Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 304
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards 10 Genotoxicity and Carcinogenicity This chapter reviews research publications and relevant review articles published since the earlier NRC (1993) report and other relevant papers not included in that review, and also considers salient earlier papers. Evaluation of the plausibility and potential for carcinogenicity is based on human epidemiologic studies, laboratory animal lifetime bioassays, shorter-term genotoxicity tests, metabolism and pharmacokinetic data, and mechanistic information. Genotoxicity tests indicate the potential for fluoride to cause mutations, affect the structure of chromosomes and other genomic material; affect DNA replication, repair, and the cell cycle; and/or transform cultured cell lines to enable them to cause tumors when implanted into host animals. In interpreting the experimental studies and the consistency among disparate tests and systems, factors to be considered include the chemical form, concentrations, duration of exposure or application, vehicle or route of exposure, presence or absence of dose response, and information that each study provides about the potential stage of cancer development at which the chemical might operate. The degree of consistency of genotoxicity tests with the epidemiologic studies and whole animal bioassays on these points was evaluated. GENOTOXICITY Genotoxicity tests comprise in vitro and in vivo assays to assess the effects on DNA and chromosomal structure and/or function. The results of these assays serve as indicators of the potential interaction of chemicals with the genetic material. Changes in chromosomal or DNA structure or
OCR for page 305
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards function may be a step in the pathway to carcinogenesis. More often, they indicate interference with the normal duplication, function, and control of cell division and genetic activity that also might result in precancer or early neoplastic processes. Genotoxicity also encompasses the ability to cause germ cell and somatic cell mutations that cause malformations, disease, and other adverse health outcomes. Many cell systems derived from various organisms have been used to the assess genotoxicity of a large array of chemicals. In evaluating the applicability of the results of these tests to human risk from fluoride ingestion, some of the key parameters are the concentrations used in the assays compared with physiologic concentrations, the form and vehicle for fluoride exposure in the assay, and existing data on overall applicability of the various assays to risk in humans. Tennant (1987) and Tennant et al. (1987) concluded that the Salmonella reverse mutation assay was the best short-term genotoxicity assay available for predicting carcinogenicity in mammals. However, Parodi et al. (1991) reviewed the results of various genotoxicity tests in comparison with animal carcinogenicity studies, and found that in vitro cytogenetic tests, particularly sister-chromatid exchange tests (SCEs), were more predictive of carcinogenicity than the Salmonella reverse mutation assay. Tice et al. (1996) subsequently reviewed relative sensitivities of rodents and humans to genotoxic agents and concluded that humans are more than an order of magnitude more sensitive than rodents to most of the genotoxic agents they examined using the genetic activity profile database. The available new genotoxicity studies of fluoride are detailed in Table 10-1. The most extensive and important additions to the genotoxicity literature on fluoride since 1993 are in vivo assays in human populations and, to a lesser extent, in vitro assays using human cell lines and in vivo experiments with rodents. These studies are discussed below. Gene Mutation Mutagenicity indicates direct action of a substance on DNA. Alterations in DNA suggest that the chemical has the potential to cause genetic effects as well as carcinogenic potential. In 1993, the existing literature did not indicate that fluoride posed a mutation hazard. The literature included assays with Salmonella (virtually all negative results), various mammalian cells lines (virtually all negative), and cultured human lymphocytes. Positive results in the human lymphocytes were seen at fluoride concentrations above 65 micrograms per milliliter (µg/mL) (parts per million [ppm]) and generally at more than 200 µg/mL, (much greater concentrations than those to which human cells in vivo typically would be exposed). No pertinent studies have been found since those reviewed in the 1993 NRC report. The committee interprets the weight of evidence from in vivo rodent studies to
OCR for page 306
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards TABLE 10-1 Summary of Recent Genotoxicity Studies of Fluoride Population or System/Method and Assay Findings Remarks References In vivo human studies Subjects (n = 746) with normal or inadequate nutrition living in regions of China with water concentrations of fluoride at 0.2, 1.0, or 4.8 mg/L. Assay: SCE in blood lymphocytes. Subjects in the 4.8-mg/L region had lower average SCEs per cell. Plasma and urine fluoride concentrations also measured; these were proportional to water concentrations. Y. Li et al. 1995 Comparison of 100 residents of North Gujarat exposed to drinking water with fluoride at 1.95 to 2.2 mg/L with 21 subjects in Ahmedabad exposed at 0.6 to 1.0 mg/L. Assay: SCE in blood lymphocytes and cell cycle proliferative index. SCE rate was significantly greater in subjects from North Gujarat, but there was no difference in the cell cycle proliferative index. Insufficient documentation of subject ascertainment or control for potential demographic confounding. Sheth et al. 1994 Phosphate fertilizer workers with inhalation exposure. Assay: chromosome aberrations, micronucleus, SCE. Exposed workers had elevation in all cytogenetic outcomes tested. Meng et al. 1995; Meng and Zhang 1997 Peripheral blood lymphocytes from inhabitants of the Hohhot region in inner Mongolia (n = 53 with fluorosis; n = 20 with no fluorosis) exposed to fluoride in drinking water at 4 to 15 mg/L compared with controls (n = 30) exposed to fluoride at < 1 mg/L. Assay: SCE and micronucleus. SCE: higher frequency in individuals with fluorosis (87% increase in SCEs), than no fluorosis (13% increase) compared with controls. Micronucleus: higher frequency in individuals with fluorosis (3.4-fold increase) than no fluorosis (1.8-fold increase) compared with controls. Insufficient documentation of subject ascertainment or control for potential demographic confounding. Wu and Wu 1995
OCR for page 307
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards Human populations with long-term residency in communities with water concentrations of fluoride at 0.2, 1.0, and 4.0 mg/L. Measured plasma and urine fluoride concentrations. Assay: SCE in blood lymphocyte. SCEs higher in 4.0-mg/L community. Follow-up study in the 4.0-mg/L community comparing residents using well water (≤0.3 mg/L) and city water (4.0 mg/L) found no difference in SCE frequency between these two groups in the 4-mg/L town. Jackson et al. 1997 Cultured peripheral blood lymphocytes from 7 female osteoporosis patients treated with disodium monofluorophosphate and NaF for 15 to 49 months (22.6 to 33.9 mg of fluoride/day). Measured serum fluoride. Assay: chromosomal aberration, micronuclei, cell cycle progression. No cytogenetic effects compared with the matched controls. Van Asten et al. 1998 Comparison of residents of South Gujarat exposed fluoride at approximately 0.7 mg/L (control village) with residents exposed at 1.5 to 3.5 mg/L (3 villages). Assay: SCE in peripheral lymphocytes. One of the high-fluoride villages had elevated SCEs. No difference was found between the other two and the control village. Insufficient documentation of subject ascertainment and demographic characteristics. Joseph and Gadhia 2000 Case series in India of osteosarcoma (n = 20) compared with population distribution regarding bone tumor fluoride concentration and p53 mutations. Assay: p53 mutation and fluoride concentrations in tumor tissue. Two (10%) cases had p53 mutants in osteosarcoma tissue, and those two had the highest bone tumor fluoride concentrations. Only patients undergoing prosthesis fitting at one hospital were selected; selection bias was possible. If replicated with systematic ascertainment, this design could indicate a mechanism for carcinogenic activity by fluoride. Ramesh et al. 2001
OCR for page 308
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards Population or System/Method and Assay Findings Remarks References In vivo animal studies Mice (B6C3F1) exposed via drinking water for 6 weeks. Measured fluoride concentrations in bone. Assay: micronuclei in peripheral red blood cells, chromosome aberrations in bone marrow. No micronuclei increase in peripheral red blood cells, and no chromosome aberration increase in bone marrow. Bone concentrations of fluoride increased with dose to >7,000 ppm. Method addresses some of the conflicts in previous in vitro and in vivo studies. Zeiger et al. 1994 Four Zucker rats, diabetic and nondiabetic males. Fluoride in water at 5 to 50 mg/L for 6 months. Assay: SCE. No SCE elevation in any exposed subgroup. Dunipace et al. 1996 Wistar rats exposed to NaF at 0, 7, and 100 mg/L in drinking water Assay: single cell gel electrophoresis (Comet assay) No increase in single-strand DNA damage. Ribeiro et al. 2004a In vitro human studies Synchronized human diploid fibroblasts. Attempt to reconcile disparate methods of classifying aberrations (e.g., gaps). Assay: chromosome aberrations. 50 ppm NaF is lowest concentration inducing aberrations. Proposes mechanism of inhibition of DNA synthesis and repair. Aardema and Tsutsui 1995 Cultured human diploid cells. NaF treatment for 2.5 hours or continuous. Assay: clastogenicity. Fluoride clastogenic at >5 ppm. No effect on ploidy. Oguro et al. 1995
OCR for page 309
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards Human diploid fibroblasts at quiescent stage treated with NaF at 1 to 10 ppm (fluoride ion at 0.45 to 4.5 ppm), 1 to 3 weeks. Assay: clastogenicity. No clastogenicity. Fluoride concentrations in range of water supplies. Tsutsui et al. 1995 Human lymphocytes from 50 individuals cultured in 10 to 30 ppm NaF. Assays: chromosomal aberration and SCE. Chromosomal aberration: 23% and 8% increased frequency of total aberrations at 20 and 30 ppm, respectively, but not at 10 ppm. SCE: no effects reported. Gadhia and Joseph 1997 Human embryo hepatocytes. Treated with NaF at 40, 80, and 160 mg/L for 24 hours. Assay: single cell gel electrophoresis (Comet assay) Lipid peroxidase and glutathione also assayed. Dose-related increase in single-strand DNA damage. Dose-related increase in lipid peroxidase, decrease in gluthathione, and increase in the percentage of apoptotic cells. Wang et al. 2004 In vitro animal studies Cell cultures of rodents, prosimians, apes, and humans. Assay: chromosome aberration. Clastogenicity of fluoride in great apes and human cells only at 42 to 252 ppm NaF. Kishi and Ishida 1993 BALB/c-3T3 mouse cells. Examined numerous chemicals, including NaF. Assay: cell transformation. 1.2 to 4.6 mM (19 to 193 ppm) NaF negative for transformation. Standard transformation assay modified to increase sensitivity Matthews et al. 1993. Rats (Sprague-Dawley) cultured bone marrow cells. NaF and KF at 0.1 to 100 µM for 12, 24, or 36 hours. Assay: cytotoxicity and SCE. Dose-response observed for cytotoxicity. No inhibition of cell proliferation. No effect on SCE. Khalil and Da’dara 1994
OCR for page 310
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards Population or System/Method and Assay Findings Remarks References Rat (Sprague-Dawley) cultured bone marrow cells. Treated with NaF and KF 0.1 to 100 µM for 12, 24, or 36 hours. Assay: chromosomal aberration and break. Weak effects at 1.0 µM, NaF and KF. Effects slightly greater for KF than NaF. Khalil 1995 Chinese hamster ovary cells. Treated with NaF at 7.28, 56, and 100 µg/mL for 3 hours. Assay: single cell gel electrophoresis (Comet assay) No increase in single-strand DNA damage. Ribeiro et al. 2004b Rat (F344/N) vertebral cells. NaF treatments 1 to 3 days. Assay: chromosomal aberration. Dose-related increases of chromosome aberrations at 0.5 and 1.0 mM for 24 and 48 hours. Potential target organ of NTP carcinogenicity studies that yielded osteosarcomas. Provides possible mechanism for carcinogenesis of vertebrae. Mihashi and Tsutsui 1996 ABBREVIATIONS: KF, potassium fluoride; NaF, sodium fluoride; NTP, National Toxicology Program.
OCR for page 311
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards indicate very low probability of a mutagenic risk for humans (NRC 1993; WHO 2002; ATSDR 2003). Chromosomal Changes and DNA Damage This section describes studies of fluoride’s effects on chromosomes and chromatids, formation of micronuclei, and DNA damage. Chromosomal alterations can include changes in chromosome number (aneuploidy) and aberrations of the chromosomes (before DNA synthesis) or chromatids (after DNA synthesis). (Nondisjunction or translocation of chromosome 21, producing Down’s syndrome, is discussed in Chapter 6 on Reproductive and Developmental Effects.) Classification of chromosome/chromatid aberrations has become standardized in recent years: some types of aberrations (e.g., chromatid gaps) are judged to be less important in evaluating effects on chromosomes than other major aberrations (e.g., breaks and translocations). SCE is not known to be on the causal pathway of any adverse health effects, but it is considered a generic indication of exposure to substances that can affect chromosomal structure, many of which are also carcinogens. The SCE assay is a helpful and widely used assay because of its greater sensitivity at lower concentrations than chromosome aberrations. Fewer cells need to be scored in order to establish with confidence whether an increase in SCEs has occurred in a specific test system. Micronuclei are DNA-containing bodies derived from chromosomal material that is left behind during mitosis. Either a faulty mitotic process or chromosomal breaks can cause this phenomenon. Micronuclei can be visualized in nondividing cells. The relatively new “Comet assay” detects single-strand DNA damage in individual cells using microgel electrophoresis. Effects on cell survival (cytotoxicity) and effects on cell division are commonly investigated and reported in the course of conducting in vitro cytogenetic studies, and they are included in the summary below. Human Cells In Vitro Interpreting the health significance of observed cytogenetic effects on human cells in culture depends on the dose, timing of application relative to the point in the cell cycle, and type of cultured cells, among other factors. As of the 1993 NRC report, the existing data of this type were inconsistent regarding the cytogenetic effects of fluorides. Since that time, Tsutsui et al. (1995) applied sodium fluoride (NaF) at or near concentrations found in water supplies (1 to 10 ppm, equivalent to 0.45 to 4.5 ppm fluoride ion) to diploid fibroblasts for up to 3 weeks and did not observe clastogenicity. Aardema and Tsutsui (1995) using a similar cell system found aberrations only above 50 ppm. The cell phases at which these effects were observed
OCR for page 312
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards suggested that the underlying mechanism of the chromosomal aberrations might be interference by fluoride with DNA synthesis and repair. In human diploid IMR90 cells, Oguro et al. (1995) observed clastogenicity only above 5 ppm NaF after short- and long-term applications. Gadhia and Joseph (1997) noted that 20 and 30 ppm NaF, but not 10 ppm, caused aberrations. No effects on SCEs were seen in their study. Recently, Wang et al. (2004) used the Comet assay to study genotoxicity in human embryo hepatocytes after treatment with NaF. They observed a dose-related increase in single-strand DNA damage at concentrations of 40, 80, and 160 mg/L. Other Mammalian Systems In Vitro Previous studies with a wide variety of test systems found cytogenetic effects in some but not all systems used (NRC 1993; WHO 2002; ATSDR 2003). Recent studies with in vitro rodent systems include those by Khalil and Da’dara (1994) and Khalil (1995). They evaluated effects on cultured bone marrow cells of Sprague-Dawley rats after exposure to NaF or potassium fluoride (KF) at concentrations ranging from 0.1 µM to 0.1 μM (up to 2 ppm fluoride) for 12 to 36 hours. They did not observe increased SCE levels at any concentration, although there was dose-dependent cytotoxicity. Both NaF and KF induced chromosomal aberrations in a dose-dependent manner between 0.1 and 100 µM. Mihashi and Tsutsui (1996) studied effects on cultured vertebral cells of F344/N rats after 1 to 3 days of 9 to 18 ppm NaF treatment and found dose-dependent increases in chromosomal aberrations based on time and concentrations. Kishi and Ishida (1993) compared activity of NaF on chromosome aberrations for a series of cell lines from rodents, prosimians, great apes, and humans. Clastogenicity by 42 to 252 ppm NaF was seen only in the great ape and human cell lines. Their work thus indicates a greater sensitivity to fluoride in human than in rodent cells. In an older study not included in the NRC (1993) report, Jagiello and Lin (1974) reported that in vitro exposure of oocytes to NaF disrupted meiotic anaphase of ewes and cows but not of mice. The effective doses were the same order of magnitude as those reported by NRC in 1993 to cause chromosome aberrations in human lymphocytes. In vivo tests performed only in mice indicated that fluoride was not genotoxic, even at high doses. Ribeiro et al. (2004b) used the Comet assay to assess effects of NaF on Chinese hamster ovary cells in vitro. No damage was observed at concentrations of up to 100 µg/mL.
OCR for page 313
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards Rodent Systems In Vivo Zeiger et al. (1994) administered NaF in drinking water for 6 weeks to B6C3F1 mice and assayed micronuclei and chromosome aberration occurrences. They observed no increases over unexposed controls. Similarly, Dunipace et al. (1996) exposed diabetic and nondiabetic Zucker male rats to fluoride concentrations up to 50 mg/L in water for up to 6 months. They found no increase in the rate of SCEs for any test group. Ribeiro et al. (2004a) exposed Wistar rats to NaF at 7 and 100 mg/L in drinking water for 6 weeks. Comet assays of peripheral blood, oral mucosa, and brain cells in vivo showed no increase in single-strand DNA damage. Nonmammalian Systems In Vivo Previous work on nonmammalian systems was sparse but did not indicate consistent cytogenetic effects. No new relevant studies have been reported. Human Cells In Vivo The NRC 1993 report noted the absence of human in vivo genotoxicity studies. Since 1993, important contributions to the evaluation of genotoxicity of fluoride have been in the area of cytogenetic studies of human populations exposed via diverse routes to various fluorides. Studies of human populations have the advantage of evaluating pertinent concentrations in a physiologically relevant context, despite the limitations inherent in all epidemiologic observational studies of not controlling for all factors that might be pertinent. Relevant studies are summarized below according to route of exposure. Ingestion Route The most well-documented in vivo human study published was that of Y. Li et al. (1995), who assayed the fluoride concentrations in water, plasma, and urine in more than 700 individuals. Six groups of 120 subjects resided in different locales with average naturally occurring fluoride concentrations in drinking water varying between 0.2 and 5 mg/L. They observed that, although plasma and urine fluoride concentrations varied with water concentrations, the groups of subjects living in the regions with higher concentrations of fluoride had lower average SCEs per cell. The study controlled for the nutritional status of the subjects. Subsequently, Jackson et al. (1997) compared SCE occurrence in lymphocytes of residents of communities with water fluoride concentrations of 0.2, 1, and 4 mg/L. Residents of the 4-mg/L
OCR for page 314
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards fluoride community had more average SCEs. In a follow-up study, there was no difference between the mean SCE level of a subsample of residents using the 4-mg/L community water and another sample of residents using 0.3-mg/L well water. The following three less-well-documented studies reported associations between cytogenetic effects and residence in areas with high natural fluoride concentrations in drinking water. Sheth et al. (1994) published a preliminary investigation of SCEs in 100 residents of Gujarat, India, with fluorosis and 21 unaffected controls. They reported higher SCE rates among the fluorosis cases as well as higher fluoride concentrations in the cases’ water. The design of this study was seriously deficient, particularly because of the possibility of selection bias; cases and controls were recruited from different areas (cases were from areas with higher naturally occurring fluoride in drinking water). Additionally, clinical criteria for case definition were not adequately documented. Wu and Wu (1995) examined peripheral blood lymphocytes in a small series (n = 53) of residents in a high-natural-fluoride area (4 to 15 mg/L) and 30 control residents from a low-fluoride area (<1 mg/L) of Inner Mongolia. SCEs and micronuclei were more frequent only among subjects with fluorosis and not among those with higher exposures who did not exhibit fluorosis. The report had a dearth of information on subject selection and on control of potential confounding factors. Joseph and Gadhia (2000) later compared residents of three villages that had drinking water concentrations of fluoride at 1.6 to 3.5 mg/L with residents of Gujarat, India, where there is fluoride in residential drinking water at 0.7 mg/L . Chromosome aberrations were strongly elevated in residents of all three of the villages. SCE rates were elevated only in residents of one of those, and the same village’s residents also demonstrated higher chromosome aberrations in mitomycinC-treated lymphocytes. Only 14 individuals were tested from each village, and the method of subject selection was not reported. Van Asten et al. (1998) found no cytogenetic effects (aberrations, micronuclei, or cell cycle progression) on cultured lymphocytes in women who had been treated with fluoride (22.6 to 33.9 mg/day) for osteoporosis for 1 to 4 years. Inhalation and/or Dermal Routes Two articles published by Meng et al. (1995) and Meng and Zhang (1997) described cytogenetic assays in phosphate fertilizer workers. Inhalation of fluoride is the principal chemical exposure in these plants. The air concentrations of fluoride ranged from 0.5 to 0.8 mg/m3 at the time of the study. Chromosomal aberrations, micronuclei, and SCEs were all elevated in exposed workers. The length of exposure did not show a dose-dependent relationship with these cytogenetic effects; those working at the plant for 5
OCR for page 329
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards guidelines. Based on tertiles for the controls, three exposure categories were expressed as 100%, 30% to 99%, and <30% of the target concentrations for fluoridated water. A unique feature of the analysis published in the literature so far was an exploratory analysis of ORs for each specific year of age. Bassin found elevated ORs for the highest tertile compared with the lowest centering on ages 6 to 8. At age 7, the respective ORs (and 95% confidence intervals) were 7.2 (1.7 to 30.0) for males and 2.0 (0.43 to 9.28) for females. For the highest tertile, graphed results for males indicated a gradual increase and then a decrease of estimated relative risk from exposure at ages 0 to 15 with peaks at age 7, with the middle tertile, compared with the lowest, showing stable ORs across all ages. For females, both the middle and highest tertiles of exposure showed relatively unchanging relative risk estimates across exposure ages. There was no analysis of cumulative exposures to fluoride, and therefore it is difficult to compare the Gelberg study, which used only cumulative exposure indices, with the Bassin work. This dissertation had a paucity of data in the results section, hampering its interpretation; for example, the report did not provide numbers of subjects in the categories upon which the ultimate analyses were based. Also, there were no data on bias potential stemming from nonparticipation of subjects due to refusal to be included or inability to locate them. Nevertheless, the higher ORs for males than for females, and the highest ORs at ages 6 to 8, during what the author describes as the “mid-childhood growth spurt for boys,” are consistent with some previous ecologic or semiecologic studies (Hoover et al. 1991; Cohn 1992) and with a hypothesis of fluoride as an osteosarcoma risk factor operating during these ages. A publication based on the Bassin thesis is expected in the spring/summer of 2006 (E. Bassin, personal communication, Jan. 5, 2006). If this paper provides adequate documentation and analyses or the findings are confirmed by another study, more weight would be given to an assessment of fluoride as a human carcinogen. A relatively large hospital-based case-control study of osteosarcoma and fluoride exposure is under way (Douglass 2004) and is expected to be reported in the summer of 2006 (C. Douglass, Harvard School of Dental Medicine, personal communication, January 3, 2006). Most of the incident cases are identified via eight participating medical centers in California, District of Columbia, Florida, Illinois, Massachusetts, Nebraska, and Ohio. The study has prospectively identified 189 incident cases of osteosarcoma and 289 hospital controls. Controls are orthopedic patients at the same hospitals as osteosarcoma patients and include patients diagnosed with malignancies other than osteosarcoma and other patients admitted for benign tumors, injuries, and inflammatory diseases. Matching criteria include gender, age,
OCR for page 330
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards and geographic characteristics. The investigation includes residence histories and detailed interviews about water consumption as well as fluoride assays of bone specimens and toenails of all subjects. The ultimate analysis and validity of this study will depend partly on the degree to which control selection is not biased in such a way as to artificially increase or decrease the likelihood of fluoride exposure compared with the general population to which this study is intended to apply. A preliminary retrospective recruitment phase of this investigation, including telephone interviews, residential history reconstruction, and an attempt to estimate dietary fluoride intakes, reported ORs of 1.2 to 1.4 that were not statistically significant (Douglass 2004). No confidence intervals were provided. The Douglass study may have limited statistical power to detect a small increase in osteosarcoma risk due to fluoride exposure, but the committee expects the forthcoming report is likely to be a useful addition to the weight of evidence regarding the presence or degree of carcinogenic hazard that fluoride ingestion might pose to osteosarcoma risk, particularly if it addresses some of the limitations of hospital-based studies that are mentioned above in the description and critique of the Bassin thesis. Kidney and Bladder Cancers The plausibility of the bladder as a target for fluoride is supported by the tendency of hydrogen fluoride to form under physiologically acid conditions, such as found in urine. Hydrogen fluoride is caustic and might increase the potential for cellular damage, including genotoxicity. The Hoover et al. (1991) analyses of the Iowa and Seattle cancer registries indicated a consistent, but not statistically significant, trend of kidney cancer incidence with duration of fluoridation. This trend has not been noted in other publications, although Yang et al. (2000) observed that the adjusted mortality rate ratios of kidney cancers among males in Taiwan was 1.55 (95% CI 0.84 to 2.84). The analogous rate for females was 1.37 (95% CI 0.51 to 3.70). Yang et al. noted statistically significant RRs in females for bladder cancer (RR = 2.79, 95% CI 1.41 to 5.55; for males RR = 1.27, 95% CI 0.75 to 2.15). The Grandjean et al. (1992) and Grandjean and Olsen (2004) historical occupational cohort study of cryolite workers in Denmark (described earlier in the section on bone and joint cancers), who were followed from 1941 to 2002, observed an elevated standardized incidence ratio (SIR) for bladder cancers (SIR = 1.67, 95% CI 1.02 to 2.59). The SIR is the ratio of observed cases of cancer to the expected number of cases based on incidence rates of the general population. Higher SIRs were seen among males employed more than 10 years, males less than 35 years old when follow-up began, and among workers observed after a minimum latency of 30 years
OCR for page 331
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards (Grandjean and Olsen 2004). In the absence of data on smoking, the authors interpreted the higher SIRs for bladder cancer than for lung cancer to suggest that smoking was unlikely to be the major cause of the elevated bladder cancer incidence. The authors proposed (2002) that excretion of fluoride compounds entailed exposure of the pertinent target tissues. As noted above, the estimated exposures of the cryolite workers were about 4-fold greater than those estimated from ingestion of fluoridated water at the MCL of 4 mg/L. However, those workers were exposed for fewer years than those involved in lifetime residency. Romundstad et al. (2000) reported on cancer among Norwegian aluminum workers exposed to polycyclic aromatic hydrocarbons and fluorides. SIRs for bladder and lung cancer were elevated among the exposed workers. However, separate effects from the two exposures could not be distinguished from this paper. Further, the authors review and compare earlier studies that used different aluminum plant processes, which support the role of polycyclic aromatic hydrocarbons in bladder cancer among the exposed cohort. It may be noteworthy that smoking did not appear to be a confounder for the risk of bladder or lung cancer among the exposed cohort. The authors state, but do not present data, that they found a “weak association” of bladder for fluoride exposures lagged less than 20 years. Oral-Pharyngeal Cancer The NCI analysis (Hoover et al. 1991) indicated an a priori interest in oral cancers. In Iowa, one of the two cancer registries they analyzed, the authors observed a trend among males in the incidence rates of oropharyngeal cancer with duration of fluoridation, but mortality analyses did not indicate an association with fluoridation. However, in an earlier study in England, oral-pharyngeal cancers among females constituted the only site-gender category for which standardized mortality ratios in England were found to be significantly elevated in areas with naturally occurring high fluoride concentrations, defined as more than 1.0 mg/L. Twenty-four site-gender combinations were examined for 67 small areas (Chilvers and Conway 1985). Uterine Cancer An association of uterine cancer (combination of cervical and corpus uteri) with fluoridation was reported by Tohyama (1996), who observed mortality rates in Okinawa before and after fluoridation was terminated, controlling for sociodemographics. This analysis is a follow-up of the positive results from a previous exploratory analysis that comprised a large number of comparisons conducted by this researcher with the same data
OCR for page 332
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards set. The only other recent publication to report on uterine cancers is that of Yang et al. (2000), who observed a mortality rate ratio of 1.25 with 95% CI of 0.98 to 1.60. Other Specific Cancers Respiratory cancers were elevated among the cohort of Danish cryolite miners for whom exposure was by the inhalation route (Grandjean et al. 1992; Grandejan and Olsen 2004; see discussions above on this cohort study). SIRs of 1.51 (95% CI 1.11 to 2.01) were observed for the cohort as a whole, with higher SIRs among those after 30 years of exposure and among males younger than 35 when follow-up began. No smoking data on the cohort were collected. Also, except for mortality among females in Taiwan (Yang et al. 2000), there has not been corroborating data from other analyses for respiratory cancers. No association between lung cancer and exposure to polycyclic aromatic hydrocarbons and fluorides was found in a study of the Norwegian aluminum industry (Romundstad et al. 2000). The NCI incidence or mortality analyses conducted by Hoover et al. (1991) observed a few suggestive increases among some subgroups for soft tissue sarcoma, non-Hodgkin’s lymphoma, colorectal cancer, and lip cancer, but those cancers were not a priori of concern as related to fluoride exposure based on biologic plausibility. All Cancers Combined A large number of mortality analyses for all cancers combined have been reported and reviewed previously (NRC 1993; McDonagh et al. 2000a), and most of those did not detect an association of combined cancer mortality with fluoridated water. Typically, studies that only report combined cancer rates are not informative for assessing possible associations between an environmental exposure and a specific cancer outcome, particularly an uncommon cancer. Thus, the committee did not use these types of studies as part of its evaluation. Other Studies Evaluated The following three studies were reviewed but were not included by the committee in the evaluation of weight of evidence of carcinogenicity of fluoride for the reasons summarized below. Takahashi et al. (2001) conducted an ecologic analysis of data from nine U.S. cities for three 5-year intervals spanning 1978-1992 combined with fluoridation data. Their analysis involved regression of log-transformed
OCR for page 333
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards cancer incidence rates on the log-transformed proportion of residents receiving fluoridated water. This paper is difficult to interpret and to compare with other studies in part because of its novel method of analysis. Unusual cancer subsites are included and major anatomical groupings typically appearing in cancer incidence reports (e.g., lymphocytic leukemia, breast, uterus) were omitted. Results were incompletely reported for subsets of data for particular cancer sites, creating issues of multiple comparisons and selective presentation. Another issue is that the ecologic exposure variable is the percentage of the population in each area with fluoridated water (or naturally occurring fluoride at 0.7 mg/L or higher). This is an aggregated form of a dichotomous variable on the individual level, which tends to bias results away from the null. There was inconsistent standardization of the outcome variable (which was age standardized) and the exposure variable (which was not), which can lead to bias. There was no adjustment for confounding by urbanization or other sociodemographic factors among the nine cities, which included widely different geographic, industrial, and demographic characteristics, and there was no population weighting by size. Finally, ecologic bias is best understood for linear or log-linear regression, making this study harder to interpret. Steiner (2002) conducted an ecologic analysis of latitude, temperature, and fluoridated water in 49 cities worldwide. When fluoride concentrations were unavailable for these cities, he substituted data from neighboring areas. Average daily temperature and latitude were also included in his models, but not simultaneously. Steiner analyzed only all cancers combined. He found a negative association between cancer incidence and fluoridation. Yiamouyiannis (1993) subtracted female from male cancer incidence rates for the United States and for New Jersey as an indication of fluoride’s carcinogenic effect among males. This paper used circular reasoning to reach a conclusion of causality; that is, it concluded that higher cancer rates in males indicate an association with fluoride on the basis of a presumed causation by fluoride of cancers in males. Because most cancers do not occur at the same rates in each gender, the committee judges it is inappropriate to subtract rates of women from those of men as a means of evaluating factors that only affect bone cancer in males. It has been suggested that differences in osteosarcoma rates found in provinces of Kenya could be related to fluoride exposure (C. Neurath, Fluoride Action Network, unpublished data, June 17, 2005). For eight provinces of Kenya, Neurath correlated enamel fluorosis prevalences reported by Chibole (1987) with osteosarcoma incidence rates reported by Bovill et al. (1985) and found a strong association. This type of fully ecologic analysis (see Appendix C) has its inherent advantages and limitations; in this instance, however, the underlying ratios of observed-to-expected osteosarcoma incidence are not reliable because Bovill et al. do not state
OCR for page 334
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards that their incidence data were adjusted for differences in the age structure of various provincial populations. Bovill et al. state that Kenya is characterized by strong contrasts of ethnicity and other demographics among its geographic regions. The provincial summaries are weighted averages of the children examined, but it is not stated if they are also weighted averages of the underlying populations. Chibole does not state how the children examined in Kenyan schools and hospitals were selected (i.e., whether the fluorosis prevalence data collected were ascertained in a manner that would accurately reflect the populations of the component provinces). Chibole’s detailed table indicates a wide range of prevalences of fluorosis within many of the provinces (e.g., from 3.7% to 69.5% in the Rift Valley province). Summary of Cancer Epidemiology Findings The combined literature described above does not clearly indicate that fluoride either is or is not carcinogenic in humans. The typical challenges of environmental epidemiology are magnified for the evaluation of whether fluoride is a risk factor for osteosarcoma. These challenges include: detection of relatively low risks, accurate exposure classification assessment of pertinent dose to target tissues, multiple causes for the effect of interest, and multiple effects of the exposure of interest. Assessing whether fluoride constitutes a risk factor for osteosarcoma is complicated by (1) how uncommon the disease is, so that cohort or semi-ecologic studies are not based on large numbers of outcomes, and (2) the difficulty of characterizing biologic dose of interest for fluoride because of the ubiquity of population exposure to fluoride and the difficulty of acquiring bone samples in nonaffected individuals. In summary, there has been partial but incomplete fulfillment of NRC’s recommendations on individual-based cancer studies in the intervening years since 1993; one analytic study of osteosarcoma has been published, but bone samples were not included. The alternative (hospital-based) design, including bone assays, from the Harvard group might be more useful in addressing this issue. EPA GUIDELINES AND PRACTICE IN SETTING MCLGS REGARDING CARCINOGENICITY The EPA Office of Drinking Water establishes MCLGs of zero for contaminants that are known or probable human carcinogens. Chemicals for which cancer hazard is judged to be absent are regulated via the reference dose (RfD) method (see Chapter 11). “Methodology for Deriving Ambient Water Quality Criteria for the Protection of Human Health (2000)” reviewed EPA’s additional practice of applying an uncertainty factor between
OCR for page 335
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards “1 and 10” to an RfD derived from noncancer health effects (EPA 2000d). This procedure has been used for substances judged to be possibly carcinogenic in humans. That methodology document also stipulates that the water concentrations estimated to result in 10−6 to 10−5 excess cancer risks should also be assessed under the RfD scenario for comparison. As of April 2005, EPA has adopted new “Guidelines for Carcinogen Risk Assessment,” which has replaced the 1986 categories with weight-of-evidence descriptors, involving textual consideration and explanation of how each category was arrived at. In addition, the Guidelines provide for consideration of mode of action and sensitive subpopulations, especially children (EPA 2005a,b). In addition to mode of action, other factors for weighing human epidemiologic studies and lifetime whole animal bioassays include data on biomarkers (genotoxicity and other assays of exposure, susceptibility, and effect) and toxicokinetics. Thus, key decisions about cancer pertinent to a MCLG for drinking water include an assessment of whether an MCLG of zero is appropriate based on the current epidemiologic, animal bioassay, and additional contributing data. If not, EPA will need to decide whether an uncertainty (safety) factor greater than 1.0 and up to 10.0 should be applied to an RfD derived from a precursor response to tumors. Some recent examples of the use by EPA of RfDs with additional safety factors imposed because of possible carcinogenic hazard, based on the July 1999 Cancer Guidelines, include the MCLG for disinfection by-products (EPA 2003c). For dibromochloromethane (DBCM), EPA imposed an additional uncertainty factor of 10 to account for possible carcinogenicity based on studies of DBCM by NTP in 1985 that showed an increase in liver tumors in both genders of mice but no increase in either gender of rats. Similarly for trichloroacetic acid (TCA), an additional uncertainty factor of 10 was added to the MCLG derived from the RfD; TCA induced liver tumors in mice but not in rats. The MCLGs for all regulated chemicals considered to be possible carcinogens has included the additional 10-fold risk management factor applied to the RfD (J. Donohue, EPA, personal commun., 2004). FINDINGS The 1993 NRC report recommended the following: Conduct one or more highly focused, carefully designed analytical studies (case control or cohort) of the cancer sites that are most highly suspect, based on data from animal studies and the few suggestions of a carcinogenic effect reported in the epidemiological literature. Such studies should be designed to gather information on individual study subjects so that adjustments can be made for the potential confounding effects of other risk factors in analyses of individu-
OCR for page 336
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards als. Information on fluoride exposure from sources other than water must be obtained, and estimates of exposure from drinking water should be as accurate as possible. In addition, analysis of fluoride in bone samples from patients and controls would be valuable in inferring total lifetime exposures to fluoride. Among the disease outcomes that warrant separate study are osteosarcomas and cancers of the buccal cavity, kidney, and bones and joints. As described above, some progress in those directions have been made, with the most comprehensive study still in progress (Douglass 2004). Fluoride appears to have the potential to initiate or promote cancers, particularly of the bone, but the evidence to date is tentative and mixed (Tables 10-4 and 10-5). As noted above, osteosarcoma is of particular concern as a potential effect of fluoride because of (1) fluoride deposition in bone, (2) the mitogenic effect of fluoride on bone cells, (3) animal results described above, and (4) pre-1993 publication of some positive, as well as negative, epidemiologic reports on associations of fluoride exposure with osteosarcoma risk. Several studies indicating at least some positive associations of fluoride with one or more types of cancer have been published since the 1993 NRC report. Several in vivo human studies of genotoxicity, although limited, suggest fluoride’s potential to damage chromosomes. The human epidemiology study literature as a whole is still mixed and equivocal. As pointed out by Hrudey et al. (1990), rare diseases such as osteosarcoma are difficult to detect with good statistical power. In animal studies, the overall incidence of osteosarcoma in male rats showed a positive trend. Based on the more recent historical control data (Haseman et al. 1998) that were closer to the time frame of the NTP study, the 4% to 5% incidence at the high dose might have exceeded the historical range. The relevance of rat osteosarcoma to humans was discussed based on the species differences in the development of long bone, the common site of human osteosarcoma (NTP 1990). Specifically, ossification of human long bones is completed by 18 years of age whereas it continues in rats throughout the first year of life (PHS 1991). Nevertheless, most of the osteosarcomas found in male rats were not in long bones. In another study (NTP 1992), that used the same strain and sex of rats, no increase in osteosarcomas was reported, even though the animals were exposed to a higher concentration of fluoride than in the earlier study. However, the primary intent of the NTP (1992) study was to test the hypothesis that ionizing radiation is an initiator of osteosarcoma and that fluoride is a promoter, and the committee thought it was noteworthy that none of the irradiated animals developed osteosarcomas. The 1993 NRC review concluded that the increase in osteoma in male and female mice (Maurer et al. 1993) was related to fluoride treatment.
OCR for page 337
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards TABLE 10-4 Evidence Summary for Carcinogenicity of Fluoride: Epidemiologic Studies and Rodent Lifetime Bioassays Cancer Site/Type Individual-Based Epidemiology Studies Ecologic Epidemiology Studies Animal Data Osteosarcoma Case-control studies ambiguous (additional comprehensive hospital-based case-control study including bone fluoride measurements is under way). Mixed. Male F344/N rats: Borderline positive. Male F344 rats: inconclusive Oral cavity NCI incidence elevated in males, but no mortality trends. Several other reports positive. Nonstatistically significant increase in male rats. Thyroid Nonstatistically significant increase in male rats. Kidney and/or bladder Occupational cohort: positive finding, inhalation route, high exposures. Some positive reports. Uterine One positive report. Respiratory Occupational cohort positive finding, inhalation route, high exposures. One positive report. TABLE 10-5 Evidence Summary for Carcinogenicity of Fluoride: Genotoxicity and Mechanistic Assays Type of Effect and Assay Strength of Evidence Mitogenesis Well established. Cytogenetic effects: human in vivo exposure, in vitro assay. Inconsistent; and the positive findings were from weak papers. Cytogenetic effects: human in vitro exposure, in vitro assay. Inconsistent. Cytogenetic effects: other mammalian systems. Inconsistent. Transformation. Inconsistent; the positive results are consistent with a promotion mechanism. DNA repair mechanism: human. Suggestive positive finding regarding tumor suppressor gene, small case series. Mutation: mammalian systems. Inconsistent. Mutation: microorganisms. Negative.
OCR for page 338
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards Although the subsequent review by AFIP considered these mouse osteomas as more closely resembling hyperplasia than neoplasia, given that osteoma is widely recognized as neoplastic, the evidence of osteoma remains important in the overall weight-of-evidence consideration. The increased incidence and severity of osteosclerosis in high-dose female rats in the NTP study demonstrated the mitogenic effect of fluoride in stimulating osteoblasts and osteoid production (NTP 1990) (see also Chapter 5). The genotoxicity data, particularly from in vivo human studies, are also conflicting; whereas three were positive on the basis of the ingestion route (Sheth et al. 1994; Wu and Wu 1995; Joseph and Gadhia 2000), all three of these reports had serious deficits in design and/or reporting, including the characterization of how the study populations were selected and whether the exposed and unexposed study subjects were comparable. Two studies (Meng et al. 1995; Meng and Zhang 1997) were positive for the inhalation route among workers in a phosphate fertilizer factory, although other contaminants cannot be ruled out as the causal factors. Contrasting negative observations by other investigators (Li et al. 1995; Jackson et al. 1997; Van Asten et al. 1998) must also be considered. RECOMMENDATIONS Carcinogenicity The results of the Douglass et al. multicenter osteosarcoma study (expected in the summer of 2006) could add important data to the current body of literature on fluoride risks for osteosarcoma because the study includes bone fluoride concentrations for cases and controls. When this study is published, it should be considered in context with the existing body of evidence to help determine what follow-up studies are needed. Further research on a possible effect of fluoride on bladder cancer risk should be conducted. Since bladder cancer is relatively common (compared with osteosarcoma), both cohort and case-control designs would be feasible to address this question. For example, valuable data might be yielded by analyses of cancer outcomes among the cohorts followed for other health outcomes, such as fractures (see Chapter 5). Genotoxicity The positive in vivo genotoxicity studies described in the chapter were conducted in India and China, where fluoride concentrations in drinking water are often higher than those in the United States. Further, each had a dearth of information on the selection of subjects and was based on small numbers of participants. Therefore, in vivo human genotoxicity studies
OCR for page 339
Fluoride in Drinking Water: A Scientific Review of EPA’S Standards in U.S. populations or other populations with nutritional and sociodemographic variables similar to those in the United States should be conducted. Documentation of subject enrollment with different fluoride concentrations would be useful for addressing the potential genotoxic hazards of fluoridated water in this country.
Representative terms from entire chapter: