Prevalence of non-insulin-dependent diabetes mellitus and related vascular diseases in southwestern arseniasis-endemic and nonendemic areas in Taiwan.

There is evidence indicating that ingestion of arsenic may predispose the development of diabetes mellitus in arsenic-endemic areas in Taiwan. However, the prevalence of diabetes and related vascular diseases in the entire southwestern arseniasis-endemic and nonendemic areas remains to be elucidated. We used the National Health Insurance Database for 1999-2000 to derive the prevalence of non-insulin-dependent diabetes and related vascular diseases by age and sex among residents in southwestern arseniasis-endemic and nonendemic areas in Taiwan. The study included 66,667 residents living in endemic areas and 639,667 in nonendemic areas, all greater than or equal to 25 years of age. The status of diabetes and vascular diseases was ascertained through disease diagnosis and treatment prescription included in the reimbursement claims of clinics and hospitals. The prevalence of non-insulin-dependent diabetes, age- and gender-adjusted to the general population in Taiwan, was 7.5% (95% confidence interval, 7.4-7.7%) in the arseniasis-endemic areas and 3.5% (3.5-3.6%) in the nonendemic areas. Among both diabetics and nondiabetics, higher prevalence of microvascular and macrovascular diseases was observed in arseniasis-endemic than in the nonendemic areas. Age- and gender-adjusted prevalence of microvascular disease in endemic and nonendemic areas was 20.0% and 6.0%, respectively, for diabetics, and 8.6% and 1.0%, respectively, for nondiabetics. The corresponding prevalence of macrovascular disease was 25.3% and 13.7% for diabetics, and 12.3% and 5.5% for nondiabetics. Arsenic has been suggested to increase the risk of non-insulin-dependent diabetes mellitus and its related micro- and macrovascular diseases.

Arsenic is a ubiquitous element in the environment, with metalloid properties. In many parts of the world, arsenic is present in drinking water from wells drilled in ground strata containing the element (IARC 1980;WHO 1981). It is also widely present in the groundwater supply in the United States (Welch et al. 1999).
Long-term exposure to ingested arsenic may induce many health effects. Biologic gradients between ingested arsenic and skin and various internal cancers have been well-documented and used to derive the maximum contamination level of arsenic in drinking water by the U.S. Environmental Protection Agency (Chen et al. 1988aMorales et al. 2000;Tsai et al. 1998). Other chronic health effects induced by arsenic have also drawn global attention, especially cardiovascular, neurologic, reproductive, and developmental hazards (Chen et al. 1997a(Chen et al. , 1999. Mortality and morbidity of vascular diseases, including peripheral vascular disease, cerebral infarction, and ischemic heart disease, have been documented to be associated with arsenic levels in drinking water in the arseniasisendemic area (Chen et al. 1988bChiou et al. 1997b;Tseng et al. 1996;Wang et al. 2002;Wu et al. 1989). The associations between long-term exposure to arsenic and microvascular diseases, including renal disease, retinopathy, and neuropathy, remain to be elucidated. Epidemiologic studies have shown a dose-response relationship between arsenic in drinking water and prevalence and mortality of diabetes mellitus in southwestern Taiwan (Lai et al. 1994;Tsai et al. 1999). Similar findings have been reported in Sweden , Bangladesh (Rahman et al. 1998), and the United States (Lewis 1999). In Taiwan, the incidence of diabetes mellitus was reported to be three to five times higher among residents in the southwestern arseniasis-endemic area compared with those in a nonendemic area (Tseng et al. 2000;Wang et al. 1997). However, no studies have been done to differentiate types of diabetes mellitus associated with arsenic. Diabetes mellitus has been documented to induce stroke, ischemic heart disease, peripheral vascular disease, nephropathy, retinopathy, and neuropathy through both macrovascular and microvascular damage (Chait and Bierman 1994;King and Banskota 1994;Krolewski et al. 1994). It remains to be revealed whether the prevalence of vascular diseases of diabetics is different between arsenic-exposed and unexposed groups. The interactive effects of arsenic and diabetes mellitus on micro-and macrovascular diseases should be closely examined.
Medical records for more than a million individuals in 1999-2000 have been released from the National Health Insurance for academic research in Taiwan. We used this database to estimate prevalence of non-insulindependent diabetes mellitus and its related vascular diseases in arseniasis-endemic and nonendemic areas in Taiwan.

Materials and Methods
National Health Insurance Database. In this study we used individual-based reimbursement claims randomly selected from the National Health Insurance Database, which was collected by the Bureau of National Health Insurance and compiled by the National Health Research Institutes in Taiwan. The National Health Insurance in Taiwan is compulsory and universal and provides comprehensive health benefits (Chiang 1997). Since 1995, more than 96% of the total population of Taiwan has been covered by the National Health Insurance system [Department of Health (DOH) 1995. Ninety-three percent of all health providers have been contracted to the Bureau of National Health Insurance, and those not contracted provide much fewer health services [DOH and British Columbia Ministry of Health (BCMH) 1999]. The copayment rate for patients is as low as 8-20%, with a fixed charge of $1.40 U.S. Thus, almost all the people, particularly those affected with chronic diseases, have been using the contracted health providers (Cheng and Chiang 1997). More than 96% of the population, in different age and gender groups, who were covered by National Health Insurance have used health services at least once through contracted health providers during (DOH 2002. Therefore, the information from the National Health Insurance Database is considered appropriate to derive accurate prevalence rates of chronic diseases such as diabetes mellitus and vascular diseases. In this study, we used individual-based reimbursement claims from the data file of clinical diagnosis and treatment and the data file of medical prescriptions. The diagnosis of disease status was made from International Classification of Diseases, Revision 9 (ICD-9) codes recorded in the clinical diagnosis and treatment and from drug names in the medical prescriptions.
In Taiwan, insulin-dependent diabetes mellitus is classified as a severe disease. A severe disease card is issued to people with such diseases, which entitles patients to free medical treatment. We used the data file of patients with severe disease cards to exclude insulin-dependent diabetics from all diabetic patients. The diabetics in this study all had the non-insulin-dependent type.
Southwestern arseniasis-endemic area. The study population in the endemic area included all those who have lived in four southwestern townships: Puttee and Ichu of Chimayo County and Peimen and Hsuehchia of Taiwan County. The water from a large proportion of artesian wells in these four townships had arsenic concentrations > 0.35 mg/L according to a national survey of more than 90% of wells in Taiwan (Chang et al. 1991). This arseniasis-endemic area has been described in detail in a previous report (Chen et al. 1988b). A total of 66,667 residents ≥ 25 years of age in the arseniasis-endemic area were included in this study.
Nonendemic area. Taiwan has 323 rural and urban townships and metropolitan precincts. In this study, the nonendemic area included 313 of them, excluding four townships in the above-defined arseniasis-endemic area and another six neighboring townships: Liuchiao of Chimayo County and Hsinying, Hsiaying, Anting, Chiang-chun, and Yenshui of Taiwan County. Because some artesian wells have arsenic concentrations > 0.35 mg/L and because a few cases of blackfoot disease were identified in these neighboring townships, they were excluded to define the nonendemic area more appropriately. Because the complete files from the original database during 1999-2000 were too large to be managed and analyzed, the National Health Insurance released a sampled database for research use, in which a proportional systematic sampling was used to retrieve one in every 500 records from the two original reimbursement data files. A total of 639,667 residents ≥ 25 years of age in nonendemic area were included in this study.
Statistical methods. The prevalence of diabetes mellitus and vascular diseases was first derived for specific age and gender groups by dividing the number of persons with a claim for a given disease by the number of persons with at least one reimbursement claim in 1999-2000. The direct method was used to calculate age-adjusted, gender-adjusted, and age-and gender-adjusted prevalence of the diseases. The general population in midyear 2000 in Taiwan was used as standard population for age and gender adjustment. Prevalence odds ratios and their 95% confidence intervals (CIs) were calculated to indicate the association between arsenic exposure and prevalence of diabetes and vascular diseases.

Results
Prevalence of diabetes mellitus. There was an increasing trend of diabetes mellitus prevalence with age in both arseniasis-endemic and nonendemic areas, as shown in Table 1. The prevalence in the arseniasis-endemic area was consistently greater than in the nonendemic area across all five age groups for both men and women. The prevalence odds ratios of diabetes in the endemic area in comparison with the nonendemic area were consistently greater in women than in men for all age groups, as shown in Figure 1. After adjustment for age and sex, the prevalence odds ratio was 2.69 (95% CI, 2.65-2.73) in the arseniasis-endemic area, using the nonendemic area as the referent.
Prevalence of vascular diseases in diabetics and nondiabetics. An increasing trend with age for microvascular disease prevalence was observed among diabetics and nondiabetics in both arseniasis-endemic and nonendemic areas for both men and women, as  Table 2. For all five age groups in both men and women, the prevalence of microvascular diseases was consistently highest among diabetics in the arseniasis-endemic area, followed by nondiabetics in the endemic area, diabetics in the nonendemic area, and nondiabetics in the nonendemic area. Similar findings were observed for macrovascular diseases (Table 3): diabetics in the arseniasisendemic area had the highest prevalence, followed by diabetics in nonendemic area, nondiabetics in endemic area, and nondiabetics in nonendemic area for most age groups in men and women. Figure 2 shows the prevalence odds ratios of microvascular diseases among diabetics of the endemic area (using the 55-64 age range as a typical example, the odds ratio is 15.6; 95% CI, 13.4-18.0), nondiabetics of the endemic area (7.0; 95% CI, 6.3-7.9), and diabetics of the nonendemic area (5.1; 95% CI, 4.4-5.9) in comparison with nondiabetics of the nonendemic area. The odds ratios tended to be higher in women than in men before 65 years of age but without statistical significance. Figure 3 illustrates the prevalence odds ratios of macrovascular diseases in the four diabetes-area groups by age and sex. The prevalence odds ratios increased from nondiabetics in the nonendemic area (for ages 55-64), to nondiabetics in the endemic area (2.0; 95% CI, 1.9-2.2), to diabetics in the nonendemic area (1.7; 95% CI, 1.6-1.8), to diabetics in the endemic area (3.7; 95% CI, 3.3-4.0) for most age groups in men and women. The odds ratios tended to be higher in women than in men before 65 years of age, with statistical significance.
The prevalence of various vascular diseases among diabetics and nondiabetics in endemic and nonendemic areas is shown in Table 4. The prevalence of all vascular diseases studied was much higher among diabetics than among nondiabetics and significantly higher in the arseniasis-endemic area than in the nonendemic area. The prevalence odds ratios of these vascular diseases in the arseniasis-endemic area compared with the nonendemic area ranged from 1.22 (95% CI, 1.10-1.35) for peripheral vascular disease to 7.21 (95% CI, 6.51-7.97) for neurologic disorder among diabetics, and from 1.34 (95% CI, 1.29-1.39) for coronary heart disease to 13.97 (95% CI, 13.38-14.58) for neurologic disorder among diabetics.

Discussion
This study confirmed the findings of our previous study (Lai et al. 1994) in which the subjects in the arseniasis-endemic area had an elevated prevalence of diabetes compared with the nonendemic area (odds ratio = 2.7 after adjustment for age and sex). In this study we found that women tended to have a higher prevalence of diabetes than did men in the arseniasis-endemic area but not in the nonendemic area. Furthermore, in the arseniasis-endemic area, women had a statistically significantly higher age-adjusted prevalence of vascular diseases than did men (9.4% vs. 7.8% for microvascular disease and 13.3% vs. 11.3% for macrovascular disease) among the nondiabetics. Also, women have been found to drink less water than men. The hypothesis of a greater vulnerability to arsenic exposure in women than in men needs further investigation.
In a national diabetes survey (Pan et al. 1998) applying both fasting glucose and oral glucose tolerance tests, the diabetes prevalence for men ≥ 65 years of age in Taiwan was 7.6%. The diabetes prevalence for the same age group of men in this study was 8.1%, which is similar to that observed by Pan et al. (1998). The diabetes prevalence was also compatible with that observed in the United States [National Health and Nutrition Examination Survey (NHANES)] (Harris et al. 1987) and in Japan (Kuzuya 1994). Diabetes prevalence by age and sex in the southwestern arseniasis-endemic area in this study is also similar to that observed in a previous survey conducted in the same endemic area (Lai et al. 1994).
Arsenic has been proposed to induce insulin-dependent and non-insulin-dependent diabetes, probably through increased oxidative stress (Longnecker and Daniels 2001;Wu et al. 2001). Oxidative stress has been found to induce the development of insulin resistance and endothelial dysfunction by the observations of normal, impaired glucose-tolerant, and diabetic subjects (Gopaul et al. 2001). Hypertension, an important component of insulin resistance syndrome, has also been found to be associated with long-term ingested arsenic exposure (Chen et al. 1995). It is essential to evaluate insulin secretion and insulin sensitivity in subjects with various degrees of arsenic exposure, taking genetic susceptibility ) into account.
Phenylarsine oxide binding to sulfhydryl groups (-SH) has been found to induce insulin resistance (Frost and Lane 1985;(Henriksen and Holloszy 1990) via the increase of cell stress and reduction of glucose Articles | Diabetes and vascular diseases related to arsenic Environmental Health Perspectives • VOLUME 111 | NUMBER 2 | February 2003  transport proteins, especially for GLUT4 and GLUT1, and glucose uptake (Jhun et al. 1991). However, phenylarsine oxide is an artificial organic form of arsenic used mainly for testing the role of the sulfhydryl group in insulin resistance. The use of natural arsenic compounds such as arsenite, arsenate, and/or methylated forms for such studies has been suggested. There are no consistent changes in glucose levels in experimental studies. Plasma glucose and triglycerides were the lowest in mice with high arsenate exposure administered via drinking water (Hughes and Thompson 1996). Enhanced glucose uptake was found in response to arsenite (100 µM for up to 180 min) in bovine chromaffin cells (Fladeby and Serck-Hanssen 1999). Nonetheless, intraperitoneal administration of sodium arsenite of 1.0 mg/kg has been found to cause significantly higher blood glucose in guinea pigs at 1 and 2 hr (Mitchell et al. 2000). Organic arsenic was found to induce the inhibition of glucose uptake (Liebl et al. 1995). Studies using longterm treatment of well-specified arsenic species are necessary for future conclusions. Arsenicinduced oxidative stress, mainly through the depletion of glutathione (Suzuki et al. 2001), has been proposed to cause both insulin resistance and atherosclerosis (del Razo et al. 2001), and the latter may be profound in hyperglycemia or diabetic states (Curcio and Ceriello 1992;Lorenzi 1992).
The finding of a strikingly increased prevalence of macrovascular diseases observed in the arseniasis-endemic area compared with the nonendemic area is consistent with our previous findings of arsenic-induced atherosclerosis (Chen et al. 1988bChiou et al. 1997b;Tseng et al. 1996;Wang et al. 2002;Wu et al. 1989). There have been few studies comparing the prevalence of specific vascular diseases in relation to arsenic exposure in diabetics and nondiabetics. Age-adjusted prevalence of cerebrovascular disease in the Lanyang Basin, a recently identified northeastern endemic area of arseniasis in Taiwan, was 15.8% for men and 13.2% for women ). The two figures were between those found in the present study: 21.4% for diabetics and 8.7% for nondiabetics. Diabetes was associated with an increased risk of cerebrovascular disease showing an age-and gender-adjusted odds ratio of 1.8 in the northeastern endemic area and 2.4 (21.35 ÷ 8.72) in the southwestern endemic area of this study. A previous study showed a dose-response relationship between ingested arsenic and peripheral vascular diseases (Wang et al. 2002). The present study demonstrated the odds ratios of 1.2 (95% CI, 1.1-1.4) in diabetics and 12.5 (95% CI,) in nondiabetics. Arsenic, mainly trivalent arsenicals (e.g., arsenite), may induce atherosclerosis through damage of endothelial cells or smooth muscle cells by intracellular-reduced glutathione or through oxidative DNA damage (Chang et al. 1991;Chiou et al. 1997a;Lynn et al. 2000;Wu et al. 2001). Further studies are necessary to test the hypothesis that arsenic induces renal (Mitchell et al. 2000) and neural (Brouwer et al. 1992;Mahajan et al. 1992) damage directly or through angiopathy.
The National Health Insurance Database used in this study consisted of reimbursement claims of all patients who had received care from contracted clinics and/or hospitals at least once in 1999-2000. Therefore, those who were not cared for by contracted hospitals or clinics during the study period were excluded Articles | Wang et al. 158 VOLUME 111 | NUMBER 2 | February 2003 • Environmental Health Perspectives    from the database. However, more than 96% of insured people had ever received care from contracted hospitals and clinics. The prevalence estimated in this study was considered reasonably correct. The disease prevalence might be overestimated if patients are more likely to visit clinicians and to be included in the database than are unaffected people. Nonetheless, the odds ratio comparing arseniasis-endemic and nonendemic areas would be valid if the frequencies of visiting clinicians were the same between two comparison areas.
Considering the rural and urban differences in lifestyles and disease patterns, residents in the rural area were considered less likely to develop cardiovascular diseases as a result of decreased prevalence of risk factors from dietary intake, obesity, and physical activity (Singh et al. 1998). However, residents in the arseniasisendemic area had a higher prevalence of cardiovascular disease despite the fact that the endemic area was more rural than was the nonendemic area in Taiwan. Thus, the vascular effect of ingested arsenic observed in this study was based on a conservative comparison.

Conclusions
This study demonstrated that residents in the arseniasis-endemic area had an increased risk of diabetes and its related vascular diseases compared with those in the nonendemic area. This study also found a larger contribution of ingested arsenic than of diabetes on the development of microvascular diseases. Future studies will be directed to mechanistic investigations of arsenic inducing non-insulin-dependent diabetes and atherosclerosis. Risk assessment of arsenic exposure for diabetes and the related vascular diseases should be integrated with the current scheme for cancer risk from arsenic.