Interpreting studies of cardiovascular mortality in rheumatoid arthritis: The importance of timing
Version of Record online: 26 NOV 2008
Copyright © 2008 by the American College of Rheumatology
Arthritis Care & Research
Volume 59, Issue 12, pages 1687–1689, 15 December 2008
How to Cite
Ward, M. M. (2008), Interpreting studies of cardiovascular mortality in rheumatoid arthritis: The importance of timing. Arthritis & Rheumatism, 59: 1687–1689. doi: 10.1002/art.24170
- Issue online: 26 NOV 2008
- Version of Record online: 26 NOV 2008
- Manuscript Accepted: 16 SEP 2008
- Manuscript Received: 11 SEP 2008
- Intramural Research Program
- National Institute of Arthritis and Musculoskeletal and Skin Diseases
Cardiovascular disease (CVD) is the major cause of death of patients with rheumatoid arthritis (RA) (1). In this issue of Arthritis Care & Research, Aviña-Zubieta and colleagues examine whether the risk of CVD mortality is increased in patients with RA, compared with that in age- and sex-matched groups in the general population (2). In a meta-analysis of published English-language studies, the authors found that the risk of CVD mortality was, on average, 50% higher among patients with RA. When component causes of death were examined, risks of death from ischemic heart disease and cerebrovascular disease were elevated to similar degrees. The authors provided not only a pooled summary estimate of risk, but in a thorough analysis examined how the estimates varied among the primary studies. Risks were lower, and rates of CVD mortality were closer to those of the general population, in studies rated to be of higher methodologic quality, as indicated by the study of community-based and inception cohorts (in contrast to clinic-based or prevalence cohorts), use of validated case definitions and confirmation of the cause of death in medical records, control for CVD risk factors, and having few patients lost to followup. In these studies, the risk of CVD mortality was estimated to be 21% higher among patients with RA.
As a meta-analysis, the study by Aviña-Zubieta et al succeeds in its goal of quantitatively summarizing the results of studies of CVD mortality in RA. The literature search was carefully performed and the statistical analysis was state-of-the-science. The authors examined their results for publication bias and for potential impact of outliers. However, the authors were sensibly cautious in their consideration that the 50% higher risk of CVD mortality among patients with RA was the true estimate, given the nature of the primary studies included in the meta-analysis. By contrasting results across primary studies, their article provides insight into factors that affect estimates of the relative risk of CVD mortality in RA that single studies cannot provide. In particular, 3 issues of timing deserve consideration, because they affect the interpretation of the results. These 3 issues are the duration of RA among patients at the time of their entry into the primary study, the timing of onset of RA relative to CVD death, and the time of study entry relative to historical time.
The duration of RA at the time patients enter a study of outcomes can have an important influence on the results. Studies of inception cohorts assemble and follow patients from a uniform and early point in the course of their RA, e.g., prospectively enrolling all patients newly diagnosed in a particular clinic for 1 year. By doing so, inception cohorts are likely to include patients with mild RA as well as those with more severe RA. In contrast, studies of prevalence cohorts assemble a cross-section of patients, regardless of RA duration, and follow them over time. Because less severely ill patients, and those who are in remission, tend to migrate out of clinics or are seen less frequently, prevalence cohorts typically have a higher number of patients with more severe RA. If the outcome occurs more often in patients with severe disease, as is the case for mortality in RA (3–8), risks will be falsely inflated in studies of prevalence cohorts. Because the spectrum of disease severity tends to be less skewed in inception cohorts, studies using these cohorts are considered to provide less biased estimates of mortality risk. In this meta-analysis, the risk of CVD mortality in patients with RA compared with the general population was substantially lower in studies of inception cohorts than in studies of prevalence cohorts (standardized mortality ratios of 1.19 and 1.56, respectively), and the risk of CVD mortality for inception cohorts was not significantly different from the null. This difference may be due in part to the inclusion of patients with milder RA in these studies.
The study of inception cohorts versus prevalence cohorts also relates to the second issue of how timing may influence the interpretation of these results, that of the relationship between the onset of RA and the time of CVD death. Studies of inception cohorts are able to capture deaths that occur soon after the onset of disease, and guard against underestimating mortality risks in conditions that are rapidly fatal. In these circumstances, prevalence cohorts represent survivors, and may convey falsely low mortality rates. Conversely, if any increase in the risk of mortality associated with a disease is typically delayed years or decades after its onset, studies of inception cohorts may underestimate risks unless the followup period is sufficiently long. Unfortunately, the hazard rate of death among patients with early, mid-duration, and late RA, relative to the general population, has not been reported, so the degree to which short-term studies of inception cohorts may underestimate the risk of mortality in RA is not precisely known. Studies have suggested that overall mortality risk begins to increase only after approximately 7 years of RA (9, 10). However, inception cohort studies with followup periods of 10 to 14 years have reported risks of overall mortality that were 28–84% higher than expected, and were therefore of sufficient length to be able to detect increased risks of this magnitude (10–13). Inception cohort studies of CVD mortality in this meta-analysis had median durations of followup that ranged from 6.9 to 25 years, suggesting that they might have been long enough to capture any increased risk. It is unknown whether the pooled standardized mortality ratio estimate of 1.19 would increase with the inclusion of studies with longer followup. CVD mortality may occur earlier in the course of RA than other causes of death, so inception cohorts may be less susceptible to this timing issue in studies of CVD mortality than in studies of overall mortality (14).
The third issue of timing is the relationship of these studies to historical time. Aviña-Zubieta et al compared results among studies that enrolled patients before or after 1987, using this date to distinguish studies that used different classification criteria for RA. No difference in standardized mortality ratios was found between early and more recent studies. Perhaps more important than distinguishing studies by classification criteria, the similarity in risks between early and more recent studies may seem to suggest that changes in RA treatment strategies over time have not resulted in improvements in CVD mortality. Plotting the standardized mortality ratios by calendar year of enrollment also indicated no temporal trend (Figure 1). Differences among studies in patient characteristics and designs may have masked changes over time (15, 16). It may also be that CVD mortality is influenced by too many factors other than RA activity or severity to expect to see a decrease in rates with more aggressive antirheumatic therapy, although studies suggest, some more convincingly than others, that treatment with methotrexate, other conventional disease-modifying medications, and biologic agents is associated with fewer CVD events and lower CVD mortality in patients with RA (17–21). The more recent cohorts included in this meta-analysis may not have had long enough exposure to more aggressive antirheumatic treatment to experience a difference in CVD mortality, or too few patients in these cohorts may have received aggressive treatment (14). Because of these factors, the lack of temporal trend in CVD mortality should not be interpreted as a failure of treatment effectiveness. To examine this question, studies should report the prevalence and types of antirheumatic treatment, ideally as person-years or percentage of time on treatment, whenever possible.
The risk of CVD mortality is most certainly higher in patients with RA than in the general population, and attention to CVD risk factor modification and educating patients about this risk is critically important (1). However, these issues of timing suggest that the true risk may be lower than the 50% increase estimated in this analysis.
- 21British Society for Rheumatology Biologics Register Control Centre Consortium, , , on behalf of the British Society for Rheumatology Biologics Register. Reduction in the incidence of myocardial infarction in patients with rheumatoid arthritis who respond to anti–tumor necrosis factor α therapy: results from the British Society for Rheumatology Biologics Register. Arthritis Rheum 2007; 56: 2905–12., , , , and the