Experience suggests that aggregations of cancer incidences on a small-area scale may provide clues to identify environmental or lifestyle risk factors. As what is expected, it is important to take notice of fluctuations in diagnosis, reporting, survival time, migration, and mobility. Since every sociodemographical event including birth, death, and migration is mandatorily registered in Taiwan, the demographical data reported by the Household Registration Offices in Taiwan are essentially complete and accurate. In Taiwan, it is mandatory to submit death certificates to household registration offices, the death certificate registry in Taiwan is basically complete. We showed the maps of ranks township-specific age-adjusted mortality rate. We found that the age-adjusted mortality rates were significantly higher in townships on the southwestern coast than those in other townships. The difference of mortality rates between southwestern coastal townships and other townships could not merely result from a variation in survival times because the overall five-year survival rate for liver cancer in Taiwan was 15% during 1987–1992 . Although the association between concentrations of iron in groundwater and HCC mortality rate did not reach statistical significance for females (P = 0.058), a statistical significance did exist for both genders in the correlation coefficient between concentrations of iron in groundwater and HCC incidence rates. The correlation coefficients were less affected (in comparison with those for mortality data) by the adjustment of land subsidence. In particular, after adjusting for land subsidence, the correlation coefficient for mortality in women was not significant and very close to the null value (0.128 only). On the other hand, after adjusting for land subsidence, the correlation coefficients for incidence data had less reduction in the absolute values and remained statistically significant. It seems that the incidence data had more stable correlations with the iron levels.
Why did the most striking cluster of mortality ASRs of HCC in males occur in the southwestern coastal regions (Changhua, Yunlin, Chiayi, and Tainan county) of Taiwan? A series of epidemiological studies on major risk factors for liver cancer in Taiwan have been carried out. Both HBV and HCV infections are major risk factors for HCC in Taiwan [7, 8, 19–24]. On the other hand, previous studies have demonstrated that the mortality rate of liver cancer is strikingly high in the Matzu islets (located near the north coast of Fukien Province of mainland China and separated from Taiwan Island by the Taiwan Strait), Penghu islets, and the endemic area of arseniasis in southwestern Taiwan; however, the seroprevalence of HBsAg among the residents of these areas was similar or slightly higher than that among general Taiwanese population [19, 25–27]. Moreover, the seroprevalence of anti-HCV antibodies in Matzu was even lower than that among general Taiwanese population . These results suggest that major risk factors in addition to HBV and HCV infections can be involved in the etiology of liver cancer, and those factors are responsible for the high liver cancer mortality in the Matzu and Penghu islets and in southwestern Taiwan. Chen et al.  revealed a geographic variation in HBV and HCV seroprevalence in Taiwan from a large-scale survey of free hepatitis screening participants. It was found that the highest anti-HCV positive rates were in Miaoli County, Chiayi County, Chiayi City, and Yunlin County while the highest HBsAg positive rates were in Keelung City and Yilan City. That is, the areas with higher HBV or HCV prevalence rate are not clustering at the areas with land subsidence, but the higher HCC mortality rates were clustered at the land subsidence. These indicate that HBV and HCV infection aren’t the only risk factors that influence the incidence of HCC.
What is the possible major risk factor besides HBV and HCV infections that involved in the etiology of liver cancer and is responsible for the most striking geographical cluster observed in southwestern Taiwan? The fact that the incidence of HCC varies greatly around the world suggests the involvement of environmental etiological factors. Some studies showed that excess uptake of iron is possibly involved in carcinogenesis [29, 30]. The experimental studies have also supported the hypothesis that iron may facilitate the development of HCC in cirrhotic and non-cirrhotic patients . Furthermore, long-term treatment of patients with chronic HCV infection with iron-reducing agents results in a significant improvement of liver function and marked reduction in the progression to HCC [32, 33]. In another study , iron deposition was found in 43% of the HCC cases and indicated that, in Myanmar patients with HCC, iron deposition might accelerate hepatocarcinogenesis, by promoting cancer cell proliferation, without affecting the Fas/FasL apoptotic system. A recently study  showed that chronic iron overload seemed to induce nuclear localization of MT and NF-jB activation, with a resultant marked acceleration of liver regeneration in ironoverload rats after PH occurring at least 12 h earlier than that in normal-diet rats. The accelerated liver regeneration was likely due to the shortening of G0–G1 transition, possibly by bypassing the signaling cascades required for the induction of hepatocyte proliferation. This accelerated response by chronic iron overload could be involved in the promotion of hepatocellular carcinogenesis under infection with hepatitis viruses. From epidemiology viewpoint, the interaction between human uptake of overload iron and HBV/HCV infection may play an important role on the incidence of HCC which worths a further study.
Groundwater is a major source of water supply in southwestern coastal Taiwan due to the shortage of surface water mainly obtained from a few creeks and rivers. The over-pumping of groundwater for aquaculture has caused serious land subsidence in southwestern coastal Taiwan . Serious land subsidence areas are located in the counties of Changhua and Yunlin, the southwestern part of the alluvial fan of Chou-Shui River (Figure 3), where the supply rates of tap water from surface water in Changhua and Yunlin are the lowest among all the counties in Taiwan (less than 36.95% and 52.51% in 1975 and 1981 respectively). On the other hand, the iron concentrations of untreated tap water pumped from groundwater in “serious land subsidence areas” and “no serious land subsidence areas” were found to be 1.04 (± 0.20) and 0.34 (± 0.05) mg/l respectively (Table 1). According to a WHO report, the concentrations of iron in drinking water are normally less than 0.3 mg/l . The above statements suggest that the residents of the serious land subsidence areas have been consuming a large quantity of groundwater with high iron concentration for drinking and water supplies for aquaculture and fishponds over several decades. Table 1 shows that, in both males and females, the average ASRs of HCC among the residents of serious land subsidence areas were significantly higher than ASRs among areas of no serious land subsidence.
Unlike arsenic, a confirmed human carcinogen, the carcinogenicity of iron remains debatable. However, increasing evidence summarized below shows that iron overload can contribute to cancer development either as an initiator or a promoter . The toxicity of iron is largely based on Fenton and Haber-Weiss chemistry wherein catalytic amounts of iron are sufficient to yield hydroxyl radicals (OH.) from superoxide (O2
.–) and hydrogen peroxide (H2O2), collectively known as “reactive oxygen intermediates” (ROIs). Importantly, ROIs are inevitable byproducts of aerobic respiration and are generated by incomplete reduction of dioxygen in mitochondria [38, 39].
Hepatocellular carcinoma patients may have higher iron stores than those who do not develop HCC . Iron can catalyze the production of oxygen radicals that may be proximate carcinogens [39, 41]. The results of a study conducted in Taiwan by Stevens et al.  were consistent with the hypothesis that increased iron stores increase the risk of primary hepatocellular carcinoma. Moreover, serum ferritin was thought to be related to body iron stores [43, 44]. The serum ferritin concentration, which is widely used clinically as an index of body iron stores, was predominantly varying by sex, blood donation, and age. The serum ferritin concentration tends to be lower among women and regular blood donors [45, 46], and it can also be expected to run parallel with physiological changes of human body. For example, female serum ferritin is increased before menopause to after menopause , but the concentration in females remains lower than that of men until after menopause [47, 48]. To sum, this difference in body iron stores is responsible for the gender difference in HCC mortality rate clustering in southwestern Taiwan.
To the best of our knowledge, this study was the first ecologic epidemiology research to examine the association of HCC mortality with the concentration of iron in groundwater in Taiwan. Therefore, there is not yet any evidence to support the correlation of iron deposit in HCC patients with the concentration of iron in groundwater would be indicated in the previous studies. In an excellent work from Soe et al.  in Japan, Prussian blue staining was performed in order to detect iron deposition, and the iron deposition was found in 43% of the HCC cases. The study showed that, in Myanmar patients with HCC, iron deposition might accelerate hepatocarcinogenesis. We believe that iron overload in human consumption through diet pathway such as by groundwater may be a potential and important risk factor for etiologic investigations of HCC. Iron can catalyze the production of oxygen radicals that may be proximate carcinogens. Moreover, iron may play a role of limiting nutrient for the growth and replication of cancer cells. Therefore, our study may provide the clue to investigate the causal role of exogenous iron from groundwater in the etiology of human liver carcinogenesis that cannot be ruled out and warrants further investigation by using Prussian blue staining or iron contents in biopsy liver specimens, or etc.