Coronary Artery Calcification Score and the Progression of Chronic Kidney Disease

Study Design and Participants

The KNOW-CKD is a nationwide prospective cohort study from nine tertiary care South Korean centers. The study rationale, design, methods, and protocol summary are detailed elsewhere (NCT01630486; http://www.clinicaltrials.gov). Briefly, The KNOW-CKD enrolled Korean patients with CKD between the ages of 20 and 75 years with G1–G5 without kidney replacement therapy (KRT) between 2011 and 2015. The KNOW-CKD study included participants with albuminuria or other early markers of kidney damage and with normal or mildly decreased GFR (CKD G1–2). CKD was defined based on the eGFR, which was calculated using the CKD-Epidemiology Collaboration equation.¹⁷ We excluded patients with any of the following: unable or unwilling to provide written consent, previous chronic dialysis or organ transplantation, heart failure (New York Heart Association class 3 or 4), cirrhosis (Child-Pugh class 2 or 3), pregnancy, or history of malignancy. All patients should annually provide blood and urine samples according to the study protocol. They were followed up regularly (1- to 3-month intervals) at the physicians’ discretion. The study was conducted by the principles of the Declaration of Helsinki, and the study protocol was approved by the Institutional Review Boards of participating centers.

The cohort comprised 2238 participants who voluntarily provided informed consent. Among these, we excluded 194 patients who did not measure CACS. As we intended to explore the clinical implication of CACS with respect to CKD progression in patients without previous coronary artery disease, we further excluded patients with previous history of coronary artery disease (n=77) and coronary artery stenting (n=31) (Figure 1). Previous coronary artery disease was confirmed by study investigators at participating centers. Finally, 1936 participants were included in the study.

Data Collection

Baseline demographics and laboratory data were retrieved from the electronic data management system (PhactaX, Seoul, Republic of Korea). Demographic data including age, sex, smoking history, etiology of CKD, comorbidities, and medications were collected at the time of screening. Height was measured using a digital stadiometer while participants stood barefoot. Body weight (kg) was measured using standard methods in light clothes without shoes. Height and weight were measured to the nearest 1 cm and 0.1 kg. Body mass index was calculated as the weight divided by height squared (kg/m²). Smoking status was divided into three categories as never, former, and current. Education level was categorized as low, less than middle school; middle, middle school; and high, more than middle school. Hypertension was defined as self-reported hypertension, systolic BP ≥140 mm Hg or diastolic BP ≥90 mm Hg, or current use of antihypertensive drugs. Antihypertensive drugs included angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, calcium channel blockers (dihydropyridine and nondihydropyridine calcium channel blockers), β-blockers, and diuretics. Diabetes mellitus was defined as a history of diabetic mellitus, fasting glucose ≥126 mg/dl, or use of glucose-lowering drugs. In this study, we included only patients with type 2 diabetes mellitus (T2DM). Lipid-lowering drugs included statins, ezetimibe, and fibrates. The Charlson comorbidity index score was used to reflect disease burden at baseline.

BP measurements were performed by a trained nurse at each clinic according to the standardized protocol as suggested by the American Heart Association.¹⁸ BP was measured after >5 min of rest in a sitting position at the clinic office using a calibrated oscillometer electronic sphygmomanometer or mercury sphygmomanometer. The electronic device was validated by comparison of the device readings (four in all) alternating with five mercury readings taken by two trained observers. The mean of three BP readings was used as the BP value for each visit.¹⁹

After overnight fasting, venous samples were collected to determine hemoglobin, BUN, creatinine, uric acid, total cholesterol, triglycerides, HDL cholesterol, LDL cholesterol, albumin, calcium, phosphate, high-sensitivity C-reactive protein (hs-CRP), and intact parathyroid hormone (intact-PTH) levels. Fibroblast growth factor-23 (FGF-23) was measured by ELISA at baseline (Immutopics, San Clemente, CA). Random urine samples from midstream collection were used to measure the urine protein-to-creatinine ratio (UPCR). We used a creatinine method that requires calibration traceable to isotope dilution mass spectrometry. CKD classification was followed by the Kidney Disease Improving Global Outcomes guidelines.²⁰

Measurement of CACS

CACS was determined by electrocardiography-gated coronary 64-slice multidetector CT scanning of the thorax according to the standard protocol. Calcification was defined as an area of hyper-attenuation greater than 130 Hounsfield units within a 1-mm² area. The quantitative CACS was calculated using the Agatston method on a digital workstation by cardiac radiologists at each participating center.⁶ They were all blinded to patient information. Participants were categorized into three groups according to baseline CACS (CACS of 0, CACS of 1–100 AU, and CACS >100 AU). In addition, CACS per 1-SD log was calculated and used as a continuous variable for the analysis.

Study End Point

The primary outcome of interest was a composite of the first occurrence of a 50% decline in eGFR from the baseline value or the onset of KFRT. KFRT was defined as CKD G5 requiring dialysis or kidney transplantation. Patients with CKD G3 and greater were closely observed for the occurrence of adverse clinical events and had visited each participating center at 1- to 3-month intervals. Patients who reached the end points were reported by each center. Then, we cross-checked between centers if the outcome events were correct. The secondary end points were the separate components of the primary outcome and all-cause mortality. The study observation period ended on March 31, 2019.

Statistical Analyses

Continuous variables were expressed as mean with SD or median with interquartile range (IQR) for skewed data. Categorical variables were expressed as numbers of participants with percentage. To compare the difference according to CAC groups, a one-way analysis of variance was used for continuous and categorical variables. The Kruskal–Wallis test was used for data with skewed distribution. The multivariable cause-specific hazard models were used to assess the association of CACS with CKD progression. To validate the results of the primary analysis, we further constructed subdistribution hazard models. In these two models, a death that occurred before reaching the primary outcome was treated as a competing risk. To be more specific, subjects experiencing a competing risk event remain in the risk set in the subdistribution model, whereas they are removed in the cause-specific model.²¹ Model 1 represents unadjusted hazard ratios (HRs). Model 2 adjusted for sex, age, Charlson comorbidity index, smoking history, education level, body mass index, and etiology of CKD. We also created model 3 after adjustment for BP and laboratory parameters such as eGFR, total cholesterol, HDL cholesterol, LDL cholesterol, hs-CRP, and medications including antihypertensive (renin-angiotensin system blockers, calcium channel blockers, β-blockers, and diuretics) and lipid-lowering drugs (statins, ezetimibe, and fibrates), in addition to factors included in model 2. In model 4, phosphate, intact-PTH, FGF-23, and UPCR were further added. Because CACS is a strong predictor of CVD and there is a significant connection between CVD and CKD, we further examined relative contribution of nonfatal CVEs to CKD progression. To this end, we constructed a time-varying model, in which the occurrence of nonfatal CVEs before the primary outcome was treated as a time-varying covariate. The nonfatal CVEs were defined as myocardial infarction, unstable angina, heart failure, and stroke. Other repeated variables such as BP, body mass index, UPCR, and eGFR were also considered as time-varying covariates. There were missing values for education level, BP, total cholesterol, HDL cholesterol, LDL cholesterol, hs-CRP, phosphate, UPCR, intact-PTH, and FGF-23, which were less than 5% of all measurements (Supplemental Table 1). We assumed missing data as missing at random. The multiple imputation by chained equations method was used to impute 10 independent copies of the data for the analyses. The results of hazard models were presented as HRs with 95% confidence intervals (CIs). The probability of kidney survival rates was estimated with cumulative incidence function and compared by Gray’s test.²² Survival time was defined as the time interval between enrollment and the first onset of clinical outcomes. Patients who were lost to follow-up were censored at the date of the last examination. To validate our findings, we performed several sensitivity analyses. First, we examined the association of CACS with the risk of CKD progression after excluding 116 (6.1%) patients with eGFR <15 ml per min per 1.73 m² at baseline because CKD G5 is considered equivalent to kidney failure requiring KRT. Second, this association was further tested after including patients with previous coronary artery disease. In this analysis, we excluded patients with prior coronary stenting (n=31) because the presence of stent can interfere with the calculation of CACS. Third, we also analyzed the association after excluding patients with calcium-based phosphate binders (n=164) because this medication can be a potential confounder or effect modifier. Finally, we further analyzed using two different CACS categories with 0, 1–400, and >400, and 0, 1–100, 101–400, and >400. A restricted cubic spline curve was used to reveal the association between log CACS and the HRs for the composite kidney outcome. In addition, we further examined effect modification among prespecified subgroups by sex (male or female), age (<60 years or ≥60 years), smoking history (never or former and current), T2DM (without or with), antiplatelet drugs such as aspirin or clopidogrel (without or with), statins (without or with), systolic BP (<140 mm Hg or ≥140 mm Hg), LDL level (<100 mg/dl or ≥100 mg/dl), severity of CKD (G1–2 or G3–5ND), and UPCR (<1.0 g/gCr or ≥1.0 g/gCr). Finally, the rate of kidney function decline per year was assessed using the slope of eGFR obtained from a generalized linear mixed model. The statistical analyses were performed using Stata, version 15.1 (Stata Corporation, College Station, TX). A P value <0.05 was considered statistically significant.

Baseline Characteristics

Baseline characteristics according to CACS categories are shown in Table 1. The mean age of the participants was 53.2 years and 60.7% were male. The mean eGFR was 51.5 (SD, 30.3) ml/min per 1.73 m² and the median CACS was 0.5 (IQR, 0.0–75.4) AU. Graphs showing distribution of CACS in all patients and patients with and without CKD progression are presented in Supplemental Figure 1. Compared with patients without CAC, those with CACS of 1–100 and CACS >100 were more likely to be older and male, and more likely to have T2DM and to be treated with antiplatelet drugs (aspirin and clopidogrel) and lipid-lowering medications. In addition, these patients had higher levels of body mass index, BP, glucose, triglycerides, hs-CRP, intact-PTH, and proteinuria, but had lower levels of total cholesterol, HDL cholesterol, LDL cholesterol, and eGFR.

Association between CACS and CKD Progression

During 8130 person-years of follow-up, the primary outcome occurred in 584 (30.2%) patients. The overall incidence rate was 5.4, 8.2, and 10.3 per 1000 person-years in patients across the CACS categories, respectively (P<0.001) (Table 2). In addition, mean time to event occurrence was significantly shorter (3.2 versus 2.9 versus 2.5 years) and cumulative adverse kidney events were significantly higher in patients with CACS of 1–100 and CACS >100 than in those without CAC (Figure 2). The unadjusted HRs for CACS of 1–100 and CACS >100 were 1.54 (95% CI, 1.27–1.86) and 1.96 (95% CI, 1.60–2.41), respectively, compared with CAC of 0. This association was consistent for sequential adjustment of demographics and laboratory parameters (models 2–4). In the final model (model 4), patients with CACS of 1–100 and CACS >100 were associated with a 1.29-fold (95% CI, 1.04–1.61) and 1.42-fold (95% CI, 1.10–1.82) higher risk of the primary outcome. Furthermore, when CACS was used as a continuous variable, the HR per 1-SD log of CACS was 1.13 (95% CI, 1.03–1.24). Another competing risk model with subdistribution hazard model yielded similar results (Supplemental Table 2). Restricted cubic spline curve analysis clearly showed a linear relationship between CACS and CKD progression (Figure 3). Finally, the slope of eGFR decline per year was significantly greater in patients with CACS of 1–100 (−3.01 ml/min per 1.73 m²; 95% CI, −3.44 to −2.58) and CACS >100 (−4.18 ml/min per 1.73 m²; 95% CI, −4.71 to −3.66) compared with CACS of 0 (−2.55 ml/min per 1.73 m²; 95% CI, −2.87 to −2.22) (Table 3).

Secondary Outcome Analyses

To evaluate what extent the risk of adverse kidney outcome with CACS was affected by CVEs, we performed time-varying analysis. During the follow-up period, nonfatal CVEs occurred in 118 (6.0%) participants. In keeping with the results of primary analysis, the HRs for CAC of 1–100 and CACS >100 were 1.31 (95% CI, 1.07 to 1.60) and 1.46 (95% CI, 1.16 to 1.85), respectively (Supplemental Table 3). Interestingly, in this time-varying model, nonfatal CVEs were associated with a 1.25-fold (95% CI, 1.02 to 1.79) higher risk of CKD progression, suggesting a significant contribution of nonfatal CVEs to the adverse kidney outcome. We further evaluated the association of CACS with the individual components of the kidney outcome and all-cause mortality. In both categorical and continuous models, the significant association mostly persisted. For the outcome of ≥50% decline in eGFR, CACS of 1–100 and CACS >100 were associated with a 1.28-fold (95% CI, 0.99 to 1.64) and 1.34-fold (95% CI, 1.00 to 1.80) increased risk for eGFR decline (Supplemental Table 4). In addition, the HRs for KFRT were 1.19 (95% CI, 0.94 to 1.52) for CAC of 1–100 and 1.54 (95% CI, 1.17 to 2.02) for CACS >100, respectively. For 1-SD higher level of log CAC, the HRs were 1.15 (95% CI, 1.03 to 1.29) for ≥50% decline in eGFR and 1.19 (95% CI, 1.07 to 1.32) for KFRT. For all-cause mortality, CAC >100 was associated with 2.34-fold (95% CI, 1.24 to 4.42) higher risk of all-cause mortality (Supplemental Table 5). In addition, the risk of all-cause mortality was 1.65-fold (95% CI, 1.28 to 2.12) increased for per 1-SD higher log of CACS.

Sensitivity Analysis

In sensitivity analysis after excluding 116 (6.1%) patients with eGFR <15 ml/min per 1.73 m² at baseline, CACS of 1–100 (HR, 1.27; 95% CI, 1.01 to 1.60) and CACS >100 (HR, 1.43; 95% CI, 1.08 to 1.89) were significantly associated with a higher risk of the primary kidney outcome (Supplemental Table 6). This association was similar in continuous modeling; for every 1-SD increase in log CACS, there was a 1.13-fold (95% CI, 1.02 to 1.26) higher risk of the progression of CKD. This association was consistent in analysis including patients with previous coronary artery disease (Supplemental Table 7) and in analysis excluding patients with calcium-based phosphate binders (Supplemental Table 8). Furthermore, in analyses using two different CACS categories with 0, 1–400, and >400, and 0, 1–100, 101–400, and >400, the graded association of CACS and risk of CKD progression remained unchanged (Supplemental Table 9).

Subgroup Analyses

To evaluate the effect modification of subgroups on the relationship between CACS and CKD progression, we tested interactions among prespecified subgroups. P values for interactions were >0.05 for the subgroups by sex, smoking, statin, proteinuria, BP, and LDL cholesterol level. However, there were significant interactions of age, T2DM, severity of CKD, and the use of antiplatelet drugs with CACS categories. The significant association between CACS and CKD progression was more pronounced in older patients and those with T2DM and nonuse of antiplatelet drugs (Figure 4).