Proton Pump Inhibitors and Hypomagnesemia in the General Population: A Population-Based Cohort Study

Abstract

Proton pump inhibitor (PPI) use has been associated with hypomagnesemia in case reports and hospital-based cohort studies. Our objective was to determine whether PPI use is associated with hypomagnesemia in the general population and whether this is also found in histamine 2 receptor antagonist (H2RA) users. Prospective cohort study. 9,818 individuals from the general population (Rotterdam Study). PPI use and H2RA use compared to no use. Serum magnesium and hypomagnesemia (serum magnesium≤ 1.44 mEq/L).Analyses were adjusted for age, sex, body mass index, kidney function, comorbid conditions, and alcohol and diuretic use. Serum magnesium level was 0.022 mEq/L lower in PPI users (n=724; 95% CI,-0.032 to-0.014 mEq/L) versus those with no use. PPI use was associated with increased risk of hypomagnesemia (n=36; OR, 2.00; 95% CI, 1.36-2.93) compared to no use. Effect modification was found between the use of PPIs and loop diuretics; in participants using loop diuretics (n=270), PPI use was associated with a further increased risk of hypomagnesemia (n=5; OR, 7.22; 95% CI, 1.69-30.83) compared to no use. The increased risk with PPIs was only seen after prolonged use (range, 182-2,618 days; OR, 2.99; 95% CI, 1.73-5.15). Including dietary magnesium intake into the model did not alter results (available for 2,504 participants, including 231 PPI users). H2RA users (n=250) also had a lower serum magnesium level (-0.016 [95% CI,-0.032 to-0.002] mEq/L) and increased risk of hypomagnesemia (n=12; OR, 2.00; 95% CI, 1.08-3.72) compared to those with no use, but no interaction with loop diuretics. Cross-sectional analysis with single serum magnesium measurement. PPI use is associated with hypomagnesemia in the general population. Prolonged PPI use and concomitant loop diuretic use are associated with a stronger risk increase. Similar but weaker associations were found in H2RA users, except for interaction with loop diuretics.