Oral bisphosphonate use and latent class trajectories of kidney function: A cohort study

Background: Oral bisphosphonates (BP) are cautioned or contraindicated in patients with moderate‐severe chronic kidney disease (CKD) due to concerns regarding their effects on renal function. There is a scarcity of data on the existing trajectories of variation in estimated glomerular filtration rate (eGFR) over time. Objectives: To identify clusters of patients with stage 3B+ CKD with different patterns of eGFR over 7 years; and to examine the association between the baseline BP use and the observed eGFR trajectories. Methods: A cohort of 105 720 people were identified from UK primary care records (CPRD) aged 40+ with baseline (first available or most recent before BP initiation) eGFR < 45 ml/min/m2. Patients with <2 eGFR measurements over 7 years or a prescription of antiosteoporosis therapy prior to baseline were excluded. Discrete time survival mixture analysis was used to identify clusters of patients with distinct kidney function trajectories over 7 years. This method addresses non‐ignorable missing eGFR due to death. Bayesian information criterion and change in log‐likelihood were used to select the optimal model. BP use was measured before ascertaining trajectories. Multinomial logistic regression was used to examine the association between the baseline BP use and the identified eGFR trajectories, adjusted for pre‐identified baseline confounders. Results: We identified six clusters representing different trajectories. Two clusters (C1, 2%, C2, 11%) experienced improving eGFR but with different accelerating rates (mean(SD) 6.08 ± 2.22 and 3.20 ± 1.60 per year, for C1 and C2 respectively). One (C3, 31%) had gradually improving eGFR (1.18 ± 1.32). Patients in C4 (35%) had either fluctuating eGFR, late‐improving eGFR or stable stage 3B eGFR. 17% (C5) of patients had a (rapid) declining eGFR (−1.23 ± 1.59), and 4% (C6) had persistent low eGFR. When compared with C3, patients in the low or declining eGFR clusters (C4, C5 and C6) were more likely to have BP users, with adjusted relative risk ratios (95% CI) of 1.74 (1.60, 1.89), 3.03 (2.78, 3.32) and 3.08 (2.68, 3.55) respectively. Conclusions: Six distinct trajectories of kidney function progression are identified in 3B+ CKD patients. BP use is almost twice as common in those with fluctuating/late‐improving/stable CKD, and 3‐fold more prevalent in those with more rapid decline in or persistently low eGFR. Further research is needed, including the validation of the identified clusters and the observed associations with BP use, in external datasets.