Design and study population
We conducted a secondary analysis of the longitudinal, multicenter, retrospective OPIOIDS (Outcomes in Patients usIng Opioids In Painful Disorders in Spain) study [20], which used electronic medical records (EMR) from the BIG-PAC registered health database (The European Network of Centers for Pharmacoepidemiology and Pharmacovigilance; http://www.encepp.eu/encepp-/viewResource.htm? id-29236) of records of an allocated population of 1.8 million from seven Spanish autonomous communities [21]. EMRs were anonymized according to Organic Law 3/2018, of December 5, on the Protection of Personal Data and guarantee of digital rights (https://www.boe.es/eli/es/lo/2018/12/05/3). The OPIOIDS study was classified by the Spanish Agency for Medicines and Health Products and approved by the Ethics Committee of the Hospital of Terrassa (Barcelona). The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement guidelines for reporting observational studies were followed [22].
We included the EMR of patients who started a new treatment with strong or weak opioids, alone or combined with other analgesics, for the treatment of moderate to severe (≥ 5 points on the numerical pain scale) chronic nociceptive pain (≥ 3 months) in any site due to OA between 1/01/2010 and 31/12/2015 [23]. Patients were followed for a maximum of 3 years. Patients had to be previously treated with ≥ 1 first-line analgesic such as acetaminophen, metamizole and/or nonsteroidal anti-inflammatory drugs (NSAIDs). EMRs were obtained during the 12 months before the index date and over the next 36 months. The index date was the initiation of a new opioid treatment, starting from the date of diagnosis of chronic nociceptive pain due to OA. Additional inclusion criteria were (a) age ≥18 years, (b) occupationally active, c) patients active in the database (≥2 health records in the computer system) a minimum of 12 months before the start of the study, (d) ensured regular monitoring (≥2 health records in the computer system from the index date), and (c) inclusion in chronic prescription program (with a record of the daily dose, time interval and duration of each treatment administered (≥2 prescriptions during the follow-up period)). Exclusion criteria were (a) displaced or out-of-area subjects; (b) permanently institutionalized patients, c) terminal illness and/or dialysis (ICD-10:N18), d) neuropathic pain/radiculopathy (ICD-10: G50-65) or cancer (ICD-10: G89.6), and e) not occupationally active (retired, unemployed, permanent work disability, early retirement and/or maternity leave).
Definition of diagnosis and study cohorts
Diagnoses were obtained from the International Classification of Diseases (10th edition) Clinical Modification (ICD-10-CM) (https://eciemaps. mscbs.gob.es). Chronic pain was defined as pain that persisted for ≥ 3 months. The site of pain was a) hip and knee (M16, M17), b) lumbar (M54.5), and other sites (M15, M18, M19, M40, M41) according to medical criteria. Two study cohorts were developed according to type of opioid (weak or strong) according to the Anatomical Therapeutic Chemical Classification System (ATC)[24]. The index date was the start of a new opioid treatment (weak or strong) from the date of diagnosis of chronic nociceptive pain due to OA. Adherence, persistence, medication possession ratio (MPR) and discontinuation definitions were as previously described in the OPIOIDS study [20].
Demographic variables, comorbidity and medication administered
Demographic and comorbidity variables were age (continuous and by range: 18-44, 45-54 and ≥55 years on the index date), sex and body mass index (BMI, Kg/m2) and the personal medical history. The Charlson comorbidity index and the number of chronic comorbidities were used as a summary variable of general comorbidity and an approximation of patient severity [25]. The medications (active ingredients) indicated for chronic OA pain were obtained (ATC N02AA01 to N02AX06) from pharmacological prescription records. The choice of medication in a specific patient was at the physician’s discretion. The type and medical specialty that initiated the prescription was determined (Supplementary table S1).
Cost of days of sick leave
The cost of days of temporary work disability from the social perspective was determined by the human capital method (mean salary of the replacement of the active patient [€ 119.44 per day not worked]), source: Spanish Institute of Statistics (INE))[26]. Days of sick leave were expressed as absolute and relative values (in patients with sick leave), and as the mean days of sick leave per patient (patients with/without sick leave [0 days of sick leave]).
Clinical effectiveness
Clinical effectiveness was obtained using the variation in pain intensity on an 11-point numerical pain scale (5-7; moderate pain, >7; severe pain) from the closest date before the start date (index date) and the end date of the study [23]. The absolute variation in the natural and relative units was calculated in percentage changes from the baseline value.
Statistical analysis
A descriptive, univariate statistical analysis was made. Qualitative data were described using absolute and relative frequencies. The 95% confidence intervals (CI) for parameter estimation were based on the total number of subjects with no missing values. ANOVA and the chi-square test for independent groups were used in the bivariate analysis. Paired tests were also used to evaluate the before-after differences for each subgroup (McNemar’s test and the Student’s t test). A survival analysis was made by estimating the Kaplan-Meier curves (log-rank test) to analyze the persistence of opioid treatment, applying a Cox proportional risk model (corrected by covariates) to determine the hazard ratio (HR) and its 95% CI between strong and weak opioids. Covariate analysis was used to correct the cost per patient of days of sick leave. The covariates included were sex, age, general comorbidity (Charlson index), time from diagnosis, treatment persistence and the MPR. A multiple linear regression model (procedure: consecutive steps) was constructed to examine the variables associated with sick leave (dependent variable: cost per patient [all] of sick leave). The analyses used IBM SPSS, version 23.0, NY, USA software (https://www.ibm.com/analytics/spss-statistics-software).