Long-term, high-dose opioid prescription for chronic non-cancer pain in primary care: an observational study

Background Opioid prescriptions for chronic pain have risen sharply over the last 25 years; harms associated with these drugs are related to dose and length of use. Aim The main aim of this study was to identify patients prescribed long-term, high-dose (LTHD) opioids in the community and to assess the prevalence of such use. Design & setting An observational study of opioid prescribing in two demographically dissimilar GP practices in North Wales, UK. Method Details of opioid prescriptions were collected for 22 841 patients, of whom 1488 (6.5%) were being prescribed opioids on the census date. Exhaustive examination of the data identified all patients who were prescribed oral morphine equivalent doses of ≥120 mg/day for ≥1 year. Results All these patients were being prescribed ≥120 mg/day, as a single drug, morphine, oxycodone, fentanyl, or buprenorphine, irrespective of opioid polypharmacy. Across both practices, 1.71/1000 patients were identified as LTHD users of opioid medication for chronic non-cancer pain (CNCP). Prevalence was similar in the two practices. Repetition of the process until January 2021 showed no change in the pattern. Conclusion This study offers confirmation that a significant group of patients are prescribed long-term opioid medication for chronic pain at doses that are unlikely to be effective in reducing pain, but are likely to have harmful consequences. The findings offer a simple, reliable, and practical method of data extraction to identify these patients individually from routinely collected prescribing data, which will help in monitoring and treating individuals and establishing the problem prevalence.


Introduction
UK opioid prescribing has increased substantially since 1990, 1,2 with similar changes internationally, 3 especially in North America. [4][5][6] There is widespread concern in the UK related to severe problems in the USA, [7][8][9][10] where there has been an epidemic of opioid use and resulting harm arising, in part, from the indiscriminate prescribing and diversion of pain medication. There is little evidence for problems of the scale and severity experienced in the USA occurring outside North America. Most opioid prescribing in the UK is for CNCP [11][12][13] despite a lack of evidence of effectiveness [14][15][16][17][18][19] and the risk of direct and indirect harms. 15,17,[19][20][21][22][23][24][25][26] Epidemiological evidence shows that opioids for CNCP are associated with worse functioning and quality of life, [27][28][29][30] and the authors' clinical experience suggests this is particularly true with high doses.
An emerging consensus suggests that opioid treatment for chronic pain should be approached with care rather than with unbridled enthusiasm. 31 However, the field must now cope with the consequences of past over-enthusiasm. It is important to find practical ways to identify patients in primary care on LTHD regimens.
Studies of opioid prescribing in the UK are summarised in Table 1. Overall prescribing may have decreased but rates remain much higher than they were 25 years ago. 2,11,32 In contrast to the decrease in overall prescribing, the prescribing of stronger drugs has increased, 2,11,33,34 as has the number of people prescribed long-term opioids. 32,35 These studies illustrate persistent concerns about use of opioid medication in the UK population. However, to properly understand and then address the problem, individual-level information is needed. Aggregated population-level statistics showing 100 prescriptions issued cannot distinguish between 100 people with one prescription each, 10 people with 10 prescriptions each, or one person with 100 prescriptions, and each of these indicates a different clinical situation. There is also a need for a method to identify individual patients on these regimens within practice populations to facilitate evaluation and intervention.
The aim of this study was to develop a practical method to identify all patients on LTHD opioid regimens within primary care and to describe patterns of opioid prescribing in two practice populations.

Method
The criteria for LTHD opioid prescribing were 1) daily oral morphine equivalent (ME) dose of 120 mg or more; 36,37 2) prescribed continuously for more than 1 year; and 3) patients with cancer were excluded. The Royal College of Anaesthetists 38 indicates that 'the risk of harm increases substantially at doses above an oral ME of 120mg/day, but there is no increased benefit' and the British Pain Society 39 has recommended that patients with CNCP being prescribed opioids at this doses or above should be referred to specialist pain services. A one-year minimum prescribing duration was chosen as it is unequivocally long-term use.
A point prevalence study was carried out in two GP practices in North Wales. Practice A was a large, fully medically staffed practice in a market town with national average levels of social deprivation, as indicated by the Welsh Index of Multiple Deprivation. 40 It was not identified, either internally or externally, as an outlier regarding opioid prescribing, lying below the mean for the Local Health Board (LHB).
During data collection in practice A, the authors were asked to provide support on opioid prescribing to a second GP practice, practice B, in an ex-mining community with higher Welsh Index of Multiple Deprivation 40 levels of social deprivation. It was entirely staffed by locum GPs and administered by the LHB. Opioid prescribing was close to the LHB mean.
Data were extracted from the practices' computerised records for all prescriptions of opioid analgesic medication, including compound drugs (see Box 1 for details), dated between 1 October 2016 and 31 January 2017 (practice A), and 17 January and 15 May 2017 (practice B). The data included: age and sex; date of prescription; drug name, strength, and dose direction; and quantity prescribed. The weeks beginning Monday 9 January 2017 (practice A), and Monday 24 April 2017 (practice B), were selected as the index week and all prescriptions issued in that week or for use in that week were identified. PRN (as required) prescriptions dated in the month preceding the index week and/or records indicated continuing use, were included. An estimate of daily dose was made. For PRN prescriptions, this was calculated by dividing the total quantity of drug prescribed by the number of days between the two prescriptions nearest the index week. This method was also used where there was a clear difference between the use calculated and the amount specified in the dose

1991-2009
England All prescriptions Five-fold increase in prescribing opioids.
Regional variations in prescribing. Co-prescribing of Z-drugs, benzodiazepines, and gabapentinoids. ME = Morphine equivalent.. a Drugs categorised as "strong" vary between studies, but always include morphine, oxycodone, and fentanyl. b This is estimated from the published text, which does not indicate the data collection period. It was assumed that there were likely to be three ways in which total ME dose might exceed 120 mg/ day:

Ruscitto
A. as a high dose of a single drug; B. as a relatively high dose of one drug with a dose of a second drug; C. as lower doses of several opioid drugs in combination.
Three sets of criteria were applied sequentially to the collected data to identify three (nonmutually exclusive) patient groups: A. those prescribed a daily dose of a single drug ≥120 mg ME; B. those prescribed a daily dose of a single drug between 60 mg ME and 119 mg ME and other opioid(s); C. those prescribed three or more different opioids. (This iteration identified very few cases not found in A or B, none of which had doses approaching 120 mg ME. It was subsequently discontinued.) All patients prescribed LTHD opioids were identified. Those with a cancer diagnosis were excluded. Comparison data on these patients for the corresponding period a year earlier were collected.
In practice A, additional data were collected from review of patient records for those prescribed LTHD opioids: indication for use of opioid; non-opioid analgesic or adjuvant medications; referral history to secondary care pain services; and longitudinal opioid prescription histories.
The two GP practices studied in this research used different database systems (Vision and EMIS) to record practice information. Different search terms were enabled in these systems to access information about drug prescriptions. In the Vision system, the search terms were: Drug Class of type Opioid Analgesics, and Drug Class of type Non-opioid and Compound Analgesics.
In the EMIS system, the search terms were: The Drug is Opioid Analgesics, and The Drug is Compound Analgesic Preparations.
These searches produced slightly differing results, but included, in both cases, all prescriptions for drugs listed by BNF [British National Formulary] as opioid analgesics, in the relevant time period. Therefore, there were false positive results but no false negatives.
Box 1 Searching practice records for prescriptions for opioid medication . The LHB policy is that opioid replacement for substance misuse must be prescribed by specialist services, not GPs; no opioid prescriptions for substance misuse were identified.
When the data had been collated, the findings were presented to a meeting of practice A's staff. Prescribers' comments were elicited and recorded. This was not possible in practice B due to discontinuity of medical staffing.
The exercise was repeated yearly in practice A in the corresponding week from 2018-2021. One repeat audit was conducted in practice B in October 2020.
Service users from PÂR-NCMH 42 were consulted during the planning of this research. All the patients on LTHD opioids were prescribed morphine, oxycodone, or fentanyl transdermal patches. In practice A:

Results
• eight on morphine (highest daily ME 619 mg), • six on oxycodone (highest daily ME 480 mg), • eleven on fentanyl (highest daily ME 540 mg).  Fourteen out of 25 also had PRN opioid prescriptions. When PRN was added, the highest daily ME was 778 mg.
In practice B: • four on morphine (highest daily ME 480 mg), • five on oxycodone (highest daily ME 320 mg), • five on fentanyl (highest daily ME 540 mg), Three out of 14 patients also had PRN opioid prescriptions.  Table S1 shows that once high doses were reached they tended to be maintained, and that in January 2021 there were a small number of LTHD buprenorphine prescriptions (n = 3; 6% of all the patients prescribed LTHD opioids from Practice A).
It was possible to determine the duration of continuous opioid use for n = 23/25 patients (the other two patients were on high-dose opioids on registration with the practice). Regular noncompound opioids were initiated 42-240 months before the index week (median 126). Initiation of doses ≥120mg ME was 13-198 months before the index week (median 106).
Findings from the first year's data were presented to a group of GPs and other staff at Practice A . They expressed surprise at the number of patients prescribed LTHD opioids, having expected there to be few or none. The consensus was that patients experience adverse drug effects at doses below 120 mg ME per day, and that a lower dose-threshold might be more appropriate. Problems with adjuvant medications were raised. Conversations about opioids were described as difficult for prescribers and patients, sometimes leading to conflict. Refusal to increase dose or strength and raising cessation or reduction of opioids were difficult topics. There was a perceived lack of alternatives. Prescribers did not feel they had full control of opioid prescribing. The GPs were asked if there were any patients prescribed LTHD opioids who experienced satisfactory levels of pain relief without negative effects; none could be identified.

Discussion Summary
This study's findings show that LTHD opioid use for CNCP can be identified in primary care records by searching for prescriptions of morphine, fentanyl, oxycodone, or buprenorphine at or above 120 mg ME per day and examining prescriptions 1 year previously. This method (see Box 2) is less laborious than that used in previous UK studies.
As indicated in Table 2, examination of opioid prescribing data for practices A and B identified all patients prescribed LTHD opioids. A rate of 1.71 per 1000 patients was calculated for the practices combined. Unadjusted extrapolation would suggest that there might be over 100 000 patients prescribed LTHD opioid medication in the UK, but this figure is not epidemiologically reliable; it is a crude estimate. These patients are at risk of harm and functional impairments with very limited pain relief.
This study's criteria were conservative: most of the literature accepts 120 mg ME as the threshold for limited analgesic benefit and high risk of harm. 38 A one-year duration was chosen pragmatically. It is likely that some patients experience negative effects at lower doses and shorter durations. Table 2 sets out data for high doses for less than 1 year and daily doses between 60 and 120 mgME.
This study shows that some patients follow LTHD regimens for many years. No judgements about drug effectiveness can be made from this study's findings, but there is a real possibility that harms outweigh benefits for these patients.
This study found frequent use of adjunctive pain medications alongside high dose opioids. Many of these have sedative effects that compound the effects of opioids.
The feedback meeting with practice A's medical staff suggested that LTHD prescribing can arise without the prescriber intending it, confirming uncertainty over the locus of control. The discussion confirmed previous findings regarding difficult doctor-patient interactions about opioids. 43,44 The prescribers were unable to identify any patients prescribed LTHD opioids who experienced good pain relief and no adverse effects on functioning.

Strengths and limitations
This study's main limitation was sample size. Small-scale studies are susceptible to sampling error, and the findings should not be over-interpreted. This study has presented purely descriptive findings. Set against this, the size of the study allowed collection and analysis of detailed data on an individual patient level with direct counts of cases rather than estimates from a sample of the population of interest. This is a necessary step in developing practical methods to generate large individual level datasets and to identify individual patients for medication review. Replication in other locations with different demographics would be helpful. The authors are particular aware of the lack of ethnic diversity in the area studied; 97.8% White in North Wales in 2015-2017. 45 These findings are indicative, but a larger study is necessary for fully generalisable findings. The method outlined in this study will facilitate larger scale studies of LTHD opioid use.

Comparison with existing literature
Two previous studies sought to identify patients prescribed high-dose opioids for CNCP. 36,37 Their methods did not measure duration, and made extensive demands on clinical and administrative staff time. This study's method was less laborious and captures the group of greatest concern.
Rates of LTHD prescribing in practice A are consistent with other evidence on potent, high-dose and long-term opioid prescribing. 2,11,32,34,35 Increasing levels of such prescribing in practice A suggest that prescriber awareness of LTHD opioids because of this study was insufficient to alter prescribing. Decreased LTHD prescribing in practice B may be a result of pharmacist prescribing support provided in the absence of permanent medical staff.
The number of patients prescribed opioids, sex ratio, and average age in practice A are similar to a recent primary care study. 35 The results for practice B are different. Research has shown that higher overall rates of opioid prescribing are associated with older-age patients. 11,13,35 Table 4 shows higher rates in people over 75 years, though the absolute number is evenly spread across those aged over 40 years.
Research has shown higher levels of opioid prescribing in deprived areas. 2,11,32,34,35,46 Accordingly, rates of overall opioid prescribing were higher in the more socially deprived practice B. However, rates of LTHD opioid prescribing were similar in both practices. The prevalence finding in this study was higher than in previous studies. This may reflect differences in method, geographical differences, change in prevalence over time, or all three. However, it does not suggest a recent reduction in high-dose opioid prescribing. It underlines the need for evidence-based interventions to reduce high-dose prescribing.

Implications for research and practice
This study's method is practical and reliably identifies patients prescribed LTHD opioids in UK primary care. It has utility in auditing such regimens and in identifying at-risk patients. It has potential to unpick the fine grain epidemiology of the problem.
The findings indicate factors that should be explored in future studies: social deprivation, age, duration of high-doses regimens, indications for opioids, adjunctive medications, and better prescribing strategies. Such studies will be important in the development of better pain management strategies in primary care (see 47 ).

Funding
There was no external funding for this research.

Ethical approval
The research was registered as audit of routinely collected anonymised data by the LHB R&D department. It did not require NHS or University ethical approval. Bangor University CBLESS Research Ethics Committee gave approval for the collection of staff meeting data. All staff gave prior informed consent.

Provenance
Freely submitted; externally peer reviewed.

Data
The dataset relied on in this article is available from the corresponding author on reasonable request.