with TOA were younger (F(1, 197) = 4.43, P = .04) and more likely to be male (c2(1, n = 199) = 6.79, P = .01) than patients without TOA. There was no significant difference in marital status between groups. Patients with TOA were more likely to have a lifetime history of a substance use disorder than those without (59.77% vs 9.82%, c2(1, n = 179) = 54.56, P = .01). Table 1 details nonopioid substance use disorders among participants. Mean admission daily morphine equivalence dosage for all participants was 132.85 mg (6175.39). Mean dose for participants with TOA was 215.93 mg (6231.44), and mean dose for participants without TOA was 71.11 mg (671.65). Data were not normally distributed (skewness = 2.73, standard error = .18), and Levene’s Table 1.
129
statistic (F = 61.75, P < .001) indicated that the variances between groups were not homogenous. Thus, groups were compared using nonparametric statistics. A MannWhitney U test demonstrated that opioid dose was higher in patients with TOA (mean rank = 122.83) than patients without (mean rank = 75.19; U = 2201.00, P = < .001, r = .43). There was no difference in opioid dose between patients with and without a lifetime history of a substance use disorder (mean = 133.59 6 153.69 vs mean = 135.20 6 191.20; F(1, 168) = .00, P = .96). Data are fully detailed in Table 2.
Nonopioid Substance Use Disorders and TOA A binary logistic regression model was built including both opioid dose and lifetime history of a substance abuse disorder as predictor variables. The limited numbers of patients with some types of substance use disorders did not permit inclusion of these subtypes in the model as individual predictors. Because patients with TOA were significantly more likely to be male and significantly younger than patients without TOA, gender and age were included as covariates. Missing data, excluded on a listwise basis, resulted in the overall exclusion of 29 cases. Post hoc power analysis19,20 indicated that the power to detect a moderate effect size14 with 4 predictors was acceptable (.99). The results yielded a final model in which both life history of a substance use disorder and morphine equivalence dosage at admission showed a highly statistically significant association with TOA (all P < .01) after adjustment for age and gender. Gender and age were not significantly associated with TOA. According to the model, the presence of a lifetime history of a substance use disorder was associated with 28 times increased odds of TOA (odds ratio [OR] = 28.5; 95% confidence interval [CI] = 10.86, 75.27). Table 3 details the results. For a more meaningful interpretation, the estimates (b) and corresponding ORs were rescaled for values of 10, 50, and 100 mg oral morphine. Each 10-mg increase was associated with a 12% increase in odds of TOA (OR = 1.12; 95% CI = 1.05, 1.18). Each 50-mg increase was associated with a 73% increase in odds of TOA (OR = 1.73; 95% CI = 1.29, 2.32). Each 100-mg increase
Participant TOA and Lifetime History of Nonopioid Substance Use Disorders NONOPIOID SUBSTANCE USE DISORDERS
No life history of a nonopioid substance use disorder Life history of a nonopioid substance use disorder Specific types of substance use disorders observed Active sedative hypnotic Active nonprescription Remission Active sedative hypnotic and active nonprescription Active nonprescription and remission Active sedative hypnotic and remission Active sedative hypnotic, active nonprescription, and remission Unclear life history of a substance use disorder NOTE. Values are n (%).
TOA (N = 87)
NO TOA (N = 112)
TOTAL (N = 199)
29 (33.33) 52 (59.77)
87 (77.68) 11 (9.82)
116 (58.29) 63 (31.66)
13 (14.94) 12 (13.79) 11 (12.64) 3 (3.45) 4 (4.60) 6 (6.90) 3 (3.45) 6 (6.90)
0 (.00) 3 (2.68) 7 (6.26) 0 (.00) 0 (.00) 1 (.89) 0 (.00) 14 (12.50)
13 (6.53) 15 (7.54) 18 (9.05) 3 (1.51) 4 (2.01) 7 (3.52) 3 (1.51) 20 (10.05)
130
Predictors of Therapeutic Opioid Addiction
The Journal of Pain
Table 2.
Characteristics of Participants With and Without a TOA
CHARACTERISTICS Female, n (%) Age, M (SD) Marital status, n (%) Single Married Other Opioid dose, M (SD)* Opioid dose, MR
TOA (N = 87)
NO TOA (N = 112)
44 (50.57) 44.53 (12.95)
77 (68.75) 48.26 (13.85)
23 (26.44) 55 (63.22) 9 (10.34) 215.93 (231.44) 122.83
21 (18.75) 74 (66.07) 17 (15.18) 71.11 (71.65) 75.19
c2, F, U
P
c2(df = 1) = 6.79 F(df = 1) = 4.43 c2(df = 5) = 6.32 – – – – U = 2201.00
.01 .04 .28 – – – – .001
Abbreviations: M, mean; SD, standard deviation; MR, mean rank. *Significance testing not conducted.
false-positive results (specificity = 1.00; PPV = 100%; NPV = 28.20%). For patients without a known history of a nonopioid substance use disorder, the optimal cut point was 78.75 mg, corresponding to a sensitivity of .88 and a specificity of .69. The PPV in identifying TOA was 46.94%, NPV was 95.08%. Youden’s index, which weights both sensitivity and specificity equally, does not necessarily provide an optimal clinical decision threshold.59 Because the personal and societal costs associated with addiction are high in comparison to the mechanisms needed for detection, sensitivity holds more clinical utility than specificity. Therefore, we also considered the criterion of 80% sensitivity, where 80% of true TOAs were detected using only data on daily opioid dosage. For the total patient sample, the threshold that detected cases of TOA 80% of the time was 46.00 mg (PPV of 54.6%; positive likelihood ratio (LR1) = 1.6, sensitivity = .80; specificity = .50.). For patients with a lifetime history of a nonopioid substance use disorder, the threshold was 42.50 mg (PPV of 86.7%; LR1 = 1.45; sensitivity = .80, specificity of = .45). For patients with no lifetime history of nonopioid substance use disorder, the cut point of 86.25 mg provided an 80% true-positive rate (PPV of 47.7%; LR1 = 3; sensitivity = .81; specificity = .73).
was associated with 3-fold increase in odds of TOA (OR = 3.00; 95% CI = 1.67, 5.41). Because of the low frequencies of patients with substance use disorders who did not also have TOA, it was not possible to create an additional logistic regression model comparing the frequencies of TOA as a function of a specific type of substance use disorder. Frequency counts (as reflected in Table 1) showed that only 3 of the 24 patients (12.5%) with some type of co-occurring active substance use disorder and 7 of the 18 patients (38.89%) with a substance use disorder in remission (and no other active substance use disorder) did not also have TOA. In the case of comorbid sedative hypnotic use disorders, there was only 1 patient of the 20 (5%) who did not also have a TOA.
Opioid Dosage ROC curves corresponding to the full sample and stratified by the presence or absence of the lifetime history of a nonopioid substance use disorder are shown in Figs 2 and 3. AUC was calculated and showed that opioid dosage was a moderately accurate predictor21 of TOA for all patients (AUC = .75; 95% CI = .68, .82), for patients with a lifetime history of a substance use disorder (AUC = .73; 95% CI = .59, .88), and for patients without a lifetime history of a substance use disorder (AUC = .84; 95% CI = .75, .94). For the entire sample, without regard to substance use history, the optimal cut point (Youden’s index) of 76.10 mg oral morphine yielded a sensitivity of .70 and a specificity of .68. At this cut point, the PPV was 62.50% and the NPV was 74.50%. For patients with a lifetime history of a substance use disorder, an optimal cut point of 117.50 mg identified only 43% of true positives but led to no
Discussion The prevalence of TOA in this study, 43.72%, was higher than in other studies documenting the prevalence of TOA in patients with chronic pain.12 This is probably a conservative estimate of the prevalence of TOA in our sample. In order to avoid inflated estimates from patients whose opioid use could not be clearly evaluated,
Binary Logistic Regression: Lifetime History of Nonopioid Substance Use Disorders and Daily Oral Morphine Equivalence Dosage as Predictors of TOA
Table 3.
PREDICTORS
b
SE b
WALD’S c2
DF
P
fb
95% CI
R2Cox
R2Nag
Gender Age Dosage (10 units)* Life history of SUD Constant
.02 -.01 .11 3.35 -2.17
.47 .02 .03 .49 .85
.00 .74 19.22 46.07 6.50
1 1 1 1 1
.97 .39
Original Reports Nonopioid Substance Use Disorders and Opioid Dose Predict Therapeutic Opioid Addiction Kelly L. Huffman,*,y Elizabeth R. Shella,* Giries Sweis,*,y Sandra D. Griffith,y,z Judith Scheman,*,y and Edward C. Covington* *Cleveland Clinic Neurological Center for Pain, yCleveland Clinic Lerner College of Medicine of Case Western Reserve University, and zCleveland Clinic Department of Quantitative Health Sciences, Cleveland, Ohio.
Abstract: Limited research examines the risk of therapeutic opioid addiction (TOA) in patients with chronic noncancer pain. This study examined TOA among 199 patients undergoing long-term opioid therapy at the time of admission to a pain rehabilitation program. It was hypothesized that nonopioid substance use disorders and opioid dosage would predict TOA. Daily mean opioid dose was 132.85 mg ± 175.39. Patients with nonopioid substance use disorders had 28 times the odds (odds ratio [OR] = 28.58; 95% confidence interval [CI] = 10.86, 75.27) of having TOA. Each 50-mg increase in opioid dose nearly doubled the odds of TOA (OR = 1.73; 95% CI = 1.29, 2.32). A 100-mg increase was associated with a 3-fold increase in odds (OR = 3.00; 95% CI = 1.67, 5.41). Receiver operating characteristic analysis revealed that opioid dose was a moderately accurate predictor (area under the curve = .75; 95% CI = .68, .82) of TOA. The sensitivity (.70) and specificity (.68) of opioid dose in predicting TOA was maximized at 76.10 mg; in addition, 46.00 mg yielded 80% sensitivity in identifying TOA. These results underscore the importance of obtaining a substance use history prior to prescribing and suggest a low screening threshold for TOA in patients who use opioids in the absence of improvement in pain or functional impairment. Perspective: This article examines TOA in patients with chronic noncancer pain undergoing longterm opioid therapy. Results suggest that patients should be screened for nonopioid substance use disorders prior to prescribing. In the absence of improvement in pain or function, there is a low threshold (50 mg daily opioid dose) for addiction screening. ª 2015 by the American Pain Society Key words: Long-term opioid therapy, chronic pain, therapeutic opioid addiction, opioid dosage, substance abuse.
S
ince the late 1980s, the prescribing of opioid analgesics to patients with chronic noncancer pain has risen dramatically.7,32,33,57 There is now a widely held belief that this practice has contributed to an epidemic of prescription opioid abuse and addiction.23,34,43 Prescription opioids are now the leading cause of unintentional drug overdose deaths.44 Thus, clinicians find themselves faced with the dilemma of balancing Received June 27, 2014; Revised October 14, 2014; Accepted October 24, 2014. Conflict of Interest/Disclosures: No funding was received for this study. All authors report no conflict of interest and have nothing to disclose. Address reprint requests to Kelly L. Huffman, PhD, Cleveland Clinic, 9500 Euclid Avenue/C21, Cleveland, OH 44195. E-mail: [email protected] 1526-5900/$36.00 ª 2015 by the American Pain Society http://dx.doi.org/10.1016/j.jpain.2014.10.011
126
the potential benefits of long-term opioid therapy with the risk of misuse or addiction. The interplay of chronic pain and addiction is a significant challenge faced by primary care physicians, pain specialists, and addictionologists alike. The decision on whether to initiate long-term opioid therapy is complicated by a dearth of reliable information on the prevalence of misuse in those who receive it. Recent estimates of the prevalence of misuse and addiction in patients with chronic pain vary widely, ranging anywhere from 0% to 50%.5,12 Our own data from the Cleveland Clinic Chronic Pain Rehabilitation Program, an interdisciplinary approach that incorporates opioid weaning, suggests that approximately one-third of patients present with a co-occurring therapeutic opioid addiction (TOA).27 This is a highly select group, given
Huffman et al that many patients were referred precisely because they were experiencing difficulties with medication use. Therefore, these data may not be reflective of the prevalence of TOA in other settings. The factors that increase the likelihood that a patient will develop TOA are not completely understood. It is generally believed that past misuse of other substances increases the likelihood that long-term opioid therapy will result in aberrant behavior12,42,50; however, there is little empirical data to support this.12,42 The majority of studies (66%) in a Cochrane review of randomized clinical trials of long-term opioid therapy not only excluded participants with substance use disorders but also did not assess whether addiction occurred.42 There is some evidence, mostly from small clinical studies, that other substance use disorders increase the risk of prescription opioid misuse or abuse.36,50,53 In a large study, Edlund et al reviewed the medical records of $15,000 veterans and found that a personal substance abuse history was the single strongest predictor of therapeutic opioid abuse or addiction.17 Further, as the authors note, this relationship was likely underestimated because substance use disorders often go undetected.13,37 Despite the known risks, patients with histories of substance abuse are more likely to be prescribed opioids, more likely to be prescribed Schedule II drugs, and more likely to be prescribed higher dosages.57 For example, a large-scale study of adult health plan enrollees demonstrated that 51.5% of patients with a prior history of an opioid use disorder were on long-term opioid therapy. In addition, 17.0% of patients with substance use disorder histories were on long-term opioid therapy versus only 3.9% of those without.57 Similarly, Morasco et al examined the prevalence of high-dose opioid use among veterans with chronic pain in a Veterans Affairs regional health network and found that patients receiving high doses had the highest rates of comorbid substance use disorders.38 There is a clear need for physicians to balance the benefits of long-term opioid therapy with the risk of addiction. The aim of the present study was to examine the associations of nonopioid substance use disorders, opioid dosage, and TOA in a population of patients thought to
The Journal of Pain
127
be representative of those at highest risk. It was hypothesized that nonopioid substance use disorders and increasing opioid dosages would be associated with increased odds of TOA. An additional aim was to examine whether opioid dosage can be used as a diagnostic indicator of opioid addiction and to determine dosage thresholds for prompting screening for TOA. It was hypothesized that opioid dosage is predictive of TOA.
Methods Participants This retrospective study examined clinical outcomes, tracked in an institutional review board–approved data registry, among adult outpatients with chronic noncancer pain treated in an interdisciplinary chronic pain rehabilitation program. The majority of patients were admitted after having failed to respond favorably to extensive treatment, such as interventional procedures, surgery, long-term opioid therapy, other medication management, and physical therapy. Many participants had additional concerns such as multiple medical conditions, TOA, a personal history of a substance use disorder, or psychological comorbidities. Thus, participants in this study were representative of those believed to be at the highest risk of developing TOA. Data were available for 352 patients admitted between January 2010 and March 2011. Cases were included in the study if patients 1) were undergoing long-term opioid therapy at the time of admission or 2) had been weaned from opioids as part of chemical dependency treatment (or by virtue of running out of opioids) directly before admission. Patients were excluded if 1) they had a history of nonmedicinal opioid use that preceded the introduction of therapeutic opioids; 2) they had abused therapeutic opioids but did not meet full criteria for TOA; or 3) the diagnosis of TOA could not be excluded or made conclusively (however, in most such cases, there were clear indications of misuse or abuse). Patients were not excluded on the basis of gender, race/ethnicity, disability status, or comorbid conditions such as chronic disease, psychiatric illness, and
Figure 1. Inclusion flow sheet. *Includes 13 participants weaned in chemical dependency treatment prior to admission, 5 participants with TOA in remission, and 3 participants who ran out of opioids prior to admission. yIncludes 12 participants on tramadol and no other opioids and 8 participants on buprenorphine. zIncludes 2 participants with known and 6 with probable nonmedicinal use of opioids preceding the introduction of therapeutic opioids. Abbreviation: COT, chronic opioid therapy.
128
Predictors of Therapeutic Opioid Addiction
The Journal of Pain
substance use disorders. Fig 1 details the inclusion/exclusion process. There were 199 participants remaining for inclusion in the study. Participants were predominantly married (64.82%, n = 129), female (60.80%, n = 121), and between ages 19 and 81 years (mean = 46.48, standard deviation = 13.62). Participants did not differ from those excluded in age (F(1, 350) = .01, P = .91), gender (c2(1, n = 352) = 2.71, P = .10), marital status (c2(5, n = 352) = 4.00, P = .55), likelihood of having a lifetime history of a nonopioid substance use disorder (c2(1, n = 309) = .00, P = .99), or rates of program completion (c2(1, n = 352) = .26, P = .61).
Measures Long-Term Opioid Therapy Long-term opioid therapy was defined as the use of legitimately obtained prescribed opioids at least 5 days per week for 90 days.24 This criterion is consistent with previous literature24 and drug adherence studies.4
TOA The term ‘‘therapeutic opioid’’ is used to distinguish prescribed opioid analgesia from recreational use, and the term ‘‘addiction’’ rather than ‘‘dependence’’ is used to avoid the ambiguity inherent in the latter term. Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revised, criteria,3 supplemented by the consensus definitions developed by the American Academy of Pain Medicine, the American Pain Society, and the American Society of Addiction Medicine,2 were used to diagnose TOA. Per consensus definitions, Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revised, diagnostic criteria of tolerance and withdrawal3 were not used as diagnostic criteria, as physical dependence is a normal response to appropriate long-term use. Diagnosis instead utilized the remaining criteria, which focus on such characteristics of addiction as loss of control, compulsive use, and continued use despite consequences. A positive diagnosis required the presence of 3 of these remaining criteria. It is important to note that the diagnosis of TOA can be difficult. For example, the diagnostic criteria refer to reduced function due to the substance; however, it can be difficult to ascertain whether impairment is due to pain, substance use, deconditioning, or some combination of the former. To be conservative in our diagnoses, failure to fulfill role expectations (employment, family, etc) was not considered adequate evidence of addiction; rather, this was attributed to pain. Examples of substance-related impairment included behaviors such as somnolence during conversations or at meals, cognitive impairment that waxed and waned with medication use, or personality changes clearly associated with medication use. Readers are referred to the consensus definitions for more information about the subtleties of diagnosing TOA in this population.2 The diagnostic process was multifaceted and included clinical interview, review of medical records, urine toxi-
cology, and collateral interviews with family. Diagnostic interviews were conducted by either a master’s-level chemical dependency counselor (E.R.S.) or a licensed psychologist (K.C.H. or G.S.) with experience in treating patients with chronic pain and co-occurring substance disorders. The final diagnosis was confirmed by a physician board certified in both pain management and psychiatry (E.C.C. or another). TOA was diagnosed only if the participant had at least initially lawfully obtained prescription opiates for a legitimate medical purpose and had no known history of recreational opioid abuse or dependence.
Nonopioid Substance Use Disorders We use the terms ‘‘substance use disorder,’’ ‘‘substance abuse,’’ and ‘‘substance dependence’’ specifically in reference to nonopioid substance use disorders, diagnosed using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revised3 criteria. All nonopioid substance use disorders, including prescription sedative-hypnotic abuse or dependence, were diagnosed using the same process used to diagnose TOA, as described above. It was not possible from our data set to determine whether the onset of a substance use disorder predated the introduction of therapeutic opioids.
Opioid Dosage Opioid dosages at the time of admission were converted to daily oral morphine equivalence dosages based on ratios derived from empirical studies,11,35 package inserts,8,9,26,30,47,49 and clinical guidelines.1,15,16,31,39,54
Treatment Participants underwent treatment in a wellestablished (>30 years) interdisciplinary 3- to 4-week chronic pain rehabilitation program. The program is a time-intensive (5 days/wk, 7:30 AM–5 PM) adult outpatient program that includes individual and group psychotherapy, physical therapy, occupational therapy, medication management, weaning of opioids and other habituating medications, substance use education, and monthly aftercare.
Data Analysis Analyses were conducted using SPSS 20 (IBM Corp, Armonk, NY)29 and R version 3.0.1 (R Foundation for Statistical Computing, Vienna, Austria).48,51 Descriptive statistics, including means, standard deviations, counts, and percentages, as applicable, were examined for patients with and without TOA. Group differences were analyzed using 1-way analysis of variance, Pearson chi-square analysis without a continuity correction, or nonparametric statistics, as appropriate. A binary logistic regression, including gender and age as controls, was used to determine whether a lifetime history of a nonopioid substance use disorder or morphine equivalence dosage at time of admission predicted increased odds of having TOA. Data were examined for outliers (63 standard deviation). Missing data were excluded on a listwise basis. Post hoc power
Huffman et al
The Journal of Pain
analyses were conducted using G*Power software (Heinrich-Heine University, Dusseldorf, Germany).19,20 Descriptive statistics were used to examine the rates of TOA as a function of the specific type of substance use disorder (ie, nonprescription, sedative hypnotic, or nonopioid substance use disorder in remission). To test whether opioid dose discriminated between patients with and without TOA, 3 receiver operating characteristic (ROC) curves were built plotting the truepositive rate (sensitivity) against the false-positive rate (1-specificity); in addition to empirical curves, smoothed ROC curves were constructed using the binormal distribution method.51 The first curve examined the relationship between opioid dose and TOA for all patients, the second for patients with a lifetime history of a substance use disorder, and the third for patients without a lifetime history of a substance use disorder. In each case, the area under the curve (AUC) and 95% confidence intervals were calculated. AUC scores were evaluated using current conventions.21 Optimal thresholds (Youden’s index) for maximizing both sensitivity and specificity were calculated,58 as well as corresponding positive predictive values (PPV) and negative predictive values (NPV).
Results Eighty-seven patients were diagnosed with TOA (43.72%). In patients with no known history of a substance use disorder, 25.00% had TOA. In patients with a known history of a substance use disorder, 82.54% also had TOA. Patients with TOA were younger (F(1, 197) = 4.43, P = .04) and more likely to be male (c2(1, n = 199) = 6.79, P = .01) than patients without TOA. There was no significant difference in marital status between groups. Patients with TOA were more likely to have a lifetime history of a substance use disorder than those without (59.77% vs 9.82%, c2(1, n = 179) = 54.56, P = .01). Table 1 details nonopioid substance use disorders among participants. Mean admission daily morphine equivalence dosage for all participants was 132.85 mg (6175.39). Mean dose for participants with TOA was 215.93 mg (6231.44), and mean dose for participants without TOA was 71.11 mg (671.65). Data were not normally distributed (skewness = 2.73, standard error = .18), and Levene’s Table 1.
129
statistic (F = 61.75, P < .001) indicated that the variances between groups were not homogenous. Thus, groups were compared using nonparametric statistics. A MannWhitney U test demonstrated that opioid dose was higher in patients with TOA (mean rank = 122.83) than patients without (mean rank = 75.19; U = 2201.00, P = < .001, r = .43). There was no difference in opioid dose between patients with and without a lifetime history of a substance use disorder (mean = 133.59 6 153.69 vs mean = 135.20 6 191.20; F(1, 168) = .00, P = .96). Data are fully detailed in Table 2.
Nonopioid Substance Use Disorders and TOA A binary logistic regression model was built including both opioid dose and lifetime history of a substance abuse disorder as predictor variables. The limited numbers of patients with some types of substance use disorders did not permit inclusion of these subtypes in the model as individual predictors. Because patients with TOA were significantly more likely to be male and significantly younger than patients without TOA, gender and age were included as covariates. Missing data, excluded on a listwise basis, resulted in the overall exclusion of 29 cases. Post hoc power analysis19,20 indicated that the power to detect a moderate effect size14 with 4 predictors was acceptable (.99). The results yielded a final model in which both life history of a substance use disorder and morphine equivalence dosage at admission showed a highly statistically significant association with TOA (all P < .01) after adjustment for age and gender. Gender and age were not significantly associated with TOA. According to the model, the presence of a lifetime history of a substance use disorder was associated with 28 times increased odds of TOA (odds ratio [OR] = 28.5; 95% confidence interval [CI] = 10.86, 75.27). Table 3 details the results. For a more meaningful interpretation, the estimates (b) and corresponding ORs were rescaled for values of 10, 50, and 100 mg oral morphine. Each 10-mg increase was associated with a 12% increase in odds of TOA (OR = 1.12; 95% CI = 1.05, 1.18). Each 50-mg increase was associated with a 73% increase in odds of TOA (OR = 1.73; 95% CI = 1.29, 2.32). Each 100-mg increase
Participant TOA and Lifetime History of Nonopioid Substance Use Disorders NONOPIOID SUBSTANCE USE DISORDERS
No life history of a nonopioid substance use disorder Life history of a nonopioid substance use disorder Specific types of substance use disorders observed Active sedative hypnotic Active nonprescription Remission Active sedative hypnotic and active nonprescription Active nonprescription and remission Active sedative hypnotic and remission Active sedative hypnotic, active nonprescription, and remission Unclear life history of a substance use disorder NOTE. Values are n (%).
TOA (N = 87)
NO TOA (N = 112)
TOTAL (N = 199)
29 (33.33) 52 (59.77)
87 (77.68) 11 (9.82)
116 (58.29) 63 (31.66)
13 (14.94) 12 (13.79) 11 (12.64) 3 (3.45) 4 (4.60) 6 (6.90) 3 (3.45) 6 (6.90)
0 (.00) 3 (2.68) 7 (6.26) 0 (.00) 0 (.00) 1 (.89) 0 (.00) 14 (12.50)
13 (6.53) 15 (7.54) 18 (9.05) 3 (1.51) 4 (2.01) 7 (3.52) 3 (1.51) 20 (10.05)
130
Predictors of Therapeutic Opioid Addiction
The Journal of Pain
Table 2.
Characteristics of Participants With and Without a TOA
CHARACTERISTICS Female, n (%) Age, M (SD) Marital status, n (%) Single Married Other Opioid dose, M (SD)* Opioid dose, MR
TOA (N = 87)
NO TOA (N = 112)
44 (50.57) 44.53 (12.95)
77 (68.75) 48.26 (13.85)
23 (26.44) 55 (63.22) 9 (10.34) 215.93 (231.44) 122.83
21 (18.75) 74 (66.07) 17 (15.18) 71.11 (71.65) 75.19
c2, F, U
P
c2(df = 1) = 6.79 F(df = 1) = 4.43 c2(df = 5) = 6.32 – – – – U = 2201.00
.01 .04 .28 – – – – .001
Abbreviations: M, mean; SD, standard deviation; MR, mean rank. *Significance testing not conducted.
false-positive results (specificity = 1.00; PPV = 100%; NPV = 28.20%). For patients without a known history of a nonopioid substance use disorder, the optimal cut point was 78.75 mg, corresponding to a sensitivity of .88 and a specificity of .69. The PPV in identifying TOA was 46.94%, NPV was 95.08%. Youden’s index, which weights both sensitivity and specificity equally, does not necessarily provide an optimal clinical decision threshold.59 Because the personal and societal costs associated with addiction are high in comparison to the mechanisms needed for detection, sensitivity holds more clinical utility than specificity. Therefore, we also considered the criterion of 80% sensitivity, where 80% of true TOAs were detected using only data on daily opioid dosage. For the total patient sample, the threshold that detected cases of TOA 80% of the time was 46.00 mg (PPV of 54.6%; positive likelihood ratio (LR1) = 1.6, sensitivity = .80; specificity = .50.). For patients with a lifetime history of a nonopioid substance use disorder, the threshold was 42.50 mg (PPV of 86.7%; LR1 = 1.45; sensitivity = .80, specificity of = .45). For patients with no lifetime history of nonopioid substance use disorder, the cut point of 86.25 mg provided an 80% true-positive rate (PPV of 47.7%; LR1 = 3; sensitivity = .81; specificity = .73).
was associated with 3-fold increase in odds of TOA (OR = 3.00; 95% CI = 1.67, 5.41). Because of the low frequencies of patients with substance use disorders who did not also have TOA, it was not possible to create an additional logistic regression model comparing the frequencies of TOA as a function of a specific type of substance use disorder. Frequency counts (as reflected in Table 1) showed that only 3 of the 24 patients (12.5%) with some type of co-occurring active substance use disorder and 7 of the 18 patients (38.89%) with a substance use disorder in remission (and no other active substance use disorder) did not also have TOA. In the case of comorbid sedative hypnotic use disorders, there was only 1 patient of the 20 (5%) who did not also have a TOA.
Opioid Dosage ROC curves corresponding to the full sample and stratified by the presence or absence of the lifetime history of a nonopioid substance use disorder are shown in Figs 2 and 3. AUC was calculated and showed that opioid dosage was a moderately accurate predictor21 of TOA for all patients (AUC = .75; 95% CI = .68, .82), for patients with a lifetime history of a substance use disorder (AUC = .73; 95% CI = .59, .88), and for patients without a lifetime history of a substance use disorder (AUC = .84; 95% CI = .75, .94). For the entire sample, without regard to substance use history, the optimal cut point (Youden’s index) of 76.10 mg oral morphine yielded a sensitivity of .70 and a specificity of .68. At this cut point, the PPV was 62.50% and the NPV was 74.50%. For patients with a lifetime history of a substance use disorder, an optimal cut point of 117.50 mg identified only 43% of true positives but led to no
Discussion The prevalence of TOA in this study, 43.72%, was higher than in other studies documenting the prevalence of TOA in patients with chronic pain.12 This is probably a conservative estimate of the prevalence of TOA in our sample. In order to avoid inflated estimates from patients whose opioid use could not be clearly evaluated,
Binary Logistic Regression: Lifetime History of Nonopioid Substance Use Disorders and Daily Oral Morphine Equivalence Dosage as Predictors of TOA
Table 3.
PREDICTORS
b
SE b
WALD’S c2
DF
P
fb
95% CI
R2Cox
R2Nag
Gender Age Dosage (10 units)* Life history of SUD Constant
.02 -.01 .11 3.35 -2.17
.47 .02 .03 .49 .85
.00 .74 19.22 46.07 6.50
1 1 1 1 1
.97 .39
