The current study enriches and extends previous cross-sectional studies of DD in RA patients, by examination of a cohort with comprehensive multi-modal CV phenotyping and with longitudinal followup in a subset. The major findings of our study are that DD was prevalent in RA patients without clinical HF, progresses on followup, and is associated with RA disease activity and with several cardiac biomarkers, but not with myocardial inflammation, myocardial microvascular dysfunction, or with subclinical atherosclerosis.
Our finding of a baseline frequency of DD of 40.7%, defined by at least one criterion for DD (and 21% by ≥ 2 criterion), is consistent with previous reports citing prevalences between 26–49% (16, 23, 24). In our study, this frequency increased to 57.9% over an interval of 3–5 years. Direct comparisons of frequencies may be misleading due to utilization of different criteria. We utilized the updated, arguably more stringent, 2016 ASE/EACVI criteria for DD which exclude DT, E/A, and isovolumic relaxation time (IVRT) that were utilized in many of the prior studies and were considered to have poor reproducibility, including our previous publication16.
In our longitudinal analyses, septal/lateral e’ wave velocities (average to form e’) declined significantly, while the prevalence of abnormal septal e’ (< 7) and lateral e’ (< 10) increased significantly, all together indicating worsening diastolic function. We also observed non-statistically significant increases in E/e’ and LAVI, also consistent with worsening DD. To our knowledge, there are only two prior longitudinal studies of DD in RA (14, 15). Our observed increase in E/e’ and trend towards increase in LAVI are consistent with that of Davis et al (14). Other contributory trends towards worsening DD included a significant increase in TR velocity, which was not analyzed in the other two studies.
Risk factors for the development of DD in the general population include age and hypertension, as well as mildly inflammatory states such as obesity and insulin resistance (7, 8, 25, 26). Concordantly in our study, we observed a consistently positive association between age and multiple individual baseline diastolic parameters (E/e’, and LAVI,) as well as with baseline composite DD. Higher SBP was associated with higher baseline E/e’ and annualized rate of change in E/e’, but not with composite DD, in our patient cohort. These results reinforce the importance of tight management of conventional CV risk factors for the prevention of CVD in RA patients. Also, elevated calcium scores (CAC > 100), though prevalent (17% at baseline), were not significantly associated with composite baseline or followup composite DD, underscoring the point that DD can occur independently of atherosclerosis.
RA associated inflammation is more severe than that of other chronic inflammatory conditions that are associated with DD and HFpEF, such as obesity and diabetes, and is hypothesized to be the primary driver of the higher prevalence of DD observed in RA vs the general population (27). To examine this hypothesis, we assessed the correlation of multiple measures of inflammation in our RA cohort with DD. Whereas prior studies of DD in RA utilized only serum markers of inflammation and/or patient self-reported measures of joint pain (13, 14), ours is the first study to include physical assessments of swollen and tender joint counts in our assessment of disease activity, and to use cardiac PET-CT to assess local (myocardial) inflammation and perfusion, in addition to serum markers of inflammation, to assess the relationship of RA disease activity with DD. While individual markers like ESR, CRP, and IL-6 did not associate with baseline or followup composite DD (only CRP was associated with annualized rate of change in E/e’; Supplementary Table 5), measures of RA disease activity that do (DAS28CRP) or do not (CDAI) incorporate CRP were both positively associated with baseline composite DD, as well as with multiple individual baseline parameters of DD (E/e’ and LAVI), even after adjusting for smoking, blood pressure, cholesterol levels and CAC levels, suggesting that inflammation contributes to DD independently of conventional CV risk factors. However, neither averaged CDAI nor DAS28CRP was a predictor of followup DD, possibly owing to the relatively small number of patients who returned for followup, and the possibility that this effect may take longer than 5 years to observe. The previous literature demonstrated correlations of diastolic function with RA duration (10–12) which we did not observe. Another, albeit indirect, way to probe the impact of disease activity on DD is to explore whether specific RA DMARD treatment(s), through subduing of disease activity, are associated with a reduced risk of DD. Interestingly, we did note a strong positive association between leflunomide use and baseline composite DD (Table 4), but given the limited number of patients who were using leflunomide at baseline (13/156) and statistical uncertainty (wide confidence interval), meaningful interpretation is limited and this observation should be confirmed in additional studies.
Our study is one of few in the literature that analyzed sensitive biomarkers of myocardial injury, including BNP and troponin-I, with respect to diastolic function. BNP and troponin-I were associated with individual parameters of diastolic function (LAVI and E/e’, respectively), but not baseline composite DD. BNP is released in response to atrial contraction and increased LA pressure, thus its positive association with LAVI (Supplementary Table 6), a surrogate measure of LA pressure, appears to be physiologically consistent. Troponin has long been heralded as a marker of myocyte injury/necrosis and prognosticator of CV associated death in non-RA patients (34). In our patients, higher troponin-I levels were associated with higher E/e’ (and thus higher LV filling pressures), indicating a possible link between subclinical myocardial damage and diastolic filling. These results suggest that cardiac biomarkers may be used in conjunction with echocardiography to identify patients at risk for DD and HFpEF, however this will require formal evaluation in future studies.
Subclinical myocarditis is prevalent in RA, as evidenced by myocardial inflammation in up to 20–40% of RA patients without clinical CVD as assessed by cardiac FDG PET/CT (17) or by late gadolinium uptake on cardiac magnetic resonance (CMR) imaging in other RA cohorts (28–31). It is postulated that persistent myocardial inflammation leads to LV wall fibrosis and stiffness (with concurrent myocardial endothelial damage, microvascular dysfunction and reduced perfusion), and then altered LV function (32). Surprisingly, we found no association in multivariable analyses of myocardial mean or max SUV with baseline or followup composite DD. Since most studies of myocardial inflammation report imaging findings at only one time point, it is unclear whether myocardial inflammation is transient or persistent and recurrent. If transient, it may not constitute a significant contributor to the development of DD.. We previously reported that 29% of the RHYTHM RA cohort had reduced MFR even after adjusting for CAC scores, and that lower MFR was associated with higher LAVI but not E/e’ at baseline (16). In the present studies, we did not identify MFR as a predictor of worsening DD over time.
The main strengths of this study are the large baseline sample size (> 150) of RA patients without clinical CVD, and the extensive cardiac phenotyping that included a comprehensive set of echocardiography measures, novel measures of myocardial inflammation and /perfusion, cardiac biomarkers, a measure of atherosclerosis, and followup echocardiography in a significant number of the original cohort. Limitations include the absence of matched non-RA controls that preclude direct comparisons, but previous data confirm a higher rate of DD in RA vs non-RA without clinical HF. While we incorporated change in RA disease activity over time by averaging disease activity scores in regression models, as well as any change in RA DMARD treatment on followup, treatment was not assigned or randomized, which limits our interpretation of these findings as associations but not causality. Although the number of patients who returned for followup was limited (about a 1/3 of the baseline cohort), we did note significant declines in several key diastolic parameters.