Benzodiazepines Predict Use of Opioids A Follow-Up Study of 17,074 Men and Womenpme_

Transcription

1 Pain Medicine 2010; 11: Wiley Periodicals, Inc. ORIGINAL RESEARCH ARTICLES Benzodiazepines Predict Use of Opioids A Follow-Up Study of 17,074 Men and Womenpme_ Svetlana Skurtveit, PhD,* Kari Furu, PhD,* Jørgen Bramness, PhD,* Randi Selmer, PhD,* and Aage Tverdal, PhD* *Division of Epidemiology, Norwegian Institute of Public Health; Norwegian Centre for Addiction Research, University of Oslo, Oslo, Norway Reprint requests to: Svetlana Skurtveit, PhD, Norwegian Institute of Public Health, PO Box 4404 Nydalen, 0403 Oslo, Norway. Tel: ; Fax: ; svetlana.skurtveit@fhi.no. Abstract Objective. To evaluate the effect of the use of benzodiazepines on prescription of opioids 4 7 years later in patients with noncancer pain. Design. A cohort of 7,991 men and 9,083 women aged 40, 45 and 60 years who reported no use of opioids in health surveys in was linked to the nationwide Norwegian Prescription Database, and their prescriptions of opioids during were analyzed. Moderate-high prescription frequency of opioids was defined as at least 12 prescriptions during the period January 2004 December Results. The unadjusted odds ratio for moderatehigh prescription frequency of opioids for individuals who had previously used benzodiazepines was 7.7 (95% confidence interval ) as compared with previous nonusers. After adjustment for musculoskeletal pain, alcohol, smoking habits, and socioeconomic variables, the odds ratio was lowered to 3.1 ( ). The analysis of the effect of benzodiazepines and chronic pain individually and in combination suggest that use of benzodiazepines is an even stronger predictor of later opioid use than self-reported chronic pain. Conclusions. Our study suggests that earlier use of benzodiazepines may predict repeated use of opioids. Before starting pain treatment with opioids, clinicians should take into consideration the possibility of substance abuse and mental health disorders. A central issue when prescribing opioids for chronic noncancer pain is to balance the risk of problematic use of these drugs with the benefits of pain relief. Key Words. Benzodiazepines; Noncancer Patients; Opioids; Prospective Cohort Study Introduction In the last few decades, opioids have been considered useful treatment option not only for cancer related and severe acute pain, but also to some extent for chronic nonmalignant pain. Guidelines recommend the use of long acting opioids, while suggesting that short acting opioids should be avoided [1 6]. However, the beneficial effect of long lasting opioid treatment has been questioned [3 6]. Opioids should be avoided in patients using other addictive drugs such as benzodiazepines, because this may increase the potential for problematic use or abuse. There are no recommendations concerning the use of weak opioids, although these are by far the most widely used opioids in Europe [7 9]. Chronic nonmalignant pain lasting 6 months or more affects between 12% and 30% of the adult western population [8,10]. Tissue damage is often followed by pain. How pain is perceived is, however, not only related to the extent of the damage, but also to other more subjective factors such as personality traits and mental state. This is even more so for musculoskeletal pain, which is not related to malignant disease or trauma. The most commonly reported locations for pain are in the back, knee, neck, hand, or leg. It has been documented that pain has an impact on employment status, family relationships and emotional status [8,10]. Breivik et al. documented that more than 20% of pain patients had been diagnosed with depression [8]. The likelihood of an individual taking pain medication may also be influenced by personality and emotional characteristics. The most commonly prescribed medicines for pain are nonsteroid anti-inflammatory drugs (NSAIDs) and weak opioids. Use of opioids for chronic noncancer pain is increasing in many countries [8,11,12]. A Danish study, based on data from 2000, indicated that 9% of those reporting chronic nonmalignant pain used weak opioids 805

2 Skurtveit et al. regularly or continuously [13]. Approximately 3% of the US general population without cancer used opioids regularly for a month or more per year [14]. However, the epidemiology of opioid-taking behaviors of noncancer patients has not been fully investigated [15]. An association between the use of analgesics and of anxiolytics or hypnotics has been suggested in crosssectional studies [7,16]. In an earlier prospective study, we have shown that earlier use of benzodiazepines was associated with increased prescription of opioids 20 years later in all those individuals who also consumed alcohol, even after controlling for several potential confounders. Only a weak association was observed between earlier use of benzodiazepines and later use of NSAIDs for the treatment of pain [17]. However, we were unable to determine whether this study result was related to different levels of pain in the two groups. In current study, the association between the use of benzodiazepines and later opioid use was investigated in different age groups, adjusting for musculoskeletal pain. Data from the Norwegian health surveys were linked to data from a prescription database 4 7 years later. We excluded patients with cancer pain and restricted the study to participants who reported no use of opioids at baseline. We adjusted for a series of other factors that may have influenced the later use of opioids. Materials and Methods Health Survey Information on Independent Variables Data from population based surveys conducted in in the three Norwegian counties Oslo, Oppland and Hedmark are included in this study [18]. These three counties cover both rural and urban regions. The surveys are part of Cohort of Norway (CONOR) [19]. The number of individuals invited and the number of participants in CONOR, as well as the procedures used, have been published elsewhere [19]. All subjects in selected age cohorts were invited to participate in the respective surveys. At the time of investigation, the subjects were aged 40 41, 45 46, and years. In these age cohorts, 21,802 men and 21,451 women were invited. Of these, 9,386 men (43.1%) and 11,244 women (52.4%) participated and agreed to the storage of blood samples and data for research purposes and agreed to have their data linked to other health registers (Figure 1). One hundred twenty-two men and 124 women who died or emigrated before January 1, 2004 were excluded from the analyses. In a second step, we removed all those whose drugs were reimbursed for the treatment of cancer (Figure 1). The participants in the health survey completed a selfadministered questionnaire covering drug use, history of cardiovascular disease and diabetes, and different lifestyle Figure 1 Flow chart for the study population. Health surveys in Oslo, Oppland and Hedmark ( ). *Subsample of the study population who had answered a second questionnaire including question regarding chronic pain. BZD = benzodiazepine. 806

3 Benzodiazepines Predict Use of Opioids Table 1 Questions on drug use and other factors with answer alternatives used in the health survey in Oslo, Oppland and Hedmark ( ) Self-Administered Questionnaire Anxiolytics How often in the course of the last 4 weeks have you taken tranquillizers medicine? Hypnotics How often in the course of the last 4 weeks have you taken sleeping medicine? Analgesics How often in the course of the last 4 weeks have you taken painkillers, on prescription? Open question on use of medicine State the name of the medicine you have taken during the last 4 weeks Alcohol consumption How often have you consumed alcohol in the course of the past year? Smoking habits Have you smoked/do you smoke daily? Physical activity Description of four levels of physical activity in spare time. Cardiovascular diseases Previous myocardial infarction, angina, stroke, diabetes Working status Are you receiving disability pension (full or part)? Musculoskeletal pain Have you suffered from pain and/or stiffness in muscles and joints in the course of the last 4 weeks? (neck/shoulders; arms/hands; upper back;lower back; hips/legs/feet; elsewhere) Have you, during the past year, suffered from pain and/or stiffness in muscles and joints that has lasted for at least 3 months? Answer Alternatives Daily; every week, but not daily; less than every week; not used during last month Daily; every week, but not daily; less than every week; not used during last month Daily; every week, but not daily; less than every week; not used during last month 4 7 times pr week; 2 3 times pr week; once pr week; 2 3 times pr.mth; about once pr mth; a few times in the past year; have not drunk alcohol this past year; have never drunk alcohol Yes, now; yes, earlier; no Mean activity (hours pr week) in the course of the past year: none;less than 1, 1 2, 3 or more Yes; no Yes; no Not troubled; somewhat troubled; very troubled Yes; no habits such as alcohol use, smoking, and physical activity (Table 1). Drug use questions consisted of specific questions related to the frequency of use of different drug categories during the previous 4 weeks, followed by an open question on drug trade names [18]. It was possible to register up to 10 different trade names in the selfadministered questionnaire. Only 0.1% of attendees in the surveys reported ten different trade names, suggesting that very few, if any, were using more. The questionnaire included a specific question on the use of analgesics ( pain medicine in questionnaire), anxiolytics ( tranquillizers ) or hypnotics ( sleeping medicine ) on prescription (Table 1) [18]. Validity on drug use questions are described earlier [20]. The participants who answered that they used anxiolytics (whether daily, weekly, or less than weekly) during the previous month in the specific question and/or who wrote trade names of benzodiazepines in the open question, are referred to below as benzodiazepine users. We chose to examine the effect of only benzodiazepines on later use of opioids, not the benzodiazepine-like zopiclone and zolpidem because they have well known negative effects that interfere with people s daily functioning [21]. Opioid users were defined as individuals who wrote in open question down the trade names of opioids used during the previous month. Regarding alcohol use, participants were divided in three categories: nondrinkers, alcohol use 3 times a week and alcohol use 4 times a week. The questionnaire included questions about the location, intensity and duration of musculoskeletal pain [18]. Participants were divided in three groups: no painful areas; some/very troubled in one to two painful areas or some/very troubled in three to six areas. Physically active 807

4 Skurtveit et al. individuals were those doing exercise for at least 3 hours a week. A physical examination comprised measurements of weight, height, and blood pressure. In addition, a second questionnaire handed out at screening included a question regarding chronic pain: Have you, during the past year, suffered from pain and /or stiffness in muscles and joints that has lasted for at least 3 months? A total of 8,152 men and 9,436 women were defined as nonusers of opioids in according to the definition (Figure 1). Of these, we had information from 7,991 men and 9,083 women on their use/nonuse of benzodiazepines at the baseline according to the definitions above. These individuals constitute our study population and were followed up for subsequent prescription of opioids in (Figure 1). In parallel, in the subgroup who had responded to the second questionnaire including a question regarding chronic pain, 6,425 men and 7,487 women were followed with respect to subsequent prescription of opioids in Information about marital status was retrieved from the Norwegian Person Registry. Norwegian Prescription Database (NorPD) Information on Dependent Variables Prescription data about opioids in were drawn from the NorPD, which covers the entire nation (4.7 million inhabitants). From January , all pharmacies in Norway have been legally obliged to send in electronic data on all prescriptions. NorPD contains information on all individuals living outside institutions who have received prescription drugs dispensed at pharmacies [22]. All prescriptions from ambulatory care, whether publicly reimbursed or not, are stored in the database. The drugs are classified according to the Anatomical Therapeutic Chemical (ATC) classification [23]. The data collected for our study were patient unique identifying number (encrypted), sex, age, the date the drug was dispensed, and drug information (brand name, package size, number of packages, ATC-code, defined daily dose [DDDs]). The reimbursement code is also recorded and this functions as a proxy for diagnosis. The opioids were defined by the ATC code N02A in the ATC-classification system [23]. The weak opioids used in Norway are codeine, dextropropoxyphene and tramadol, whereas the remaining opioids (ketobemidone, morphine, fentanyl, buprenorphine, hydromorphone, oxycodone, pethidine, fentanyl, pentazocine) are categorized as strong opioids. For each prescription number of DDD s dispensed are recorded. A DDD is defined as the assumed average maintenance dose per day for a drug used on its main indication in adults. DDD s for opioids are described earlier [24]. In Norway in 2007, average and median number of DDDs per prescription of opioids are 19 and 17 DDDs, respectively. Moderate to high prescription frequency of opioids was defined as at least 12 prescriptions of opioids during This cut-off was based on the number of prescriptions for the highest users (1 2 percentiles) received during the study period of 4 years. Data from the health surveys and NorPD were linked using the unique 11-digit identification number, assigned to all individuals living in Norway. The record-linkage was approved by the Norwegian Data Inspectorate. Statistics Chi-square test was used to assess equality of proportions across the groups of drug use. Odds ratio (OR) with 95% confidence intervals (CI) were estimated by logistic regression. All analyses were carried out using SPSS 14.0 for Windows (SPSS Inc., Chicago, IL, USA). The level of significance was set to P < For the fit of the logistic regression model the Hosmer Lemeshow test was applied [25]. We ran logistic regression on strata with and without chronic pain. The logistic regression model is multiplicative. Rothman has argued that interaction on an addictive scale better reflects biologic interaction [26]. An additive model is also more relevant from the view of public health as it takes the absolute level of risk into account. By defining a logistic regression model including the individual effects of two exposures and their joint effect relative to the group not exposed to any of the two, it is possible to assess additive interaction [26,27]. Relative excess risk due to interaction (RERI) with 95% confidence interval is presented [27,28]. If the confidence interval does not cover 0, there is a significant additive interaction (P < 0.05). Results The characteristics of the study population are shown in Table 2. The proportion of benzodiazepine users was higher for women (N = 479; 5.3%) than for men (N = 289; 3.6%). There was a higher prevalence of nondrinkers, current daily smokers and physically inactive people among users of benzodiazepines. There was also a higher prevalence of unmarried individuals, of people receiving disability pensions and people reporting history of cardiovascular disease among the users of benzodiazepines. Benzodiazepine users were more likely than nonusers to report musculoskeletal pain in three or more locations in the body. The incidence of moderate to high prescription frequency of opioids 4 7 years after baseline measurements was higher for men and women with a history of using benzodiazepines (Table 3). Weak opioids were dominant. Codeine combinations made up 79% of all these prescriptions followed by tramadol (12%) (data not shown). Opioids in combination with benzodiazepines were more commonly dispensed during to those who reported 808

5 Benzodiazepines Predict Use of Opioids Table 2 Baseline characteristics for men and women aged 40, 45, and 60 years who were nonusers of opioids when surveyed in Men Use of Benzodiazepines ( ) Women Use of Benzodiazepines ( ) Baseline characteristics Yes (N = 289 ) No (N = 7,702 ) P-value Yes (N = 479 ) No (N = 8,604 ) P-value Nondrinkers (%) *** *** Drinkers (4 and more times per week) (%) ** n.s Current daily smokers (%) ** *** Ex-smokers (%) NS ** Seldom/never physically active (%) * *** Age, 60 years (%) ** *** Unmarried or formerly married (%) *** *** Disability pension (%) *** *** Cardiovascular disease (%) *** *** Musculoskeletal pain 1 2 regions (%) ** *** Musculoskeletal pain 3 6 regions (%) *** *** * P value < 0.05; ** P-value < 0.01; *** P-value < (19%) had at least one missing value. 743 (10%) had at least one missing value. 91 (19%) had at least one missing value. 1,037 (12%) had at least one missing value. NS = not significant. use of a benzodiazepine at baseline in than those who did not, for both men and women (Table 3). The unadjusted OR for later being a moderate to high user of prescription opioids was 7.7 for individuals who used benzodiazepines at the baseline, compared with nonusers (Table 4). After adjustment for reported baseline musculoskeletal pain, lifestyle, socio-economic variables the OR was reduced to 3.1. People who were current daily smokers, who were on disability pension or who reported three or more types of musculoskeletal pain had an adjusted OR of 2.5 or higher. Disability pension and the Table 3 Moderate to high prescription frequency of opioids retrieved from the Norwegian Prescription Database in for men and women who were user and nonusers of benzodiazepines when surveyed in Men Use of Benzodiazepines ( ) Women Use of Benzodiazepines ( ) Prescription in Yes (N = 289) No (N = 7,702) P value Yes (N = 479) No (N = 8,604) P value Moderate to high prescription of opioids, N (%) of these; N (%) use of benzodiazepines in (5.9) 71 (0.9) *** 38 (7.9) 91 (1.1) *** 14 (82.4) 33 (46.5) ** 34 (89.5) 46 (50.5) *** ** P value < 0.01; *** P value < Moderate to high prescription frequency of opioids was defined as at least 12 prescriptions of opioids during study period NS = not significant. 809

6 Skurtveit et al. Table 4 Number (N) and percent (%) subsequent moderate-high users of opioid among participants in population based studies , who reported nonuse of opioids at baseline Moderate to High Frequency* of Opioid Prescription In the Model N(%) Unadjusted OR (n = 17,074 ) OR (95%CI) Adjusted OR (n = 15,149) OR (95%CI) Use of Benzodiazepines No 162 (1.0) 1.0 (referent) 1.0 (referent) Yes 55 (7.2) 7.7 ( ) 3.1 ( ) Age (1.1) 1.0 (referent) 1.0 (referent) (1.7) 1.7 ( ) 1.1 ( ) Sex Men 88 (1.1) 1.0 (referent) 1.0 (referent) Women 129 (1.4) 1.2 ( ) 1.0 ( ) Married Yes 100 (0.9) 1.0 (referent) 1.0 (referent) No 117 (1.8) 2.0 ( ) 1.5 ( ) Alcohol Up to 3 times a week 183 (1.2) 1.0 (referent) 1.0 (referent) Nondrinkers 18 (1.5) 1.2 ( ) 0.7 ( ) 4 and more times a week 13 (1.4) 1.2 ( ) 1.0 ( ) Smoking No smoking 42 (0.6) 1.0 (referent) 1.0 (referent) Current daily smokers 126 (2.5) 4.4 ( ) 2.8 ( ) Ex-smokers 45 (0.9) 1.7 ( ) 1.3 ( ) Physically active Yes 158 (1.1) 1.0 (referent) 1.0 (referent) No 59 (2.2) 1.9 ( ) 1.3 ( ) Disability pension No 114 (0.8) 1.0 (referent) 1.0 (referent) Yes 82 (3.4) 7.5 ( ) 3.3 ( ) Cardiovascular disease No 191 (1.2) 1.0 (referent) 1.0 (referent) Yes 23 (2.7) 2.6 ( ) 1.2 ( ) Musculoskeletal symptoms No 16 (0.4) 1.0 (referent) 1.0 (referent) 1 2 regions 67 (0.9) 2.6 ( ) 2.5 ( ) 3 6 regions 128 (2.5) 7.2 ( ) 4.7 ( ) Association between use of benzodiazepines in and other factors with moderate to high prescription frequency of opioids in Unadjusted and adjusted odds ratio (ORs) with 95% confidence interval (CI). * Moderate to high prescription frequency of drug was defined as at least 12 prescriptions of opioids during period (11%) had at least one missing value. Adjusted for all variables in the model; Hosmer and Lemeshow test (chi-square 6.0; P = 0.65). frequency of musculoskeletal pain, controlling for the effects of other factors, are even greater correlates to later opioid use than benzodiazepine use as judged by the odds ratio. But the risk difference is greater for benzodiazepine use, the prevalence of moderate to high prescription frequency of opioids in being 7.2% in benzodiazepine users at baseline compared with % in nonusers. Unmarried people had 50% higher odds than married people of becoming a moderate to high user of opioids. There was no effect of alcohol use in the overall analysis. The adjusted model seems to fit well as the Hosmer Lemeshow goodness of fit test was 6.0 with P value

7 Benzodiazepines Predict Use of Opioids Chronic pain was included in analyses of a subsample who had answered a supplementary questionnaire at screening. Higher proportion of women (41.8%) than men (32.2%) reported chronic pain. Among people reporting no use of benzodiazpines at baseline the incidence of moderate to high frequency of opioid prescription was only 0.6% in people reporting no chronic pain, increasing to 1.6% in people reporting chronic pain (Table 5). Thus the prevalence difference or excess prevalence due to chronic pain is 1%. In people reporting no chronic pain, the prevalence increased from 0.6% in nonusers of benzodiazepines to 4.1% in users. The prevalence difference or excess prevalence due to use of benzodiazepines is 3.5%.The prevalence is particularly high, 10%, in people reporting both use of benzodiazepines and chronic pain at baseline. The excess rate due to interaction is 4.9%. This means that the combined effect of use of benzodiazepines and chronic pain is 4.9 percentage points greater than the sum of the individual effects ( [ ] = [10-1.6] - [ ] = 4.9). The corresponding RERI is 9.7% ( [ ] = 9.7) with 95% CI This means that there is a significant positive interaction on the additive scale. The odds ratio for opioid use by chronic pain only was almost the same in unadjusted (OR = 2.8) and adjusted analyses (OR = 2.3), whereas the odds ratio for benzodiazepines only, and in combination with chronic pain, was markedly reduced after adjustment, from 7.1 to 4.4 and from 18.6 to 8.0, respectively (Table 5). There was no significant additive interaction in the adjusted model (RERI = 2.4 with 95% CI ). The Hosmer Lemeshow test statistic was 9.0 with P-value 0.35 suggesting adequate fit. In analyses of people with and without chronic pain separately adjusted odds ratios for opioid use by use of benzodiazepines were 4.8 (95% CI ) in people with no chronic pain and 3.6 (95% CI ) in people with chronic pain (not shown). Thus, there was a significant effect of use of benzodiazepines in both groups. Discussion In this prospective study, the use of benzodiazepines was associated with a higher prescription frequency of opioids 4 7 years later. The study suggests that the probability of being prescribed opioids is associated with previous use of benzodiazepines, independent of the patients reported musculoskeletal pain/chronic pain. The prevalence of moderate to high prescription frequency of opioids was particularly high among individuals reporting both chronic pain and use of benzodiazepines at baseline. Before adjustment for confounders there was a significant additive interaction. The analyses of the effect of benzodiazepines and chronic pain individually and in combination suggest that use of benzodiazepines is an even stronger predictor of later opioid use than self-reported chronic pain. The association with benzodiazepines was, however, markedly reduced after adjustment. Benzodiazepine users had a Table 5 Prevalence (%) and odds ratio (ORs) with confidence intervals (CI) of receiving a moderate/high prescription frequency of opioids Moderate to High Frequency of Opioid Prescription * No Chronic Pain Chronic Pain Prevalence Difference (%) Adjusted OR (n = 4,563) Prevalence N (%) (95% CI) (95% CI) N (%) (95% CI) (95% CI) Unadjusted OR (n = 5,192) Adjusted OR (n = 8,098) Prevalence Unadjusted OR (n = 8,720) Use of benzodiazepines No 50 (0.6) 1.0 (Referent) 1.0 (Referent) 71 (1.6) 2.8 ( ) 2.3 ( ) 1.0 Yes 12 (4.1) 7.1 ( ) 4.4 ( ) 33 (10.0) 18.6 ( ) 8.0 ( ) 5.9 Prevalence difference (%) Presentation of individual effects of both exposures (use of benzodiazepines; chronic pain) and their joint effect, each relative to no exposure. Reference category no use of benzodiazepines and no chronic pain. * Moderate to high prescription frequency of drug was defined as at least 12 prescriptions of opioids during study period Adjusted for age, sex, civil staus, alcohol, smoking, physical activity, disability pension, cardiovascular disease. Relative excess risk due to interaction (RERI) for unajusted model = 9.7 (95% CI ). RERI for adjusted model = 2.4 (95% CI ). 811

8 Skurtveit et al. higher prevalence of disability pension, cardiovascular disease and reported musculoskeletal pain than nonusers. Thus, chronic illness at baseline may partly explain the association between use of benzodiazepines and later use of opioids. There was, however, still a substantial effect of benzodiazepines after adjustment, suggesting an independent effect. Probably there is a complex relationship between the independent variables and the use of opioids. Whatever the mechanism, we can, however, conclude that benzodiazepine use is an important predictor of later use of opioids. This study confirms results from our earlier study on the association between use of benzodiazepines and moderate to high prescription frequency of opioids [17]. In the current study, we had a shorter observation period, 4 6 years compared with 20 in the earlier study. We were also able to adjust for reported symptoms of musculoskeletal pain. Different ages of participants in the studies should also to be mentioned. Other studies on opioid use and risk factors have been cross sectional and often based on self reported use of medicine. We are not aware of other longitudinal analyses in the general population that were able to adjust for risk factors that were present before opioid treatment was started. As expected, self-reported musculoskeletal symptoms or chronic pain were important predictors of subsequent moderate to high opioid use. A cross-sectional population study from Denmark found a higher proportion of opioid users in groups reporting pain [13]. In our study, receiving disability pension and being a current daily smoker were strong predictors for later use of moderate-high amount of opioids. This is in agreement with what has been documented in an earlier cross-sectional study [13]. One-year period prevalence shows that 9.7% of the Norwegian population was dispensed at least one opioid analgesic for noncancer related pain in 2007 [24]. Among persons with noncancer pain 77% received only one dispension per year and 1 2 percentiles received 12 and more prescription during the period of 4 years [24]. Use of strong opioids is less common among noncancer patients. The prevalence of weak opioids was 9.3% compared with 0.3% for strong opioids [29]. In Norway, there is a high consumption of short-acting codeine (12.9 DDD/ 1,000/day) and more than 70% of the opioids in our study were codeine compounds. Other weak opioids for the treatment of moderate pain are tramadol and dextropropoxyphene, but they are less frequently used in Norway [29]. Cross sectional study on codeine use in Norway in 2006 showed that codeine use in the entire Norwegian population was mainly sporadic, but that a relatively large sub-group of users were dispensed repeated prescriptions of the drug in combination with other potential drugs of abuse such as benzodiazepines and carisoprodol [7]. The use of benzodiazepines at baseline could influence the decision to start opioid medication in different ways. The use of benzodiazepines could be regarded as a proxy of psychiatric disorders. Studies have shown that psychiatric patients are more prone to pain [30] and that common psychiatric disorders (depression and anxiety) may predict initiation and continuation of opioid use in patients with chronic pain [31]. For benzodiazepines, there is also a concern about long-term use and addiction problems. Long-term therapeutic use of benzodiazepines in the population is quite common [32 34]. It is also common to use benzodiazepines and opioids in combination [7,35]. This combined use is regarded as particularly problematic for addiction potential. This is a prospective study, using an endpoint recorded in a national register. This design limits recall bias. However, the end point is based on dispensed drugs, and we do not know if the drugs were actually taken. Furthermore, this study included data on drug use from ambulatory care only, as the NorPD does not include person-specific data from inpatient care. Because our study has a maximum age for participants of less than 67 years, this potential bias is likely to be minimal. Although many potential confounders were registered at baseline, it is not possible to take all confounders into account. For example, we do not know the indication for the prescription of opioid analgesics. It could also be argued that the participants habits may have changed since the screening took place 4 7 years before the endpoint was recorded. Change in risk factors during follow-up is a problem inherent in all follow-up studies with only baseline information. Also, the information on drug use during the 4 weeks at baseline was self-reported. Underreporting of both benzodiazepines and opioids at baseline would be likely to bias the odds ratio estimates towards the null. The attendance rates were 43% in men and 52% in women. Further losses resulted from missing data on benzodiazepine and/or opioid use at baseline. We investigated selection bias by comparing dispensing of benzodiazepines and/or opioids in the period in the attendees, responders (attendees who answered the questions about opioids and benzodiazepines at baseline), and nonresponders with the general population in the same age groups. The period prevalence of benzodiazepines and/or opioids was slightly lower in the attendees than in the general population. The prevalence was further slightly lower in responders. The prevalence was generally somewhat higher in nonresponders than in responders. It is a more likely nonresponse from the most frequent drug users or even drug abusers. Even if the prevalence of drug use is lower among attendees there is not necessarily a selection bias in the association between use of benzodiazepines and opioids. To study possible selection bias we have analyzed the association between use of benzodiazepines and use of opioids in the prescription database in the period in attendees, responders, nonresponders and the general population in the same age groups. The percentage of users of benzodiazepines (at least one prescription dispensed in ) who also used opioids during the period was similar in all groups and similar to the general population, slightly above 50%. This indicates limited bias in the reported association between use of benzodiazepines at baseline and later use of opioids. 812

9 Benzodiazepines Predict Use of Opioids A dose-response relationship between the frequency of benzodiazepine drug use and total mortality has been reported [16]. Excluding individuals who died between the baseline and the establishment of NorPD would probably affect the estimates of association. As users of benzodiazepines have higher mortality rates than nonusers, the bias would be towards the null. We chose to examine the effect of benzodiazepines on later use of opioids, excluding the benzodiazepine-like zopiclone and zolpidem, which are the most commonly used hypnotics [29]. Although recent literature suggests an addictive potential for these drugs similar to that of traditional benzodiazepines [36], these substances are now more commonly used as first-choice hypnotics. We wanted to restrict our analysis to the traditional benzodiazepines, which have well known abuse potential. Opioids may cause cognitive and psychomotor impairment and are potential drugs of abuse [11,37,38]. However, a total of twelve or more prescriptions of opioids over 4 years does not necessarily indicate a problematic use of opioids, and does not indicate abuse or addictive use per se. The concomitant use of benzodiazepines and opioids in those patients who do not report musculoskeletal symptoms or chronic pain warrants further investigation. Finally, there are no randomized controlled trials on long term effects and consequences of opioid use in noncancer patients [39]. It could be argued that opioids have benefits in the short term, but in the long run may cause harm. It is therefore important to investigate predictors for moderate to high or long term use of opioids in noncancer patients [15]. Before starting pain treatment with opioids, clinicians should take into consideration the possibility of substance abuse or mental health disorders. A central issue when prescribing opioids for chronic noncancer pain is to balance the risk of problematic use of these drugs with the benefits of pain relief. Our study suggests that earlier use of benzodiazepines may predict repeated use of opioids. Conflict of Interest None of the authors have any conflict of interests to report. Acknowledgments The data collection was performed by the National Health Screening Service of Norway (now Norwegian Institute of Public Health). This paper is a part of the project The epidemiology of prescription drug use. A record-linkage study in Norway, which is financially supported by the Norwegian Research Council. References 1 Norwegian Medicines Agency. Terapianbefaling: Bruk av opioider ved behandling av langvarige, non-maligne smertetilstander en oppdatering. [Therapeutic recommendations: Use of opioids in treatment of non-malignant pain Un update]. legemiddelverket.no/templates/interpage aspx. [updated 2010 March 25]. 2 Chou R, Fanciullo GJ, Fine PG, et al., Opioid treatment guidelines. Clinical guidelines for the use of chronic opioid therapy in chronic noncancer pain. J Pain 2009;10: Breivik H. Appropriate and responsible use of opioids in chronic non-cancer pain. Eur J Pain 2003;7: Jovey RD, Ennis J, Gardner-Nix J, et al. Use of opioid analgesics for the treatment of chronic noncancer pain a consensus statement and guidelines from the Canadian Pain Society, Pain Res Manag 2003;8(suppl A):3A 28A. 5 Kalso E, Allan L, Dellemijn PLI, et al. Recommendations for using opioids in chronic non-cancer pain. Eur J Pain 2003;7: Trescot AM, Boswell MV, Atluri SL, et al. Opioid guidelines in the management of chronic non-cancer pain. Pain Physician 2006;9: Bachs L, Bramness JG, Engeland A, Skurtveit S. Repeated dispensing of codeine is associated with high consumption of benzodiazepines. Nor J Epidemiol 2008;18: Breivik H, Collett B, Ventafridda V, Cohen R, Gallacher D. Survey of chronic pain in Europe: Prevalence, impact on daily life, and treatment. Eur J Pain 2006;10: Hamunen K, Laitinen-Parkkonen P, Paakkari P, et al. What do different databases tell about the use of opioids in seven European countries in 2002? Eur J Pain 2008;12: Sjogren P, Ekholm O, Peuckmann V, Gronbaek M. Epidemiology of chronic pain in Denmark: An update. Eur J Pain 2009;13: Joranson DE, Ryan KM, Gilson AM, Dahl JL. Trends in medical use and abuse of opioid analgesics. JAMA 2000;283: Sullivan MD, Edlund MJ, Fan MY, et al. Trends in use of opioids for non-cancer pain conditions in Commercial and Medicaid insurance plans: The TROUP study. Pain 2008;138: Eriksen J, Sjogren P, Bruera E, Ekholm O, Rasmussen NK. Critical issues on opioids in chronic non-cancer pain: An epidemiological study. Pain 2006;125: Sullivan MD, Edlund MJ, Steffick D, Unutzer J. Regular use of prescribed opioids: Association with common psychiatric disorders. Pain 2005;119:

10 Skurtveit et al. 15 Simoni-Wastila L. Increases in opioid medication use: Balancing the good with the bad. Pain 2008;138: Hausken AM, Skurtveit S, Tverdal A. Use of anxiolytic or hypnotic drugs and total mortality in a general middle-aged population. Pharmacoepidemiol Drug Saf 2007;16: Skurtveit S, Furu K, Bramness JG, Tverdal A. Benzodiazepine use in all alcohol consumers predicts use of opioids in patients 20 years later-a follow-up study of 13,390 men and women aged years. Pharmacoepidemiol Drug Saf 2008;17: Norwegian Institute of Public Health. Norwegian Health surveys. Available from: eway/default.aspx?pid=238&trg=mainarea_; 5811&MainArea_;5811=5895:0:15,4225:1:0:0::0:0 (accessed March 2010). 19 Naess O, Sogaard AJ, Arnesen E, et al. Cohort profile: Cohort of Norway (CONOR). Int J Epidemiol 2008;37: Furu K, Skurtveit S, Rosvold EO. Drug use questions in Norwegian health surveys- response rate and agreement between specific and open-ended questions. Nor J Epidemiol 2003;13: Barker MJ, Greenwood KM, Jackson M, Crowe SF. Cognitive effects of long-term benzodiazepine use: A meta-analysis. Cent Nerv Syst Drugs 2004;18: Furu K. Establishment of the nationwide Norwegian Prescription Database (NorPD) new opportunities for research in pharmacoepidemiology in Norway. Nor J Epidemiol 2008;18: WHO Collaborating Centre for Drug Statistics Methodology. Norwegian Institute of Public Health. Guidelines for ATC Classification and DDD Assignment. Oslo: Norwegian Institute of Public Health; Fredhem OM, Skurtveit S, Breivik H, Borchgrevink PC. Increasing use of opioids from 2004 to 2007 Pharmacoepidemiological data from a complete national prescription database in Norway. Eur J Pain 2010;14: Hosmer DW, Lemeshow S. Applied Logistic Regression, 2nd edition. New York: John Wiley & Sons, Inc; Rothman K. Epidemiology. An Introduction. New York: Oxford University Press; Andersson T, Alfredsson L, Kallberg H, Zdravkovic S, Ahlbom A. Calculating measures of biological interaction. Eur J Epidemiol 2005;20: Hosmer DW, Lemeshow S. Confidence interval estimation of interaction. Epidemiology 1992;3: Norwegian Prescription Database. [updated 2008; cited. Available from: (accessed March 2010). 30 Garcia-Cebrian A, Gandhi P, Demyttenaere K, Peveler R. The association of depression and painful physical symptoms a review of the European literature. Eur Psychiatry 2006;21: Sullivan MD, Edlund MJ, Zhang L, Unutzer J, Wells KB. Association between mental health disorders, problem drug use, and regular prescription opioid use. Arch Intern Med 2006;166: Isacson D. Long-term benzodiazepine use: Factors of importance and the development of individual use patterns over time a 13-year follow-up in a Swedish community. Soc Sci Med 1997;44: Isacson D, Carsjo K, Bergman U, Blackburn JL. Longterm use of benzodiazepines in a Swedish community: An eight-year follow-up. J Clin Epidemiol 1992;45: van Hulten R, Isacson D, Bakker A, Leufkens HG. Comparing patterns of long-term benzodiazepine use between a Dutch and a Swedish community. Pharmacoepidemiol Drug Saf 2003;12: Solomon DH, Avorn J, Wang PS, et al. Prescription opioid use among older adults with arthritis or low back pain. Arthritis Rheum 2006;55: Hajak G, Muller WE, Wittchen HU, Pittrow D, Kirch W. Abuse and dependence potential for the nonbenzodiazepine hypnotics zolpidem and zopiclone: A review of case reports and epidemiological data. Addiction 2003;98: Hojsted J, Sjogren P. Addiction to opioids in chronic pain patients: A literature review. Eur J Pain 2007;11: Bachs LC, Engeland A, Morland JG, Skurtveit S. The risk of motor vehicle accidents involving drivers with prescriptions for codeine or tramadol. Clin Pharmacol Ther 2009;85: Kalso E, Edwards JE, Moore RA, McQuay HJ. Opioids in chronic non-cancer pain: Systematic review of efficacy and safety. Pain 2004;112: