In this study, we found that younger age and depression were associated with higher odds of use of NSAIDs, and male gender with lower odds of use of NSAIDs for the treatment of index TKA pain, 2 and 5 years after primary TKA. BMI ≥ 40 was associated with higher NSAID use 2 years after primary TKA. Female gender, younger age and anxiety were associated with higher odds of use of opioid pain medication for the treatment of index TKA pain at both 2 and 5 years after primary TKA.
Our study has many novel findings that add to the current knowledge. The prevalence of use of pain medication at 2-year (8.5% NSAIDs and 1.4% opioids), or 5-year (9% NSAIDs and 1.4% opioids) follow-up after primary TKA in our study is similar to that reported previously in a cohort of primary THA patients from our clinical registry at the same time-points (12 to 13% NSAIDs or 2 to 3% opioids) . Loss to follow-up and non-response bias may explain this discrepancy in use of pain medication after THA vs. TKA, since poorer pain outcomes are expected after TKA than after THA, based on other clinical outcomes studies of pain. The proportions are lower, as expected, than the previously reported overall 14% prevalent use of opioids 1 year after TKA  and 49% for use of any NSAID 2 years after TKA . The use of pain medication in both these studies represented the overall use of NSAIDs or opioids compared to use of these medications for persistent index arthroplasty pain. Our study outcome captures patients with inadequate relief of index arthroplasty pain leading to persistent use of either NSAIDs or opioid pain medications. This represents a clinically important outcome for patients undergoing arthroplasty. Studies are needed to examine use of these medications at long-term follow-up after TKA.
Preoperative depression predicted use of NSAIDs for index TKA pain at 2 and 5 years after primary TKA. In an unadjusted analysis, one previous study reported that female gender and younger age were associated with higher use of opioid pain medication in TKA . Our study confirms this finding in a multivariable-adjusted analysis, and extends this observation to 2 and 5 years after primary TKA. Depression predicts pain at 1 year  and pain and function at 2-year  and 5-year follow-up after primary TKA . To our knowledge, there are no published studies that have examined whether depression is associated with use of pain medication 2 to 5 years after primary TKA. We speculated that depression might be associated with use of NSAIDs and opioids, but our study found that depression predicted NSAIDs use and anxiety, but not depression, predicted opioid use after primary TKA. These analyses were adjusted for multiple demographic and clinical factors suggesting that this is an independent association. This finding suggests that screening for depression and anxiety prior to surgery may help identify patients at-risk for poorer pain outcomes. Further studies need to examine if preoperative optimization of depression and anxiety can reduce the use of pain medications after primary TKA.
The association of male gender with lower likelihood of use of NSAIDs and opioid pain medication merits some discussion. Lower odds of use of NSAIDs and opioids in men vs. women with primary TKA is similar to lower analgesic use reported for men compared to women in national cohorts of patients in the US and Sweden [27, 28]; our study extends this finding to populations undergoing primary TKA.
The association of older age, compared to patients 60 years and younger, with lower odds of use of NSAIDs and opioids is interesting. Other studies have reported lower use of pain medication or use of lower doses and/or greater benefit with the same doses in older individuals compared to younger controls [29–34]. This may be related to higher perceived risk of associated adverse events and contraindications, higher pain thresholds and/or patient preference for use or non-medical interventions in the elderly compared to younger patients.
These findings must be interpreted in conjunction with our previously published findings of pain and function outcomes in this cohort of patients. Higher Deyo-Charlson index, female gender, higher BMI and older age were associated with worse functional limitation , while female gender, higher comorbidity and younger age, but not BMI, were associated with worse pain outcomes  after primary TKA. In the current study, we found that female gender and younger age were associated with higher odds of use of NSAIDs and opioid pain medications. The differences in factors associated with each of the three outcomes, namely the use of pain medication, moderate-severe pain and moderate-severe functional limitation, should not be surprising. Conceptually, these domains have some overlap, but these not the same or simply surrogates for each other.
Our study has several strengths. We used multivariable analyses to assess predictors of the use of NSAIDs and opioid pain medication after primary TKA in a large sample of patients from a prospective US clinical joint registry. Most previous studies were limited to 1- to 2-year follow-up, did not ask about use of medication specifically for knee pain, and did not control for important covariates that can impact the use of pain medication (including depression and anxiety). The use of pain medication after primary TKA is under-studied in the arthroplasty literature.
Our study has several limitations. Non-response and referral bias may limit generalizability to general populations. The response rate of 65% at 2 years and 56% at 5 years, is similar to the average response rate of 60% in large surveys of this size . Despite controlling for several important confounders and covariates, residual confounding is possible. Currently there is no national US joint registry, and therefore an analysis of data collected over a considerable time period from a large volume medical center (such as ours) is the next best approach. We used the diagnostic codes for depression and anxiety from medical records, and these are known to be under-coded. A similar limitation applies to other diagnoses in the Deyo-Charlson index. Such misclassification bias would have made our estimates more conservative compared to true associations. The actual associations may have been even stronger, had all patients been screened for depression using validated instruments (Center for Epidemiological Studies Depression (CES-D) scale or Beck's Depression Inventory), or an examination by a psychologist. Despite a large sample size, few patients reported use of opioid medication for the index primary TKA pain, with approximately 10 to 15 patients per variable available for analysis in the multivariable-adjusted model. These results must not be over-interpreted and need to be confirmed in future studies. Non-responders to the survey may be more likely to have poorer outcomes and a higher prevalence of use of pain medication, therefore, the actual use of these medications may be higher than that reported here. Our analyses did not account for use of preoperative pain medication, which might influence postoperative use.