Background: For more than 40 years, the long-term effect of lithium maintenance therapy on renal function has been debated. We aimed to assess the effect of lithium maintenance therapy on estimated glomerular filtration rate (eFGR) in patients with affective disorders, and explore predictors for a decrease in eGFR.
Methods: This population-based cohort study included adult patients (18-65 years of age at baseline) in Tayside (Scotland, UK) who had recently started on lithium maintenance treatment between Jan 1, 2000, and Dec 31, 2011 (retrospectively assigned to the lithium group) or those with exposure to other first-line drugs used in the treatment of affective disorders (quetiapine, olanzapine, and semisodium valproate) during the same period (retrospectively assigned to the comparator group). Patients had to have at least 6 months of (incidence) exposure to lithium or any of the comparator drugs, at least two eGFR values available in the observation period (one at baseline and at least one after ≥6 months post baseline). We excluded patients with previous exposure to lithium or one of the comparator drugs, those with a previous diagnosis of schizophrenia or other psychotic disorder, those with glomerular disease, tubulo-interstitial disease, or chronic kidney disease stages 4-5 at baseline, and those who had undergone renal transplant before exposure. Maximum follow-up was 12 years. Data were provided by the University of Dundee Health Informatics Centre, who have access to health-related population-based datasets containing data for every patient registered with a regional family doctor. Each patient has a unique ten-digit identifier, the Community Health Index, enabling us to link laboratory tests, dispensed community prescriptions, Scottish Morbidity Records, and mortality records to the patient. All data were anonymised according to Health Informatics Centre standard operating procedures. The primary outcome was the change per year in the eGFR, adjusted for age, sex, and baseline eGFR, and analysed by random coefficient models.
Findings: 1120 patients (305 exposed to lithium and 815 to comparator drugs) qualified for inclusion, providing 13 963 eGFR values over 12 years. The mean duration of exposure to lithium was 55 months (SD 42; range 6-144). Mean annual decline in eGFR (adjusted for age, sex, and baseline eGFR) was 1·3 mL/min per 1·73 m(2) (SE 0·2) in the lithium group, which did not differ significantly to that in the comparator group (0·9 mL/min/1·73 m(2) [SE 0·15]). After adjustment for additional confounders, the monthly decline in eGFR attributable to lithium exposure amounted to 0·02 mL/min per 1·73 m(2) (SE 0·02, p=0·30). As a post-hoc secondary outcome, we estimated the annual decline in eGFR for the lithium group to be 1·0 mL/min per 1·73 m(2) (SE 0·2), which again did not differ significantly to that in the comparator group (0·4 mL/min/1·73 m(2) [SE 0·2]. Modelling identified significant predictors for eGFR decline as age, baseline eGFR, comorbidities, co-prescriptions of nephrotoxic drugs, and episodes of lithium toxicity; however, duration of exposure to lithium and mean serum lithium level were not significant predictors for eGFR decline.
Interpretation: Our analysis suggests no effect of stable lithium maintenance therapy (lithium levels in therapeutic range) on the rate of change in eGFR over time. Our results therefore contradict the idea that long-term lithium therapy is associated with nephrotoxicity in the absence of episodes of acute intoxication and that duration of therapy and cumulative dose are the major determinants of toxicity.