Reduced Kidney Function, Albuminuria, and Risks for All-cause and Cardiovascular Mortality in China: A Population-based Cohort Study

Chronic kidney disease (CKD), defined by reduced glomerular filtration rate (GFR) or presence of markers for kidney damage, has rapidly emerged as a severe global health problem, with the population prevalence of CKD ranging from 10% to 16% in various regions [1‐4]. Meanwhile, recent studies revealed that independent, strong, and graded associations were observed between clinical prognosis and two important markers of CKD: reduced GFR and increased urinary albumin excretion [5], suggesting the inclusion of albumin in CKD staging in order to improve risk stratification [6].

In China, the largest developing country, the prevalence of CKD is 10.8%, meaning 119.5 million adults have reduced estimated glomerular filtration rate (eGFR) and/or albuminuria [7]. Understanding the consequences of CKD based on relevant markers is important for making relevant decisions regarding prevention and treatment of CKD. However, most studies were from developed countries, and studies based on Asian general populations are limited. Previously, Wen et al. reported an inverse association between eGFR and risk of all-cause as well as cardiovascular disease (CVD) mortality, among patients receiving regular health check-ups in Taiwan. Positive urine proteinuria, which was determined by trace and higher levels in the dipstick test, was also found to be significantly associated with an increased risk of adverse outcomes [4]. Nonetheless, the high incidence of renal replacement therapy in Taiwan, and different etiology of end-stage renal disease in Taiwan compared with mainland China, limit generalizability of the study [8, 9].

To address the limitations of the existing studies, we have initiated the present study, on a nationally representative sample of the Chinese adult population, to evaluate the association between two markers of CKD based on a single measurement and mortality.

The mean age of the population at baseline was 49.6 ± 15.2 years, and 42.7% of the participants were male. The mean eGFR was 92.0 ± 18.7 mL/min per 1.73m² and the median level of ACR was 6.6 (IQR: 3.1, 13.6) mg/g. Compared to those without reduced kidney function, participants with eGFR < 60 mL/min per 1.73m² were older, with a higher percentage of nephrotoxic medication users, history of MI or stroke, hypertension and diabetes, lower percentage of male, high-school education or above, and current smokers, as well as higher levels of BMI, TG, and LDL-C. Similar patterns were observed between participants with ACR of ≥30 mg/g and those with <30 mg/g. However, the difference for each risk factor was smaller between people with and without kidney damage (Table 1).

The median follow-up duration was 6.1 (IQR: 4.7, 7.5) years. During follow-up, 878 deaths occurred, and 35.6% of them (n = 313) were cardiovascular mortality. The overall and cardiovascular mortality rates were 3.2 and 1.1 per 1000 person-years, respectively. Compared to those without indicators of CKD, participants with advanced eGFR or ACR levels had a higher mortality rate (all p-values of log-rank test <0.001 for comparisons among eGFR or ACR categories) (Table 2 and Figure 1).

Compared to those with eGFR ≥ 90 mL/min/1.73 m², participants with eGFR of 60–89 and <60 mL/min/1.73 m² was associated with increased risk of all-cause mortality, with HRs of 2.17 (95% CI: 1.88–2.50) and 4.50 (95% CI: 3.59–5.64), respectively, in the univariable analysis. However, when age and sex were adjusted, the significant associations disappeared. A similar pattern of association was observed between advanced eGFR categories and cardiovascular mortality. Regarding albuminuria, significant associations with all-cause mortality were observed through univariable and multivariable analyses. Albuminuria A3 (≥300 mg/g) was associated with a higher risk of mortality than albuminuria A2 (30–299 mg/g), and compared with albuminuria A1. Similar findings were seen for cardiovascular mortality. However, the associations of albuminuria A2 (30–299 mg/g) with cardiovascular mortality were not significant after adjusting for covariates (Table 3). We modeled eGFR and ACR in linear spline and found a higher risk of all-cause mortality associated with eGFR < 90 mL/min/1.73 m², but the associations were non-significant after adjusting for covariates. Reduced risk of all-cause mortality was associated with eGFR > 90 mL/min/1.73 m² in both univariable and multivariable adjusted analyses. However, for ACR, a significantly higher risk of all-cause mortality can be seen in both univariable and multivariable adjusted analyses, when ACR is greater than 10 mg/g. No significant associations can be found when ACR is less than 10 mg/g (Figure 2). Similar findings can be detected for cardiovascular mortality (Figure 3). Sensitivity analysis by using the eGFR estimated by a modified MDRD equation obtained similar results (data not shown).

Significant interactions between level of eGFR and age were detected in the associations for both all-cause mortality (p-value for interaction < 0.001) and cardiovascular mortality (p-value for interaction = 0.02). No interactions were detected between ACR categories and age, and ACR categories and the two CKD indicators. Hence, we initiated association analyses stratified by age groups (≥65 years or <65 years). The associations among the participants aged less than 65 years were more pronounced than in those aged over 65 years. For example, the fully-adjusted HR for all-cause mortality was 1.64 (95% CI: 1.02–2.64) for eGFR <60 mL/min/1.73 m² among people aged less than 65 years, compared to 0.86 (95% CI: 0.61–1.21) for eGFR ≥90 mL/min/1.73m² among people aged more than 65 years (Additional file 1: Tables S1 and S2).

In the joint analysis of eGFR and ACR, we observed a similar pattern of increased mortality rates for ACR ≥ 30 mg/g compared with ACR <30 mg/g across all levels of eGFR and vice versa. However, there were no significant associations of joint categories of eGFR and ACR with both all-cause and cardiovascular mortality (Table 4).

In our study conducted among a general Chinese population, increased ACR was found to be independently associated with increased risk of both all-cause and cardiovascular mortality. Reduced eGFR was only found to be associated with increased risk of the adverse outcomes in univariable analysis, but the association disappeared after adjusting for age and sex. Furthermore, the associations were modified by age, and seem to be more prominent among participants aged < 65 years than in those aged ≥ 65 years.

An increased level of urinary albumin excretion is an early sign of kidney damage and persistent albuminuria, typically ACR ≥ 30 mg/g lasting for more than three months, has been included as a diagnostic standard for CKD [14]. The association of increased ACR with both all-cause mortality and cardiovascular mortality in the general population has been well described in the analyses of the CKD Prognosis Consortium(CKD-PC) [5]. In the above meta-analysis, ACR greater than 5 mg/g or dipstick urine protein 1+ and above were associated with increased risk of adverse outcomes. However, previous studies of Asian populations did not employ ACR as a measurement for albuminuria [15, 16], except for some smaller studies [17, 18]. There is an individual study from the meta-analysis based on population from Taiwan, where people have the same ethnicity as those in mainland China. Wen et al. reported an increased risk of both all-cause and CVD mortality associated with minimal or overt proteinuria in populations with eGFR ≥ 60 mL/min/1.73 m² (CKD stages 1 and 2), compared with people without CKD [4]. In our study, the quantitative measurement of ACR was used to represent albuminuria, and we found results consistent with the above mentioned studies. Furthermore, we found that the effect of increased ACR on mortality was independent of traditional CVD risk factors and eGFR levels. Regarding the spline analysis, we selected the same knots as the above mentioned CKD-PC meta-analysis, but used 10 mg/g instead of 5 mg/g as the reference point due to the limited sample size in our study. However, we did not replicate the protective effect for mortality associated with ACR of less than 10 mg/g [5]. As the effect of ACR could be partly due to its representativeness of other metabolic disorders, e.g. hypertension and diabetes mellitus, it could be assumed that the significant association found in our study could be a reflection of the pathogenic process due to albuminuria itself, e.g. endothelial dysfunction and systemic inflammation [19, 20]. A previous national survey of CKD in China has reported a much higher prevalence of albuminuria than of reduced eGFR [7]. Hence, it may be important to place more resources in the screening and intervention of albuminuria among the general population.

Previous studies consistently demonstrated an increased risk of adverse outcomes associated with reduced kidney function among the general population [5, 21, 22]. Our study, however, failed to fully replicate the previous findings among a large sample of a general Chinese population, as the significant association between reduced eGFR and adverse outcomes turned out to be null after adjusting for the age of the population. In our study, a small percentage of the population had impaired kidney function (eGFR < 60 mL/min/1.73 m²) (n = 2090, 4.4%). Furthermore, the present study found a very high percentage of elderly among CKD versus non-CKD patients (62.3% versus 14.7%). Therefore, the non-significant association between reduced eGFR and mortality may be due to the limited number of people with exposure and a significant confounding effect of age.

Furthermore, we found that, aside from being an important confounding factor, age is also an effect modifier for eGFR. Several studies have shown that the effect of kidney insufficiency on the occurrence of cardiovascular disease and death is modified by age. For example, Raymond et al. used the central laboratory system data of 106,366 participants in Coventry, UK, and found that the relative risk of mortality in a population of 60 years of age and above was increasing more slowly than that in the younger population and the value of relative risk was also smaller among elderly adults than younger adults [23]. A US study conducted among veterans suggested the association of mortality risk with impaired kidney function was attenuated in people older than 75 years [24]. However, based on a large-scale study of a Norwegian general population, Hallan et al. found that the association between reduced kidney function and cardiovascular mortality tended to be stronger in participants aged over 70 years than in those aged less than 70 years [25]. Our study supports the former finding that reduced eGFR exerts a more obvious impact on mortality among a younger population than among an older population. The reason for the interaction may be due to the high mortality rate and high prevalence of co-morbidities of CKD (e.g. hypertension and diabetes) among the elderly, which could make the effect of reduced eGFR alone less significant.

Although our study has the advantage of a large sample size of a general population, as well as longitudinal validation of adverse outcomes, some limitations should be noted. First, biomarkers including eGFR and ACR were measured in different study centers. Although measurement was calibrated with the standard sample from the central laboratory and under tight quality control, some variation among different centers might still exist. Second, the two race CKD-EPI equation was not developed for Asian populations and may limit its accuracy for estimating GFR in the Chinese population. Nevertheless, it might perform as well as the modified MDRD equation for the Chinese population, as shown by an evaluation study among 977 adult Chinese individuals [26]. Third, given the likely under-reporting rate of death, the mortality rate among different groups of participants in our study might be underestimated. Fourth, the number of study participants in late CKD stages was limited, with less than 1% of the total population in stages 3B to 5. Only 11.6% of the population had ACR > 30 mg/g. As a result, there was not enough study power to evaluate the full effect for all five stages of CKD. Fifth, we used baseline status for analyses, and subsequent measurements for changes in risk factors were not conducted. At the same time, albuminuria and reduced kidney function determined by a single measurement did not conform to the ≥3 month persistence criteria to diagnose CKD.

Our study was a further attempt to study a large Asian population to determine the relationship between albuminuria and mortality. However, the association of reduced kidney function with mortality was not clear.