Myocardial infarction and mortality following joint surgery in patients with rheumatoid arthritis: a retrospective cohort study

Rheumatoid arthritis (RA) is the most common form of inflammatory arthritis affecting up to 1.3 % of people worldwide and approximately 2 % of Australians [1, 2]. RA is associated with excess cardiovascular mortality and morbidity, which contributes significantly to the burden of this disease [3‐5].

In an earlier study investigating mortality following a first acute cardiovascular event we found RA patients had increased 30-day case fatality after acute myocardial infarction (MI) compared to the general population [4]. Subsequent interrogation of hospital data from three Victorian tertiary referral centres showed that in patients who experienced MI, a preceding medical illness or surgical event was more substantially common among patients with RA than in those without RA (31 % vs 16 %, respectively) [6]. In other words, in almost a third of the RA patients who had MI, the MI had been preceded and possibly precipitated by a medical or a surgical insult. This led us to postulate that patients with RA might have an increased burden of silent cardiovascular disease, which is uncovered by a period of physiologic stress. To examine this further we wished to measure the risk of MI and other adverse cardiovascular outcomes in patients with RA compared to the general population following some form of physiologic insult. We chose joint surgery to represent our physiologic insult given that is a relatively standardized procedure that is not uncommonly performed in patients with RA.

Hence, the primary aim of this retrospective cohort study was to compare the risk of MI and death within 6 weeks and 12 months of any form of joint surgery in patients with RA versus patients without RA. The limited research into perioperative cardiovascular risk among patients with RA is primarily based on examination of cardiovascular complications following total knee joint replacement (TKJR) and total hip joint replacement (THJR), so we additionally performed subgroup analyses in these groups to facilitate direct comparison [7‐10].

During the study period 240,571 individuals underwent a total of 308,589 joint surgery procedures, of which 3,654 (1.2 %) were performed on 2,219 people (0.92 %) with RA. Among the RA group, hip and knee joint arthroplasty were the most commonly performed surgical procedures (55.5 %), whereas in the non-RA group these types of surgery accounted for 26.6 % of all procedures. See Additional file 2 for further details of type of joint surgery episodes by RA status. The demographic details and clinical features of the RA and non-RA groups are shown in Table 1. Compared to those without RA, the patients with RA were older and more likely to be female, of lower socioeconomic status, less likely to be smokers and had a greater burden of comorbid disease, including hypertension, diabetes, vascular disease, pulmonary disease and renal disease. There were fewer elective and more public patient admissions among people with RA compared to those without RA, but there were no differences between the groups in accessibility to goods and services, or the numbers with an English-speaking background. See Additional file 3 for summary of patient level data.

Table 2 shows the number of events and odds ratios (OR) of MI and mortality following joint surgery in the RA and non-RA groups. Unadjusted models showed those with RA had greater odds of experiencing MI or death at both 6 weeks and 12 months compared to the non-RA group. After adjusting for age, sex, comorbidities and other potential confounders the OR for MI within 6 weeks of joint surgery in the RA patient group remained elevated but was no longer statistically significant (OR 1.50, 95 % CI 0.96–2.33) (Table 2). The adjusted OR for all other 6-week and 12-month outcomes remained statistically higher in the RA group as follows: all-cause death within 6 weeks (adjusted OR 1.85, 95 % CI 1.09–3.13) and 12 months (adjusted OR 2.18, 95 % CI 1.66–2.86), cardiovascular death within 6 weeks (adjusted OR 1.90, 95 % CI 1.07–3.37) and 12 months (adjusted OR 2.30, 95 % CI 1.65–3.22) and MI within 12 months (adjusted OR 1.70, 95 % CI 1.27–2.28).

The results from the subgroup analysis are given in Table 3. Among the THJR group, the adjusted OR for CV death within 6 weeks and 12 months, and all-cause death within 12 months were significantly higher for those with RA. Following TKJR surgery, the adjusted OR for MI and death at 6 weeks and 12 months were no different between the groups. Following all other types of joint surgery, the adjusted OR for MI at 6 weeks and all 12-month outcomes was significantly higher for those with RA. However, the adjusted OR for death within 6 weeks was no different between the RA and non-RA group.

This large population-based study investigated the 6-week and 12-month risk of MI and mortality among RA patients undergoing any form of joint surgery, with subgroup analysis of total hip and knee arthroplasty. Patients with RA had adjusted odds of cardiovascular and all-cause mortality at 6 weeks after surgery that were nearly twice those of patients without RA. In our subgroup analysis, patients with RA had significantly higher odds of MI within 6 weeks and 12 months following joint surgery other than TKJR and THJR. An increased risk of early cardiovascular mortality was also demonstrated when analysis was limited to THJR surgery but not in patients with RA who underwent TKJR or other types of joint surgery. Outcomes at 12 months were almost universally worse for patients with RA, with cardiovascular and all-cause mortality significantly elevated in the cohort overall, and in the subgroup analysis among those who had undergone THJR surgery and joint surgery other than hip and knee arthroplasty. MI within 12 months of surgery was increased in the RA group overall, and was elevated but not statistically significantly so for the patients with RA who underwent total hip or knee joint arthroplasty.

Our findings of an increased risk of short-term death in the overall joint surgery cohort are consistent with a previous study. Soohoo et al. reviewed data from 138,399 patients undergoing primary THJR in California from 1995 to 2005, and as in our study, found that patients with RA had a significantly increased risk of short-term (90-day) mortality compared to patients with osteoarthritis (OR 1.88, 95 % CI 1.17–3.03) [19]. However, in most studies, as with the findings from our subgroup analysis of TKJR and THJR surgery, there was no difference in all-cause short-term mortality following hip and knee arthroplasty [7‐9, 20‐23]. A meta-analysis by Ravi et al. reported no evidence of excess 90-day mortality in patients with RA following THJR (OR 1.40, 95 % CI 0.82–2.39), or TKJR (OR 0.86, 95 % CI = 0.66–1.12) [21]. Subsequent to the meta-analysis, a number of studies have examined complications, including cardiovascular events and mortality, in patients with RA, following total joint arthroplasty. All studies reported no difference in the odds of short-term mortality [7, 8, 22, 23].

Very few studies have investigated long-term risk of mortality. We found a significant difference in the odds of 12-month mortality between the RA and non-RA groups in the total cohort or in the THJR subgroup analysis. Michaud et al. analyzed 7 years of data from US veterans to examine the complications and mortality in RA patients following total hip or knee arthroplasty [22]. Unlike our findings, there was no difference identified in mortality at 12 months (OR 1.30, 95 % CI 0.87–1.96) in the 839 patients with RA compared to those with osteoarthritis. However in their analysis of mortality up to 8 years post-surgery they reported an increased risk among those with RA compared to those with osteoarthritis (hazard ratio 1.22, 95 % CI 1.00–1.49).

Short-term risk of cardiovascular complications other than death has been assessed in a number of studies. In our total cohort of patients who underwent joint surgery and among those who underwent THJR and TKJR we found there was no difference in short-term risk of MI between the RA and non-RA groups. Similarly, in patients with RA compared to those with osteoarthritis, Michaud et al. did not observe any difference in the odds of cardiovascular events (including MI and cerebrovascular accident) within 30 days. In another study of 351,103 TKJR episodes between 2006 and 2010, of which 11,755 patients had RA, Stundner and colleagues also did not find any difference in the odds for combined perioperative adverse events (which included 30-day mortality, MI, infections and other complications) in patients with RA compared to those with osteoarthritis (adjusted OR 0.98, 95 % CI 0.92–1.05) [7]. In contrast, Stundner et al. examined US data from 157,775 THJR episodes between 2006 and 2010, of which 5,400 were in patients with RA, and found that patients with RA had higher overall odds of perioperative complications, including cardiovascular complications (adjusted OR 1.15, 95 % CI 1.05–1.25), compared to people with osteoarthritis [8].

Among those who underwent surgery other than THJR and TKJR, we found people with RA had an increased risk of MI within 6 weeks. This group of patients were mainly undergoing knee arthroscopy (69 %) and surgery on the shoulder (16 %) with the remainder undergoing hand/wrist, elbow and ankle surgery (8 %) and a small number undergoing revision arthroplasty of the hip or knee (6 %). This increased risk of MI within a group predominantly undergoing a presumably less significant hemodynamic insult is of interest and suggests the possibility that even relatively minor physiologic stress could precipitate a cardiovascular event in a predisposed patient with RA. Our data do not specifically address this issue, however.

The differences in findings from our study compared to other studies may be explained by a number of factors, including the differences in study time periods examined, the prevalence of RA, the variation in the postoperative time period examined (for example 6 weeks versus 90 days), combination of cardiovascular complications into one measure in a number of studies (whereas we looked at MI on its own) and differences in demographic characteristics of the study population. In particular, the Michaud study was a predominantly male population compared to our study, which involved predominantly female patients with RA.

Our study adds to the findings of the above mentioned studies by reporting on outcomes for patients with RA following joint surgery other than hip and knee arthroplasty. We found that joint surgery was not related to short-term MI but was significantly associated with short-term mortality and long-term MI and mortality. It is unknown whether this finding is related to the overall increased risk of CV events and mortality among patients with RA [24]. Further research should consider an investigation of CV outcomes and mortality in patients with RA undergoing surgery and those not receiving surgery, to determine if there is an additive risk associated with surgery.

This study utilized 10 years of hospital administrative data for all types of joint surgery within a single Australian state and included outcome information for 308,589 procedures, including 3,654 involving patients with RA. There are, however, several limitations that should be noted. The classification of patients with RA was based on coded diagnosis of RA in the hospital record, which was not collected specifically for research purposes. We utilized a 2 year retrospective period to increase the chances of accurately detecting RA cases, however, mild cases of RA may have been missed. The use of hospital administrative data to ascertain RA status may result in misclassification. We may have misclassified people with mild RA as false negatives (that is, no diagnosis of RA), and only captured patients with more severe or active RA who had been hospitalized. However, this is likely to lead to a bias toward the null hypothesis and is therefore a more conservative estimation of the true effect. We did not have access to medication data or outpatient administrative data to further validate the RA diagnosis. However, findings from an Australian study into the quality of ICD-10 coding of Victorian hospital discharge data reported the sensitivity and positive predictive value (PPV) for identifying RA as 91 % (95 % CI 74–100 %) and 77 % (95 % CI 54–100 %), respectively [25]. In addition, a recent Canadian validation study reported the optimal algorithm for identifying patients with RA from health administrative data was one hospitalization code ever or three physician diagnosis codes from claims with one or more RA code by a specialist in a 2-year period [26]. The comparability of Canadian and Australian coded administrative hospital data for identifying RA is unknown. Our classification of comorbidities was also dependent on coded data, which may have inaccuracies but is unlikely to be differentially distributed between RA and non-RA cases, and therefore unlikely to introduce systematic bias into the results [27].

We were interested in assessing all-cause mortality as well as cardiovascular mortality. The cause of death data were available in text form only and did not allow us to use a more standardized approach (for example, using ICD-10 codes) to define cardiovascular death, or differentiate between an underlying and associated cause of death. We accept this is a limitation of the study as it may result in misclassification of this outcome. However, this is unlikely to be unequally distributed between the RA and non-RA group and again unlikely to introduce systematic bias into the results.

Information on RA severity, activity, treatment, preventive care and use of cardiovascular medicines was not available, and we therefore unable to examine the influence of these factors on postoperative mortality. Our dataset does not capture non-fatal events that occur outside the Victorian hospital setting, such as those that occur in the community and do not result in hospitalization or hospitalization in other Australian states. Consequently we may have underestimated the true burden of postoperative MI in our cohort. Finally, with the growing use of biological disease-modifying anti-rheumatic drugs, the incidence of joint surgery has been decreasing and long-term prognosis has been improving for people with RA [28, 29]. As such, the odds of mortality and morbidity presented in this study may differ from current measures of risk, as the profile of people with RA receiving joint surgery is evolving over time.

In conclusion, at 6 weeks following joint surgery, patients with RA had almost double the odds of all-cause and CV mortality compared to those without RA. At 12 months post joint surgery, the RA group had significantly higher odds of MI, all-cause death and CV death compared to those without RA. Understanding the risks of postoperative MI and mortality can help patients with RA make informed decisions about undergoing joint surgery procedures. Further studies to determine the contribution of subclinical cardiovascular disease, RA disease severity, functional status and RA treatments to this increased mortality are warranted.