Interaction between oxidative stress and high-density lipoprotein cholesterol is associated with severity of coronary artery calcification in rheumatoid arthritis

Authors


Abstract

Objective

To test the hypothesis that oxidative stress is increased in patients with rheumatoid arthritis (RA) due to increased inflammation and contributes to the pathogenesis of atherosclerosis.

Methods

The independent association between urinary F2-isoprostane excretion, a measure of oxidative stress, and RA was tested using multiple linear regression models in 169 patients with RA and 92 control subjects, frequency matched for age, race, and sex. The relationship between F2-isoprostane excretion and coronary calcium, a marker of atherosclerosis, was examined in multivariable proportional odds logistic regression models that also assessed the interactions between oxidative stress and low-density lipoprotein and high-density lipoprotein (HDL) cholesterol.

Results

F2-isoprostane excretion was significantly higher in patients with RA (median 2.75 [interquartile range (IQR) 1.60–4.06] ng/mg creatinine) than in control subjects (median 1.86 [IQR 1.25–2.62] ng/mg creatinine; adjusted P = 0.006). In patients with RA, F2-isoprostanes were positively correlated with body mass index (P < 0.001), but not with disease activity or mediators of inflammation such as the Disease Activity Score in 28 joints or serum tumor necrosis factor α, interleukin-6, and C-reactive protein concentrations in adjusted multivariable models (P > 0.05 for all). In patients with RA, F2-isoprostanes significantly modified the effect of HDL cholesterol on coronary calcification (P = 0.02 for interaction) after adjustment for age, sex, and race. As F2-isoprostane levels increased, HDL lost its protective effect against coronary calcification.

Conclusion

Oxidative stress measured as F2-isoprostane excretion was higher in patients with RA than in control subjects. Among patients with RA, higher F2-isoprostane excretion and HDL cholesterol concentrations interacted significantly and were positively associated with the severity of coronary calcification.

INTRODUCTION

Oxidative stress occurs when there is an imbalance between reactive oxygen species relative to antioxidants (1). Consequent free radical–mediated tissue injury is thought to play an important role in the pathogenesis of many inflammatory and degenerative diseases, including atherosclerosis and rheumatoid arthritis (RA) (2, 3).

Increased oxidative stress is hypothesized to be important in the pathogenesis of RA, and to both initiate and propagate inflammation (3). Furthermore, inflammation and relative hypoxia in the joints promote additional oxidative stress (3, 4). Although oxidative stress is considered important in RA, there are few studies that have addressed this hypothesis directly. These studies were performed in a small number of patients and used a variety of measures of oxidative stress (5–12). One of the problems in this area of research has been that in vivo measures of oxidative stress have lacked sensitivity and specificity (13). The discovery of F2-isoprostanes, prostaglandin-like compounds generated in vivo by nonenzymatic free radical–mediated oxidation of arachidonic acid (14), provided a reliably stable measure of oxidative stress in vivo (1). To our knowledge, no large study has examined the relationship between F2-isoprostanes and the clinical characteristics of RA.

In addition to its role in chronic autoimmune inflammatory diseases, oxidative stress may also contribute to the pathogenesis of atherosclerotic cardiovascular disease. In the general population, increased concentrations of F2-isoprostanes are associated with coronary artery calcification and carotid intima-media thickness, noninvasive measures of atherosclerosis that predict long-term cardiovascular outcomes (15, 16), and also with the presence and severity of coronary artery disease (17).

Patients with RA have increased coronary atherosclerosis (18). Increased oxidative stress is proposed as one of the mechanisms underlying accelerated atherosclerosis in RA (3), but this hypothesis has not been addressed directly. There are several mechanisms by which oxidative stress could accelerate atherosclerosis (2); oxidative modification of low-density lipoprotein (LDL) and high-density lipoprotein (HDL) cholesterol is of particular interest in RA and systemic lupus erythematosus (SLE) because concentrations of oxidized LDL are elevated and HDL appears to be modified so that it is proinflammatory and proatherogenic rather than antiinflammatory (19, 20), and is associated with atherosclerosis (21).

Because there is currently limited information about the role of oxidative stress in the pathogenesis of inflammation, and none about its role in atherosclerosis in RA, we examined the hypothesis that oxidative stress, measured by F2-isoprostane excretion, is higher in patients with RA than in control subjects, and is associated with inflammation, traditional cardiovascular risk factors, and coronary atherosclerosis. We also hypothesized that oxidative stress could interact with LDL and HDL cholesterol and modify their effects on coronary atherosclerosis.

SUBJECTS AND METHODS

Patients and control subjects.

We enrolled 169 patients with RA and 92 control subjects, frequency matched for age, race, and sex, through advertisements, referral from local rheumatologists, and from a volunteer database maintained by the General Clinical Research Center at Vanderbilt University, recruited between 2001 and 2005. The patients were ages >18 years and fulfilled the American College of Rheumatology (formerly the American Rheumatism Association) classification criteria for RA (22); the control subjects did not have RA or any inflammatory disease. These subjects have participated in ongoing studies of cardiovascular risk factors in RA and detailed study methods have been described (18). The study was approved by the Vanderbilt University Institutional Review Board and all of the subjects gave written informed consent.

Clinical and laboratory measurements.

Clinical information, laboratory data, and coronary calcium scores were obtained as described (18, 23). Briefly, urinary F2-isoprostane excretion was quantified using gas chromatography mass spectrometry and expressed as ng/mg creatinine (24). Coronary calcium was measured by electron beam computed tomography scanning with an Imatron C-150 scanner (GE/Imatron) and was quantified as described by Agatston et al (25). RA disease activity was measured using the Disease Activity Score in 28 joints (DAS28) (26). Radiographic joint damage was measured using the Larsen score in 94 patients as described previously (27). The Framingham risk score, a composite score of traditional cardiovascular risk factors that includes blood pressure, smoking status, serum lipid concentrations, age, and sex, but not diabetes mellitus, was calculated (28). Obesity was quantified using body mass index (BMI) in kg/m2 and insulin resistance was measured using the homeostatic model assessment (HOMA) index (29), calculated as: [serum insulin (μU/ml) × glucose (mmoles/liter)]/22.5. Peripheral blood neutrophil and monocyte counts and serum C-reactive protein (CRP), glucose, triglycerides, and HDL and LDL cholesterol concentrations were measured by the Vanderbilt University Medical Center Clinical Laboratory. LDL concentrations were calculated using the Friedewald equation (30). Before 2003, the laboratory did not use a high-sensitivity CRP assay, and low concentrations were reported as <3 mg/liter; in 40 patients with RA who had CRP concentrations <3 mg/liter, concentrations were measured by enzyme-linked immunosorbent assay (ELISA; Millipore). For technical reasons, HDL and F2-isoprostane measurements were not obtained in 1 patient each in the RA and control groups. Serum concentrations of tumor necrosis factor α (TNFα), interleukin-6 (IL-6), and serum amyloid A (SAA) were measured by multiplex ELISA (Millipore).

Statistical analysis.

Descriptive statistics were calculated as the median with the interquartile range (IQR). The distribution of F2-isoprostane excretion was compared in patients with RA and control subjects using the Wilcoxon rank sum test. The independent association between disease status (RA versus controls) and F2-isoprostanes was assessed using a multiple linear regression model with urinary F2-isoprostane excretion as a dependent variable and disease status as a predictor variable, adjusted for traditional cardiovascular risk factors such as age, sex, race, BMI, hypertension, diabetes mellitus, and current smoking status. Age was assessed for nonlinear effects because it had the strongest predictive potential for atherosclerosis (31). F2-isoprostane excretion values were log-transformed to normalize the distribution of regression residuals. The antilog of the regression coefficient for the disease status variable was taken to reflect the percent change in F2-isoprostanes with 95% confidence intervals (95% CIs).

Among the patients with RA, we evaluated the association between clinical and disease-associated factors and F2-isoprostane excretion. These factors included 1) traditional cardiovascular risk factors (age, sex, BMI, smoking, hypertension, diabetes mellitus, serum HDL and LDL cholesterol, triglycerides, glucose, HOMA index, and the Framingham risk score) and 2) disease-associated indices and inflammatory mediators (drug use, DAS28 score, Larsen score, disease duration, peripheral blood neutrophil and monocyte counts, and serum TNFα, IL-6, CRP, and SAA concentrations). The Wilcoxon rank sum test or Spearman's rank correlation coefficient (ρ) was used to assess unadjusted associations. The independent relationship was assessed by multiple linear regression using F2-isoprostanes as the outcome variable, adjusted for age, race, sex, hypertension, smoking, and BMI.

The independent association between F2-isoprostanes, HDL and LDL cholesterol concentrations, and coronary calcium score was examined using a proportional odds logistic regression model, which is also known as the ordinal logistic regression method and is applicable to an ordered response variable (32, 33). Ordinal logistic regression is also applicable to skewed continuous dependent variables such as the coronary calcium score by using the ranks of the variable. F2-isoprostane excretion and serum HDL and LDL cholesterol concentrations were each used as the predictor variable and covariates for adjustment included age, race, sex, hypertension, BMI, diabetes mellitus, current smoking, and statin use.

To assess the interaction between F2-isoprostanes and HDL and LDL cholesterol on the outcome of coronary calcification, we conducted separate proportional odds models with an interaction term in the model using a cross-product between either HDL or LDL cholesterol and F2-isoprostanes (HDL × F2-isoprostanes or LDL × F2-isoprostanes), adjusted for age, race, and sex, and then further adjusted for BMI, current smoking, hypertension, diabetes mellitus, and statin use. Odds ratios (ORs) were expressed per IQR difference with 95% CIs. All of the statistical analyses used R, version 2.7.1 (online at: http://www.r-project.org).

RESULTS

The demographic and clinical characteristics of the patients with RA and the control subjects are summarized in Table 1. The 2 groups were similar with regard to age, race, sex, and BMI. Patients with RA were more likely to smoke, be insulin resistant, and have hypertension. As we have reported previously (18), the coronary calcium score was higher in patients with RA. The distribution of F2-isoprostane excretion in the 2 groups is depicted in Figure 1. Median F2-isoprostane excretion rates were significantly higher in patients with RA than in control subjects (2.75 [IQR 1.60–4.06] ng/mg creatinine versus 1.86 [IQR 1.25–2.62] ng/mg creatinine; P < 0.001), although there was considerable overlap. When adjusted for age, sex, race, hypertension, BMI, diabetes mellitus, and current smoking status, RA remained significantly associated with higher F2-isoprostane excretion rates (β = 0.22 [95% CI 0.06–0.38], P = 0.006). By taking the antilog, the regression coefficient can be interpreted as indicating that having RA leads to a 25% (95% CI 6–46%) adjusted increase in mean F2-isoprostane excretion compared with control subjects.

Table 1. Clinical factors and urinary F2-isoprostane excretion in patients with RA and control subjects*
 Controls (n = 92)RA patients (n = 169)P
  • *

    Values are the median (interquartile range) unless otherwise indicated. RA = rheumatoid arthritis; BMI = body mass index; anti-TNF = anti–tumor necrosis factor; HDL = high-density lipoprotein; LDL = low-density lipoprotein; HOMA = homeostatic model assessment; DAS28 = Disease Activity Score in 28 joints.

Age, years53.0 (44.8–59.2)54.0 (45.0–63.0)0.41
Men, %37.030.80.31
White race, %84.888.20.44
BMI, kg/m227.0 (24.6–31.8)28.3 (24.0–33.2)0.44
Current smokers, %8.724.30.002
Hypertension, %39.153.30.03
Diabetes mellitus, %4.311.20.06
Statin use, %13.012.40.89
Disease duration, years3 (2–18)
Current corticosteroid use, %54.4
Current methotrexate use, %71.0
Current antimalarial use, %24.9
Current anti-TNF agent use, %20.7
HDL cholesterol, mg/dl45.0 (38.5–54.0)43.0 (37.0–54.0)0.65
LDL cholesterol, mg/dl122.0 (104.0–145.0)112.5 (88.8–135.2)0.02
Triglycerides, mg/dl103.0 (73.0–135.5)110.5 (79.8–158.0)0.22
Glucose, mg/dl89.0 (83.0–94.2)87.0 (83.0–94.0)0.70
HOMA index0.83 (0.54–1.79)2.36 (1.19–4.47)< 0.001
Framingham score12.0 (7.0–14.0)13.0 (9.0–16.0)0.10
Coronary calcification score0.0 (0.0–18.7)1.9 (0.0–150.3)0.02
DAS28 score3.88 (2.64–4.84)
F2-isoprostane excretion, ng/mg creatinine1.86 (1.25–2.62)2.75 (1.60–4.06)< 0.001
Figure 1.

Distribution of urinary F2-isoprostane excretion in patients with rheumatoid arthritis (RA) and control subjects. P < 0.001 by the Wilcoxon rank sum test. P = 0.006 when adjusted for age, sex, race, hypertension, body mass index, diabetes mellitus, and current smoking status. The box plot shows individual data points with the median (solid horizontal line) and interquartile range (box). Cr = creatinine.

Among patients with RA, F2-isoprostane excretion was higher in women than men and in current smokers compared with nonsmokers (Table 2). There was no significant difference in F2-isoprostane excretion between patients with and without diabetes mellitus, patients with or without hypertension, or drug use.

Table 2. F2-isoprostane concentrations in patients with RA according to categorical demographic and clinical characteristics*
 NF2-isoprostane, median (IQR) ng/mg creatinineP
  • *

    RA = rheumatoid arthritis; IQR = interquartile range; anti-TNF = anti–tumor necrosis factor.

  • By the Wilcoxon rank sum test.

Sex  0.006
 Male522.06 (1.56–3.11) 
 Female1163.23 (1.85–4.54) 
Current smoking  < 0.001
 Yes403.58 (2.85–5.01) 
 No1282.44 (1.52–3.75) 
Hypertension  0.22
 Yes902.58 (1.42–4.07) 
 No783.02 (1.78–4.01) 
Diabetes mellitus  0.72
 Yes192.28 (1.68–4.00) 
 No1492.81 (1.61–4.06) 
Current corticosteroids  0.75
 Yes912.76 (1.71–4.03) 
 No772.72 (1.50–4.31) 
Current methotrexate  0.96
 Yes1192.74 (1.60–4.00) 
 No492.81 (1.63–4.21) 
Current antimalarials  0.87
 Yes1262.79 (1.54–4.07) 
 No422.75 (1.67–4.05) 
Current anti-TNF  0.25
 Yes343.19 (2.12–3.98) 
 No1342.72 (1.55–4.06) 
Current statin  0.14
 Yes212.24 (1.56–3.38) 
 No1472.88 (1.62–4.20) 

The correlation between F2-isoprostanes and clinical variables in patients with RA (including after adjustment for age, race, sex, BMI, hypertension, and smoking) are shown in Table 3. F2-isoprostane excretion was negatively correlated with age (ρ = −0.33, adjusted P = 0.02) and positively correlated with BMI (ρ = 0.35, adjusted P < 0.001). HDL cholesterol (ρ = 0.04, adjusted P = 0.03) and peripheral blood monocyte counts (ρ = 0.15, adjusted P = 0.09) were marginally correlated with F2-isoprostane excretion. Among markers of inflammation and disease activity, the Larsen score was not associated with oxidative stress in the univariate correlation (ρ = 0.02, P = 0.87), but after adjustment for potential confounders there was a weak correlation (adjusted P = 0.04) (Table 3). Other markers such as TNFα (adjusted P = 0.78), IL-6 (adjusted P = 0.41), CRP level (adjusted P = 0.66), and DAS28 score (adjusted P = 0.52) were not significantly associated with F2-isoprostane excretion (Table 3).

Table 3. Relationship between F2-isoprostane excretion and clinical factors in patients with rheumatoid arthritis*
 ρUnadjusted PAdjusted P
  • *

    BMI = body mass index; HDL = high-density lipoprotein; LDL = low-density lipoprotein; HOMA = homeostatic model assessment; DAS28 = Disease Activity Score in 28 joints; TNFα = tumor necrosis factor α; IL-6 = interleukin-6; CRP = C-reactive protein; SAA = serum amyloid A.

  • Univariate Spearman's rank correlation coefficient.

  • Adjusted P values are reported from multiple linear regression models, adjusting for age, race, sex, hypertension, BMI, and smoking.

  • §

    N = 94.

  • Log-transformed.

Cardiovascular risk factors   
 Age−0.33< 0.0010.02
 BMI0.35< 0.001< 0.001
 HDL cholesterol0.040.610.03
 LDL cholesterol0.040.560.34
 Triglycerides0.010.940.17
 HOMA index0.090.260.55
 Framingham score−0.130.100.74
 Disease duration−0.030.680.08
Disease-related indices and inflammatory mediators   
 DAS28 score0.020.750.52
 Larsen score§0.020.870.04
 Neutrophil count0.160.040.51
 Monocyte count0.150.060.09
 TNFα−0.070.400.78
 IL-60.020.820.41
 CRP level0.060.410.66
 SAA−0.060.470.43

In patients with RA, after adjusting for age, race, and sex, there was no significant association between coronary calcium and F2-isoprostane excretion (OR 1.49 [95% CI 0.92–2.43], P = 0.11) and HDL (OR 0.83 [95% CI 0.55–1.25], P = 0.38) or LDL cholesterol (OR 1.12 [95% CI 0.72–1.74], P = 0.61). A model assessing coronary calcium and the interaction between serum HDL cholesterol concentration and F2-isoprostane excretion was statistically significant (P = 0.02, adjusted for age, race, and sex). When further adjusted for additional factors, including BMI, current smoking, hypertension, statin use, and diabetes mellitus, the significance was marginally attenuated (P = 0.07 for interaction) (Figure 2). When F2-isoprostane excretion was low, higher HDL concentrations were associated with a lower risk of coronary calcification. However, as F2-isoprostane excretion increased HDL was no longer protective against coronary calcification. F2-isoprostanes did not interact significantly with serum LDL cholesterol concentrations in relation to coronary calcium scores after adjusting for age, race, and sex (P = 0.13).

Figure 2.

Interaction plot of F2-isoprostanes and serum high-density lipoprotein (HDL) concentrations and the risk of coronary calcium. High, medium, and low oxidative stress are referred to as the first, second, and third tertiles of log–F2-isoprostane concentrations, respectively. The figure shows the predicted risk of increasing calcification from the proportional odds model, including interaction between log–F2-isoprostanes and HDL, adjusted for age, sex, race, hypertension, smoking, body mass index, diabetes mellitus, and statin use. Predicted probabilities of HDL on coronary calcium by each level of oxidative stress (high, medium, low) were obtained by using the median values of each tertile of log–F2-isoprostanes (P = 0.07 for interaction). The interaction is significant when adjusted for age, race, and sex (P = 0.02).

DISCUSSION

The major finding of this study is that oxidative stress is increased in patients with RA, independent of risk factors associated with increased oxidative stress. Furthermore, oxidative stress may have an indirect impact on coronary atherosclerosis in patients with RA through an interaction with HDL cholesterol and modification of the protective effects of HDL on atherosclerosis.

RA is accompanied by activation of neutrophils and macrophages. Activation of such cells leads to the induction of oxidative bursts, which culminate in increased levels of reactive oxygen species, and tilts the balance of the redox system in a prooxidant direction (3). Oxidative stress in RA can damage cartilage (34) and modify IgG immunoglobulins to form IgG advanced glycation end products (IgG-AGEs), which can induce arthritogenic anti–IgG-AGE autoantibodies (35). Our finding of increased oxidative stress in RA is concordant with other studies (5–12). The strengths of our study included studying a large number of well-characterized patients with RA and using a state-of-the-art measure of oxidative stress. Therefore, we were able to define the relationships between oxidative stress and mediators of inflammation. Higher RA disease activity and more inflammation would be expected to be associated with increased oxidative stress; however, we found that disease activity indices such as the DAS28 score, concentrations of inflammatory cytokines such as TNFα and IL-6, or acute-phase reactants such as CRP level and SAA were not associated with increased oxidative stress. In some studies, TNFα and IL-6 have been associated with oxidative stress in patients with RA (36, 37). Although disease activity in our patients was relatively low and the majority was receiving disease-modifying antirheumatic drug therapy, inflammatory biomarkers such as IL-6 and TNFα were elevated compared with the control subjects (38). Since treatment with a TNFα inhibitor is reported to have decreased oxidative stress (39, 40), it is possible that more severe inflammation may be associated with oxidative stress in RA and that it may improve with treatment. However, in our study, F2-isoprostane excretion did not differ between patients receiving anti-TNF agents or corticosteroids and those who were not. Another possibility to account for the lack of association between oxidative stress and systemic measures of inflammation is that most of the oxidative stress occurs locally in the joints, where neutrophils and macrophages play a more important role than lymphocytes. Our results suggest that oxidative stress may be associated with radiographic bone damage; however, the results were from a subgroup of our patients and the significance was marginal, and thus the finding requires cautious interpretation.

Increased oxidative stress has been noted to be associated with increased BMI (41), smoking (42), age (43), and blood pressure (17). We found that F2-isoprostane excretion was higher in patients who currently smoked, but was not associated with blood pressure. F2-isoprostane excretion was negatively correlated with age. This finding is not unique to patients with RA; F2-isoprostanes were also inversely associated with age in the Framingham cohort (41) and in a large cohort of healthy subjects from 3 European countries (44). The Framingham study also showed a lack of multivariable association between oxidative stress and blood pressure (41). We found a significant association between F2-isoprostane excretion and HDL concentrations only after adjustment for potential confounders; this occurred in part because the association between F2-isoprostane excretion and HDL concentrations was negatively confounded by an inverse association between HDL and BMI.

Oxidative stress is thought to play a key role in the pathogenesis of atherosclerosis and has been implicated as an explanation for premature atherosclerosis that occurs in inflammatory rheumatic diseases such as SLE and RA (45). Many traditional cardiovascular risk factors are associated with increased oxidative stress, including hypertension, dyslipidemia, smoking, and obesity (1, 17). Increased oxidative stress was also associated with increased coronary calcium in healthy young adult populations (15). Since RA is associated with increased cardiovascular mortality (46) and premature coronary atherosclerosis (18, 47), it was important to define the relationship between increased oxidative stress and accelerated atherosclerosis. Our findings suggest that F2-isoprostane excretion is not independently associated with coronary calcium score in RA, but rather contributes to increasing the risk of coronary calcification by adversely modifying another cardiovascular risk factor, HDL cholesterol.

A key early process in the pathogenesis of atherosclerosis is the oxidation of LDL cholesterol that leads to the accumulation of oxidized LDL in the vessel wall, and consequently an inflammatory response and the formation of atheroma (48). The effects of oxidative processes on HDL cholesterol are less well defined. Recently, inflammatory rheumatic diseases and the accompanying increased risk of coronary atherosclerosis have been associated with the presence of a dysfunctional form of HDL, termed proinflammatory (or proatherogenic) HDL (49). Proinflammatory HDL is thought to be formed when antiatherogenic components of HDL such as apolipoprotein A-I or paraoxonase 1 are replaced by proatherogenic components such as SAA, ceruloplasmin, and oxidized lipids (45). Therefore, HDL in its proinflammatory form fails to protect against atherosclerosis.

Our results showing that increased oxidative stress is associated with a loss of protective effect of HDL cholesterol against coronary calcification are consistent with the finding of increased proinflammatory HDL in rheumatic diseases (20, 21). Measurement of proinflammatory HDL concentrations requires fresh plasma, and thus we were unable to determine the relationship between F2-isoprostanes and proinflammatory HDL concentrations. Our findings suggest that in patients with RA, high HDL concentrations might not necessarily lead to lower coronary risk and that factors such as oxidative stress may influence the cardioprotective capacity of HDL. Recent studies showing that HDL is the major lipoprotein carrier of F2-isoprostanes (50) are concordant with this observation. Additional studies to define the relationship between increased oxidative stress and proinflammatory HDL in patients with RA will be of interest.

Our study had some limitations. The study was cross-sectional and the concentrations of HDL and F2-isoprostanes may vary over time. Furthermore, coronary calcification, an excellent measure of the amount of atherosclerosis present in vessels, was the outcome. The outcome of greatest significance is incident cardiovascular events; however, that outcome will require a large prospective study. Also, high-sensitivity CRP level would have been ideally measured by the same method in all of the patients. Finally, a larger sample size would have improved statistical power and allowed adjustment for a greater number of potential confounders.

In conclusion, oxidative stress measured as urinary F2-isoprostane excretion was increased in patients with RA compared with control subjects. Oxidative stress modified the relationship between HDL and coronary calcification so that increased oxidative stress at higher HDL concentrations was associated with a greater severity of coronary calcification.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Stein had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Rho, Chung, Oeser, Stein.

Acquisition of data. Oeser, Solus, Raggi, Milne, Stein.

Analysis and interpretation of data. Rho, Chung, Solus, Gebretsadik, Shintani, Raggi, Stein.

Acknowledgements

The authors thank Jason D. Morrow, MD (deceased), for advice in the planning and performance of the study.

Ancillary