Pharmacoepidemiology evaluation of bumetanide as a potential candidate for drug repurposing for Alzheimer's disease

Abstract

Introduction: Bumetanide, a loop diuretic, was identified as a candidate drug for repurposing for Alzheimer's disease (AD) based on its effects on transcriptomic apolipoprotein E signatures. Cross-sectional analyses of electronic health records suggest that bumetanide is associated with decreased prevalence of AD; however, temporality between bumetanide exposure and AD development has not been established.

Methods: We evaluated Medicare claims data using Cox proportional hazards regression to evaluate the association between time-dependent use of bumetanide and time to first AD diagnosis while controlling for patient characteristics. Multiple sensitivity analyses were conducted to test the robustness of the findings.

Results: We sampled 833,561 Medicare beneficiaries, 60.8% female, with mean (standard deviation) age of 70.4 (12). Bumetanide use was not significantly associated with AD risk (hazard ratio 1.05; 95% confidence interval, 0.99-1.10).

Discussion: Using a nationwide dataset and a retrospective cohort study design, we were not able to identify a time-dependent effect of bumetanide lowering AD risk.

Highlights: Bumetanide was identified as a candidate for repurposing for Alzheimer's disease (AD). We evaluated the association between bumetanide use and risk of AD. We used Medicare data and accounted for duration of bumetanide use. Bumetanide use was not significantly associated with risk of AD.

Keywords: Alzheimer's disease; drug repurposing; loop diuretics; pharmacoepidemiology.

FIGURE 1

Study design. The figure summarizes the sample selection and study design. First, we selected individuals who were continuously enrolled in Medicare fee‐for‐service for at least 1 year. Index date was defined as the date of the first prescription for a drug of interest (loop diuretics, non‐loop diuretics, and non‐diuretic antihypertensives). The 360‐day period preceding index date was the baseline period used for the definition of covariates. Medication use was assessed starting on index date and throughout all the follow‐up period available for a patient. To enable sufficient time for medications to have a detectable effect on outcomes, we applied a 360‐day lag period after index date, and only started to collect outcome events 360 days after index date (start of the outcome observation period). The start of the outcome observation period was used as time zero for survival analyses. Patients who had a diagnosis of the outcome or were censored before the start of the outcome observation period were excluded from analyses, as they would not have time at risk.

FIGURE 2

Adjusted hazard ratios of Alzheimer's disease and all‐cause dementia for ever use of medications of interest. Ever use of of medications was defined with time‐varying indicator variables, which denoted whether an individual had used a drug of interest at any point of time prior to the period of assessment. In other words, once an individual used a drug, the indicator variable for ever used remained 1 throughout follow‐up. Non‐loop diuretics included diuretic drugs that are not loop diuretics (thiazides and potassium‐sparing diuretics). Non‐diuretic antihypertensives include ACE inhibitors, ARBs, calcium channel blockers, and beta‐blockers. The model was adjusted for age, sex, race, low‐income subsidy, Medicaid eligibility, and all chronic conditions listed in Table 1 with the exception of hypertension. ACE, angiotensin‐converting enzyme; ARBs, angiotensin II receptor blockers.

FIGURE 3

Adjusted hazard ratios of AD and all‐cause dementia, per 1‐year increment in drug use. Cumulative use of medications was defined with continous time‐dependent variables, and was measured at each interval as the cumulative number of 30‐day intervals that each subject had used a specific medication. To improve interpretability, the variable was expressed per 1‐year increments. For example, the hazard ratio of AD associated with bumetanide can be interpreted as follows: For each 1‐year increment in the use of bumetanide, the hazards of AD increased by 0.1% (95% CI, –1.6% to 1.9%). Non‐loop diuretics included diuretic drugs that are not loop diuretics (thiazides and potassium‐sparing diuretics). Non‐diuretic antihypertensives include ACE inhibitors, ARBs, calcium channel blockers, and beta‐blockers. The model was adjusted for age, sex, race, low‐income subsidy, Medicaid eligibility, and all chronic conditions listed in Table 1 with the exception of hypertension. AD, Alzheimer's disease; ACE, angiotensin‐converting enzyme; ARBs, angiotensin II receptor blockers; CI, confidence interval.