Proton Pump Inhibitors and Hypomagnesemia in the General Population

A Population-Based Cohort Study

Brenda C.T. Kieboom, MD; Jessica C. Kiefte–de Jong, RD, PhD; Mark Eijgelsheim, MD, PhD; Oscar H. Franco, MD, PhD; Ernst J. Kuipers, MD, PhD; Albert Hofman, MD, PhD; Robert Zietse, MD, PhD; Bruno H. Stricker, MMed, PhD; Ewout J. Hoorn, MD, PhD


Am J Kidney Dis. 2015;66(5):775-782. 

In This Article


The 4 key findings of this prospective population-based study were: (1) the demonstration that PPI use is associated with increased risk of hypomagnesemia in the general population, (2) the higher risks with prolonged PPI use, (3) the higher risk with concomitant loop diuretic use, and (4) the identification of similar but weaker associations with the use of H2RAs. In the paragraphs that follow, these findings are discussed in more detail and placed in the context of previous studies of this topic.[10,11,15]

In several case reports and case series, it was already noted that PPI-induced hypomagnesemia primarily occurs in patients who use PPIs long term (ie, >3 months).[1,4,26,27] Our study confirms these observations and is in agreement with the prediction that small deficits in magnesium balance may eventually lead to hypomagnesemia. For example, 40 mg of omeprazole in healthy volunteers reduced magnesium absorption by only 1%,[28] but it was predicted that this could lead to an 80% depletion of magnesium stores over the course of 1 year.[29] The slow depletion of intracellular magnesium stores may place individuals who use PPIs or H2RAs long term at risk for frank hypomagnesemia, especially in combination with additional factors such as diuretics or intercurrent illness.[4]

Although PPI-induced hypomagnesemia is caused by reduced intestinal magnesium absorption,[26,30,31] we found that the risk of hypomagnesemia increased with concomitant use of loop diuretics. Loop diuretics may compromise the kidney magnesium reabsorption necessary to compensate for intestinal magnesium loss. A study of critically ill patients also observed that combined use of PPIs and diuretics was associated with a greater decrease in serum magnesium level.[11] In that study, this effect was also primarily seen with loop diuretics, which were used almost twice as frequently as thiazide diuretics.[11] The fact that we only observed effect modification with loop diuretics is surprising. Thiazide diuretics can also cause hypomagnesemia by increased renal magnesium wasting,[32,33] and thiazide diuretics were used twice as frequently as loop diuretics in our population (providing sufficient power for this analysis). This suggests that quantitatively, magnesium reabsorption in the loop of Henle (where loop diuretics act) is more important than in the distal convoluted tubule (where thiazide diuretics act) to compensate for increased intestinal magnesium loss. Alternatively, this may indicate differences in compensatory mechanisms. It is well known that diuretic use results in a compensatory increase in reabsorption in nephron parts that are not blocked by the diuretic.[34] It was recently shown that PPIs cause a minor but significant decrease in transient receptor potential melastatin type 6 (TRPM6) in the kidney,[35] which is located in the distal convoluted tubule. This could imply that TRPM6 can no longer compensate for increased fecal and urinary magnesium losses induced by the combined use of PPIs and loop diuretics. In contrast, paracellular magnesium reabsorption in the loop of Henle may remain intact to prevent hypomagnesemia with the combined use of PPIs and thiazide diuretics.

An unexpected result was the association between the use of H2RAs and magnesium level. Although this association was not found for patients admitted to an intensive care unit,[11] a similar association was found in a sensitivity analysis in a large cross-sectional study in a health maintenance organization database in Israel, including more than 95,000 individuals.[10] The effect size in this Israeli study was also smaller than for PPI-induced hypomagnesemia, and the association was only observed for moderate and not for severe hypomagnesemia. The investigators interpreted this finding as residual confounding because H2RA use has not been associated with hypomagnesemia previously. Although we cannot exclude this possibility in our study, we believe that it may reflect a true association that has remained clinically unnoticed due to the smaller effect size (although the risk of hypomagnesemia was similar with the use of PPIs and H2RAs). The lack of effect modification with loop diuretics may explain why no cases of severe hypomagnesemia have been reported in the literature with H2RA use, although publication bias is difficult to exclude.

What are the clinical implications of this study? We expect that few participants had symptomatic hypomagnesemia because only 3 participants had a serum magnesium concentration < 1.00 mEq/L (the level below which symptoms usually occur[27]). A recent study from Canada also showed that the absolute risk of hospitalization for PPI-induced hypomagnesemia is very low (1 excess hospitalization in >70,000 outpatients treated with a PPI for 90 days).[15] That said, our study clearly identified a subgroup of PPI users who are at risk of developing symptomatic hypomagnesemia, including those who use PPIs long term or combine PPIs with loop diuretics. We therefore believe that the warning issued by the US Food and Drug Administration in 2011 to measure serum magnesium prior to initiation and periodically afterward[36] primarily pertains to this subgroup. Furthermore, although the 0.024-mEq/L lower serum magnesium level in PPI users is not clinically significant in terms of symptoms, it may indicate a total-body magnesium deficit that has been associated with adverse outcomes in previous epidemiologic studies.[5,8,37]

The strength of this study is that to our knowledge, it is the first performed in a general population using systematically measured serum magnesium concentrations. This minimizes bias associated with clinical settings. However, our study also has limitations. Because we only had 1 serum magnesium measurement per participant, we were not able to study the course of serum magnesium levels after initiation of PPI treatment. We therefore studied the relationship between serum magnesium levels and PPI use over different strata of use, but because this concerns different participants, confounding could still have had a role. We tried to address this confounding by indication by adjusting for dietary intake, which did not alter our results. Another limitation is the fact that we did not have information for over-the-counter use of PPIs. However, we believe this will have had little impact on our results because our latest inclusion date was before the date that PPIs were allowed to be sold over the counter in the Netherlands. Even if participants were using nonprescribed PPIs, this differential misclassification would have resulted in a dilution of the effect, which would make the association between PPIs and hypomagnesemia even stronger.

In conclusion, this study confirms the association between PPI use and risk of hypomagnesemia in the general population. The risk of hypomagnesemia is further increased when PPI use is prolonged (>6 months) or combined with the use of loop diuretics. Health care professionals should consider monitoring serum magnesium levels periodically in patients expected to be on prolonged treatment or those who combine PPIs with medication that may cause hypomagnesemia, such as diuretics. Although H2RAs had similar effects on serum magnesium level and risk of hypomagnesemia, these effects were weaker and further studies are needed before the precautions for PPIs should also be recommended for H2RAs.