Lwwus_ajp_200716 763.769

A Population-based Cohort Study on Chronic Pain: Per Sjøgren, MD, DMSC,* Morten Grønbæk, PhD,w Vera Peuckmann, PhD,z and Ola Ekholm, PhDw pharmaceutical companies have been the driving force, Objectives: The aims of this study were 2-fold: (1) to investigate the rather than scientific data on efficacy and safety. Caution consequences of opioid use in individuals with chronic pain in the about opioid treatment of chronic pain has long been based Danish population, and (2) to investigate the development of and on the fear of addiction and diversion of opioids into recovery from chronic pain from 2000 to 2005.
society.4 However, other important clinical issues such as Methods: Data derived from the Danish Health Interview Survey in physical dependency, tolerance development, cognitive 2000, which were linked on the individual level with register-based disorders, opioid-induced hyperalgesia, dysfunction of the follow-up data. The survey was based on a county-stratified immune and reproductive systems, and even increased random sample of 16,684 individuals, out of which 10,434 mortality may give rise to concerns.4–10 Guidelines for res- individuals (62.5%) completed a face-to-face interview and ponsible use of opioids in chronic noncancer pain condi- returned a self-administered questionnaire. In addition, a sub- tions reflect concerns over these problems.11–14 Pain clinics sample of the sample in 2000 was reinvited to a follow-up survey and centers seem to follow these guidelines, and in these in 2005. In total, 3649 individuals (61.7%) of this subsamplecompleted the interview and returned the questionnaire at baseline facilities opioid doses can be kept stable for years in the in 2000. At follow-up, 2354 of these participants completed the majority of patients.15,16 However, outside the specialized interview and returned the self-administered questionnaire. Re- treatment facilities the guidelines may either not be spondents with cancer diagnosis were excluded from all analyses.
Respondents with chronic pain were identified as having chronic/ In epidemiologic surveys excluding cancer patients long-lasting pain more than 6 months.
Eriksen et al2 showed that 3% of the Danish population Results and Discussion: The annual incidence for the development used opioids on a regular or continuous basis, and that the of and the recovery from chronic pain was 2.7% and 9.4%, opioid usage was significantly associated with reporting of respectively. Increasing age up to 64 years, short education, poor high pain intensity, poor functional capacity, and health- self-rated health, high body mass index, and physical strain at work related quality of life.17 Owing to the cross-sectional nature were predictors of chronic pain. The odds of recovery from chronic of this study, causality could not be proven.17 pain were almost 4 times higher among individuals not using This study is based on data from the Danish Health opioids compared with individuals using opioids. In addition, use Interview Surveys in 2000 and 2005. The Health Interview of strong opioids was associated with poor health-related quality Surveys are nationwide surveys of adult Danish citizens of life. Furthermore, the results indicated that individuals with (16y or older), which have been carried out in 1987, 1994, chronic pain using strong opioids pain had a higher risk of deaththan individuals without chronic pain (HR: 1.67; 95% CI: 1.03- 2000, and 2005. The main purpose of these surveys is to 2.70). However, this study cannot exclude disease severity as the describe the status and trends in health and morbidity in the primary cause of increased mortality.
adult Danish population and factors that influence healthstatus.18 Owing to the fact that the survey in 2000 and in Key Words: chronic non-cancer pain, cohort study, epidemiology, 2005 was based on the very same basic questions regarding chronic pain: “Do you have chronic/long-lasting pain lasting 6 months or more?” this cohort study was muchmore accurate and reliable than the cohort from 1994 to2000, in which the pain intensity verbal rating scale (withthe recall period of 4 weeks included in the SF-36) was used Denmarkhashadanextremelyhighusageofopioidsfor to identify chronic pain.19 The aims of this study were 2- years, mainly prescribed for chronic noncancer pain fold: (1) to investigate the consequences of opioid use in the conditions.1–3 Clinical needs, recommendations from pain individuals with chronic pain in the Danish population, clinicians, and massive sales promotion activities from the and, (2) to investigate the development of and recoveryfrom chronic pain from 2000 to 2005.
Received for publication October 23, 2009; revised May 3, 2010; From the *Section of Acute Pain Management and Palliative Medicine, Rigshospitalet; wNational Institute of Public Health, University ofSouthern Denmark, Copenhagen, Denmark; and zDepartment of Anaesthesiology and Department of Palliative Medicine, RWTHAachen University Hospital, Aachen, Germany.
The Danish Health Interview Surveys were designed Reprints: Per Sjøgren, MD, DMSc, Section of Acute Pain Management and carried out by the National Institute of Public Health; and Palliative Medicine, Rigshospitalet, Blegdamsvej 9, DK-2100 however, the specific pain questions were developed by our Copenhagen ø, Denmark (e-mail: [email protected]; p.sjogren@ pain research group. Data from the Danish Health Inter- Copyright r 2010 by Lippincott Williams & Wilkins view Survey in 2000 linked with individual-level register Clin J Pain  Volume 26, Number 9, November/December 2010 Clin J Pain  Volume 26, Number 9, November/December 2010 data on vital status (ie, death or emigration) were used atc_ddd_index/). Weak opioids in Denmark are codeine, to investigate the relationship between (opioid-treated) tramadol, and dextropropoxyphene. All other opioids are chronic pain and mortality. In the baseline survey in 2000, a county-stratified random sample of 16,684 indivi- Information on long-standing diseases (circulatory duals was drawn from the Danish Civil Registration System diseases, infectious and parasitic diseases, and mental (each Dane has a unique personal registration number).
disorders) derived from an open-ended question “Do you Data in the Danish Health Interview Surveys were collected have any long-standing disease, disorder or illness, long- through personal interview at the respondents’ home and standing effects of injury, any functional impairment, or after the interview, the respondents were asked to complete any other long-standing health problem?” An affirmative a self-administered questionnaire. In total, 10,434 indivi- answer led to questions about the specific nature of the duals (62.5%) completed the interview and returned the disease. The diseases were classified according to the self-administered questionnaire at baseline. Respondents International Classification of Disease (ICD-10). Indivi- with a self-reported earlier or present cancer diagnosis were duals with diabetes were identified on the basis of responses excluded from the analyses (369 individuals). Hence, the final study population consisted of 10,065 individuals. The Self-reported height and weight were used to calculate Danish Civil Registration System was used to obtain the Body Mass Index (BMI). The physical working environ- information on vital status and the date of change of vital ment was assessed by a question regarding the physical strain status. Observation time was calculated from the interview of the main occupation among actively employed 16 to 64 date until death, emigration, or 26 November 2008 (end of years of age, and the 4 response categories were categorized into 3 groups: low (mainly sedentary work that does not To investigate the association between chronic pain require any physical effort), medium (work that is largely and potential risk factors, a subsample consisting of 5912 carried out standing or walking but otherwise does not individuals from the survey in 2000 was used. This require any physical effort), and high (standing or walking subsample was also used to examine the relationship work with much lifting or carrying, or heavy or rapid work between development of or recovery from chronic pain that is strenuous). Finally, actively employed 16 to 64 year and potentially associated factors. In total, 3649 individuals olds were asked if they often (more than twice a week) are (61.7%) of this subsample completed the interview and exposed to any of these factors at work: working while bent returned the self-administered questionnaire. Five years later over or in a twisted position; repetitive motion; heavy objects (in 2005), 3430 of these participants were available when the (at least 10 kg) to be carried or lifted.
cohort was reexamined (219 were lost to follow-up because ofdeath or emigration). In total, 2354 individuals completed the interview and returned the self-administered questionnaire at The Cox proportional hazards model was assessed to follow-up. Respondents with a self-reported earlier or present investigate the association between chronic pain (opioid- cancer diagnosis were also excluded from these analyses (112 treated and nonopioid-treated) and mortality after adjust- individuals) and, hence, the final follow-up study population ment for potentially confounding factors. The covariates included were gender, the international standard classifica-tion of education, marital status, BMI, smoking behavior, regular use of antidepressants, regular use of anxiolytics, Respondents with chronic pain were identified through self-reported circulatory diseases, infectious and parasitic the question “Do you have chronic/long-lasting pain lasting diseases, diabetes, and mental disorders. In the analysis, age 6 months or more?” The question concerning chronic pain was used as the underlying time scale, thus treating age at was asked in the self-administered questionnaire at both interview as the time of delayed entry. The proportional baseline and follow-up. Educational status was classified hazard assumption was checked graphically. The results are according to The International Standard Classification of presented as hazard ratios (HR) with 95% confidence Education, that combines school and vocational education.
Self-rated health was assessed by the question: In general, The incidence of new/recovered cases of chronic pain how would you characterize your health?: Really good; per 1000 person-years was calculated with the assumption of a date of development/recovery in the middle of the The Short Form 36 (SF-36) was also included in the follow-up period. Multiple logistic regression analysis was self-administered questionnaire.20,21 The SF-36 is a 36-item carried out to estimate the association between chronic pain survey that measures 8 dimensions of health (bodily pain; at follow-up and the possible risk factors. Multiple logistic general health; mental health; physical functioning; role regression analysis was also carried out to investigate the limitation owing to emotional problems; role limitations relationship between recovery from chronic pain at follow- owing to physical health; social functioning; vitality).
up (among individuals with chronic pain at baseline) and Higher scores on the SF-36 (range 0 to 100) indicate better potential associated factors. The results are presented as gender-adjusted and age-adjusted odds ratios (OR) with Usage of self-reported medications was obtained by an open-ended question asking whether the respondent reg- At follow-up, age-standardized mean scores of the ularly or continuously takes any medication. The self- 4 groups concerning pain status were estimated for each reported use of medications was categorized according to SF-36 domain (no chronic pain 2000 – no chronic pain The Anatomical Therapeutic Chemical (ATC) Classifica- 2005; no chronic pain 2000—chronic pain 2005; chronic tion System. In the ATC classification system, the drugs pain 2000—no chronic pain 2005; chronic pain 2000— are grouped into different groups according to the organ chronic pain 2005). Furthermore, cross-sectional data from or system on which they act and their chemical, pharmaco- 2000 were used to estimate age-standardized mean scores logic, and therapeutic properties (http://www.whocc.no/ according to the chronic pain status and the use of opioids in Clin J Pain  Volume 26, Number 9, November/December 2010 A Population-based Cohort Study on Chronic Pain 2000 for each domain. The Danish population in 2005 wasused as the standard population in the SF-36 analyses. Allstatistical analyses were done using the SAS version 9.1.
Table 1 shows the baseline sociodemographic char- acteristics of the sample and the study population for thequestionnaire follow-up study. The proportion of men isslightly lower in the final study population than in theoriginal sample. As expected, the elderly are more likely tobe lost at follow-up than younger individuals. During 81,965person-years of follow-up in the health interview survey in2000, 782 deaths occurred. A statistically significant associa-tion was found between (opioid-treated) chronic pain andmortality (P=0.0427). The results showed that individuals FIGURE 1. Hazard Ratios (HR) and 95% confidence intervals for with chronic pain using strong opioids pain had a higher risk all-cause mortality according to the chronic pain status and theuse of opioids in 2000.
of death than individuals without chronic pain (HR: 1.67;95% CI: 1.03-2.70) (Fig. 1). The results also showed that therisk of death was higher among individuals with chronic pain changes (from baseline to 5 years later) in the SF-36 domain not using opioids compared with individuals without chronic scores did not indicate a poorer health-related quality of life pain (HR: 1.21; 95% CI: 1.02-1.44). However, the analysis for opioid users than nonopioid users.
did not indicate a higher risk of death among individuals There was a clear association between combined with chronic pain using weak opioids compared with indivi- school and vocational education and the development of chronic pain. The odds for reporting chronic pain were Table 2 shows that the estimated incidence rate for higher among individuals with shorter education com- developing chronic pain in Denmark was 26.9 per 1,000 pared with participants with 15 or more years of education person-years (26.8 for men and 27.0 for women). The (Table 2). The table indicates that there was no association incidence rate increased with age up to the age of 64 and between marital status and developing chronic pain then decreased subsequently. Table 3 shows the overall (P=0.961). Table 2 also shows, that persons, who rated incidence rate for recovering from chronic pain was 94.2 their health as fair, bad, or very bad at baseline, were more per 1000 person-years in Denmark. The table shows that likely to develop chronic pain in the follow-up period (OR: the pain recovery was significantly associated with the use 2.45; 95% CI: 1.63-3.71). Furthermore, the table shows that of opioids. The odds for reporting recovery from chronic obese persons (BMIZ30) were more likely to develop pain at follow-up were almost 4 times higher among chronic pain than persons with a BMI of less than 25.
individuals not using opioids at the baseline compared with The age-standardized SF-36 mean scores for each individuals using opioids. In addition, an analysis among domain in 2005 are shown in Figure 2. The figure shows individuals with chronic pain and a fair, poor, or very poor that individuals without chronic pain at both baseline and self-rated health at the baseline showed that opioid users follow-up have the highest mean score in all 8 subscales.
were more likely to report a fair/poor self-rated health at Individuals with chronic pain at both baseline and follow-up follow-up than nonopioid users (OR: 3.89; 95% CI: 1.45- have the lowest mean scores in each domain, indicating a 10.46) (data not shown). However, analyses of the mean poor physical and mental health-related quality of life.
TABLE 1. Sociodemographic Characteristics of the Sample and the Study Populations for the Questionnaire Follow-up Study *Individuals with an earlier or present cancer diagnosis are excluded.
Clin J Pain  Volume 26, Number 9, November/December 2010 TABLE 2. Incidence Rate per 1000 Person-years and Odds Ratios Regarding Potential Risk Factors for the Development of Chronic PainAmong Individuals With No Chronic Pain at Baseline Combined school and vocational education* Individuals with high physical strain at work had 1.65 Figure 3. The figure shows that individuals without chronic (95% CI: 1.07-2.56) higher odds for developing chronic pain in pain at baseline have the highest mean score in all the 8 the follow-up period than individuals with low physical strain subscales. Individuals with chronic pain and taking strong at work (Table 4). Moreover, individuals reporting to work opioids in 2000 have the lowest mean scores in each while bent over or in a twisted position more than 2 times a domain, indicating a poor physical and mental health- week were more likely to develop pain in the follow-up period (OR: 1.70; 95% CI: 1.19-2.41) than individuals working whilebent over or in a twisted position less than 3 times a week.
Age-adjusted SF-36 mean scores according to chronic Randomized controlled studies of long-term opioid pain status and use of opioids in 2000 are shown in treatment in chronic noncancer pain patients are generally TABLE 3. Incidence Rate per 1000 Person-years and Odds Ratios for Recovery From Chronic Pain Among Individuals With Chronic Painat Baseline Working while bent over or in a twisted positionwYes Heavy objects (at least 10 kg) to be carried or liftedYes *Adjusted for the potential confunders gender, age, combined school and vocational education, BMI, and self-rated health.
Clin J Pain  Volume 26, Number 9, November/December 2010 A Population-based Cohort Study on Chronic Pain must question if the controlled randomized trial is theoptimal form of evidence for assessing opioid treatment ofchronic noncancer pain given the artificiality of the trialsetting, the tendency of trials to select “ideal” patients, andthe lack of generalizability to the general population that isbeing treated outside trials. To assess the consequences andthe broader role of liberal opioid consumption in westernsocieties, attention must be given to different sources ofinformation such as population-based studies.2,18 To our knowledge, very little data exist regarding opioid use and mortality in individuals with chronicnoncancer pain. However, a 2000 to 2001 national surveyfrom the US of medical examiners’ reports of deathsattributable to prescription of oxycodone use23 and a reportfrom Utah24 documenting a dramatic increase in accidentalpoisoning death owing to prescription opioids, are worri- FIGURE 2. Age-adjusted SF-36 mean scores at follow-up (2005) some. Furthermore, a study in opioid dosing trends and according to chronic pain status at baseline (2000) and follow-up motality from 1996 to 2002 in Washington State workers found that the general increase in opioid use and the shiftfrom weaker to stronger opioids were associated with of short duration22 and long-term follow-up studies are few an increase in workers’ deaths attributable to accidental and often carried out in meticulously selected patients.15,16 overdose of prescription opioids.10 These authors also Although these studies have mainly positive outcomes the speculate that the increase in opioid dosing could be experience outside the frames of carefully controlled and ascribed to the development of pharmacologic tolerance or time-limited studies has not been entirely positive, and the opioid-induced hyperalgesia. However, these data from the limitations of current evidence in terms of assessing the US may have little to do with the findings in our consequences of the extensive and liberal use of opioids population-based cohort study and owing to latency of in noncancer pain seem to be critical.17 Furthermore, one the Danish Causes of Death Registry, the causes of death TABLE 4. Incidence Rate Per 1000 Person-years and the Results of Multiple Logistic Regression Analyses Showing Odds Ratios forPotential Work-related Physical Risk Factors of Chronic Pain Development Among Actively Employed 16-64 y Old *P<0.05.
BMI indicates body mass index.
Clin J Pain  Volume 26, Number 9, November/December 2010 pain in 2000 and 2005. In the survey 1994 to 2000, the painintensity verbal rating scale (with the recall period of 4weeks included in the SF-36) was used to identify chronicpain.19 The estimated annual incidence recovery rate from chronic pain was 9.4%/year. We have formerly reported asomewhat lower annual pain recovery incidence rate of8.7%, however, the abovementioned limitations of theearlier study should be taken into account.19 A noteworthyfinding of this survey is that the odds of recovery fromchronic pain was 4-fold decreased in individuals usingopioids, and, in contrast to earlier cross-sectional studiesfrom our research group,2,17 causality could be establishedin this cohort study.
FIGURE 3. Age-adjusted SF-36 mean scores at baseline accord- In accordance with the former surveys by our research ing to chronic pain status and the use of opioids in 2000.
group, we found that high age, short education, poor self-rated health, and high BMI were predictors of chronic are not yet available for this study. Thus, we can only pain.2,19 However, in contrast to earlier surveys by our speculate that some of the long-term consequences of group and others, we could not identify the female gender opioid use may be involved. Addiction, opioid-induced and marital status as predictors of pain.19,25 Other investi- hyperalgesia, and cognitive dysfunction may cause de- gators have reported that psychological distress is strong pressed mood and poor judgement involving suicide and predictor of chronic pain.28,32 Owing to the investigational hazards, and dysfunction of the immune and reproductive design, associations between psychological distress and systems may course increased morbidity and mortality chronic pain could not be evaluated in this survey.
for example owing to infections.4,5,7–9 However, accidental In a cross-sectional study by Eriksen et al,2 it was overdosing may also be among the death causes in our shown, that high physical job strain was associated with reporting of long-term/chronic pain. However, neither Furthermore, our study also indicated that chronic physical strain of job nor heavy workload was found to pain itself may increase mortality (Fig. 1). During a be significant predictors for development of or recovery 12-years follow-up population study from the south from chronic pain.19 In this survey, more detailed questions of Sweden, a significantly increased mortality was found regarding the impact of physical strain at work indicated in individuals with widespread chronic noncancer pain.25 that high physical strain at work predicted development of Similar to our study, the causes of death could not be chronic pain (Table 4). Furthermore, health-related quality elucidated in the Swedish study, but the authors suggested of life as measured by SF-36 was reduced most severely in that the influence of distress and pain on the immune those individuals suffering from chronic pain during the system may be the cause.25 Finally, in our study and in the other studies, chronic pain and opioids may likely be A major strength of this study is that it is based on involved in some of the covariates we adjusted for in the large national representative survey with an adequate statistical model for example high BMI, smoking behavior, response rate. However, nonresponders may pose problems self-reported circulatory diseases, infectious and parasitic in all studies based on survey data. Hence, we compared diseases, diabetes, and mental disorders.26 mortality rates among responders and nonresponders in the We have no ready explanation for the finding that the baseline survey in 2000. We found a lower mortality rate so called weak opioids (in Denmark: tramadol, codeine, among responders (12.8 per 1000 person-years) than among and dextropropoxyphene) superimposed on chronic pain nonresponders (19.8 per 1000 person-years). Furthermore, did not contribute to increased mortality (Fig. 1). On the we found that the elderly were more likely to be lost at basis of the study by Franklin et al,10 it could be speculated follow-up than younger individuals in the questionnaire that development of tolerance/opioid-induced hyperalgesia follow-up study. These findings were as expected and there and other potential consequences owing to dose increase is no indication that nonresponse has seriously biased the have been limited in this group and some of the benefits of results of this study. It may be argued that self-reporting of improved analgesia may be preserved. However, further opioid use may be unreliable, however, a recent study, based on data from the Danish Health Interview Survey in Prospective, longitudinal studies are mandatory to 2000, showed a good agreement (Cohen k value: 0.62; 95% investigate the incidence and/or recovery rates of chronic CI: 0.58-0.67) between self-reported use of opioids and pain in the general population to study the causes and national prescription records.33 A k value between 0.61 and effects of the chronic pain.19,27–31 Most of the studies have a limited number of participants and have almost always In conclusion, the annual incidence for development followed persons reporting pain at baseline. Few studies of and recovery from chronic pain was 2.7% and 9.4%, have estimated the numbers of new or recovered cases of respectively. Increasing age up to 64 years, short education, chronic pain.19,31 In this study, we found an average annual poor self-rated health, high BMI, and physical strain at incidence rate for developing chronic pain of approximately work were predictors of chronic pain. The odds of recovery 2.7%, which is slightly higher than the annual incidence we from chronic pain were almost 4 times higher among formerly have reported from 1994 to 2000 in the Danish individuals not using opioids compared with individuals population.19 However, in this survey, the incidence rate is using opioids. Furthermore, chronic pain and use of strong considered more accurate and reliable than the former, as it opioids was associated with poor health-related quality of was based on the very same questions regarding chronic life (both physical and mental). In addition, chronic pain Clin J Pain  Volume 26, Number 9, November/December 2010 A Population-based Cohort Study on Chronic Pain and strong opioid use seem to be a risk factor for mortality, 18. Ekholm O, Hesse U, Davidsen M, et al. The study design and although this study cannot exclude disease severity as the characteristics of the Danish national health interview surveys.
primary cause of increased mortality.
Scand J Public Health. 2009;37:758–765.
19. Eriksen J, Ekholm O, Sjøgren P, et al. Development of and recovery from long-term pain. A 6-year follow-up study of a cross-section of the adult Danish population. Pain. 2004;108: 1. Joranson DE, Ryan KM, Gilson AM, et al. Use, trends in medical use and abuse of opioid analgesics. JAMA. 2000;283:1710–1714.
20. Ware JE, Snow KK, Kosinski M, et al. SF-36 Health Survey 2. Eriksen J, Jensen M, Sjøgren P, et al. Epidemiology of chronic Manual and Interpretation Guide. Boston, MA: New England non-malignant pain in Denmark. Pain. 2003;106:221–228.
Medical Center, The Health Institute; 1993.
3. Jarlbaek L, Andersen M, Kragstrup J, et al. Cancer patients’ 21. Bjørner JB, Thunedborg K, Kristensen TS, et al. The Danish share in a population’s use of opioids. A linkage study between SF-36 health survey: translation and preliminary validity a prescription database and the Danish Cancer Registry.
studies. J Clin Epidemiol. 1998;51:991–999.
J Pain Symptom Manage. 2004;27:36–43.
22. Kalso E, Edwards JE, Moore RA, et al. Opioids in chronic 4. Højsted J, Sjøgren P. Addiction to opioids in chronic pain non-cancer pain: systematic review of efficacy and safety. Pain.
patients: a literature review. Eur J Pain. 2007;20:451–455.
5. Mao J. Opioid-induced abnormal pain sensitivity: implications 23. US Department of Justice, 2002. Drug Enforcement Agency.
in clinical opioid therapy. Pain. 2002;100:213–217.
Summary of medical examiners reports on oxycodone- 6. Ballentyne JC, Mao J. Opioid therapy for chronic pain. N Engl related deaths, May 16. http://www.deadiversion.usdoj.gov/ drugs_concern/oxycodone/oxycodone.htm.
7. Vallejo R, de Leon-Casasola O, Benyamin R. Opioid therapy 24. Caravati EM, Grey T, Nangle BRT, et al. Increase in and immunosuppression. A review. Am J Ther. 2004;11:354–365.
poisoning deaths caused by non-illicit drugs-Utah, 1991-2003.
8. Rajagopal A, Vassilopoulou-Sellin R, Palmer JL, et al.
http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5402a1.htm.
Symptomatic hypogonadism in male survivors of cancer with chronic exposure to morphine. Cancer. 2004;100:851–858.
25. Andersson HI. The course of non-malignant chronic pain: 9. Sjøgren P, Christrup LL, Petersen MA, et al. Neuropsycho- a 12-year follow-up of a cohort from the general population.
logical assessment of chronic non-malignant pain patients treated in a multidisciplinary pain centre. Eur J Pain. 2005;9:453–462.
26. Zhu K, Devine A, Dick IM, et al. Association of back 10. Franklin GM, Mai J, Wickizer T, et al. Opioid dosing trends and mortality in Washington State Workers’ Compensation, mobility, and quality of life in elderly women. Spine. 2007;32: 1996-2002. Am J Industrial Med. 2005;48:91–99.
11. Kalso E, Allan L, Dellemijn PLI, et al. Recommendations for using 27. Waxman R, Tennant A, Helliwell P. A prospective follow-up opioids in chronic non-cancer pain. Eur J Pain. 2003;7:381–386.
study of low back pain in the community. Spine. 2000;25: 12. The Pain Society. Recommendations for the appropriate use of opioids for persistent non-cancer pain. A consensus statement 28. Croft PR, Lewis M, Papageorgiou AC, et al. Risk factors for prepared on behalf of the Pain Society, the Royal College of neck pain: a longitudinal study in the general population. Pain.
Anaesthetists, the Royal College of General Practitioners and the Royal College of Psychiatrists. March 2004. www.british 29. McBeth J, Macfarlane GJ, Hunt IM, et al. Risk factors for persistent chronic widespread pain: a community-based study.
13. Trescot AM, Helm S, Hansen H, et al. Opioids in the management of chronic non-cancer pain: an update of 30. Bergman S, Herrstro¨m P, Jacobsson LTH, et al. Chronic American Society of the Interventional Pain Physicians’ widespread pain: a three year follow-up of pain distribution (ASIPP) Guidelines. Pain Physician. 2008;11:5–62.
and risk factors. J Rheumatol. 2002;29:818–825.
14. Chou R, Fanciullo GJ, Fine PG, et al. Clinical guidelines for 31. Elliott AM, Smith BH, Hannaford PC, et al. The course of the use of chronic opioid therapy in chronic noncancer pain.
chronic pain in the community: results of a 4-years follow-up 15. Jensen MK, Thomsen AB, Højsted J. 10-year follow-up of chronic 32. Croft PR, Papageorgiou AC, Ferry S, et al. Psychological non-malignant pain patients: opioid use, health related quality of distress and low back pain: evidence from a prospective study life and health care utilization. Eur J Pain. 2006;10:423–433.
in the general population. Spine. 1995;20:2731–2737.
16. Portenoy RK, Farrar JT, Backonja MM, et al. Long-term use 33. Nielsen MW, Søndergaard B, Kjøller M, et al. Agreement of controlled-release oxycodone for noncancer pain: results of between self-reported data on medicine use and prescription a 3-year registry study. Clin J Pain. 2007;23:287–299.
records vary according to method of analysis and therapeutic 17. Eriksen J, Sjøgren P, Bruera E, et al. Critical issues on opioids group. J Clin Epidemiol. 2008;61:919–924.
in chronic non-malignant pain: an epidemiological study. Pain.
34. Altman DG. Practical Statistics for Medical Research.

Source: http://www.dspmc.univr.it/documenti/Avviso/all/all521980.pdf