Prevalence of coronary artery calcification and its association with mortality, cardiovascular events in patients with chronic kidney disease: a systematic review and meta-analysis

Abstract Purpose: To date, the prevalence and prognostic role of coronary artery calcification (CAC) in patients with chronic kidney disease (CKD) have been investigated in several studies, but have yielded conflicting results. The aim of this meta-analysis is to derive a more precise estimation of CAC prevalence in CKD patients and its association with cardiovascular events and mortality. Methods: The relevant literature was identified and evaluated from inception until July 2018 through multiple search strategies on PubMed, Embase, and Web of Science. Cross-sectional or cohort (baseline data) studies reporting CAC prevalence were included. Data extracted from eligible studies were used to calculate effect estimates (ESs) and 95% confidence intervals (95%CI). We searched databases for observational studies that explored baseline CAC and subsequent cardiovascular or all-cause mortality risk in CKD patients. Results: The meta-analysis included 47 studies; 38 of these were included in the final analysis of CAC prevalence. The pooled prevalence of CAC in random effect model was 60% (95%CI 53–68%). CAC was positively associated with an increased risk of all-cause mortality (hazard ratio [HR] 3.44; 95%CI 2.40–4.94), cardiovascular mortality (HR 3.87; 95%CI 2.06–7.26), and cardiovascular events (HR 2.09; 95%CI 1.19–3.67), when comparing individuals in the top CAC score group to those in the bottom CAC score group. Conclusions: The pooled prevalence of CAC is highly prevalent. CAC is independently associated with all-cause and cardiovascular mortality risk as well as cardiovascular events among CKD patients. In view of the high heterogeneity, larger clinical trials are still needed.


Introduction
Deterioration in renal function is associated with marked increase in cardiovascular mortality. More than 50% of deaths in patients with end stage renal disease (ESRD) are attributable to cardiovascular disease [1]; patients with chronic kidney disease (CKD) are at higher risk of coronary artery calcification (CAC). CAC is known to be associated with cardiovascular events as well as cardiovascular and all-cause mortality [2][3][4][5]. However, another study found that vascular calcification did not independently predict mortality in predialysis patients [6]. Emerging studies found that CAC was common in predialysis patients [7]; another study showed that CAC was more prevalent in dialysis patients, up to 93% [8].
CAC can be effectively quantified using electronbeam or multi-detector computed tomography (EBCT/MDCT), by measuring total calcium levels. CAC among CKD patients is defined as an Agatston score >0. Uremia-related risk factors, such as higher plasma calcium and phosphate, as well as increased oxidative stress, contribute to CAC progression.
CAC is present in early CKD patients, but is more common in dialysis patients. Better understanding of the epidemiology of CAC in CKD populations would potentially contribute to the formulation of strategies in further clinical intervention. Although several studies have reported positive association between CAC and CKD, the prevalence of CAC varies considerably. The results of these studies were inconsistent, which may be due to inadequate statistical power, publication bias, ethnic differences or uncorrected multiple hypothesis testing. Therefore, to overcome the limitations of individual studies, a meta-analysis was performed to overcome the limitations of individual studies, to explore the prevalence of CAC among predialysis, dialysis, and renal transplant patients along with studying the impact of CAC on clinical outcomes.

Methods
The systematic review and meta-analysis were performed according to the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [9].

Search strategy
A comprehensive search was carried out in PubMed, Embase, and Web of Science databases from inception of the study until July 2018. The search terms used were as follows: 'coronary artery calcification' or 'vascular calcification' and 'chronic kidney disease' or 'chronic renal failure' or 'chronic kidney failure' or 'hemodialysis' or 'peritoneal dialysis' or 'uremia' and 'death' or 'mortality'. The search keywords were searched both as medical subject headings (MeSH) and text words, without restrictions on ethnicity or geographic area. References of included studies were searched for eligible articles.

Inclusion and exclusion criteria
Articles were included when they met the following criteria: (1) observational studies (cross-sectional or cohort study); (2) studies with a clear definition of CKD; (3) CAC was assessed using a calcification score; (4) the studies provided the prevalence of CAC, or sufficient data to calculate it. Exclusion criteria were as follows: (1) comments, review articles, meetings, letters, case reports, meta-analyses and unrelated or animal studies.

Data extraction and quality assessment
According to the MOOSE guidelines, we attempted to extract the following information: first author name, publishing year, ethnicity, study type, point prevalence, sample size, age, percentage of men, dialysis duration, Figure 1. Flowchart of the study selection process. hazard ratio (HR) and 95%CI of mortality and cardiovascular events. Data of selected studies were independently extracted by two reviewers (X.W. and X.X.). In case of disagreement, the issue was resolved by a third investigator (J.Zh.). When original important data were missing, we contacted corresponding authors to obtain the relevant data. We did not impute missing data.
The quality of included studies was independently assessed by the two reviewers (X.W. and X.X.), using a special assessment [10,11]. Studies with a score of 0-6 were considered as poor quality, and those with a score of 7-13 were considered as high quality. A set of 13 criteria (study attrition, study participation, outcome measurement, confounding measurement, and analysis) was predefined.

Statistical analysis
We used I 2 statistic to identify heterogeneity. If I 2 50% or p < 0.05, a random effect model was applied; otherwise, a fixed effect model was used. In addition, low, moderate, and high levels were nominally applied to define I 2 values as 25%, 50%, and 75%, respectively [12]. In addition, subgroup analyses were conducted when the heterogeneity was high. We conducted univariate meta-regression analysis to explore the effects of disease-related covariates on the CAC prevalence estimates.
HR and 95%CI were extracted from studies that reported proportional hazard regression, and a natural logarithm scale was conducted.
Begg's test and funnel plots were performed to evaluate publication bias. p Values of Begg's test <0.05 and asymmetry of funnel plots showed the possibility of publication bias.
Finally, p value <0.05 was considered to be significant difference. Statistical analyses for this article were conducted using STATA version 12.0 (Stata Corporation, College Station, TX).
The data of kappa of agreement during the literature search were analyzed using SPSS Statistics version 17.0 (SPSS Inc., Chicago, IL). The data of kappa >0.60 were considered to indicate good agreement. The predialysis stages include CKD 1-5 stages.

Study selection and quality assessment
The flowchart of study selection for inclusion and exclusion is presented in Figure 1. Of the eligible 38 CAC prevalence studies [7,8,, there were 25 cross-sectional studies, 13 cohort studies in the final analysis, as seen in Table 1. The mean age of the participants ranged from 48 to 68.7 years. The articles included 17 predialysis, two renal transplantation, and 19 dialysis studies. The reported diagnostic methods for CAC were uniform across all the included studies. Among the publications, 37 studies were in English and one was in Chinese.
The average score of quality assessment was 8.71 in CAC prevalence studies. Of the 38 prevalence studies, nine studies were considered poor quality. In 12 CAC prognosis articles (Table 2), the average score of quality assessment was 9.46, and only two studies were of poor quality. The data of kappa of agreement during quality assessment were 0.747, indicating good agreement.

CAC prevalence
As shown in Figure 2, the overall prevalence of CAC among CKD patients was 60% (95%CI 53-68%). Significant heterogeneity was observed in the metaanalysis (I 2 ¼ 99%, p < 0.01). A random effect model was applied since the heterogeneity was high.

Subgroup analysis and meta-regression analysis
In most cases, significant heterogeneity was observed, so we performed a subgroup analysis. The pooled estimates of CAC prevalence in different subgroups are shown in Table 3. A significant difference was found between older age (60 years) and younger age (<60 years).
We used meta-regression to explore the source of heterogeneity. We found that age (r ¼ 1.009, p ¼ 0.032) and dialysis duration (r ¼ 1.005, p ¼ 0.021) were positively associated with CAC prevalence, but there was no association between CAC prevalence and the proportion of men in the studies (r ¼ 0.996, p ¼ 0.136).

Sensitivity analyses and publication bias
No significant reporting bias was observed among the 38 studies of CAC prevalence (Begg's test p ¼ 0.513). Funnel plots were shown in Figure S1. There was no obvious bias of literature about CAC on mortality (Begg's test p ¼ 0.592). Funnel plots were shown in Figure S2. We did not perform the evaluation for publication bias of literature about CAC on cardiovascular events, as the limited number of studies was included in the meta-analysis. The sensitivity analysis demonstrated that none of the articles significantly affected the pooled prevalence, prognosis.

Discussion
Our meta-analysis of 38 selected studies investigated the prevalence Of CAC among patients with CKD. In our study, we found that CAC was highly prevalent among CKD patients, the pooled prevalence being 60% (95%CI, 53-68%).
According to our study, nearly two-thirds of CKD patients are diagnosed with CAC. Compared to studies on dialysis, there are far fewer studies about kidney transplant recipients. In consequence, sparse data and smaller sample size are partially responsible for the less precise estimated prevalence in the settings. Future studies are needed to explore the prevalence of CAC among renal transplant recipients. Due to the insufficient literature, CAC prevalence at different CKD stages could not be determined. Large clinical trials are needed to explore the prevalence of CAC at different CKD stages.
As higher heterogeneity was observed in the study, subgroup and meta-regression analyses were performed to detect the potential source of heterogeneity. Many studies have found the high prevalence of CAC in CKD patients across different regions and our meta-analysis validated these previously published results. Our analysis divided the studies based on region: Asian countries were compared with the rest of the world. This demonstrated that region was a potential source of heterogeneity. Subgroup analysis demonstrated that smaller sample sizes of these studies also contributed to the heterogeneity. The prevalence of studies published before 2010 is higher than that of the studies published 2010-2018. This implies that more attention is being paid to improve uremia factors such as CKD-mineral and bone disorder (MBD), and that treatment of CKD continues to improve. It was important that CAC prevalence was lower in cross-sectional studies, as well-conducted cohort studies provided a higher level of evidence than cross-sectional studies. Univariate meta-regression found that age and dialysis duration may be potential sources of heterogeneity, suggesting that     older age was related to CAC [58]. In agreement with previous findings, we found that longer dialysis duration accelerated the progression of CAC [59].
The relationship between CKD and CAC prevalence is not fully understood, but several possible reasons are listed as follows. First, CKD-MBD is a common risk for CAC, particularly with higher plasma calcium and phosphate as well as severe secondary hyperparathyroidism. These higher levels are positively associated with vascular calcification. Second, calcification inhibitors are down-regulated in uremic factors [60]. Third, chronic inflammation is common among patients with CKD, and positively associated with CAC in CKD patients. Homocysteine and C-reactive protein (CRP) have been shown to up-regulate the expression of the inflammation process [61].
To our knowledge, this present meta-analysis is the first to explore prognostic role of CAC among CKD    patients. We analyzed 12 studies, and found that CAC resulted in worse outcome. Mortality risk and cardiovascular events were higher among patients with coexisting CAC and CKD. We found a low number of studies that analyzed cardiovascular events and mortality and only low to moderate heterogeneity was found; therefore, we did not conduct subgroup analysis and meta-analysis. Ten studies, reported the all-cause mortality risk (the highest CAC score group vs. the lowest CAC score group), were analyzed to detect the potential source of heterogeneity. In subgroup analysis, we found that patients receiving renal replacement therapy had higher allcause mortality, and CKD stages were the source of heterogeneity. We did not demonstrate that other risk factors were responsible for the heterogeneity; the reasons may be smaller samples, limited data and short follow-up visits.
In CKD patients, calcification is found both in the intimal and medial layers of blood vessels. The mechanism of CAC that causes cardiovascular events and mortality is not fully understood. The possible reasons are listed as follows. Intimal calcification may lead to distal emboli, progressing to acute coronary syndromes [62]. Medial calcification induces stiffness of the coronary wall, resulting in reduction of myocardial perfusion. Medical calcification in the peripheral arteries increases myocardial afterload, causing left ventricular hypertrophy and cardiac failure [63,64].
Some limitations of our study have to be acknowledged. First, the results indicate considerable heterogeneity; the differences in study design, age, region, sample size, year of publication, CKD stages, and modality of dialysis were responsible for this variation. The short follow-up interval and the lower number of events recorded could affect the results, especially for cardiovascular and all-cause deaths. Second, some of the studies have small sample size, making it impossible to provide precise stratification according to CKD stages. Third, observational studies, depict associations, but do not establish cause and effect relationships. Finally, this article is not registered online to avoid repetition, which is one limitation of our study. However, we have kept searching the latest literature to track the progression of studies and avoid duplication of effort during the study period.

Conclusions
CAC is prevalent in patients with CKD. Severe CAC significantly and rapidly deteriorates the prognosis among CKD patients, and CAC significantly affects cardiovascular and all-cause mortality. Future large prospective studies are necessary to explore CAC prevalence at different stages of CKD. Moreover, long-term follow-up visits are necessary to understand the effect of CAC on patients at different CKD stages. Future researches should explore the possible mechanism and therapeutic interventions to halt CAC.

Disclosure statement
No potential conflict of interest was reported by the authors.