Associations between pre-surgical daily opioid use and short-term outcomes following knee or hip arthroplasty: a prospective, exploratory cohort study

Australia [1, 2], along with other countries [3, 4], is currently considered to have a major pharmaceutical opioid problem. Hospitalisation due to opioids [5] and mortality due to overdose [6] have increased since 2000–01. Whilst the Royal Australian College of General Practitioners consider opioids an important part of the armoury for managing pain, their use is cautioned for their association with dose-dependent harm including dependence, withdrawal, falls, and cognitive effects, and rather should be administered as a last resort [7]. It is well-recognised that people suffering from chronic knee or hip arthritis are prescribed strong opioid therapy [8‐10], and this is despite i) an absence of strong evidence of clinically relevant benefit over active or inactive control therapies [11, 12], and ii) concern over their safety [13]. Moreover, many people – up to 24% as estimated by a recent systematic review [14] - with end-stage arthritis and awaiting total knee or hip arthroplasty (TKA, THA) are taking opioids prior to surgery and this appears to have implications for recovery. Retrospective studies indicate pre-surgical opioid use is both a strong risk factor for persistent use post-surgery [15‐17] and is associated with higher risk of readmission, revision surgery and worse outcomes generally including periprosthetic joint infection [18‐23]. The aforementioned systematic review concluded that pre-surgical opioid use (versus no use) has a moderate effect on absolute index joint scores (pain, function or both) 6–58 months post-surgery though relative improvements are similar [14]. Very little is known about index joint scores and even health-related quality of life scores within 6-months of surgery. Understanding early recovery is important given that long-term recovery may be confounded by other factors, especially if people continue to take opioids long-term after surgery.

Given the above concerns, opioid tapering is recommended for chronic opioid users prior to elective surgery [18, 24], thus, tapering may be a useful strategy for chronic opioid users prior to arthroplasty. However, prior to recommending what is likely to be a resource-intensive, multi-pronged approach [24, 25], evidence from prospective trials, controlling for important confounders, are needed to accurately depict both the extent of the problem and the associated harms.

The aims of this prospective, exploratory study were multiple. Using a single-centre, Australian public hospital cohort, we aimed to:

521 people underwent primary TKA or THA (n = 381 TKA; n = 140 THA; n = 503 unilateral) over the 10-month study period (23rd July 2018 to 29th May 2019). Cohort derivation and retention to 12-weeks post-surgery are summarised in Fig. 1. 96.5% (n = 503) and 86.8% (n = 452) were available for follow-up at 4- and 12-weeks post-surgery (one patient died acutely). Compared to those retained, those LTFU at 12-weeks were similar in most key characteristics: female sex [66.7 (LTFU) vs 65.3% (retained), p = 0.82], procedure [TKA 76.8 vs 72.6%, p = 0.46], pre-surgery BMI [32.3 (6.1) vs 33.0 (6.8), p = 0.33], pre-surgery Oxford scores [17.7 (8.8) vs 18.2 (8.2), p = 0.69], pre-surgery EQVAS [70 (IQR 29) vs 65 (IQR 30), p = 0.76], age [70.2 (10.2) vs 67.5 (9.5) yr, p = 0.04], and regular opioid users pre-surgically [15.9 vs 15.7%, p = 0.98].

Table 1 summarises the cohort according to the exposure variable. 15.7% (95% CI 12.6 to 18.9) of the cohort were regular users of opioids pre-surgery. A significantly greater proportion of the pre-surgical opioid user group were undergoing THA, had an ASA score of 3 or 4, did not require an interpreter, had a mental health condition, and were a current smoker. The pre-surgery opioid user group also had a significantly higher mean BMI and a lower mean Oxford score. Education level was also significantly different between the groups with a greater proportion in the group taking opioids pre-surgically reporting a higher level of education attainment. Opioid users also had a slightly shorter wait time for surgery.

Table 2 summarises the mean, median or proportions as appropriate for each outcome. Table 3 summarises the results for unadjusted and adjusted analyses excluding the OMED results. The results of the complete models, including specification of the covariates for each model, are provided in the Additional files 2, 3, 4, 5, 6, 7, 8, 9 (Tables 1S, 2S, 3S, 4S, 5S, 6S, 7S, 8S). Due to the lower than expected frequency of regular opioid users pre-surgery, the maximum number of covariates in the models was restricted to nine. Approximately 25% of the cohort (predominantly non-English speakers) did not complete PROMS pre-surgery, reflecting the voluntary and unsupervised nature of completion using a centralised digital system. Consequently, PROM data were not used as covariates for any outcomes.

Observations from this exploratory, prospective study corroborate data obtained elsewhere using retrospective study designs, but also provide new, potentially contrary insights about the early post-surgical risks associated with opioid use pre-surgery. We observed, as have others, that people undergoing primary TKA or THA who are regular opioid users pre-surgery have higher odds of ongoing opioid use post-surgery [15‐17]. A novel finding was the greater daily consumption of opioids acutely by regular opioid users even after accounting for other factors. Few others have captured opioid use acutely [32, 33]; confounding factors such as LOS or complications have not been accounted for so comparisons are difficult. We acknowledge (and the same uncertainty applies to previous studies [32, 33]) that it is unknown whether the greater consumption acutely is clinician-driven (i.e. is pre-emptive as the anaesthetist a priori prescribes higher doses to those perceived as opioid tolerant) or patient-driven (i.e. is reactionary, with the patient simply requiring more to alleviate pain given their possible dependency).

In contrast to previous investigations [18‐23, 33], chronic opioid use pre-surgery was not strongly associated with complications after accounting for other factors. The directional changes (higher rates amongst pre-surgical opioid users) were consistent with previous research, however, so a larger sample may have secured a significant result. Differences in follow-up duration, however, may also explain the discrepancy. Longer follow-up times are themselves confounded by time so other factors (for example, persistent opioid use) may be contributing to complications seen in studies with long follow-up, and our follow-up was too short to gauge insights into prosthesis longevity, for example. The degree of opioid dependence pre-surgery may also be a factor. Recent large studies from the United States (US) using insurance claims data demonstrate that adverse events or readmissions are more common in those who used opioids for longer periods pre-surgery [19, 22]. At the very least, our data suggest that minor, non-joint complications may be driving an association between chronic pre-surgical use and acute complications. Whilst such events appear to increase the costs of care in the form of investigation costs, their clinical relevance, certainly for the long-term success of the surgery, may be less concerning than joint-specific issues. That the extra risk with pre-surgical opioid use appeared to be associated with only minor complications may explain why we observed a similar LOS across the groups whilst others have demonstrated that pre-surgery opioid use is a predictor of a longer LOS [32‐34].

We did not find that pre-surgery opioid use was associated with worse patient-reported index joint scores or health-related quality of life in the early sub-acute period. That we were unable to include pre-surgery PROMs in the modelling (due to a high proportion of missing data) does not likely explain this particular lack of association as patients in the pre-surgical opioid group had worse scores prior to surgery (significantly worse for the Oxford scores) and this would likely have meant this group had a larger relative improvement in their scores at 12-weeks. It is possible, then, we would have found that pre-surgical use was associated with greater improvement. An aforementioned systematic review concluded opioid use pre-surgery does undermine joint-specific scores at 6-months or more post-surgery [14], but only in terms of absolute recovery and not relative change. A recent non-randomised, retrospective study concluded that PROMs were better 6–12 months post-surgery amongst chronic opioid users who reduced their dose prior to surgery compared to those who did not; and that the former were similar to an opioid naive group [35]. We contend then that the association between pre-surgical opioid use and PROMS post-surgery, like the association with complications, is complex and in part depends on the length of follow-up but also possibly on the level of dependence or the degree of tapering that has occurred (if any) pre-surgery. Our study only established ‘use’ pre-surgery and not ‘how much’ and for ‘how long’.

It is of interest to note that almost 34% of chronic opioid users remained opioid users at 12-weeks post-surgery and this is at a time when symptomatic and functional recovery has already undergone large and clinically relevant improvements [36‐38]. It is also noteworthy that many reported the ongoing use for problems other than their index joint. A recent population-based Australian study observed that amongst a cohort of people who were prescribed opioids for non-cancer pain, 2.6% were persistent users 12-months after the initial prescription [39]. Considering these observations together, we contend that the rate we observed at 12-weeks is alarming given that many of our patients were using opioids at the time they were waitlisted for surgery almost 12-months prior. If arthroplasty recipients are at higher risk of persistent opioid use than persons suffering from other forms of non-cancer pain, arthroplasty services are arguably obliged to educate recipients and their families as well as their general practitioners about the need to monitor use and have a plan for weaning [40]. This is in addition to acute care practices designed to spare opioid analgesia [41].

Pooled data from the US estimate that approximately 24% of people undergoing TKA or THA are chronic opioid users pre-surgery [14]. Our data indicate the proportion (~ 16%) to be slightly lower. The disparity in observed rates may in part be related to how pre-surgical opioid use was defined; we note that in the aforementioned review the definition of ‘opioid use pre-surgery’ varied widely, incorporating use within 6-weeks of surgery or any use within 2-years of surgery [14]. There may be other contributing factors as well. The present study was conducted in a local health district with vast ethno cultural diversity and constitutes one of the most multicultural communities in Australia [42]. This is significant because Australian data suggest lower consumption of pharmaceutical and other drugs among CALD communities [43], consistent with international literature on minority communities [44, 45]. Our cultural and language diversity also likely explains our arguably counterintuitive observation that a greater proportion of people who were opioid users pre-surgery had a higher level of education than non-users. Approximately 30% of the cohort were non-English speakers and lack of English proficiency is often associated with low level education in our CALD population. The majority of the non-English speakers were in the non-user group, hence, we see a higher level of education amongst the pre-surgical opioid users (mostly English speakers). Another explanation for our comparatively low rate of pre-surgical opioid use is that opioid use may be region-specific. Data from the US data suggest that opioid use prior to arthroplasty varies by region from 8.9 to 26.4% [46]. An additional or alternative explanation is that because the pooled data in the systematic review were based on data entirely from populations within the US where the level of opioid prescription observed is higher generally [47], our differences may simply be reflective of this phenomenon.

Strengths of our study include the provision of novel data from an Australian perspective capturing almost 100% of those screened with little loss to follow-up. The prospective design allowed us to include relevant covariates not included in retrospectively designed studies such as need for an interpreter or complication status (for PROMs outcomes), and we included non-English speakers, improving the generalisability of our results. Limitations are several and include that we do not know whom was opioid dependent pre-surgery and we could not test for a dose-response as data on dose pre-surgery was unclear. We relied on patient-reported use of opioids both prior to surgery and after discharge. The sample was small compared to the administrative datasets used in retrospective studies and this may have affected our power to control for many covariates, especially those with low rates of occurrence such as alcohol use and bilateral surgery. That said, the incidences in both groups were similar. We were unable to include PROMS as covariates in our modelling, we examined multiple outcomes without adjusting significant probability values, and our results were limited to a single centre and the short term.

This exploratory study confirms previous observations that people who take daily opioids pre-surgery have significantly greater odds for greater opioid consumption acutely and ongoing use post-surgery. Given the sample size limitations here and the many covariates that need to be considered in modelling, adequately powered prospective studies are required to confirm whether pre-surgical opioid use is or is not associated with poorer joint and quality of life scores or a complication in the short-term.