This large study of an unselected osteoarthritis population undergoing hip or knee replacement found that higher age, female gender, obesity, number of comorbidities (according to the modified Charlson Comorbidity Index score), preoperative use of analgesics, and unilateral knee replacement (compared to simultaneous bilateral procedure) were associated with a higher probability of using analgesic drugs 1 year after surgery, whereas individual comorbid conditions had little or no effect. The associations related to obesity, gender, operated joint, and preoperative use of analgesics were similar for different types of analgesic agents whereas the associations with different comorbid conditions were more mixed. Moreover, the results suggest that NSAIDs are avoided in patients with multimorbidity, especially cardiac disease, and in older patients, which is supported by current guidelines . The present study expands on the earlier literature by analyzing the impact of several comorbidities and clinical factors on the postoperative consumption of all analgesic drugs and by including not only NSAIDs and opioids, but also acetaminophen.
Our main finding is that obesity was associated with a higher risk ratio for the consumption of all the studied analgesic drugs both pre- and postoperatively. Previously, a higher BMI has been shown to predict NSAID use after THA and TKA [18,19,20] and opioid use after TKA [14, 19]. Our study shows that obesity predicts the use of all analgesic drugs (acetaminophen, NSAIDs, and opioids) after hip or knee replacement. Similar to the findings of earlier studies [18,19,20], drug use was more frequent especially when patient BMI exceeded 35 kg/m2. Obese patients report more pain after joint replacement [7, 18], which is one possible explanation for this finding. Obesity has also been associated with analgesic use in the general population [30, 31].
The number of comorbidities measured with CCI score predicted the use of acetaminophen and opioids, but not NSAIDs. Earlier, Hansen et al.  reported an association between higher CCI score and chronic opioid use after TKA. We were, however, unable to provide a good explanation for this finding. A history of malignancy, a component of the CCI with a score of 2, could be one possible explanation, but a CCI score of 1 also predicted analgesics use, and the results considering CCI in the logistic regression were identical, even when patients with a history of malignancy were excluded. Additionally, a history of malignancy was not independently associated with analgesic use.
Out of the separate comorbidities, patients with neurodegenerative disorders used acetaminophen (but not NSAIDs or opioids) more often. This may be related to the adverse events of opioids and NSAIDs in long-term use in the elderly [22,23,24, 32, 33]. As expected, patients with cardiac disease used fewer NSAIDs than other patients [23, 24]. Patients with epilepsy use more opioids in the general population , possibly due to a higher prevalence of painful conditions, and we found a similar finding in joint replacement recipients. Interestingly, although diabetes is associated with persistent postoperative pain , the disease was not associated with analgesic consumption, especially when their higher BMIs were taken into account . Earlier, controversial results have been reported. Namba et al.  reported an association with diabetes and increased use of opioids after knee replacement (OR 1.03 [95% CI 1.01–1.05]), whereas in two other studies [9, 15], no differences were found after knee or hip replacement regarding diabetes. These differences may be explained by over twofold higher prevalence of diabetes in the study by Namba et al. than in the other two and the present study.
Patients who underwent knee surgery used acetaminophen, NSAIDs, and opioids more often than patients who underwent hip surgery [12, 15, 20]. Patients with simultaneous bilateral knee surgery used more analgesics preoperatively (compared to unilateral operation) but used fewer opioids and acetaminophen postoperatively. These findings may reflect patient selection bias as those patients with previous opioid use or signs of a history of pain sensitization are not candidates for simultaneous bilateral knee arthroplasty. It is not known whether a simultaneous or staged bilateral operation should be preferred in bilateral osteoarthritis , and a comparison of these procedures was not possible in the present study due to the chosen exclusion criteria (another joint replacement 2 years before or after index surgery; therefore, patients with staged bilateral operations were excluded). Similar to the findings of earlier studies [9, 21], UKA (compared to TKA) was associated with lower RRa for the use of opioids and acetaminophen. Interestingly, preoperative clinical state, based on HHS/KSS, was not associated with postoperative overall analgesic use although better KSS Function score was associated with lower RRa for the use of acetaminophen and opioids.
Preoperative pain has been associated with postoperative analgesic use , and although it is a component of KSS Knee Score and HHS, these were not associated with postoperative analgesic use. Preoperative pain was not possible to analyze separately in this study.
In line with earlier studies [12, 14, 15, 21, 26], preoperative analgesic use was associated with greater postoperative opioid use. In this study, we also found a similar association with the use of acetaminophen and NSAIDs. Furthermore, opioid consumption was greater in patients with higher CCI score, epilepsy, obesity (BMI > 35 kg/m2), knee vs hip surgery, unilateral knee replacement vs bilateral operation, TKA vs UKA, and those who used analgesics preoperatively. Previously, increased opioid consumption after hip and knee replacement has been associated with TKA vs THA, preoperative opioid or other analgesic use, psychiatric disorders (especially depression/anxiety), tobacco use, cardiac disease, younger age, greater affected joint pain, other pain sites, TKA vs UKA, use of walking aids, and female gender [9, 12,13,14,15, 17,18,19, 21, 26]. In addition, obesity and the number of comorbidities (based on CCI score) have been associated with increased opioid use after knee replacement [14, 19]. The new findings in this study were that obesity, epilepsy, and overall comorbidity (based on modified CCI score) were associated with greater opioid use after both hip and knee replacement.
To the best of our knowledge, factors associated with postoperative consumption of acetaminophen have not been analyzed in previous literature. In this study, its use was in general mostly associated with the same factors as the overall analgesic use. This is expected as acetaminophen has been considered the basis of pain management  and it was the most common analgesic also in this study. Increased use in patients with higher age, number of comorbidities, and those with neurodegenerative disease is likely explained by avoidance of NSAID and opioid use in these patient groups.
The main strength of our study is the inclusion of all analgesic drug groups (acetaminophen, NSAID, opioid) in a large sample of unselected joint replacement patients. We included only primary joint replacements and only osteoarthritis as an indication for surgery. Other indications were excluded because, for example, patients with rheumatoid arthritis tend to have polyarticular involvement more often, which serves to hamper the analysis of the use of analgesic drugs. Patients with revision or other joint replacement during the follow-up period were excluded because the perioperative peak in the consumption of analgesic drugs related to the latter operation would have been a potentially confounding factor considering the analysis of postoperative analgesic use after index surgery. According to our previous analysis , 9–12 months after surgery appears to be a suitable time point for analyzing the level of postoperative analgesic use. Because the use patterns are similar after hip and knee replacement and also among different comorbid conditions, the materials could be analyzed as a whole to increase statistical power. All operations were performed in a single orthopedic hospital with standardized perioperative care, anesthesia, and analgesia. Medication data include all the prescribed medications that were dispensed in Finland. The presence of comorbidities was based on a nationwide Reimbursement Register that covers the most important comorbidities, although a clear limitation is that the register does not cover mental health problems other than psychotic disorders. However, we included the use of antidepressants in the adjusted model, along with several other possible confounders.
The limitations of the study include the lack of information on OTC drugs and the fact that in a register study based on redeemed prescriptions, it is not possible to find out the indication for the analgesic drug (index joint or other pain sites) nor whether the patient has taken the drug or not. For these reasons, the actual use of acetaminophen and NSAIDs (that are available OTC) might be higher than reported here, but it is unlikely that this would have affected the use patterns or associated factors. In the general population, higher analgesic use has been found on women, elderly, smokers, obese, and patients with chronic diseases or with several medications [30, 31, 37], which should be noted in the interpretation of the results since indication for the analgesic is not known. Although patients with major complications leading to revision were excluded, we were not able to analyze all postoperative complications. Moreover, preoperative pain, the intensity of pain, and the prevalence of persistent pain were not examined in this study.