- That opioid utilisation has increased in Wales in the last 10 years and varies between different Health Boards, population demographic and type of prescriber
- That there is association between individual patients’ use of opioids and their use of health care resources and that this relationship can be determined by the opioid being used, persistence of use and the doses prescribed
This study aims to:
- Describe the trends in opioid prescribing in the Welsh Primary Care Population, over a 10 year period (2005 – 2015)
- Determine any causal or potential association between opioid use, non-pain morbidities and the utilisation of other health care services, in primary, community or acute care settings
- Develop an economic model that enables the full costs of opioid use in chronic, non-cancer pain to be more accurately determined, in order to inform the development of more effective services for pain management and improved use of medicines
To describe the trends in opioid prescribing
- Opioid utilisation and prescribing patterns evaluation
- To evaluate the prescribing data of opioid prescriptions and to explore the type of prescriber (medical or non-medical), geographic (Health Board), patient age and deprivation level variations
- To investigate the impact of major clinical guidance and recent legislative changes on opioid prescribing pattern
- Opioid prescribing persistence in chronic, non-malignant pain management
- To determine the persistence of opioid prescribing and assess whether the persistence and pattern of prescribing can be a proxy for measuring adherence, drug diversion and misuse
- To determine the potential factors contributing to persistence of long-term opioid prescribing including the nature of Chronic Pain (CP) being treate
- Clinical outcomes related to long-term opioid utilisation
- To assess adverse effects which are associated with short-term and long-term opioid use
- To compare the mortality rate between opioid and non-opioid users in patients with chronic, non-malignant pain conditions
To determine associations between opioid use, morbidity and health care utilisation:
- Association between opioid use and health care utilisation
- To assess the frequency of primary and secondary healthcare referral or attendance by patients on long-term opioids therapy compared to the general population (patients not receiving opioids or who have a pain diagnosis on record)
- To assess if any association is determined by the type of opioid medication (e.g. weak or strong opioid) being prescribed, the doses administered or the duration of use
To evaluate the cost of opioid use in chronic, non-cancer pain:
- Estimation of the ‘true cost’ of opioid prescribing
- To develop economic descriptions of the impact of health care utilisation associated with long-term opioid therapy which may then be used in the development of strategies to change how opioid prescribing is used in CP
Pain is generally defined as ‘an unpleasant sensory and emotional experience, associated with actual or potential tissue damage, or is described in such terms’. (Merskey H, Bogduk N, 1994) Chronic pain (CP) is accepted to be pain that has persisted for longer than three months and can be more accurately described as chronic, non-malignant pain (CNMP) or chronic non-cancer pain (CNCP), which helps to further differentiate it from pain associated with end-of-life morbidities, tumours or chemotherapy.
Chronic pain may result from nociceptive or neuropathic aetiologies and some people will display elements of both. People living with CP will commonly have comorbidities such as sleep disturbance, anxiety and depression in addition to physical debilitation. (Martelli MF et al. 2004) Severe CP also adversely affects employment, relationships and people’s general health, for example daily back pain is associated with a higher incidence of coronary events (National Pain Audit, 2012). The complex interplay of bio-psychosocial factors results in CP causing a reduction in the quality of life of sufferers more than almost any other condition (Chief Medical Officer, 2009; Welsh Government, 2015).
The reported prevalence of CP varies considerably depending on the type of pain being recorded and the population being studied. The Chief Medical Officer’s report 2008 gave an estimate of 7.8 million people in the United Kingdom being affected by moderate to severe pain of 6 months duration or more. This equates to over a third of households having someone in pain at any given time (Chief Medical Officer, 2009). The high economic burden of CP in addition to poor quality of life has resulted in increasing attention being paid to its treatment over the last 10 years.
In Wales as in the rest of the UK, it is estimated that around 13% of the population are likely affected by chronic pain (Breivik H et al, 2006). The Welsh Government published a commissioning directive in 2008, aimed at improving access to multi-disciplinary, multi-modal pain services – the intimation being that services were at that time being developed around the skills and experience of the professionals running the services rather than patient need (Welsh Government, 2008).
The use of opioids to treat pain has been established for thousands of years and commonly accepted for the treatment of cancer and acute pain e.g. pain associated with trauma or surgery (Kalso E et al, 2004). Over more recent times, the use of opioids in the treatment of CP has increased. Some attribute the change in attitude to opioids to a sentinel paper by Portenoy and Foley in the late 1980s that appeared to open the door to a more liberal approach to prescribing in non-cancer pain (Portenoy, 1986) despite the available evidence being limited to short-term efficacy and side-effects. It is thought that opioids are likely to have less effect over time periods longer than 12 weeks and beyond certain doses the harms of continuation outweigh any therapeutic benefits (Opioids Aware, 2015).
Concerns over the risk of dependence, tolerance and addiction have been raised (Chou et al, 2009; Franklin, GM, 2014).
Despite this, over the last 10 to 15 year period, Wales has seen an increase in analgesic prescribing, which fits with general trends in the UK, Europe and the United States. (Zin et al., 2014; Ruscitto et al., 2014; Wright et al., 2014) Fredheim, 2010; Olsen et al., 2006). In the UK, a doubling of opioid prescriptions was seen in the five-year period to 2011, (NTA, 2011). Although the exact reasons for this trend have still to be fully elucidated more recent studies have demonstrated the increase to be associated with use in CP (Zin et al., 2014).
Patients’ adherence to opioids, the incidence of adverse-effects, risks of dependence and addiction and the impact on patients’ general health are not clearly defined in a Welsh or UK-wide population. The British Pain Society initially published good practice guidelines for prescribing opioids in persistent pain in 2004 (BPS, 2004) and this is now updated to an online, evidence-based resource (Opioids Aware, 2015). The effect in practice of such initiatives is unclear beyond what can be elucidated from prescribing data.
The current study attempts to begin to address the overall use of opioids in the Welsh population and to develop an economic model to examine the impact of existing patterns and potential benefits of future change. The effectiveness and safety of long-term opioid use in the management of chronic pain will be explored as well as any association with differences in other health care utilisation or clinical morbidity in those receiving opioids for CP compared to those who do not.
Plan of investigation
This study is in two parts; firstly an observational study to analyse opioid utilisation and patient characteristics as well as to assess the association between long-term opioid exposure and clinical outcomes including the use of non-pain related health care services in Wales.
Secondly, a cost of illness analysis that accounts for all facets of opioid prescribing and which will allow a realistic estimation of the cost of chronic pain and the use of opioid medication.
Prescribing trends study
Cross sectional study design: This will be achieved via a quantitative, observational study conducted using a cross-sectional (with repeated cross-sections) and longitudinal cohort study designs, during a ten-year period from 2005 – 2015. Data will be accessed via the Secured Anonymised Information Linkage (SAIL) Databank, which provides hospital episode statistic (HES) data in addition to primary care prescribing and other health care interactions.
All patients aged 18 years or over (adult), at any point in the study period (2005 – 2015) and prescribed any of the identified opioid drugs during that period, will be included in the cross-sectional study. Identification of patients will be solely by the prescription of opioids, using specific drug product codes from the SAIL datasets.
The commonly prescribed opioid or opioid-containing drugs selected as study drugs for this purpose include; codeine phosphate, codeine and paracetamol (co-codamol), dihydrocodeine, dihydrocodeine and paracetamol (co-dydramol), buprenorphine, dextropropoxyphene, dextropropoxyphene and paracetamol (co-proxamol), tramadol, tramadol and paracetamol, morphine, oxycodone, hydromorphone, fentanyl, tapentadol, meptazinol and pethidine. Prescriptions will be identified by product code and then dose and duration data will be calculated. The outcome measures will be summed and repeatedly calculated to generate monthly and annual data.
Included patients’ opioid prescription records will be measured repeatedly during the study period (2005 – 2015) and then collated to generate a range of prescribing measures in aggregated time-series and monthly/annual time-series data in order to undertake trend analysis.
Measures on utilisation of study drugs include the number of prescriptions, the dose and the cost of opioids described. Total doses of drugs will be adjusted by defined daily dose (DDD) and morphine equivalent dose (MEQ). Costs will be ascribed using data from national tariffs for prescription medication.
In this study, there is no particular selection of comparison and instead the data will be stratified by Health Board and type of prescriber (doctor or non-medical prescriber). Aggregated utilisation and time-series data will be stratified according to Health Board, demographic of the local population and the profession of the prescriber to determine any variation.
Various legislative changes e.g. controlled drug legislation post-Shipman, change in controlled drug scheduling of tramadol in 2014, introduction of new drug-driving legislation in 2015 will be used as marker points. Any changes in opioid prescribing after those dates may then be attributed to the guidance or legislation changes.
Cohort study design: All patients aged 18 or over (adult) at any point in the study period (2005 – 2015) and with coded chronic pain conditions will be included in the cohort study. All patients with chronic cancer pain or a cancer diagnosis at any point in their medical history will be excluded from this arm of the study. (Due to the potential problems with coding in primary care in particular, in order to ensure that patients receiving opioids for chronic, non-cancer pain only are included in the cohort study, it will be necessary to identify all patients who have ever received a cancer diagnosis and then remove their data from this section of analysis).
Chronic pain conditions include back pain, chronic low back pain, spinal pain, neck pain, osteoarthritis, inflammatory arthritis (ankylosing spondylitis, psoriaritic arthropathy, rheumatoid arthritis), painful diabetic neuropathy, post-herpetic neuralgia, trigeminal neuralgia and phantom limb pain. For clinical outcomes, only patients with complete records including hospital episode statistics (HES) will be included in the final analysis.
Adult patients with specific chronic, non-cancer pain conditions will be selected from 2005 at the beginning of the study period and grouped into either non-opioid users (control group), short-term opioid users (<6 months) or long-term opioid users (>6 months).
In order to control confounding for clinical outcomes, the intervention and control groups will be matched by age, gender, Charlson Comorbidity Index (CCI) (Quan, 2011), inclusion day, opioid history (2004-2005) and deprivation score.
Patients will be followed from the date of the first prescription of any of the study drugs – within the study period to six months after the final prescription covering date or the end of the study period. This mirrors the selections of Zin et al and would allow comparison with combined data for Wales and England at a later time (Zin et al., 2014).
Prescribing persistence is defined as, ‘the duration of time from initiation to discontinuation of therapy’ (Cramer et al., 2008). Prescription of study drugs, prescribed dose and duration of prescribing will be analysed for individual patients. The duration of prescribing will be used to calculate exposure days, prescription covering days and to then estimate the medication possession ratio (MPR). The MPR will be further adjusted by covariates such as socio-demographic characteristics (e.g. age at diagnosis, sex, deprivation scores), smoking and alcohol use, co-morbid conditions, number of drugs prescribed and disease-related pain conditions. The prescribed opioid dose will be used to evaluate individual opioid consumption. The trends in dosing of opioids will also be adjusted by covariates.
Adult patients with specified chronic pain indications are included and the corresponding opioid drug utilisation data will be analysed. Evaluating prescribing persistence and medication possession ratio (MPR) will identify patients who have continuous short-term or long-term exposure to any of the study drugs. Patients identified with a CP indication but without an opioid prescription will be entered into the control group.
Continuous use is assumed when the gap between each opioid prescription-covering period is less than 30 days. Short-term use refers to the accumulated period of opioid prescription covering less than 180 days (6 months). Long-term use refers to the accumulated prescription period covering more than 180 days (6 months). The intervention (long-term users) and control groups will be selected by 1:1 matching in age, gender, CCI, inclusion day, opioid use history (2004 – 2005) and deprivation score.
Prescription data for the study of opioid drugs for individual patients will be collected; number of items and where possible, dose data and duration of prescribing will be used to assess prescribing persistence. The medication possession ratio (MPR) (calculated from number of days of medication supplied within the refill period) for different opioids will be compared between different types of CP.
Clinical outcomes will be analysed using read codes for the most common adverse affects (ADE) of opioids (constipation, nausea, dizziness and drowsiness and vomiting) and will be used to search the records of all adult patients with a CP indication and opioid prescription. Proxy measures such as anti-emetic and laxative prescriptions will be evaluated. Comparison will be made of ADEs and their frequency between long-term and short-term opioid users and further comparison made between those and the control group in order to determine ADE rates between the groups and to establish whether an association can be made between opioid use and duration of use.
Mortality rates of long-term and short-term opioid users will be calculated and compared to each other and to the mortality rate within the control group.
Association between opioid use, morbidity and health care utilisation
Study design: Health care utilisation in all sectors including those related to managing pain or opioid related outcomes will be captured using ICD-10 and OPCS-4 coding in accordance with the data recording methods for the Patient Episode Database for Wales (PEDW), which collects hospital data. Long-term ADE such as endocrine, immune dysfunction, sexual dysfunction and infertility will be examined, by proxy measures if necessary such as GP and outpatient attendances. Time to events will also be measured. Data will be grouped according to long-term or short-term opioid use and compared to the control group of patients not receiving opioid prescriptions.
Health care utilisation data in patients receiving opioid prescriptions will be further analysed by categorising into ‘weak’ and ‘strong’ opioids as per the World Health Organisation (WHO) analgesic ladder and then further by dose banding of <20mg, 20-49mg, 50-99mg, 100-199mg and >200mg daily MEQ and as described in previous work (Dunn et al., 2010; Gomes et al., 2011). Duration of prescribing will be a secondary measure to analyse the association between prescribed opioid dose and healthcare interactions.
To Estimate the cost of opioid use in chronic, non-cancer pain:
- Estimation of the cost of opioid prescribing
- To develop economic descriptions of the impact of health care utilisation associated with long-term opioid therapy which may then be used in the development of strategies to change how opioid prescribing is used in CP
Sample size / Power calculations
It has been estimated that 13% of the UK population suffer with moderate to severe pain (Breivik, 2006), which, equates to 400,000 people in Wales, a country with a population of around 3 million (Statistics Wales, 2015). In the European Union (EU), 13% of patients with persistent pain appear to be prescribed weak opioids and 3% strong opioids (Langley, 2011).
Using the estimated prevalence of chronic pain above, this suggests 52000 patients receive prescriptions for weak opioids and 12000 patients, strong opioids in Wales. However, there has been a marked increase in prescribing of opioid analgesics within primary care in Wales from 1 million items in 2007 to 1.4million items in 2013, an increase of nearly 40%, meaning that these figures could be an underestimate.
The SAIL database in Wales as a whole represents over 70% of Primary Care Practices (General Practitioners). Therefore, the Law of Large Numbers and Central Limit Theorem dictate that it should be expected to estimate these effects accurately. In fact, the Law of Truly Large Numbers may be used even to detect differences in rare events, if it is found to be relevant to the study.
In clinical trials of short duration, over half (51%) of patients taking opioids within clinical trials experience at least one treatment emergent adverse effect (Moore, 2005). Recent data from the United States (US) suggest doses of morphine 200mg (or equivalent dose of an alternative opioid), were associated with a nearly 3-fold increase in the risk of opioid-related mortality (Gomes, 2011).
Since long-term opioid exposure information is limited in Wales, there are currently no estimates with which to determine power-based sample sizes for detecting the association between long-term opioid exposure and clinical outcomes. Similarly, no estimates can be made for power-based sample sizes for quantifying any apparent causal relationship.
Data / Statistical Analysis
Records with missing dates of birth, diagnosis and drug, which consequently, are unable to identify age, disease conditions and drug utilisation, will be excluded from the analysis. However, as appropriate, then a progressive procedure will be in place to either incorporate or exclude missing data.
Missing data will be included in the sensitivity analysis; firstly it will be flagged and then used to compare analyses in which missing data is dropped with analyses relying on multiple imputations (using STATA Imputation by Chained Equation (ICE) procedure in Stata14. (Stata.com, 2015) This will enable the establishment of randomness and leverage of any missing data.
All datasets will be stored on SAIL using authenticated, password protected servers. Although access can be gained remotely, data extraction from the server is not possible without written authority from SAIL. The database mainframe is situated in a locked and highly secured building. Initial data extraction can only be performed by a trained coder with the appropriate level of security clearance from SAIL. All data is anonymised.
Extracted data will be double encrypted before being made available to the researcher in order to maintain its security. Individual SAIL records will not be divulged outside of the research group (researcher, supervisors and informatics support).
Information processing and data analysis will be performed on mainstream applications such as IBM SPSS©, in addition to specialised statistical and econometric packages like STATA 14 (Stats Corp LP, 2014. USA).
Baseline descriptive analysis will be undertaken for all variables in the study, in order to ensure data and variables are included and analysed appropriately. Simple statistical analysis methods (e.g. Analysis of Variance (ANOVA)) will be used to compare average drug utilisation across different groups. Linear trend analysis will be used on time-series data before further time-series analysis is undertaken.
For the longitudinal data, either static (cross-sectional modelling for correlation and treatment effects, including generalised linear-regression modelling) or dynamic (repeated cross-sectional and longitudinal methods including generalised estimating equations), seemingly unrelated regression and panel data methods will be used. The Cox proportional hazard model will be used to compare outcome events and time to events between groups and the results reported as hazard ratio (HR) and 95% confidence interval (95% CI).
- Breivik, H., Collett, B., Ventafridda, V., Cohen, R., & Gallacher, D. (2012). Survey of chronic pain in Europe: Prevalence, impact on daily life, and treatment. European Journal of Pain, 10(4), 287–287.
- British Pain Society. Opioids for persistent pain: Good practice. 2004
- Chief Medical Officer. Pain: Breaking through the barrier. Department of Health, 2009; 1 – 8.
- Chou R, Ballantyne JC, Fanciullo GJ, Fine PG, Miaskowski C. Research gaps on use of opioids for chronic non-cancer pain: Findings from a review of the evidence for an American Pain Society and American Academy of Pain Medicine Clinical Practice Guideline. Journal of Pain. 2009; 10 (2); 147 – 59
- Cramer, J. A., Roy, A., Burrell, A., Fairchild, C. J., Fuldeore, M. J., Ollendorf, D. A., & Wong, P. K. (2008). Medication Compliance and Persistence: Terminology and Definitions. Value in Health, 11(1), 44–47. http://doi.org/10.1111/j.1524-4733.2007.00213.x
- Dunn, K. M., Saunders, K. W., & Rutter, C. M. (2010). Overdose and prescribed opioids: Associations among chronic non-cancer pain patients. Annals of Internal …. http://doi.org/10.1059/0003-4819-152-2-201001190-00006
- Fairman, K., & Motheral, B. Evaluating Medication Adherence: which measure is right for your program? Journal of Managed Care Pharmacy, 2000. 6 (6), 499–506.
- Franklin GM. Opioids for chronic non-cancer pain: A position statement of the American Academy of Neurology. Neurology. 2014; 83 (September); 1277 – 1284.
- Fredheim, O. M. S., Skurtveit, S., Breivik, H., & Borchgrevink, P. C. (2010). Increasing use of opioids from 2004 to 2007 – Pharmacoepidemiological data from a complete national prescription database in Norway. European Journal of Pain, 14(3), 289–294. http://doi.org/10.1016/j.ejpain.2009.05.006
- Furlan AD, Sandoval JA, Mailis-Gagnon A, Tunks E. Opioids for chronic non-cancer pain: a meta-analysis of effectiveness and side-effects. Canadian Medical Association Journal. 2006; 174 (11): 1589 – 94.
- Gomes, T., Mamdani, M. M., Dhalla, I. A., Paterson, J. M., & Juurlink, D. N. (2011). Opioid Dose and Drug-Related Mortality in Patients With Nonmalignant Pain. Archives of …, 171(7), 686–691. http://doi.org/10.1001/archinternmed.2011.117
- https://www.rcoa.ac.uk/faculty-of-pain-medicine/opioids-aware/clinical-use-of-opioids/effectiveness-for-long-term-pain (Accessed November 2015)
- Kalso, E, Edwards JE, Moore RA, McQuay HJ. Opioids in non-cancer pain:systematic review of efficacy and safety. Pain. 2004; 112 (3); 372 – 80)
- Langley PC. The prevalence, correlates and treatment of pain in the European Union. Current Medical Research and Opinion. 2011; 27 (2): 463 – 80
- Martelli MF, Zasler ND, Bender MC, Nicholson K. Psychological, neuropsychological and medical considerations in assessment and management of pain. Journal of Head Trauma Rehabilitation. 2004; 19 (1); 10 – 28.
- Merskey H, Bogduk N. Classification of chronic pain – descriptions of chronic pain syndromes and defintiions of pain terms. 2 ed. Seattle: International Association for Study of Pain; 1994.
- Moore RA, McQuay HJ. Prevalence of opioid adverse effects in chronic, non-malignant pain: systematic review of randomised trials of oral opioids. Arthritis Research and Therapy. 2005; 7 (5): R1046 – 51.
- NationalPain AuditFinal Report. (2012). NationalPain AuditFinal Report, 1–66.
- Olsen, Y., Daumit, G. L., & Ford, D. E. (2006). Opioid Prescriptions by U.S. Primary Care Physicians From 1992 to 2001. The Journal of Pain, 7(4), 225–235. http://doi.org/10.1016/j.jpain.2005.11.006Portenoy RK, Foley KM. Chronic use of opioid analgesics in non-malignant pain: report of 38 cases. Pain. 1986; 25 (2); 171 -86
- Quan, H., Li, B., Couris, C. M., Fushimi, K., Graham, P., Hider, P., et al. (2011). Updating and Validating the Charlson Comorbidity Index and Score for Risk Adjustment in Hospital Discharge Abstracts Using Data From 6 Countries. American Journal of Epidemiology, 173(6), 67
- READING BETWEEN THE LINES: The annual profile for substance misuse 2014-15. (2015). READING BETWEEN THE LINES: The annual profile for substance misuse 2014-15, 1–86.
- Ruscitto, A., Smith, B. H., & Guthrie, B. (2014). Changes in opioid and other analgesic use 1995-2010: Repeated cross-sectional analysis of dispensed prescribing for a large geographical population in Scotland. European Journal of Pain, 19(1), 59–66. http://doi.org/10.1002/ejp.520
- Welsh Assembly Government. Designed for people with chronic conditions; Service development and commissioning directives; Chronic non-malignant pain. 2008, 1 – 46
- Welsh Government & Statistics Wales. (2015). 2014 Mid-year Population Estimates for Wales (No. SB 34/2015) (1–8). Cardiff.
- Wright, E. A., Katz, J. N., Abrams, S., Solomon, D. H., & Losina, E. (2014). Trends in Prescription of Opioids From 2003–2009 in Persons With Knee Osteoarthritis. Arthritis Care & Research, 66(10), 1489–1495. http://doi.org/10.1002/acr.22360
- Zin, C. S., Chen, L. C., & Knaggs, R. D. (2014). Changes in trends and pattern of strong opioid prescribing in primary care. European Journal of Pain, 18(9), 1343–1351. http://doi.org/10.1002/j.1532-2149.2014.496.x