An Investigation of Hypomagnesemia Among Ambulatory Urban African Americans

The standard for hypomagnesemia at the hospital reference laboratory where the specimens were analyzed was a serum magnesium level of 1.5 mg/dL. This level represents a value 2 standard deviations below the mean value among 1000 Red Cross volunteer donors. Levels less than and equal to 1.5 mg/dL were defined as demonstrating hypomagnesemia. Using this criterion, we split the subjects in our study into hypomagnesemic (eg, serum magnesium · 1.5 mg/dL) and normal (eg, serum magnesium >1.5 mg/dL) groups.

Data Analyses

Descriptive data analysis was performed to explore the prevalence of hypomagnesemia and the relationship between the dependent variable (serum magnesium levels) and various independent variables (demographics, clinical status). Odds ratios and 95% confidence intervals were used to assess the likelihood of hypomagnesemia with selected independent variables. In addition, SAS software⁹ was used to construct hierarchical logistic models to predict hypomagnesemia. Demographics and health behavior indicators entered the model first. Indicators of hypertension and diabetes were subsequently added to the analysis, thus capturing the potential contribution of comorbidity to serum magnesium levels. A dichotomous measure of hypertensive and diabetic comorbidity was used.

Results

A total of 120 patients was enrolled. Before the analyses, 1 patient was eliminated from the study after it was determined that the patient’s medical regimen included magnesium supplements. This yielded a total of 119 patients for analysis. Demographic breakdowns revealed majorities of women (n = 84, 71%) and blacks (n = 105, 87%). Most subjects were older than 40 years (n = 75, 63%). A majority reported having medical insurance through Medicaid or Medicare (n = 85, 71%). Among these subjects, 59 had previously had a diagnosis of either hypertension (n = 47; 39%) or diabetes (n = 38, 32%).

The overall prevalence of hypomagnesemia was 20% (n = 24). Table 2 summarizes univariate analyses for selected demographic and medical characteristics. We did not find magnesium status to be related to age, sex, race, hyperlipidemia, diabetes, hypertension, renal disease, or smoking status. Hypomagnesemia was increased 6-fold among subjects with a history of alcoholism (odds ratio [OR] = 6.0; 95% confidence interval [CI], 1.41 - 6.91). Hypomagnesemia was also increased among subjects with 1 or more comorbid medical conditions, including hypertension, diabetes, alcoholism, renal disease, and reactive airways disease (OR = 4.69; 95% CI, 1.37 - 17.65).

A logistic model was constructed that included terms for sex, age, race, alcoholism, diabetes, and hypertension; however, the only significant term was alcoholism (OR = 7.8; 95% CI, 1.82 - 33.6). The final model explained 19% of the variance in serum magnesium levels.

Discussion

Previous studies of serum magnesium have focused nearly exclusively on inpatient populations. Although this approach simplified collection of serum specimens for analyses, it overlooked examination of the magnesium status for the majority of patients who are typically seen as outpatients. Our study was unique in undertaking an examination of serum magnesium level in an ambulatory setting. Results from this study are consistent with prevalence rates noted by Mather and colleagues,⁶ who reported hypomagnesemia among 25% of patients with diabetes.

Although these findings should be considered suggestive and certainly worthy of further investigation, the observation of an overall 20% prevalence of hypomagnesemia among clinically stable outpatients was unanticipated. The prevalence of low magnesium levels noted in this study of ambulatory patients may have important clinical implications. It is possible that magnesium supplementation may improve clinical status among these patients with hypomagnesemia. It is uncertain whether low magnesium levels are reflective of poor overall health status among these subjects, or result from chronic diseases, poor nutritional status, a combination of these factors, or from other factors.

It is known that low magnesium results in a variety of nonspecific clinical manifestations. Unfortunately, there is no distinct constellation of features associated with a low magnesium state, and for this reason low magnesium levels are not clinically obvious. It is likely that complex interrelationships exist among various clinical and temporal factors that relate to magnesium homeostasis.

Limitations

Since this represented an exploratory study of 120 patients, statistical power is limited, as evidenced by broad confidence intervals for calculated odds ratios. Although several of the results in this study fail to achieve statistical significance, they may have clinical relevance. Two major nutritional and epidemiologic studies have shown decreased magnesium intake among African Americans.^3,10 Moreover, numerous smaller studies of magnesium supplementation report significant improvements in blood pressure measurements.^11-16 The American Diabetes Association has issued a consensus statement recommending that magnesium be replaced if patients with diabetes are proved to have low levels of serum magnesium.¹⁷