Kidney disease progression and the factors influencing progressive loss of kidney function in a primary care population. A retrospective database analysis
Monitoring kidney function using estimated glomerular filtration rate (eGFR) is recommended in people with, or at risk of, chronic kidney disease (CKD). Current guidelines suggest increasing the intensity of monitoring according to eGFR, urine albumin levels, comorbidities and changes in therapy, but the evidence base is weak. To inform recommendations for the timing of eGFR monitoring in primary care, we set out to create a statistical model for kidney disease progression and the factors influencing this progressive loss of kidney function in a primary care population.
We used the Clinical Practice Research Practice Datalink (CPRD) to construct an open cohort study of all adult patient records from April 2005 to March 2014. Patients were eligible for inclusion if they had three or more serum creatinine tests on record. We excluded patients who, in the 12 months before study entry, were pregnant, had a renal transplant, were receiving dialysis, or were living kidney donors. Follow-up ended at the study end date, unless preceded by death, transfer out of CPRD, pregnancy, renal transplantation/donation, or dialysis. To model kidney function over time, a hidden Markov model (HMM) was fitted to four cohorts of patients based on their baseline albuminuria status. The HMM estimates the relationship between observed values of eGFR and the unobserved values of GFR over time. Models were adjusted for annually updated age, sex, read codes for chronic heart failure or cancer at baseline. The average time spent in stage 3b (the sojourn time) and five and ten year risks of renal failure (eGFR < 15 ml/min/1.73^2) and all-cause death are presented.
Of 1,973,064 patients, 1,921,949 had no recorded urine albumin at baseline, 37,946 had tested and normal urine albumin levels (< 3 mg/mmol), 10,247 had microalbuminuria (3 – 30 mg/mmol) and 2,922 had macroalbuminuria (> 30 mg/mmol) at baseline. The sojourn time (95% C.I.) in stage 3b CKD was 25.1 (22.3 to 27.1) years for patients with untested urine albumin, 15.7 (14.6 to 16.8) years in patients with normal urine albumin, 7.3 (6.7 to 8.1) years for patients with microalbuminuria and 4.5 (3.9 to 5.3) years in patients with macroalbuminuria. For a male, age 60, with eGFR 45-59 ml/min/1.73^2 the five-year risk (95% CI) of end-stage renal failure is 0.017 (0.012 to 0.026)% and the 10 year risk 0.1 (0.07 to 0.15)%, whereas the 5 and 10 year risk of death is 6.8 (6.1 to 7.6)% and 13.5 (12.3 to 14.8)%.
Our findings have implications for scheduling appointments for monitoring eGFR. As true change in kidney function is slow, monitoring intervals should be extended for low-risk groups. This model will combine with a cost-effectiveness analysis to provide evidence-based recommendations for monitoring eGFR in primary care.