|Home | About | Journals | Submit | Contact Us | Français|
Rheumatoid arthritis (RA) is associated with an increased cardiovascular (CV) burden similar to that of diabetes mellitus (DM). This risk may warrant pre-operative CV assessment as is performed for patients with DM. We aimed to determine if the risk of perioperative mortality and CV events among patients with RA differed from those of unaffected patients and those with DM.
We used 1998 to 2002 Nationwide Inpatient Sample of the Healthcare Cost Utilization Project (HCUP-NIS) data to identify elective hospitalizations of patients undergoing non-cardiac surgery. Surgical procedures were categorized as low risk, intermediate risk, and high risk of CV events using established guidelines. Logistic models provided the adjusted odds of study endpoints in RA, DM, or both relative to neither condition.
Among 7,756,570 patients with a low risk, intermediate risk, or high risk non-cardiac procedure, 2.34%, 0.51%, and 2.12% had a composite CV event, respectively, and death occurred in 1.47%, 0.50%, 2.59% respectively. Among those with an intermediate risk procedure, death was less likely in RA than DM patients (0.30% vs. 0.65%; p <0.001), but the difference in mortality among those with low risk or high risk procedures was not significant. Patients with RA were less likely to have a CV event than patients with DM with procedures of low risk (3.38% vs. 5.30%; p <0.001) and intermediate risk (0.34% vs. 1.07%; p <0.001). In adjusted models, RA was not independently associated with an increased risk of perioperative mortality or CV event.
RA was not associated with adverse perioperative CV or mortality risk, suggesting a lack of need for a change from current perioperative clinical care.
Persons with Rheumatoid Arthritis(RA) have a higher risk of all-cause and CV-specific mortality as compared to non-RA population.1,2 In addition, individuals with RA are at an increased risk for myocardial infarction, congestive heart failure, stroke, and peripheral vascular disease.3–6 Moreover, there is evidence of clinically silent cardiovascular disease in individuals with RA.7 Since the differences in traditional cardiovascular risk factors do not entirely explain this elevated risk, the underlying etiology of the increased CV risk is likely to be multi-factorial. In addition to the traditional CV risk factors, the effects of systemic inflammation, plaque instability, impaired coronary flow reserve, elevated thrombotic markers, and insulin resistance may play a role.7–15 Recently, studies have compared the elevated CV risk in RA to that observed in patients with diabetes mellitus (DM).16,17 Accordingly, due to the increase in CV disease burden, the need for CV risk assessment and management in RA has been recognized.18
Cardiac complications are the major cause of perioperative morbidity and mortality.19,20 The risk of a perioperative coronary event or cardiac-specific mortality is nearly two-fold higher in patients with known or suspected coronary or atherosclerotic disease relative to those without.21–23 The presence of increased atherosclerotic burden in patients with RA adds complexity in the perioperative risk assessment. Additionally, physical inactivity and limitation due to RA may make angina an insensitive marker of CAD, as physical exertion may not be at a sufficient level for patients to develop symptoms of angina. Indeed, if RA is a "diabetic equivalent", then the associated CV risk would need to be incorporated into the preoperative risk stratification.
Currently, it is to be determined whether the increased atherosclerotic burden of RA translates into an increased risk of adverse perioperative events, including CV events and mortality. Additionally, given the recent literature drawing similarities between the CV risk in RA and DM in the non-operative setting, it is of clinical interest to determine whether these similarities extend to the perioperative setting. Such similarities would imply the need for a more comprehensive approach to preoperative clearance with a focus on CV risk in patients with RA. We aimed to determine and compare the risk of perioperative CV events and mortality during elective hospitalizations among patients with RA, DM, both, and neither diagnoses. We hypothesized that patients with RA would have a risk of perioperative CV events and mortality higher than that of patients without RA or DM, similar to that of patients with diabetes.
The Nationwide Inpatient Sample (NIS) of the Healthcare Cost and Utilization Project (HCUP), sponsored by the Agency for Healthcare Research and Quality (AHRQ), is a survey design-based database of discharge information for inpatient care from non-federal (excludes Veterans Hospitals and other federal facilities), non-rehabilitation acute-care short-term hospitals.24 The NIS is an annual sample of hospital discharges providing national estimates of the characteristics of patients, diagnoses, and hospital-based procedures performed in U.S. acute-care hospitals. Each year of NIS data used in our analysis included approximately 6.8 million hospital discharges based on a sample from more than 984 hospitals across 22 states. The number of hospitals and states included in our analysis increased annually from 984 hospitals among 22 states in 1998 to 995 hospitals from 35 states in 2002. The data for this study, including the annual NIS datasets, Procedure Class software tool, Clinical Classifications Software tool, and NIS Trends Supplemental Files (NIS-Trends) for pooled analysis of multiple years of NIS data, was provided by the Nationwide Inpatient Sample (NIS), Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality (AHRQ).24–27
The characteristics of the subpopulation of interest included: hospital discharges between 1998 through 2002 with a non-cardiac principal procedure during the hospitalization, age 18 years and older, and the hospitalization designated as elective. We included only elective hospitalizations because this group is most relevant for the question of pre-operative clearance, and to ensure the greatest likelihood that the procedure preceded the perioperative CV event.
Medical diagnoses of RA, DM, and other comorbid disease states, including hypertension, congestive heart failure without pulmonary edema, valvular heart diseases, coronary artery disease, and chronic kidney disease, were abstracted from among 15 possible NIS discharge diagnosis fields, by ICD-9 medical diagnosis codes (see Appendix A). Likewise, the principal procedure during the hospitalization was abstracted using the ICD-9 code listed as the principal procedure by NIS. The Clinical Classifications Software (CCS), a software tool developed by the AHRQ, was used to cluster the principal procedures into meaningful procedure categories (see Appendix B).25 The CCS defined principal procedure categories were then further categorized according to cardiac risk level into low, intermediate, or high cardiac risk guided by the American College of Cardiology/American Heart Association Task Force published clinical practice guidelines.28 For example, endoscopic, ophthalmologic, and procedures on the breast were categorized as low risk while orthopedic, prostatic, other intra-abdominal procedures were categorized as intermediate risk. High risk surgery included peripheral vascular surgery or non-cardiac procedures within the cardiothoracic cavity. Finally, any secondary procedure was categorized using the Procedure Class software tool and a tally count variable was created for minor and major therapeutic secondary procedures.27
The additional covariates included as possible confounders included age, sex, comorbid diseases including hypertension, heart failure without pulmonary edema, valvular heart disease, coronary artery disease, and chronic kidney disease, and any secondary minor or major therapeutic procedures.
The study outcomes included all-cause mortality during the hospitalization and a composite CV event during the hospitalization, which was defined as the occurrence of an acute myocardial infarction, non-ST elevation MI, CHF with pulmonary edema, or acute cerebrovascular accident.
A four level disease status categorical variable was generated to categorize individuals based on the presence or absence of the diagnosis of RA and DM (neither RA nor DM present, RA present and DM absent, RA absent and DM present, both RA and DM present). Associations in categorical data were determined using the design-adjusted Rao-Scott Chi Square test. Logistic regression was used to determine the odds of the composite CV outcome and all-cause mortality among hospitalized individuals with RA alone, DM alone, both RA and DM, as compared to those without either RA or DM. Bivariate and logistic regression analyses were performed separately for each level of procedure risk using subpopulation analyses. Two supplementary analysis, one to exclude case-mix in which we conducted separate procedure-specific analysis to determine the odds of adverse perioperative events among elective hospitalizations with a principal procedure of total knee replacement. Secondly, because of the multi-factorial and potentially non-atherosclerotic etiology of decompensated heart failure, we conducted separate analyses excluding acute heart failure as a component of the composite CVD endpoint.
The potential cofounders included age as a continuous variable, and categorical variables including sex, hypertension, coronary artery disease, valvular heart disease, congestive heart failure without pulmonary edema, chronic kidney disease, minor and major therapeutic secondary procedures. A significance level of 0.05 with a two-sided test was used for all hypotheses. All statistical analyses were conducted using STATA 11.1 (STATA Corporation, College Station, TX) accounting for the survey design features of the NIS in order to provide population-based estimates.
From 1998 to 2002 there were an estimated 7,756,570 discharges for an elective hospitalization in which a non-cardiac principal procedure was performed in adults aged ≥18 years. The characteristics of the hospitalized patients with a low risk, intermediate risk, and high risk principal procedure by disease status are displayed in Tables 1, ,22 and and3,3, respectively. As displayed, RA patients were significantly older and more likely female as compared to those with DM and those without either DM or RA regardless of the procedure risk category. In general, comorbid diseases of coronary artery disease, hypertension, congestive heart failure, and chronic kidney disease were significantly more common in patients with diabetes than those with RA alone or neither RA or DM. One exception to this general pattern was that the proportion of those with DM and those with RA who had hypertension did not significantly differ among those with high risk procedures. Also, the difference in the proportion of valvular heart disease in patients with DM and those with RA was not statistically significant among the low risk and intermediate risk categories.
The frequency of perioperative all-cause mortality by procedure risk level and disease status is displayed in Table 4. Among patients with DM alone, those who underwent a low (unadjusted odds ratio [OR]: 1.54, 95% Confidence Interval [CI]:1.23,1.93) or an intermediate risk principal procedure (unadjusted OR: 1.34, 95% CI: 1.19,1.51) were more likely to suffer an in-hospital death as compared to those without RA or DM; this risk was attenuated after sequentially adjusting for age and gender, potential confounders and secondary procedure as displayed in Table 5. Persons with RA alone with a procedure of any risk level were not at a greater risk of all-cause mortality as compared to those without RA or DM.
The frequency of perioperative composite CVD events by procedure risk level and disease status is displayed in Table 4. In addition to group comparisons shown on Table 4, additional group comparisons between persons with RA and DM compared with non-RA and DM revealed that among persons without RA but with DM who underwent low risk procedures had greater composite CVD events while the difference among persons with intermediate or high risk procedures was not significant. The risk of composite CV events among those with a low, intermediate, or high risk principal procedure are displayed in Table 6. In an unadjusted logistic model, patients with RA alone who underwent a low risk principal procedure had at a nearly two-fold greater odds of a composite CV event as compared to those without RA or DM and persons with DM alone were at more than 2.5-fold greater odds of a composite CV event. The increased odds of perioperative CV events observed in those with RA alone was significantly attenuated after adjusting for age and gender while the increased composite CVD risk in persons with DM remained statistically significant in a fully adjusted model. A similar pattern of association was observed among persons with intermediate risk procedure. Among persons with a high risk principal procedure, neither those with RA alone nor DM alone had a significantly greater risk of a composite CV event relative to those without RA or DM.
The results of our supplementary analyses did not alter the conclusion based on our primary analysis. First, the results of procedure-specific analysis to determine the odds of adverse perioperative events among elective hospitalizations with a principal procedure of Total Knee Replacement were consistent with that of the procedure group-based analyses. Relative to persons without RA or DM, persons without diabetes but with RA(adjusted OR: 0.38, 95% CI: 0.05,2.92) or diabetics without RA(adjusted OR: 1.10, 95% CI: 0.60,2.02) were not at increased risk of all-cause mortality while persons with both RA and DM were at significantly increased risk(adjusted OR: 6.49, 95% CI: 1.45,29.04) of all-cause mortality. Relative to persons without DM or RA, persons with RA but without DM were not at greater risk of composite CVD(adjusted OR: 0.47, 95% CI: 0.14,1.53). In contrast, persons with DM were at increased risk of composite CVD events which was observed among diabetic without RA(adjusted OR: 1.51, 95% CI: 1.06,2.14) and with RA(adjusted OR: 3.50, 95% CI: 1.10,11.19).
In this large population-based study of hospitalized individuals, we sought to determine the association between the absence or presence of either or both diagnoses of RA and DM with perioperative CV events and all-cause mortality. In our analyses, we found that hospitalized RA patients who underwent a non-cardiac principal procedure were not at an increased risk of in-hospital short-term perioperative adverse CV events or mortality compared with those without RA or DM. These findings were consistent across all three levels of procedural risk and after controlling for comorbid diseases and additional secondary hospital-based procedures. To our knowledge, this is the first study to investigate the association between RA and perioperative mortality and CV risk.
Recent studies comparing the risk of CV disease in RA with DM have found an elevated risk of CV events in RA comparable in magnitude to DM.16,17 In the perioperative setting, we compared the prevalence of CV comorbid disease and the risk of perioperative CV events among hospitalized persons with either RA, DM, neither, or both conditions. In comparison to persons without RA or DM, persons with RA did indeed have a higher prevalence of CV comorbid diseases including coronary artery disease, hypertension, and congestive heart failure. Regarding the risk of adverse perioperative events, our results were consistent with previous literature in that persons with DM remained at an increased risk of adverse perioperative events.29,30 However, despite the higher prevalence of CV comorbid disease, our study did not find persons with only RA to have an elevated risk of CV events. Given the high prevalence of CV comorbid disease in persons with RA in our study population, we addressed the possibility of model over-adjustment by including a multivariable model adjusting for only age and sex. However, the results of this intermediate multivariable model did not reveal that over-adjustment accounted for the lack of association between RA and adverse perioperative events.
The results of our supplementary analyses were in large part consistent with our primary analyses. In order to account for the possibility of procedure case-mix within each procedure risk category, we conducted a separate procedure-specific analysis to determine the odds of adverse perioperative events among elective hospitalizations with a principal procedure of total knee replacement. While the result of this supplementary analysis was consistent among persons with RA alone, it did suggest that the subgroup of persons with both RA and DM who may be at heightened risk for adverse perioperative events. Due to the multi-factorial and possible non-atherosclerotic etiology of decompensated heart failure we conducted additional analyses excluding acute heart failure as a component of the composite CVD endpoint results of which did alter study conclusion.
Given the large body of evidence suggesting an increased burden of CV disease in RA, we consider some limitations and potential explanations for our somewhat surprising findings. First, our study in limited due to factors related to its observational and cross-sectional design. Secondly, some possible explanations for our results include a differential pre-hospitalization preoperative risk assessment and modification. Specifically, patients with active or aggressive RA would probably be less likely to undergo an elective non-cardiac surgical procedure. Also, RA patients may not have been referred for an elective surgical procedures because the recognition of the overall increase in risk of CV events leading to a selection process that excluded RA patients from elective surgical procedures. Given that much of the epidemiologic research highlighting the association of RA with CV disease has been carried out in the past decade, and post-dates the time of this dataset (1998–2002) this possibility seems unlikely. Likewise, patients with DM who at greater perioperative risk may have been identified prior to an elective surgery because of pre-operative scrutiny, therefore underestimating the risk in these patients relative to an unselected group of patients with DM. Third, a number of RA-specific factors which were not available in the NIS dataset, such as the disease duration, magnitude of systemic inflammation, use of disease modifying drugs, as well as functional status may have had an impact on the CV disease profile.4, 31–33 Fourth, one must consider the possibility of coding errors or misclassification when using ICD-9 codes for abstraction of diseases and procedures.34–36 For instance, the misclassification of a non-inflammatory arthritis as RA may have led to an attenuation of the association between disease and perioperative CV events. Fifth, while a mix of procedures within each procedure risk group, especially if these differed by diagnosis, may have introduced residual confounding, the results of our supplementary analysis make this possibility less likely. Lastly, there was a lack of an anticipated graded increase in the rate of adverse perioperative events across low, intermediate, and high risk procedures which may have been due to above mentioned possibilities of differential pre-operative screening and limitations inherent to use of ICD-9 codes and other limitations such as the lack of additional information on procedure-related factors, including indication and duration.
The strengths of our study include the large population-based sample of patients, the comparison with both unaffected controls and patients with DM, and examination of risks within surgical risk strata. It serves as the initial step in attempting to quantify the significance of the increased atherosclerotic burden of patients with RA in the perioperative period serving well those who encounter this task routinely in clinical practice. Several risk stratification indices have been developed to aid clinicians in preoperative risk stratification with their components including various recognized CV disease states, such as coronary heart disease, congestive heart failure and valvular heart disease.28 In addition, these indices include characteristics such as age, DM, chronic kidney disease, and hypertension. Accumulating evidence for the CV disease burden in RA prompted our investigation of the risk associated with RA in the perioperative setting.18 Our findings did not reveal a significant association RA and mortality or perioperative CV events. The results of our cross-sectional observational study we present here do not support the need for a change in the currently practiced preoperative screening for CV disease in persons with RA being evaluated before elective non-cardiac procedures. However, our result reflect short-term in-hospital adverse outcomes and do not exclude the possibility of an increased risk of adverse post-operative events after longer follow-up time after surgery. therefore, prospective studies are warranted to further elucidate the relationship between RA and perioperative CV risk.
Funding Support: The study was supported in part by the Intramural Research Program, National Institute of Arthritis, Musculoskeletal, and Skin diseases, National Institutes of Health.
Disclosure: No conflict of interest