High serum magnesium is associated with lower risk of hepatocellular carcinoma among patients with nonalcoholic fatty liver disease

Abstract

Background:

Nonalcoholic fatty liver disease (NAFLD) is a major contributor to the rising incidence of hepatocellular carcinoma (HCC). Magnesium is a major cation in cellular activities. Epidemiological data on magnesium level and its relation to HCC are sparse. This study aimed to examine the associations between serum levels of magnesium and the risk of HCC among patients with NAFLD.

Methods:

A total of 26,053 patients with NAFLD were identified in the University of Pittsburgh Medical Center Electronic Health Records from 2004 through 2018. After an average of 5.15 years of follow‐up, 395 patients developed HCC after the first measurement of serum magnesium. Cox proportional hazards regression model was used to calculate hazard ratios (HRs) and 95% CIs of HCC incidence associated with quartile levels of serum magnesium after adjustment for age, sex, race, body mass index, diuretics use, history of type 2 diabetes, history of hypertension, history of hyperlipidemia, and tobacco smoking.

Results:

Patients with NAFLD who developed HCC had a significantly lower mean ($\pm$standard deviation) serum magnesium (0.769 $\pm$ 0.131 mmol/L) than those who remained free of HCC (0.789 $\pm$ 0.125 mmol/L;$p\hspace{0.17em}=\hspace{0.17em}.003$). Compared with the lowest quartile, the HRs (95% CIs) of HCC second, third, and fourth quartiles of serum magnesium were 0.87 (0.67–1.12), 0.77 (0.57–1.04), and 0.73 (0.56–0.96), respectively, after adjustment for multiple potential confounders ($P_{\text{trend}}\hspace{0.17em}=\hspace{0.17em}.02$).

Conclusion:

This finding suggests higher levels of serum magnesium were significantly associated with decreased risk of HCC among patients with NAFLD.

INTRODUCTION

Since the mid‐1980s, the mortality rate of primary liver cancer has doubled in the United States.¹ Globally, the projected number of deaths from primary liver cancer is 1 million in 2030.^2,3 The major subtype of primary liver cancer is hepatocellular carcinoma (HCC).⁴ Approximately 75% to 90% of primary liver cancer cases are classified as HCC.⁵ There are large geographical variations in HCC incidence rate: the lowest rate (<5.0 per 100,000 persons) has been observed in North America and North Europe, whereas the highest rate (>20.0 per 100,000 persons) is in sub‐Saharan Africa and Eastern Asia.⁶ Chronic infection with hepatitis B virus (HBV) and/or hepatitis C virus (HCV), heavy alcohol drinking, and cigarette smoking have been found to be major risk factors for HCC.^7,8 Nonalcoholic fatty liver disease (NAFLD), as one of the most common chronic liver conditions in the developed countries,⁹ is an emerging risk factor for HCC.¹⁰ NAFLD encompasses a wide spectrum of liver conditions from nonalcoholic fatty liver and nonalcoholic steatohepatitis to fibrosis and cirrhosis, which ultimately lead to the development of HCC.^11,12 The global prevalence of NAFLD is estimated to be 30% to 40% for men and 15% to 20% for women.¹³ NAFLD is expected to overtake HBV and HCV and become the leading cause of HCC by 2025.^14,15

Magnesium is the fourth most abundant cation in the human body¹⁶ and is critical in many cellular processes such as DNA replication and repair, ion transportation, and cellular proliferations.¹⁷ The major source of magnesium is food and water.¹⁸ Approximately 24% to 76% of the magnesium intake can be absorbed in the small or large intestine and the remaining magnesium is excreted in the feces.¹⁹ It has been estimated that up to 20% of the population in developed countries have magnesium deficiency,²⁰ likely because of low magnesium intake. Diuretic medications may cause a higher level of magnesium excretion.^21–23 Magnesium deficiency is often associated with liver disease.²⁴ A low level of magnesium in serum and liver tissue may lead to the progression of liver diseases because of the dysfunction of the mitochondrial, inflammatory response, defective protein kinase C translocation, or oxidative stress.²⁴ Recent studies have suggested that a positive feedback loop in which a lower magnesium concentration in cirrhosis patients may result in an improper translocation of protein kinase Cε (an enzyme that can regulate fibrinogen deposition in the liver), which further reduces cytoplasmic magnesium level in liver cells, increases collagen deposition and fibrinogen, and aggravates cirrhosis.^24–26 In the National Institutes of Health‐American Association of Retired Persons Diet and Health Study (NIH‐AARP), higher magnesium intake derived from the 24‐hour food recall was found to be associated with a lower risk of HCC.²⁷ However, there have been no epidemiological studies on serum levels of magnesium, which is more objective and biologically relevant, in relation to HCC risk. Previous reports found that serum magnesium level was lower in patients with cirrhosis and HCC than those without HCC in an Italian population.²⁸ However, this study was a cross‐sectional study design with a small sample size (total of 291 patients) and diverse underlying risk factors for HCC including HBV, HCV, and alcohol abuse. Therefore, the association between serum magnesium and HCC risk warrants further investigation in a large prospectively designed cohort study.

Using the electronic health records (EHRs) from a large health care system in the Commonwealth of Pennsylvania (Pennsylvania), USA, we investigated the association between serum levels of magnesium and the risk of HCC development in a large cohort of patients with NAFLD without a diagnosis of major underlying causes of HCC such as alcohol abuse and chronic infection with HBV and/or HCV.

MATERIALS AND METHODS

Study population

We conducted a retrospective cohort study among participants with NAFLD using data from the University of Pittsburgh Medical Center (UPMC) Health Insurance Plan (UPMC NAFLD Cohort Study). UPMC is a large regional healthcare provider including more than 40 hospitals located in western Pennsylvania that provide healthcare services to more than 3 million patients. The data of this study were requested by Health Record Research Requestion (R3), provided by the University of Pittsburgh Biomedical Informatics Services. We requested deidentified EHRs of all UPMC Health Plan participants who were aged 40 to 89 years when they received health care services in the UPMC Healthcare System from January 1, 2004, to December 31, 2018. The study was approved by the University of Pittsburgh Human Research Protection Office (https://www.hrpo.pitt.edu/) with internal review board approval number STUDY2103003 according to the Declaration of Helsinki Ethical Principles for Medical Research Involving Human Subjects. Because this study was based on the secondary analysis of the existing deidentified data set derived from EHRs of the University of Pittsburgh Medical Center, individual informed consent was exempted.

Inclusion and exclusion criteria for the NAFLD cohort

We used the International Classification of Diseases 9th and 10th Clinical Modifications (ICD‐9‐CM and ICD‐10‐CM) to define eligible patients in the UPMC NAFLD Cohort Study according to the most recent expert panel consensus statement for administrative coding in eletronic health record‐based research of NAFLD.²⁹ We included patients with nonalcoholic fatty liver, nonalcoholic steatohepatitis, and compensated and decompensated cirrhosis (see ICD codes in Table S1). We excluded patients who had alcoholic liver disease, alcohol use disorder, somatic consequences of alcohol, autoimmune liver disease, alpha‐1‐antitrypsin deficiency, secondary or unspecified biliary cirrhosis, drug use disorder except for nicotine/caffeine, hemochromatosis, Budd‐Chiari syndrome, viral hepatitis, unspecified chronic hepatitis, or Wilson disease (see ICD codes in Table S2). Overall, there were 47,165 patients with NAFLD.

The primary exposure variable for the present analysis was serum magnesium, whereas the primary outcome was HCC, which was defined by either ICD‐9‐CM code 155.0 or ICD‐10‐CM codes C22.0 and C22.8. Any patient with NAFLD without a magnesium measurement was excluded ($n\hspace{0.17em}=\hspace{0.17em}19,176$). In addition, we excluded 1936 patients whose earliest serum magnesium was measured less than or equal to 30 days before or after the HCC diagnosis or the last date of follow‐up to minimize the potential impact of diagnostic and medical procedures for HCC. The present analyses included 26,053 patients with NAFLD including 395 HCC cases (Figure 1).

FIGURE 1.

Flow chart for the construction of the final analytic data set for the UPMC NAFLD Cohort Study, 2004–2018. NAFLD indicates nonalcoholic fatty liver disease; UPMC, University of Pittsburgh Medical Center.

All covariates for the present analysis were also extracted from patients’ EHRs and included age, sex, race, body mass index (BMI), smoking status, and histories of type 2 diabetes, hyperlipidemia, and hypertension. Use of diuretics can increase the urinary excretion of magnesium and reduce its circulating level.^24,30 Thus, the duration of use of major diuretics including thiazide diuretics, loop diuretics, and potassium‐sparing diuretics before the magnesium measurement was a covariate for adjustment.

Statistical analysis

The ${\chi }^2$ and pooled two‐sample t‐tests were used to examine the difference in the distributions of categorical and continuous variables, respectively, between HCC cases and non‐HCC cases. Person‐years at risk for each patient were calculated from the date of magnesium measurement to the date of HCC diagnosis, death, or the last encounter with any UPMC Healthcare system facility, whichever occurred first. The Cox proportional hazards regression method was used to calculate hazard ratios and their 95% CIs for HCC development associated with higher quartiles of serum magnesium with adjustment for various covariates described above. The proportional hazard assumption was examined by using time‐dependent covariates in the model and no violation was found. A statistical test for linear trend was conducted based on the ordinal value of serum magnesium in quartiles (i.e., 1, 2, 3, and 4). A product term for serum magnesium (quartile of magnesium) with the diagnosis of type 2 diabetes mellitus, or use of diuretics or other selected risk factors was additionally included in the Cox regression to assess their potential modifying effect on the magnesium‐HCC risk association. Similar analyses were performed in the subgroups of patients stratified by BMI (<30 kg/m² vs. ≥30 kg/m²), use of diuretics (never, ever), or history of type 2 diabetes (no vs. yes). To minimize the potential impact of the underlying liver condition on serum magnesium level, we conducted stratified analysis for patients stratified by a widely used fibrosis score in the clinic: FIB‐4 ($\le 2.67$vs.$>2.67$), which was calculated as $\frac{Age\hspace{0.17em}(years)\times \hspace{0.17em}AST\hspace{0.17em}(U/L)}{Platelet\hspace{0.17em}Count\hspace{0.17em}({10}^9/L)\times \hspace{0.17em}\surd ALT\hspace{0.17em}(U/L)}.$All statistical analyses were conducted using SAS software, version 9.4 (SAS Institute, Cary, North Carolina). All $p$ values reported are two‐sided and $p$ values < .05 were considered statistically significant.

Availability of data and materials

The data sets generated and/or analyzed during the current study are not publicly available because of privacy or ethical restrictions but are available from the corresponding author upon reasonable request.

RESULTS

After an average of 5.15 years of follow‐up, 395 patients with NAFLD developed HCC of 26,053 patients with NAFLD. Patients with NAFLD who developed HCC had a higher mean age and a lower mean BMI than their counterparts without HCC (Table 1). After excluding patients with missing values, percentages of men, ever smokers, and those with diabetes were significantly higher, whereas the histories of hyperlipidemia and hypertension were significantly lower in HCC cases than non‐HCC cases. There was a higher percentage of fibrosis (i.e., FIB‐4 >2.67) in HCC cases than in non‐HCC cases after excluding those with missing values. Overall, approximately 12% of patients with NAFLD were prescribed with diuretics for 2 weeks or longer, but there was no difference in diuretic prescriptions between patients with NALFD and with HCC and without HCC.

TABLE 1.

Distribution of chrematistics, selected risk factors for HCC in subjects by the incident HCC status: the UPMC NAFLD Cohort Study, 2004–2018.

Characteristics	Incident HCC	Free of HCC	p a
Number of subjects	395	25,658
Age (years), mean $\pm$ SD	65.3 $\pm$ 10.9	60.0 $\pm$ 11.9	<.001
BMIa (kg/m2), mean $\pm$ SDb	31.8 $\pm$ 7.5	34.0 $\pm$ 8.0	<.001
Sex, n (%)
Male	200 (50.6)	10,400 (40.5)	<.001
Female	195 (49.4)	15,258 (59.5)
Race, n (%)
White	373 (94.4)	23,540 (91.7)	.054
Non‐White	22 (5.6)	2118 (8.3)
Cigarette smoking status, n (%)
Never smoker	134 (33.9)	10,885 (42.4)	<.001
Ever smokers	190 (48.1)	12,408 (48.4)
Missing	71 (18.0)	2365 (9.2)
History of type 2 diabetes, n (%)
No	257 (65.1)	17,271 (67.3)	.344
Yes	138 (34.9)	8387 (32.7)
History of hyperlipidemia, n (%)
No	253 (64.0)	13,710 (53.4)	<.001
Yes	142 (36.0)	11,948 (46.6)
History of hypertension, n (%)
No	157 (39.8)	8889 (34.6)	.035
Yes	238 (60.3)	16,769 (65.4)
Diuretic users
Never or using <14 days	356 (90.1)	22,555 (87.9)	.389
Using 14–364 days	17 (4.3)	1275 (5.0)
Using ≥365 days	22 (5.6)	1828 (7.1)
FIB‐4 score (age <65/age ≥65 years)
<1.30/<2.00	108 (27.3)	14,710 (57.3)	<.001
1.30–2.67/2.00–2.67	75 (19.0)	4894 (19.1)
>2.67/2.67	128 (32.4)	3495 (13.6)
Missing	84 (21.3)	2559 (10.0)

Abbreviations: BMI, body mass index; FIB‐4, fibrosis‐4 score; HCC, hepatocellular carcinoma; NAFLD, nonalcoholic fatty liver disease; SD, standard deviation; UPMC, University of Pittsburgh Medical Center.

^aDerived from the t‐test (for continuous variables) or ${\chi }^2$ test (categorical or nominal variables).

^bA total of 1675 participants’ BMI values were missing.

The mean (standard deviation) serum magnesium concentration in HCC cases was 0.769 (0.131) mmol/L, which was significantly lower than controls at 0.789 (0.125) ($p\hspace{0.17em}=\hspace{0.17em}\hspace{0.17em}.003$). Compared with the lowest quartile, the hazard ratios (95% CIs) of HCC for the second, third, and fourth quartiles of serum magnesium were 0.87 (0.67–1.12), 0.77 (0.57–1.04), and 0.73 (0.56–0.96), respectively ($p_{\text{trend}}\hspace{0.17em}=\hspace{0.17em}.020$) (Table 2). There was a statistically significant 12% reduction in HCC risk (hazard ratio, 0.88; 95% CI, 0.80–0.97) per one‐standard deviation magnesium increment ($p_{\text{trend}}\hspace{0.17em}=\hspace{0.17em}.11$).

TABLE 2.

The association between serum magnesium and risk of developing HCC: the UPMC NAFLD Cohort Study, 2004–2018.

Serum Magnesium (Range in mmol/L)	Total no. of subjects	Total no. of person‐years	Incident HCC cases	HR (95% CI)a	p
Quartile 1 (<0.700)	6441	33,668.2	120	1.00
Quartile 2 (0.700–0.782)	7176	38,886.3	112	0.87 (0.67–1.12)	.279
Quartile 3 (0.782–0.864)	4670	25,541.3	66	0.77 (0.57–1.04)	.089
Quartile 4 (>0.864)	7766	35,962.6	97	0.73 (0.56–0.96)	.026
p for trend					.020
Continuous (per SD)				0.88 (0.80–0.97)	.011

Abbreviations: HCC, hepatocellular carcinoma; HR, hazard ratio; NAFLD, nonalcoholic fatty liver disease; SD, standard deviation; UPMC, University of Pittsburgh Medical Center.

^aAdjusted for age (years), sex, race, body mass index (kg/m²), history of type 2 diabetes, history of hypertension, history of hyperlipidemia, cigarette smoking status, and diuretic use.

We further conducted analyses on subgroups of patients with NAFLD stratified by categories of BMI, FIB‐4 scores, use of diuretics, and history of type 2 diabetes. The inverse associations between serum magnesium quartiles and HCC risk in all subgroups except for BMI < 30 kg/m² were not statistically significant, primarily because of a small sample size with fewer number of HCC cases within each subgroup (Table 3). On the other hand, we did not detect any statistically significant heterogeneity in the magnesium‐HCC risk association between the two subgroups by BMI, FIB‐4, use of diuretics, or type 2 diabetes.

TABLE 3.

The association between serum magnesium and risk of developing HCC, stratified by BMI levels, FIB‐4 score, diuretic users, and history of type 2 diabetes: the UPMC NAFLD Cohort Study, 2004–2018.^a

		Q1	Q2	Q3	Q4	pTrend	pInteraction
BMI < 30	No. cases	45	44	30	37	.039	.706
	HR (95% CI)	1.00	0.86 (0.57–1.31)	0.84 (0.52–1.33)	0.62 (0.40–0.97)
BMI ≥ 30	No. Cases	59	51	25	45
	HR (95% CI)	1.00	0.83 (0.57–1.21)	0.65 (0.40–1.04)	0.81 (0.55–1.20)	.207
FIB‐4 ≤ 2.67	No. Cases	50	53	34	46
	HR (95% CI)	1.00	0.9 5 (0.64–1.40)	0.89 (0.57–1.38)	0.82 (0.55–1.24)	.327	.611
FIB‐4 > 2.67	No. Cases	42	35	24	27
	HR (95% CI)	1.00	0.87 (0.55–1.37)	1.00 (0.60–1.67)	0.64 (0.39–1.05)	.119
Diuretic never users	No. Cases	97	96	55	80
	HR (95% CI)	1.00	0.90 (0.68–1.19)	0.77 (0.55–1.07)	0.75 (0.55–1.01)	.037	.869
Diuretic ever users	No. Cases	23	16	11	17
	HR (95% CI)	1.00	0.74 (0.39–1.40)	0.81 (0.39–1.68)	0.69 (0.37–1.31)	.309
No history of T2DM	No. Cases	67	71	52	67
	HR (95% CI)	1.00	0.83 (0.59–1.16)	0.83 (0.58–1.19)	0.71 (0.51–1.00)	.065	.922
T2DM history	No. Cases	53	41	14	30
	HR (95% CI)	1.00	0.96 (0.64–1.45)	0.61 (0.34–1.10)	0.80 (0.51–1.26)	.179

Abbreviations: BMI, body mass index; FIB‐4, fibrosis‐4 score; HCC, hepatocellular carcinoma; NAFLD, nonalcoholic fatty liver disease; T2DM, type 2 diabetes; UPMC, University of Pittsburgh Medical Center.

^aAdjusted for sex, age (years), race, BMI (kg/m²), history of T2DM, history of hypertension, history of hyperlipidemia, cigarette smoking status, and diuretic use, if applicable. Subjects who had missing information on BMI (total$n\hspace{0.17em}=\hspace{0.17em}1675$) and FIB‐4 score (total$n\hspace{0.17em}=\hspace{0.17em}2643$) were excluded from the subgroup analyses.

DISCUSSION

To our knowledge, this was the first study that examined serum magnesium levels and the risk of HCC in a prospective cohort of patients with NAFLD. The present study demonstrated that higher serum magnesium levels measured at least 30 days before HCC diagnosis was significantly associated with a lower risk of HCC in a large cohort of patients with NAFLD in a large healthcare system in the Commonwealth of Pennsylvania. Patients with NAFLD in the highest quartile of serum magnesium (>0.864 mmol/L) experienced a statistically significant 27% (95% CI, −44 to −4) risk reduction in HCC incidence.

In recent years, NAFLD has become a major and fastest‐growing chronic liver disease in the United States and other developed nations, in which it emerges as the main underlying cause of HCC.^31,32 With a projected dramatic increase in NAFLD‐related HCC cases,³² there is an urgent need to identify novel markers that can be used for surveillance and risk reduction in HCC development for this particular population. Magnesium deficiency is often found in patients with advanced liver diseases such as liver fibrosis and cirrhosis.²⁴ For example, magnesium level has been found low in patients with liver cirrhosis, likely from the decreased synthesis of albumin, which is a critical element in the body’s magnesium supply chain.²⁴ At the same time, magnesium deficiency has a negative impact and exacerbates the progression of liver disease, likely because of its adverse impact on the mitochondrial bioenergetic process, where an impaired mitochondrial function would reduce the ATP production in the hepatocytes, increase oxidative stress in the liver, and further enhance liver damage.^24,33,34

A number of large prospective cohort studies have examined the association between dietary magnesium intake and risk of various diseases. A recent meta‐analysis including a total of more than 1 million individuals and more than 52,000 deaths found that dietary magnesium intake was associated with a lower risk of all‐cause and cancer mortality, but not with cardiovascular disease mortality.³⁵ The US National Health and Nutrition Examination Survey III including 12,504 patients who completed liver ultrasound examination for hepatic steatosis found that every 100‐mg increase in intake of magnesium was associated with a 49% reduction in the risk of mortality from liver diseases ($p\hspace{0.17em}=\hspace{0.17em}\hspace{0.17em}.05$).³⁶ In the NIH‐AARP cohort study that included 536,359 participants with a median follow‐up time greater than 15 years, the highest quartile of magnesium intake (499 $\pm$ 131 mg/day) was associated with a statistically significant 35% lower risk of HCC incidence.²⁷ Epidemiological studies on the association between serum magnesium and HCC risk are limited. A case‐control study of 130 HCC cases and 161 non‐HCC cases within a cohort of cirrhotic patients because of all causes in Italy reported that cirrhotic patients with HCC had significantly lower levels of serum magnesium than patients with cirrhosis without HCC, suggesting that magnesium might offer certain protection against the development of HCC in patients with cirrhosis.²⁸ In general, our results derived from a study with prospective design and large sample size were consistent with those in previous studies as described previously.

Our study had several strengths. We retrospectively constructed a cohort of patients with clinical diagnosis of NAFLD, on whom serum magnesium concentration was determined before HCC diagnosis. The cohort consisting of NAFLD patients only eliminated a potential confounding effect of different underlying etiologies of liver diseases such as alcoholic, autoimmune, and viral causes on the observed magnesium‐HCC risk association. A large sample size (>26,000 subjects) with a relatively long follow‐up (>5 years from the measurement of magnesium to HCC diagnosis), yielding 395 incident cases of HCC, provided sufficient statistical power for robust estimates of the association between serum magnesium level and HCC risk.

We also note several limitations in our study. The present analysis was based on the existing data derived from patients’ EHRs. We excluded approximately 40% of total patients with NAFLD without a serum magnesium measurement from the present analysis. The potential selection bias is unknown in the observed results. The duration of follow‐up (5.15 years from magnesium measurement to the censoring date or HCC diagnosis) might not be long enough to observe all potential cases of HCC within the study population. In addition, we do not know when and how many study subjects left UPMC Healthcare System. A single‐time measurement of magnesium was used in the analysis, which might not represent the long‐term in vivo level of magnesium for a given individual. Nevertheless, the random variation in exposure measurement (i.e., magnesium concentration) usually results in an underestimate of the association between exposure and outcome (i.e., HCC risk) toward the null. Missing or underreporting of certain variables such as cigarette smoking and low levels of alcohol consumption might result in somewhat residual confounding on the observed magnesium‐HCC risk association, although we excluded patients with heavy alcohol use (i.e., alcoholic liver disease and alcohol use disorder) from the present study.³⁷ Last, although most commonly used, less than 1% of the total human body magnesium pool is in the circulation, whereas more than 50% is stored in the bone and the remaining in other tissues.³⁸ A low level of serum magnesium could be the result of low nutrient intake. The major food sources of magnesium are legumes and seafood. If our findings are confirmed in future studies, dietary supplementation of magnesium may provide a potential means for the prevention of the NAFLD progression to HCC.

In summary, the present study demonstrates that a high level of serum magnesium was significantly associated with decreased risk of HCC in a cohort of patients with NAFLD within a large healthcare system in Western Pennsylvania. Our findings provide additional evidence in support of the beneficial role of in vivo magnesium, over dietary magnesium intake, against the development of HCC in patients with NAFLD. Future studies are warranted to elucidate the biological mechanism for the protective role of magnesium on liver health before clinical intervention with magnesium could be implemented for patients with NAFLD.