Diabetes and Trajectories of Estimated Glomerular Filtration Rate: A Prospective Cohort Analysis of the Atherosclerosis Risk in Communities Study

Abstract

Objective: To characterize long-term kidney disease trajectories in persons with and without diabetes in a general population.

Research design and methods: We classified 15,517 participants in the community-based Atherosclerosis Risk in Communities (ARIC) study by diabetes status at baseline (1987-1989; no diabetes, undiagnosed diabetes, and diagnosed diabetes). We used linear mixed models with random intercepts and slopes to quantify estimated glomerular filtration rate (eGFR) trajectories at four visits over 26 years.

Results: Adjusted mean eGFR decline over the full study period among participants without diabetes was -1.4 mL/min/1.73 m²/year (95% CI -1.5 to -1.4), with undiagnosed diabetes was -1.8 mL/min/1.73 m²/year (95% CI -2.0 to -1.7) (difference vs. no diabetes, P < 0.001), and with diagnosed diabetes was -2.5 mL/min/1.73 m²/year (95% CI -2.6 to -2.4) (difference vs. no diabetes, P < 0.001). Among participants with diagnosed diabetes, risk factors for steeper eGFR decline included African American race, APOL1 high-risk genotype, systolic blood pressure ≥140 mmHg, insulin use, and higher HbA_1c.

Conclusions: Diabetes is an important risk factor for kidney function decline. Those with diagnosed diabetes declined almost twice as rapidly as those without diabetes. Among people with diagnosed diabetes, steeper declines were seen in those with modifiable risk factors, including hypertension and glycemic control, suggesting areas for continued targeting in kidney disease prevention.

Figure 1

Distribution of annual unadjusted and adjusted eGFR slopes from best linear unbiased predictions by diabetes status. Short dashed line, no diabetes; dashed line, undiagnosed diabetes; and solid line, diagnosed diabetes. Adjusted for the following characteristics at baseline and their interactions with time, continuous variables centered at their means: age (reference: 54.67 years), sex (reference: male; and female), race–center (reference: Forsyth County, NC–white; and Forsyth County, NC–black, Jackson, MS–black, Minneapolis, MN–white, and Washington County, MD–white), systolic blood pressure (reference: 121.22 mmHg), hypertension medication use (reference: no; and yes), BMI (reference: 27.68 kg/m²), HDL (reference: 51.60 mg/dL), prevalent coronary heart disease (reference: no; and yes), smoking status (reference: never; and former and current), annual family income (reference: <$25,000; and ≥$25,000), and educational status (reference: high school; and less than high school, vocational school, college, and graduate/professional school).

Figure 2

Distribution of annual unadjusted and adjusted eGFR slopes from visit 1 to visit 2 from best linear unbiased predictions by diabetes status. Short dashed line, no diabetes; dashed line, undiagnosed diabetes; and solid line, diagnosed diabetes. Adjusted for the following characteristics at baseline and their interactions with time, continuous variables centered at their means: age (reference: 54.67 years), sex (reference: male; and female), race–center (reference: Forsyth County, NC–white; and Forsyth County, NC–black, Jackson, MS–black, Minneapolis, MN–white, and Washington County, MD–white), systolic blood pressure (reference: 121.22 mmHg), hypertension medication use (reference: no; and yes), BMI (reference: 27.68 kg/m²), HDL (reference: 51.60 mg/dL), prevalent coronary heart disease (reference: no; and yes), smoking status (reference: never; and former and current), annual family income (reference: <$25,000; and ≥$25,000), and educational status (reference: high school; and less than high school, vocational school, college, and graduate/professional school).