High Serum Iron Is Associated with Increased Cancer Risk

Although interest in linking high body iron with cancer risk in humans has spanned several decades, results from epidemiologic studies have been inconclusive. The most recent study from Sweden, with a sample size of more than 200,000, reported no association between serum iron and cancer risks at most sites (1). In earlier studies, both negative (2–5) and positive associations (6–8) have been described. Even among those studies reporting a positive correlation, the observed risk varied by cancer type and by gender (6–8). A recent clinical trial conducted at VA hospitals in the United States was a milestone study, in which a reduction of serum iron levels attained through phlebotomy was positively correlated with decreased cancer risk (9). The overall trend and the positive result reported were encouraging, but the study suffered from limited statistical power. The trial could only be considered a pilot because the expected iron–cancer dose–response relationship was not observed and consistent cancer types were not reported.

Though epidemiologic studies have yielded mixed results, there is a well-known iron–cancer relationship among patients with hemochromatosis, who are prone to develop liver cancer (10–12). As the basic mechanism related to iron-induced carcinogenesis seems well established (13–17), more definitive epidemiologic studies are needed. Using the serum iron data available from a large cohort of more than 300,000 adults in Taiwan, we prospectively assessed whether high serum iron was associated with increases in cancer occurrences or in cancer-related deaths among initially cancer-free subjects.

The study cohort consisted of 309,443 adults (145,088 men and 164,355 women) who were at least 20 years of age and had no history of cancer at the time of recruitment. The subjects participated in a standardized medical screening program between 1997 and 2008, with a median follow-up period of 7.07 years. The self-paying screening program was run by a private firm (MJ Health Management Institution, Taiwan) that attracted paying participants from all over Taiwan because of its known quality services, operational efficiency, and accessible key facilities. The majority of cohort members would come back for repeated examinations in subsequent years, but only test results from the initial examination were used in our analysis. A detailed description of the cohort has been documented elsewhere (18).

Each subject completed a self-administered questionnaire with medical history, lifestyle, and demographic information. Regarding smoking status, individuals were classified as nonsmokers, current smokers, or ex-smokers. Regular drinkers were those consuming two or more alcoholic beverages at least three times per week. Leisure time physical activity (LTPA) volume, the product of intensity (metabolic equivalent, MET) and duration of exercise (hours) were classified into three levels: inactive (<3.75 MET h/wk), somewhat active (15–29 min/d or 3.75–7.49 MET h/wk), and fully active (≥30 min/d or ≥7.5 MET h/wk), with the fully active category comprised of individuals who met the current LTPA recommendation (19). Anemia was defined as a hemoglobin level below 13 g/dL for men and below 12 g/dL for women. HBV (+) subjects were individuals with a positive hepatitis B surface antigen test and HCV (+) subjects were individuals with a positive hepatitis C antibody test.

Serum iron was measured by a Nitroso-PSAP method, which used acid to dissociate the Fe(III)–transferrin complex, and Fe(II) was then assessed with an Abbott Architect C8000 automatic biochemistry analyzer.

Informed consent was obtained to authorize data processing and analysis. Ethical reviews were approved by the Institutional Review Boards at the National Health Research Institutes (Taiwan). Individually identifying data were removed and remained anonymous during the entire study process.

Using the unique national identification numbers, subjects were each matched with the National Cancer Registry and National Death File between 1997 and 2008. A total of 8,060 cancer incidents and 3,066 cancer-related deaths were identified.

Cox proportional hazard analysis was used to calculate HRs. The HRs were adjusted for 10 confounding variables, including six continuous and four categorical variables. The continuous confounding variables were age, BMI, systolic blood pressure, total cholesterol level, C-reactive protein level, and hemoglobin level; the categorical variables were gender, smoking (nonsmoker, ex-smoker, and current smoker), drinking (never or occasional drinker, regular drinker), physical activity (inactive, somewhat active, and fully active), HBV (hepatitis B surface antigen carrier status), and HCV (hepatitis C antibody carrier status). Attributable fraction (AF) estimates the proportion of cancer cases, which could be avoided if high serum iron were eliminated. AF = (R₁–R₀)/R₁ = (RR–1)/RR, where R₁ is the cancer incidence among high serum iron, R₀ is the cancer incidence among of the reference group, and RR (relative risk) = R₁/R₀. In this study, AF was estimated as (HR-1)/HR. Relative excess risk due to interaction (RERI) was used to determine whether the combined effect of two exposures, high serum iron and each risk factor, was larger than the sum of the single effects from each exposure as a reflection of synergistic interaction. The value of RERI > 0 indicates a positive synergy, RERI = 0, no synergy, and RERI < 0 antagonistic interaction (20).

All statistical tests were two-sided, with the alpha level set at 0.05. Analyses were performed with SAS, version 9.2.

One quarter of the cohort (26.2%) had high serum iron, defined as ≥120 μg/dL (Table 1). Serum iron levels were consistently higher in males than in females (means of 109 and 88 μg/dL, respectively), regardless of age. By adjusting age and education levels of the cohort to those of Taiwan, the national prevalence of high serum iron was estimated. It was found that nationally, twice as many males as females had high serum iron (35% and 19%, respectively; Fig. 1), a figure similar to that observed in this cohort.

We observed a dose-dependent relationship between high serum iron and cancer, with a 4% increase in all cancer incidence risk for each 10 μg/dL increment of serum iron above 80 μg/dL (Fig. 2 and Supplementary Fig. S1). On the other hand, a serum iron level below 60 μg/dL also conferred a significantly increased cancer risk [HR, 1.18; 95% confidence interval (CI), 1.08–1.29]. As a result, when using 60 to 79 μg/dL as a reference level, the plot of cancer risk against serum iron revealed a J-shaped relationship with two major increases in risk observed, one above 80 μg/dL and one below 60 μg/dL.

Subjects with high serum iron (≥120 μg/dL) had a 25% increase in cancer incidence risk (HR, 1.25; 95% CI, 1.16–1.35; Table 2) and a 39% increase in cancer-related mortality risk (HR, 1.39; 95% CI, 1.23–1.57; Supplementary Table S2). There was a 2.49-fold increase in liver cancer incidence (HR: 2.49, 95% CI: 1.97–3.16) for males and females combined who had high serum iron and a 31% increase in breast cancer incidence (HR, 1.31; 95% CI, 1.01–1.70) for females with serum iron ≥140 μg/dL. Although the results were adjusted for HBV and HCV, which are the two major risk factors for liver cancer, additional analyses were conducted on sub-cohorts stratified by HBV status, with all HCV (+) subjects excluded. With HBV (–) and HCV (–) subjects shown in Table 3, their liver cancer risk was increased by approximately 3-fold (HR, 3.19; 95% CI, 2.23–4.57). Compared with the HBV (–) population with a serum iron level of 60 to 79 μg/dL, cancer risks were increased among subjects either with high serum iron alone (HR, 1.14 for all cancer and 3.19 for liver cancer) or with HBV alone (HR, 1.33 for all cancer and 12.14 for liver cancer). However, a synergistic increase was found when both risks were present (HR, 2.12 for all cancer and 24.38 for liver cancer; Table 3).

A forest plot of the data in Fig. 3 shows that the increase in cancer risk was significant regardless of gender, smoking status, drinking, physical activity, or HBV infection. In addition, the increased risk was still observed when the first 5 years of cancer cases were excluded (Fig. 3).

Fig. 4 shows the cancer risk of smoking, drinking, and inactivity individually and in combination with high serum iron in a multivariate analysis. The effect of smoking, drinking, or inactivity combined with high serum iron produced a roughly 50% extra-increase in the already-elevated cancer risks, from 1.13 to 1.74 for smoking (RERI, 0.27; 95% CI, 0.12–0.42), 1.14 to 1.61 for drinking (RERI, 0.24; 95% CI, 0.08–0.39), and 1.17 to 1.85 in combined risks (RERI, 0.36; 95% CI, 0.10–0.61). These interactions between these risks and high serum iron were synergistic and statistically significant (RERI > 0 and the 95% CI did not contain 0).

Given the increased HR of 1.25 for those with high serum iron, the AF from high serum iron was 20%, based on the formula AF = (HR-1)/HR.

In this study, high serum iron is found to be a common disorder and a strong marker for cancer incidence or mortality. The relationship between serum iron and cancer risk was a J-shaped one, with higher cancer risk at both ends, at lower than 60 μg/dL and higher than 120 μg/dL (Fig. 2). At higher end, there was a dose–response relationship with cancer risk increased by 4% for every 10 μg/dL increment above 80 μg/dL with 60 to 79 μg/dL as a reference level. For serum iron above 120 μg/dL, a 25% increase in cancer incidence and 39% increase in cancer-related mortality was found. At even higher serum iron levels (above 140 μg/dL), cancer risk increased by 37% (HR, 1.37; 95% CI, 1.26–1.50) for both genders combined and by 46% (HR, 1.46; 95% CI, 1.30–1.65) for males. This increase is a magnitude that is comparable with or higher than the increases in cancer risk associated with lifestyle factors such as tobacco smoking (HR, 1.67; ref. 21), obesity (HR, 1.14; ref. 22), or physical inactivity (HR, 1.20; ref. 19) in this study cohort.

Under the definition of high serum iron as ≥120 μg/dL, we found that as many as one third of males (35%) and one fifth of females (18%) fell into this classification. This is a large proportion of the general population with increased cancer risk. To assess whether this high prevalence is unique to Taiwan, we compared the numbers from Taiwan with those from the United States. The U.S. data came from the National Health and Nutrition Examination Survey (NHANES) 2001 to 2006 (23). The U.S. prevalence of high serum iron was 21% for males and 14% for females, which are figures that are substantially lower than those observed in Taiwan. Nevertheless, the proportion of individuals with high serum iron in the United States (1/5 males and 1/7 females) was still remarkable.

To our knowledge, this study is the first large population study to epidemiologically show the association of high serum iron with increased risk for all cancers combined and for liver cancer and breast cancer specifically. The same association was observed for selected nonliver cancers, which were defined as cases from all sites excluding cases of liver cancer, stomach cancer, colorectal cancer, and pancreatic cancer from all cancer cases. When these sites with excesses were combined, we found the overall excess risk to be statistically significant. It is evident that the positive association between serum iron levels and cancer risk was not limited to liver cancer. We also conducted a number of sensitivity analyses to strengthen our findings. We separated smokers from nonsmokers as well as HBV carriers from noncarriers, and we found that the increased cancer risk from high serum iron persisted in all groups (Table 3 and Supplementary Table S1). We also excluded cancer cases that occurred in the first 5 years and found that the association was still observed (Fig. 2). The fact that as many as 10 variables were controlled for in our Cox model lends further support to the independent effect of high serum iron on cancer risk. We also analyzed cancer-related mortality and found results similar to those for cancer incidence (Supplementary Table S2). Our results support the notion that body iron plays an important role in human carcinogenesis (14, 16, 17).

The role of iron overload in liver cancer development has been well documented for genetic hemochromatosis (10–12) and nonhemochromatosis (24) as well as for HBV (+) conditions (25, 26). However, we are the first to report a definite association between high serum iron and a large increase in risk of liver cancer in a population at average risk. Although hemochromatosis is associated with increased liver cancer risk, the prevalence of hemochromatosis in Taiwan is very low, approximately 0.3% to 0.5% (27, 28). This prevalence translates into 3 to 5 cases per thousand individuals among Caucasians, with estimates being lower for Asians (29, 30). Especially when compared with the 18% to 35% of subjects with high serum iron in this study, the prevalence of hemochromatosis is so low that its effect on the development of liver cancer could be ignored in this large prospectively followed cohort.

The association between high serum iron and elevated liver cancer risk remained regardless of HBV or HCV infection, but a strong synergistic interaction between high serum iron and HBV (+) existed when both were present. The increased risk for liver cancer (HR, 24.38; 95% CI, 17.14–34.67) conferred by high serum iron and HBV (+) acting together far exceeded the sum of the increased risks from high serum iron (HR, 3.19; 95% CI, 2.23–4.57) and HBV [(+) (HR, 12.14; 95% CI, 7.98–18.46)] acting alone (Table 3). Incidentally, the prevalence of HBV in this cohort was high—17% for males and 12% for females—though this is commonly observed in Asian countries, including Taiwan (31, 32). The significant increase in cancer risk associated with high serum iron in the non-HBV carrier cohort supports the idea that high serum iron plays a role in increasing the liver cancer and all cancer risks, which are independent of HBV status.

In our study, we found a significant 31% increase in breast cancer risk among women with serum iron levels higher than 140 μg/dL. This finding is consistent with the prevailing hypothesis that iron overload was known to increase breast cancer risk (33, 34). Epidemiologically, this finding of a breast cancer risk increase in women before or after menopause is new, as the existing studies have focused mainly on older age groups (1, 35). Because high serum iron is a risk factor for breast cancer, oral iron supplementation for women may be contraindicated. If our results are clinically confirmed, iron salts should be reconsidered as an ingredient in the vitamin pills regularly consumed by women without overt anemia.

The increased cancer risk associated with high serum iron was enhanced by three lifestyle risk factors: smoking, drinking, and inactivity (RERI, 0.36; 95% CI, 0.10–0.61). We used RERI to determine the presence of synergy between two or more risks. As each of these risks, high serum iron and each of the three life style risks, had the property of increasing cancer risk, one wonders to whether the combined effect was just a sum of the two or significantly larger than the sum. When found to be significantly larger, by the RERI test, the combination was said to have synergistic interaction. In this study, the presence of one of the three risk factors, along with high serum iron increased cancer risk over and above the sum of their individual risk increases (Fig. 4). A reasonable assumption from this result is that an effort to reduce some of these lifestyle factors could substantially reduce cancer risks. As there are limited options for reducing high serum iron, we believe that the pursuit of modifying life style risks should be considered for every patient. Of course, encouraging periodic blood donations may be another effective way to reduce cancer risk, as demonstrated in a pilot study (9).

We looked into the differences between the recent Swedish study (which found a null association) and our study, and we found that the selection of the reference group was critical. We observed an increased cancer risk (HR, 1.18; 95% CI, 1.08–1.29) among those with low serum iron. Colorectal or stomach cancer is known to be associated with iron deficiency anemia from chronic blood loss. As a result, the correlation of cancer risk with serum iron level was not linear but J-shaped, with the lowest serum iron level (<60 μg/dL) having a significantly increased cancer risk. As such, the lowest quartile of serum iron level should not be used as a reference group, as doing so could mask the increased cancer risk among those with higher serum iron (Fig. 2). A rough calculation using the second quartile of the Swedish study as a reference instead of the first quartile found that the cancer risk ratio in the highest quartile of serum iron level would have changed from 0.96, indicating no association or a slightly decreased risk, to 1.06, indicating a slightly increased risk. We could not determine the significance of this new ratio without access to the original data. However, this observation is compatible with a J-shaped association in the Swedish study.

There are limitations to this study. First, we focused our analysis on serum iron, one of several biomarkers of iron status, which may or may not correlate with body iron stores (4, 36). A recent meta-analysis of 59 epidemiologic studies concluded that biomarkers of iron stores tend to be negatively correlated with cancer risk (37). It should be noted that our study did not aim to assess the correlation of cancer risk with body iron stores, but with serum iron. Without adjusting for body iron stores, we were able to establish a link and a dose–response relationship between serum iron levels and the risk for cancer incidence and mortality. As serum iron is a relatively inexpensive and widely available test, tests can be ordered and high serum iron could be found and interpreted in daily practice, making our results clinically relevant. A second study limitation was that the average follow-up time from when serum iron was tested was 7 years, which could be viewed as too short a time for cancer to develop. Questions may arise to whether there was a sufficient length of time for the elevated serum iron to affect cancer development, as the mean follow-up time was approximately 7 years, with the longest being 14 years. When the prevalence of high serum iron was examined across age groups, we observed high serum iron levels at all ages, with higher ones in earlier years. A third limitation is that the conclusions from this study may not be applicable to non-Asians because the cohort came from Taiwan. However, similar results were found from several studies on Caucasians (6, 7), whereas reductions of cancer risks by phlebotomy in blood donors were reported from non-Asian countries in America and Europe (9, 38). In addition, the patterns of serum iron levels were grossly similar between this cohort and that of NHANES (23). A fourth limitation is that, we used single point measurements of serum iron, but serum iron levels can vary with time. In fact, diurnal variation of serum iron levels has been observed in a given day (39). We compared results for those who had two tests of serum iron level and found they were highly correlated, as evidenced from the way the values were distributed and the similarity of cancer risks they demonstrated, with HR at 1.25 (95% CI, 1.16–1.35) for the first test and 1.22 (95% CI, 1.09–1.36) for the second test, in the Supplementary Table S3. Thus, the use of one single test provided sufficiently stable information for cancer risk prediction. Although serum iron is continuously metabolized by the liver, it is replenished constantly and kept relatively stable. Other than blood loss, the body has no mechanism for excretion of excess iron, and iron homeostasis is closely regulated within each individual. A final limitation is that the cohort came from paying participants who are considered to be at higher socioeconomic status than the average Taiwanese population. However, regarding the prevalence of high serum iron, there are no data to indicate that higher socioeconomic status had any bearing on serum iron levels.

In summary, high serum iron (≥120 μg/dL), a common occurrence observed in one third of men and one fifth of women, was associated with increased risks of incidence and mortality from all cancers combined in this cohort, and a dose–response relationship was observed. Liver cancer and breast cancer were the specific sites of significant risk increases.

No potential conflicts of interest were disclosed.

Data in this research came from MJ Health Resource Center. Authorization Code: MJHRFB2014001C. The MJ Health Resource Foundation is responsible for the data distribution. Conclusions were those from authors and did not represent the views of MJ Health Resource Center.

Conception and design: C.P. Wen, D.P.H. Hsieh, C.Y. Hsu, X. Wu

Development of methodology: C.P. Wen, M.K. Tsai, C.Y. Hsu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.A. Hsiung

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.P. Wen, J.H. Lee, Y.-P. Tai, S.B. Wu, M.K. Tsai, D.P.H. Hsieh, C.Y. Hsu, X. Wu

Writing, review, and/or revision of the manuscript: C.P. Wen, C. Wen, M.K. Tsai, D.P.H. Hsieh, H.-C. Chiang, C.A. Hsiung, C.Y. Hsu, X. Wu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.P. Wen, M.K. Tsai, H.-C. Chiang, C.Y. Hsu

Study supervision: C.P. Wen, M.K. Tsai, D.P.H. Hsieh, H.-C. Chiang

This work was supported in part by the Taiwan Department of Health Clinical Trial and Research Center of Excellence (MOHW103-TDU-B-212-113002).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.