Prescription patterns and predictors of unmet pain relief in patients with difficult-to-treat osteoarthritis in the Nordics: analyses from the BISCUITS study

Study design and population

A multinational observational cohort study was conducted using data on healthcare visits, medication use, demographics, and socioeconomics. Data were collected as part of ‘BISCUITS’ and all data sources and the linkage process in the BISCUITS study are described elsewhere [submitted/if accepted: reference to “sjpain-2021-0212”].

All adult patients (≥18 years old) with a primary diagnosis of OA (ICD-10 codes: M15–M19) in specialty care between 1 January 2011 and 31 December 2012 were included in the study population. Patients with a diagnosis of OA or cancer (ICD-10 codes: C00–C43 or C45–C97) within the 3 years prior to inclusion were excluded to create a patient population representative of an incident and newly diagnosed group seeking specialty care for the first (recorded) time for OA without malignant pain. The index date was set to the inclusion date for patients fulfilling the criteria and information during the 3 years pre-inclusion was also retrieved.

Included patients were followed for up to 5 years post-index or until the following censoring events: death, a primary cancer diagnosis, or a joint replacement; whichever occurred first. Patients with cancer were censored as we aimed to study pain medication use due to non-malignant pain. Patients with a joint replacement were also censored to not capture post-operative pain requiring high use of pain medication. Patients with an orthopedic assessment or a less invasive surgery (e.g., arthroscopy) were not censored.

Subgroups

Four subgroups were pre-defined on the assumption that patients included in these subgroups were expected to be difficult-to-treat, have a potential high unmet need of pain relief and/or have a problematic treatment journey. These groups were ‘Multiple comorbidities’, ‘Frequent healthcare visits’, ‘Multiple pain medications’, and ‘Contraindication to an NSAID’. See Appendix S1, Supplementary Table S1 (online supplement) for a list and detailed description of all subgroups.

Study outcomes and variables

Patient characteristics on demographics, comorbidities and medication use were gathered at or before index. Prescription patterns were studied by describing annual medication use after the index date. Unmet pain relief (defined as using several types of opioids, persistent opioid use or long-term NSAID use) was used as the outcome in regression analyses. All patient characteristics, prescription patterns, unmet pain relief and predictors in the regression analyses are described further in Appendix S1.

Statistical analysis

Continuous variables are presented with mean and standard deviation, and categorical variables with frequency and percentage. A two-tailed t-test was used to test for statistically significant differences in continuous variables and a chi-square test for categorical variables. The statistical significance level was set at 95% in statistical tests and in estimation of confidence intervals. Statistical tests were not performed to test differences between countries. Predictors of unmet pain relief were tested using survival analysis graphs (Kaplan-Meier [21]) and Cox proportional-hazards regression [22] with unmet pain relief, as defined earlier, as the outcome. Patients with an outcome before the index date were excluded from this analysis. All data management and statistical analyses were performed using RStudio v1.3 (RStudio Team, PBC, Boston, US) and Stata v16 (StataCorp, College Station, US).

Patient characteristics

The total number of patients was 288,174 (62,300, 70,857, 54,544, and 100,473 in Denmark, Finland, Norway and Sweden, respectively). For all patients, ranges per country were relatively consistent for most of the patient characteristics (Table 1). The combined mean age was 63.5 years at the index date and 58% were female. Most patients did not have a diagnosis recorded in the Elixhauser comorbidity index based on all visits made in specialty care, but 29.5% had at least one comorbidity. A diagnosis of depression or anxiety was relatively rare (2.1%). In total, 45.3% were opioid users, of which 7.1% were persistent users and over 20% were long-term NSAID users.

Of the four subgroups of interest, the largest subgroup was ‘Multiple pain prescriptions’ which represented 33.1% of patients (Table 2). In the subgroups ‘Multiple comorbidities’ and ‘Frequent healthcare visits’, 13.0 and 12.3% of patients were included, respectively. The smallest subgroup was ‘Contraindication to a NSAID’, including 2.8% of all patients. The overlap of patients between the subgroups are presented in Appendix S2, Supplementary Figures S1–S3.

Patients in the subgroups ‘Multiple comorbidities’, ‘Frequent healthcare visits’, and ‘Contraindications to NSAIDs’ had substantially more comorbidities (by the Elixhauser index) (p-value <0.001), depression, and anxiety diagnoses (p-value <0.001), and the former two subgroups were more likely to have a contraindication to an NSAID than the overall group comprising all patients (p-value <0.001). Patients in these groups were also older than the overall group (p-value <0.001). Opioid use (p-value <0.001), except for the subgroups ‘Multiple comorbidities’ in Finland and ‘Contraindication to an NSAID’ in Finland, Norway and Sweden and persistent opioid use (p-value <0.001) were higher in these difficult-to-treat subgroups than the overall group with the highest ranges per country among those with multiple pain prescriptions for opioid use (63.8%) and persistent opioid use (15.0%). Among the subgroup of patients contraindicated for NSAID use, 13.4% were long-term NSAID users (at least two prescriptions within 3 months) compared to 23.6% in the overall group (p-value <0.001).

Prescription patterns

The percentage of patients with an NSAID (Figure 1A) or opioid (Figure 2A) prescription was highest in the first year following inclusion for all countries and subgroups, and decreased in subsequent years.

Figure 1:

Prescription patterns – NSAID. Error bars indicate 95% confidence interval.

NSAID, nonsteroidal anti-inflammatory; DDD, Defined daily dose.

Figure 2:

Prescription patterns – opioids. Error bars indicate 95% confidence interval.

NSAID, nonsteroidal anti-inflammatory drug; OMEQ, oral morphine equivalents.

NSAID use was the highest in the group with multiple pain prescriptions and lowest among those with contraindications to NSAIDs, in all countries. The ‘Frequent healthcare visits’ and ‘Multiple comorbidities’ subgroups had a lower percentage of patients with an NSAID prescription than the overall group in Finland, Norway, and Sweden (p-value <0.05) (similar percentages in Denmark). Dosing (in daily defined dose [DDD]) of NSAIDs (Figure 1B) was the highest in the first year and decreased in subsequent years, reflecting the same pattern as number of NSAID users over the same time period.

In contrast to NSAID use, all subgroups consistently had a higher number of opioid users than the overall OA group (p-value <0.001). Five years after inclusion, between 35 and 50% of patients in the subgroups were dispensed opioids, compared to 20–25% of patients in the overall group (p-value <0.05). Even though the percentage of opioid users decreased over time, dosing in oral morphine equivalents (OMEQ) (Figure 2B) was flat or increasing for patients that were dispensed opioids. The increase in dosing occurred in all Norwegian subgroups and the subgroups ‘Frequent healthcare visits’ and ‘Multiple pain prescriptions’ in Denmark and Finland. OMEQ doses for all subgroups and the overall group in Denmark was higher than for the other countries.

The number of persistent opioid users (Figure 3A) directly reflect the patterns of dosing in OMEQ as the definition of persistent use is the amount of dispensed OMEQ. More patients were dispensed several types of opioids each year (Figure 3B) in the subgroups than in the overall group (p-value <0.05) and this percentage decreased over time.

Figure 3:

Persistent opioid use and use of several types of opioids. Error bars indicate 95% confidence interval.

NSAID, nonsteroidal anti-inflammatory drug.

Patients were censored due to a joint replacement, cancer, or death (Appendix S2, Supplementary Figure S4). The percentage of patients that underwent a joint replacement was the highest in the first year. In total, between 30 and 40% of patients had received a joint replacement after 5 years and were censored for this reason. The percentage of patients with cancer was highest among the group with frequent healthcare visits in year one (p-value <0.05). Deaths were highest in the group with a contraindication to an NSAID (p-value <0.05) and lowest in the overall group as well as in those dispensed multiple pain prescriptions (p-value <0.05).

Predictors of unmet pain relief

Predictors of unmet pain relief and prescription patterns

The association between comorbidities, in particular depression and anxiety, and unmet pain relief, defined as persistent opioid use, using several types of opioids, or long-term NSAID, was statistically very strong and in line with a previous study [23]. Disability pension was another strong predictor of all definitions of unmet pain relief, indicating a high comorbid and socioeconomic burden in some patients with OA.

Compared to the overall OA population, all difficult-to-treat subgroups had higher opioid use in all years following the index date. This may indicate that we successfully defined groups of patients with additional pain management challenges compared to the overall OA population. It was shown that the percentage of patients on opioids decreased over time, while the strength of doses in dispensed opioids increased, indicating that patients remaining on long-term opioid use required continuous dose escalation as part of their pain treatment regimen. Continuous dose escalation is considered very problematic for these patients as it is against any recommendations for OA management [9].

Between 5 and 15% of those with a contraindication to an NSAID were prescribed an NSAID during the 5-year study period despite having a diagnosis contraindicated to NSAID use. This subgroup was dispensed as much OMEQs as the other subgroups and higher than the overall group. If these patients are prescribed opioids instead of NSAIDs for pain relief, it may highlight the challenges associated with long-term pharmacological treatment of this patient population in terms of substantial unmet need and demand for alternative therapies.

The proportions of patients treated with NSAIDs and opioids found in this study were consistent with those reported in earlier studies showing similar high levels of medications use among European patients with OA [18, 20, 24] and among US patients [25, 26]. In Denmark, the dispensed doses of opioids were higher compared to Sweden, Finland and Norway, in line with another study comparing opioid use in the general population in Denmark, Sweden and Norway [27].

The first visit in specialty care is often linked to orthopaedic assessment as observed in a Dutch study [15] and subsequent surgery, which could also be seen in our study given the high proportion of patients censored due to a joint replacement in the first year after diagnosis. The patterns of joint replacements from index to 5 years after were similar between all subgroups and the overall OA group. This may indicate that other factors than the ones used for defining difficult-to-treat subgroups are driving the decision to proceed with this procedure. As no gold standard exists for joint replacement referral [28], it remains unclear which factors that were associated with joint replacement.

Difficult-to-treat subgroups

Patients in the difficult-to-treat subgroups were dispensed higher doses of opioids in the later years of the analyses compared to the first year. This indicates that pre-operative opioid doses increased in patients waiting longer for surgery or that these groups of patients were ineligible for a joint replacement and with a substantial need for pain relief. Both reasons are troublesome as high pre-surgery use of opioids is a risk factor for continued pain and opioid use post-surgery, and the list of negative health effects of long-term opioid use is long [29, 30].

The results reflect a discrepancy between current treatment guidelines [5], [6], [7], [8], [9] and routine clinical practice, especially regarding long-term use of opioids. However, the long study period spans multiple iterations of clinical treatment guidelines for the management of OA. Current guidelines increasingly account for the patient’s medical profile, including risks of opioid and NSAID use; thus, these findings should be considered based on these recent changes.

The differences between the difficult-to-treat subgroups and the overall OA group may indicate a sufficient sensitive identification to differentiate the difficult-to-treat patients in the cohort of OA patients referred to specialty care. However, the differences also highlight the heterogeneity within the overall OA population. Assessing the full OA population may diffuse the true heterogeneity of patients with OA and addressing the subpopulations may enable us to understand more about managing the illness.

Strength and limitations of study

A major strength of the present study is the high degree of validity, mandatory reporting, completeness, data quality and long follow-up in the Nordic national health registers.

Limitations include the inability to capture non-prescription medication, adherence to dispensed medication and the indication of prescribed medications. The OA diagnosis included in the regression analyses is the diagnosis used at the index date and does not consider the possibility of OA in multiple joints. Furthermore, no indication of pain severity or treatment satisfaction was available for the patients.

Future studies should focus on complete patient journeys from the initial contact in primary care through to treatment by specialists, prescription patterns in relation to joint replacements, and poor outcomes or adverse events because of problematic analgesic use.