Association between albuminuria, incident cardiovascular events, and mortality in persons without hypertension, diabetes, and cardiovascular disease

Epidemiological studies have reported associations between elevated urine levels of albumin (albuminuria) and risk of cardiovascular (CV) events.^1–3 A meta-analysis showed that albuminuria could improve CV risk prediction beyond established risk factors,¹ and albuminuria was recently suggested as a surrogate endpoint for the progression of kidney disease.⁴ In current clinical practice, albuminuria screening is predominantly performed in the staging of kidney disease or in persons with hypertension or diabetes as a marker for end-target organ damage.⁵ It remains unclear if albuminuria screening for CV risk assessment should be extended to persons without diabetes and hypertension. Our aim was to investigate the association between albuminuria and 10-year risk of a composite CV outcome in adults without hypertension or diabetes in the UK Biobank study.

Among participants with urinary biomarker data (n = 483 993), we excluded those who had withdrawn consent, an outcome event before baseline, and prevalent hypertension, diabetes, or a history of CV disease. Our study was approved by the UK Biobank (project 42176) and the Swedish Ethical Review Authority (Dnr.: 2019-02328). The composite CV outcome combined the first event of CV death, acute myocardial infarction, stroke, or heart failure. We calculated the urine albumin-to-creatinine ratio (ACR) and Cox regression was used with censoring at the first event, death, or the last date of follow-up (31 March 2017). Analyses were adjusted for CV risk factors: age, sex, ethnicity, smoking, systolic and diastolic blood pressure, total cholesterol, high-density lipoprotein (HDL)-cholesterol, low-density lipoprotein (LDL)-cholesterol, aspirin, lipid-lowering medication, glycosylated haemoglobin type A1c (HbA1c), body mass index, and creatinine-estimated glomerular filtration rate (eGFR). Missing values were imputed by predictive mean matching. In secondary analyses, we compared the following risk subgroups: current vs. non-smokers, ever vs. never smokers, impaired vs. normal kidney function (eGFR > 60 mL/min/1.73 m²), presence vs. absence of dyslipidaemia (lipid-lowering treatment), triglycerides ≥2.0 mmol/L or LDL-C ≥4.1 mmol/L, high-normal vs. normal blood pressure (>120/80 mmHg), and HbA1c in the pre-diabetic vs. normal rage (≥39 mmol/mol). We compared the predictive performance of albuminuria with that of eGFR with regard to risk discrimination (Harrell's C), explained variance, and model fit (likelihood ratio test).

We included 198 637 persons aged 54 ± 8 years (62.7% women) without hypertension, diabetes, or CV disease at baseline (a flowchart is provided as Supplementary material online, Figure S1). Sample characteristics are shown in Table 1 and incidence rates for outcomes during up to 10 years’ follow-up (mean 7.1, median 7.0, range 5.4–10.0 years) are displayed in Supplementary material online, Table S1. More than 5% of participants had prevalent microalbuminuria or macroalbuminuria (5.6% and 0.1%, respectively). Raised urine ACR was associated with increased 10-year risk of the composite CV outcome following adjustment for CV risk factors [hazard ratio (HR) per log unit in ACR, 1.21, 95% confidence interval 1.14–1.29, P = 9.0 $\times$ 10⁻¹⁰] (Figure 1A and Supplementary material online, Table S1). The increase in risk was more evident in ACR levels above the microalbuminuria threshold (Figure 1B). Similar associations were observed for CV mortality, all-cause mortality and the individual outcomes myocardial infarction, stroke, and heart failure. A heatmap (Supplementary material online, Figure S2) of HRs for albuminuria in combination with dyslipidaemia and smoking indicates ACR as an independent risk factor in addition to established risk markers. In the comparison between subgroups (Figure 1C and Supplementary material online, Table S2), raised ACR was strongly associated with increased CV risk, regardless of smoking, kidney function, dyslipidaemia, blood pressure, or pre-diabetes. To compare ACR to eGFR as an alternative marker of kidney function, we assessed both as added components of the ACC/AHA’s Atherosclerotic CV Disease (ASCVD) model, and the ESC’s Systematic Coronary Risk Evaluation (SCORE) model for 10-year CV risk and CV mortality risk, respectively (Supplementary material online, Table S3). For the composite CV outcome, ACR added marginally more risk discrimination information to the ASCVD model than eGFR (increase in C-statistic 0.0007 for ACR, 0.0004 for eGFR, and 0.0013 when both were added, Supplementary material online, Table S3). Adding ACR led to a significantly stronger improvement of model fit compared to eGFR. The case numbers for CV mortality were limited and analyses too underpowered for meaningful interpretation (Supplementary material online, Table S3).

Table 1.

Sample characteristics of 198 637 persons without hypertension, diabetes mellitus, or cardiovascular disease at baseline of the UK Biobank

Female gender, n (%)	124 559 (62.7)
Age (years), mean ± SD	53.5 ± 8.0
Urine ACR (mg/mmol), mean ± SD	13.7 ± 9.4, median 9.4 (interquartile range 5.8–15.6)
Macroalbuminuria, n (%)	231 (0.1)
Microalbuminuria, n (%)	11 110 (5.6)
Ever smoker, n (%)	83 071 (41.8)
Current smoker, n (%)	22 859 (11.5)
Lipid treatment, n (%)	8736 (4.4)
Aspirin, n (%)	9039 (4.6)
Total cholesterol (mmol/L), mean ± SD	5.7 ± 1.0
LDL-cholesterol (mmol/L), mean ± SD	3.6 ± 0.8
HDL-cholesterol (mmol/L), mean ± SD	1.5 ± 0.4
Triglycerides (mmol/L), mean ± SD	1.5 ± 0.9
BMI (kg/m2), mean ± SD	25.9 ± 4.1
HbA1c (mmol/mol) (%), mean ± SD	34 ± 3 (5.8 ± 0.4)
Systolic blood pressure (mmHg), mean ± SD	124 ± 10
Diastolic blood pressure (mmHg), mean ± SD	75 ± 7
eGFR (mL/min/1.73 m2), mean ± SD	93 ± 12
eGFR <30/30–60/60–90/>90 mL/min/ 1.73 m2, n (%)	16 (<0.1)/1834 (0.9)/66 774 (33.6)/130 013 (65.5)

Urine albumin was measured with a Randox Bioscience immuno-turbidimetric assay (analytical range 6.7–200 mg/L, coefficient of variation 1.85–2.08%). Urine creatinine was measured with a Beckman Coulter ion selective electrode assay (analytical range 2–200 mmol/L; coefficient of variation 2.09–2.10%). Samples were processed on a Beckman Coulter AU5400 analyser. All measurements were carried out centrally by the UK Biobank organisation.

ACR, albumin-to-creatinine ratio; BMI, body mass index; CV, cardiovascular; eGFR, estimated glomerular filtration rate; HbA1c, glycosylated haemoglobin type A1c; HDL, high-density lipoprotein; LDL, low-density lipoprotein; MI, myocardial infarction; SD, standard deviation.

Figure 1.

(A) Association between urine albumin-to-creatinine ratio and 10-year incidence of cardiovascular outcomes and mortality. Hazard ratios (95% confidence interval) per natural log unit increase in urinary albumin-to-creatinine ratio estimated in Cox regression adjusted for age, sex, ethnicity, ever smoking, current smoking, systolic blood pressure, diastolic blood pressure, total cholesterol, high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol, aspirin treatment, lipid medication, glycosylated haemoglobin type A1c, body mass index, and estimated glomerular filtration rate. (B) Spline curve illustrating the association between urine albumin-to-creatinine ratio and 10-year incidence of the composite cardiovascular endpoint. Hazard ratios (95% confidence interval) per natural log unit increase in urinary albumin-to-creatinine ratio (polynomial smoothing spline with four degrees of freedom) estimated in Cox regression adjusted for age, sex, ethnicity, ever smoking, current smoking, systolic blood pressure, diastolic blood pressure, total cholesterol, high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol, aspirin treatment, lipid medication, glycosylated haemoglobin type A1c, body mass index, and estimated glomerular filtration rate. The composite outcome includes cardiovascular mortality, non-fatal MI, heart failure, and stroke. (C). Association between urine albumin-to-creatinine ratio and 10-year risk of the composite cardiovascular outcomes in high- and low-risk strata according to established cardiovascular risk factors. Hazard ratios (95% confidence interval) per natural log unit increase in urinary albumin-to-creatinine ratio estimated in Cox regression adjusted age, sex, ethnicity, ever smoking, current smoking, systolic blood pressure, diastolic blood pressure, total cholesterol, high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol, aspirin treatment, lipid medication, glycosylated haemoglobin type A1c, body mass index, and estimated glomerular filtration rate. The composite outcome includes cardiovascular mortality, non-fatal MI, heart failure, and stroke. CI, confidence interval; CV, cardiovascular; eGFR, estimated glomerular filtration rate; HR, hazard ratio.

In summary, we found that microalbuminuria was present in more than 5% of adults without diabetes or hypertension. Worse albuminuria was associated with greater CV event risk, and the associations were similar in apparently healthy persons with or without conventional CV risk factors, suggesting that albuminuria may be a risk factor already present prior to the development of other CV risk factors. When added to an established risk score, albuminuria improved the prediction of CV disease and outperformed eGFR. Our results point to the importance of albuminuria for the development of CV disease and show the potential of albuminuria screening even in low-risk persons, as it identifies a subgroup of high-risk persons who would not otherwise have been identified by conventional risk factors. The limited power, insensitivity of the C-statistic to added predictors,⁶ and small improvements in model fit with unclear clinical relevance do not allow firm conclusions about the superiority of albuminuria over eGFR in our study. Previous studies have demonstrated the cost-effectiveness of albuminuria screening for CV and renal disease in persons with hypertension and diabetes.⁷ Whether extending albuminuria screening to all adults without hypertension and diabetes is cost-effective is less well-studied, but there is some evidence that it may be cost-effective in persons over 50 years,⁸ and if performed at 10-year intervals.⁹

Strengths of our study include the large contemporary sample with high-quality phenotyping. A single‐void urine sample test is reliable and useful in screening contexts,¹⁰ but there is substantial day-to-day variation in ACR and reliance on spot samples may have diluted associations and overestimated the prevalence of microalbuminuria. The relatively insensitive albuminuria assay may have limited our ability to detect associations between normal range albuminuria and CV outcomes, and 71% of participants had undetectable albuminuria. Generalizability is limited to persons of European descent.

Our study provides new evidence that the predictive ability of albuminuria for CV risk appears to hold true even in persons without hypertension and diabetes. Our findings challenge the current paradigm and suggest that albuminuria screening could also be considered in apparently healthy persons. Replication in independent population-based cohorts and cost-effectiveness studies are warranted to ascertain the clinical importance of our findings.

Supplementary material

Supplementary material is available at European Journal of Preventive Cardiology online.