Exposure to peroxisome proliferators: reassessment of the potential carcinogenic hazard.

Melnick (1) recently suggested that because peroxisome proliferation is not established as an obligatory step in the carcinogenicity of peroxisome proliferators (PPs), the proposal that the peroxisome proliferator di(2-ethylhexyl)phthalate (DEHP) poses no carcinogenic risk to humans (2) due to species differences in peroxisome proliferation should be viewed as an unvalidated hypothesis (1). In this context, Melnick (1) raised the recent downgrading by the International Agency for Research on Cancer (IARC) of DEHP (3) to “not classifiable as to its risk to humans (group 3)” based on their conclusion that it produces rodent liver tumors by a mechanism involving peroxisome proliferation, which they judged to be not relevant to humans (3). As illustrated by Melnick (1), there is a large body of data to correlate the phenomenon of rodent liver peroxisome proliferation with rodent liver cancer but “published studies have not established peroxisome proliferation per se as an obligatory pathway on the carcinogenicity of DEHP” (1). This focuses attention on the need, as also suggested by O’Brien et al. (4), for a fundamental review of how PPs induce liver cancer in rodents and the relevance of these rodent tumors for humans. There are two distinct usage patterns for PPs: as drugs such as clofibrate for the treatment of hypolipidemia (5) and in nonclinical applications such as the plasticiser DEHP. Most PPs are carcinogenic to the rodent liver, and the task of assessing their human carcinogenic potential has fallen to different regulatory agencies, depending on the primary usage pattern of the particular PP in question. There seems to be little regulatory concern regarding the safety of the clinical PPs, yet continuing uncertainty regarding the safety of the nonpharmaceutical PPs. This presents an untenable situation that we suggest is unjustified. Hypolipidemic fibrates such as clofibrate and gemfibrozil have been used extensively over the past 20 years to treat cardiovascular disease and are enjoying a revival due to recent reconfirmation of efficacy (6). However, by 1980 several of these agents had become associated with a rodent-specific response known as hepatic peroxisome proliferation (7), a property shared by a number of nonpharmaceutical chemicals (8). Additionally, a link between rodent liver peroxisome proliferation and an increased risk of rodent liver carcinogenesis was emerging (7). Nonetheless, clinical side effects of the fibrate PPs are rare, and analyses of causes of death during treatment show no evidence of an adverse effect and no evidence of an increase in malignant disease compared to the normal population (5). Specifically, the carcinogenic risk to humans of gemfibrozil and clofibrate has been formally evaluated by IARC (9,10), and in the case of clofibrate, for which most clinical data exist, IARC (9) concluded that “the mechanism of liver carcinogenesis in clofibrate-treated rats would not be operative in humans.” This conclusion was based on the observation that clofibrate causes peroxisome proliferation and cell proliferation in rodent but not human hepatocytes and on the results of extensive epidemiologic studies, particularly the World Health Organization trial on clofibrate that included 208,000 man-years of observation (11,12). Further, a meta-analysis (13) of the results from six clinical trials on clofibrate also found no excess cancer mortality (9). It therefore appears that there are no remaining concerns about the human carcinogenic potential of the clinical PPs and that the rodent liver effects have been set aside as probable laboratory curiosities. However, this is not true for the nonpharmaceutical PPs; several regulatory agencies continue to be concerned about their carcinogenic potential to the human liver. The unease of these agencies is due to their belief that in the absence of a definitive mechanism of PP-induced rodent liver carcinogenesis, it is not possible to make a clear statement on the human safety of these chemicals. Nonetheless, there are now several strong lines of evidence that PPinduced rodent liver carcinogenesis is not relevant to humans, which supports the conclusion drawn by IARC for the clinically used PPs and the recent IARC decision to downgrade DEHP from group 2B (possibly carcinogenic to humans) to group 3 (3).These lines of evidence are as follows: • Direct genetic toxicity has been eliminated as a common mechanism of carcinogenic action for PPs in general (14). Thus, rodent hepatocarcinogenicity must occur via a nongenotoxic mechanism that correlates with peroxisome proliferation, although, as pointed out by Melnick (1), the hepatocarcinogenicity is unlikely to be caused by peroxisome proliferation per se, as initially suggested (15). • There are marked species differences in the induction of peroxisomes, with human hepatocytes being resistant (8,16,17). These data provide evidence that the phenomenon of PP-induced peroxisome proliferation is rodent specific. • PPs suppress rodent hepatocyte apoptosis (18–20) and induce rodent hepatocyte replication (8). This duality of effects provides a plausible mode of rodent carcinogenic action based on liver growth perturbation (21,22). As well as being resistant to peroxisome proliferation, human hepatocytes are also resistant to PP-mediated induction of replication and suppression of apoptosis (8,16,17). Whatever the precise mechanism by which PPs induce rodent liver cancer, rodent liver peroxisome proliferation, induction of the peroxisomal gene acyl CoA oxidase (ACO) (23), hypertrophy (24), and carcinogenicity (25) are all mediated through activation of the peroxisome proliferator-activated receptor (PPARα). This is illustrated dramatically by the absence of all of these responses in PPARα knockout mice treated with the PPs DEHP or Wyeth-14,643 (24–26). Although human liver expresses around 10-fold less PPARα mRNA than the rodent liver (27,28), evidence suggests that the hypolipidemic effects of the fibrate drugs in humans are also mediated by activation of PPARα, leading to regulation of the apolipoprotein (Apo) genes such as ApoA1 (29). Thus, PPARα levels in human liver may be sufficient to mediate PPinduced hypolipidemia, but insufficient to activate the gene battery associated with rodent peroxisome proliferation and cancer (30). In addition to these quantitative data, there are species differences in the molecular sequence of the PPARα response elements (PPREs) located upstream of genes associated with rodent peroxisome proliferation such as ACO. In the rat, ACO responds to PPs via a functional PPRE, whereas the equivalent gene in humans does not (31,32). Thus, the human ACO gene does not respond to PPs even in the presence of excess PPARα (31–33). Similarly, recent data have shown that PPARα cannot induce the battery of peroxisome proliferation-associated genes in human hepatoma cells (33,34). Conversely, the human ApoA1 gene is responsive to fibrate PPs, whereas the equivalent rat gene is not (35). Such findings isolate the human hypolipodaemic effects of PPs from the rodent cancer effects. Although the precise mechanism of the carcinogenic action of PPs in the rodent liver remains to be determined, all of the phenomena associated with this response of rodent hepatocytes (peroxisome proliferation, ACO gene expression, induction of cell proliferation, and the suppression of apoptosis) are absent in human hepatocytes. This body of data provides a plausible mode of carcinogenic action for the rodent liver,

Melnick (1) recently suggested that because peroxisome proliferation is not established as an obligatory step in the carcinogenicity of peroxisome proliferators (PPs), the proposal that the peroxisome proliferator di(2-ethylhexyl)phthalate (DEHP) poses no carcinogenic risk to humans (2) due to species differences in peroxisome proliferation should be viewed as an unvalidated hypothesis (1). In this context, Melnick (1) raised the recent downgrading by the International Agency for Research on Cancer (IARC) of DEHP (3) to "not classifiable as to its risk to humans (group 3)" based on their conclusion that it produces rodent liver tumors by a mechanism involving peroxisome proliferation, which they judged to be not relevant to humans (3). As illustrated by Melnick (1), there is a large body of data to correlate the phenomenon of rodent liver peroxisome proliferation with rodent liver cancer but "published studies have not established peroxisome proliferation per se as an obligatory pathway on the carcinogenicity of DEHP" (1). This focuses attention on the need, as also suggested by O'Brien et al. (4), for a fundamental review of how PPs induce liver cancer in rodents and the relevance of these rodent tumors for humans.
There are two distinct usage patterns for PPs: as drugs such as clofibrate for the treatment of hypolipidemia (5) and in nonclinical applications such as the plasticiser DEHP. Most PPs are carcinogenic to the rodent liver, and the task of assessing their human carcinogenic potential has fallen to different regulatory agencies, depending on the primary usage pattern of the particular PP in question. There seems to be little regulatory concern regarding the safety of the clinical PPs, yet continuing uncertainty regarding the safety of the nonpharmaceutical PPs. This presents an untenable situation that we suggest is unjustified.
Hypolipidemic fibrates such as clofibrate and gemfibrozil have been used extensively over the past 20 years to treat cardiovascular disease and are enjoying a revival due to recent reconfirmation of efficacy (6). However, by 1980 several of these agents had become associated with a rodent-specific response known as hepatic peroxisome proliferation (7), a property shared by a number of nonpharmaceutical chemicals (8). Additionally, a link between rodent liver peroxisome proliferation and an increased risk of rodent liver carcinogenesis was emerging (7). Nonetheless, clinical side effects of the fibrate PPs are rare, and analyses of causes of death during treatment show no evidence of an adverse effect and no evidence of an increase in malignant disease compared to the normal population (5). Specifically, the carcinogenic risk to humans of gemfibrozil and clofibrate has been formally evaluated by IARC (9,10), and in the case of clofibrate, for which most clinical data exist, IARC (9) concluded that "the mechanism of liver carcinogenesis in clofibrate-treated rats would not be operative in humans." This conclusion was based on the observation that clofibrate causes peroxisome proliferation and cell proliferation in rodent but not human hepatocytes and on the results of extensive epidemiologic studies, particularly the World Health Organization trial on clofibrate that included 208,000 man-years of observation (11,12). Further, a meta-analysis (13) of the results from six clinical trials on clofibrate also found no excess cancer mortality (9).
It therefore appears that there are no remaining concerns about the human carcinogenic potential of the clinical PPs and that the rodent liver effects have been set aside as probable laboratory curiosities. However, this is not true for the nonpharmaceutical PPs; several regulatory agencies continue to be concerned about their carcinogenic potential to the human liver. The unease of these agencies is due to their belief that in the absence of a definitive mechanism of PP-induced rodent liver carcinogenesis, it is not possible to make a clear statement on the human safety of these chemicals. Nonetheless, there are now several strong lines of evidence that PPinduced rodent liver carcinogenesis is not relevant to humans, which supports the conclusion drawn by IARC for the clinically used PPs and the recent IARC decision to downgrade DEHP from group 2B (possibly carcinogenic to humans) to group 3 (3).These lines of evidence are as follows: • Direct genetic toxicity has been eliminated as a common mechanism of carcinogenic action for PPs in general (14). Thus, rodent hepatocarcinogenicity must occur via a nongenotoxic mechanism that correlates with peroxisome proliferation, although, as pointed out by Melnick (1), the hepatocarcinogenicity is unlikely to be caused by peroxisome proliferation per se, as initially suggested (15). • There are marked species differences in the induction of peroxisomes, with human hepatocytes being resistant (8,16,17). These data provide evidence that the phenomenon of PP-induced peroxisome proliferation is rodent specific. • PPs suppress rodent hepatocyte apoptosis (18-20) and induce rodent hepatocyte replication (8). This duality of effects provides a plausible mode of rodent carcinogenic action based on liver growth perturbation (21,22). As well as being resistant to peroxisome proliferation, human hepatocytes are also resistant to PP-mediated induction of replication and suppression of apoptosis (8,16,17).
Whatever the precise mechanism by which PPs induce rodent liver cancer, rodent liver peroxisome proliferation, induction of the peroxisomal gene acyl CoA oxidase (ACO) (23), hypertrophy (24), and carcinogenicity (25) are all mediated through activation of the peroxisome proliferator-activated receptor (PPARα). This is illustrated dramatically by the absence of all of these responses in PPARα knockout mice treated with the PPs DEHP or Wyeth-14,643 (24-26).
Although human liver expresses around 10-fold less PPARα mRNA than the rodent liver (27,28), evidence suggests that the hypolipidemic effects of the fibrate drugs in humans are also mediated by activation of PPARα, leading to regulation of the apolipoprotein (Apo) genes such as ApoA1 (29). Thus, PPARα levels in human liver may be sufficient to mediate PPinduced hypolipidemia, but insufficient to activate the gene battery associated with rodent peroxisome proliferation and cancer (30). In addition to these quantitative data, there are species differences in the molecular sequence of the PPARα response elements (PPREs) located upstream of genes associated with rodent peroxisome proliferation such as ACO. In the rat, ACO responds to PPs via a functional PPRE, whereas the equivalent gene in humans does not (31,32). Thus, the human ACO gene does not respond to PPs even in the presence of excess PPARα (31-33). Similarly, recent data have shown that PPARα cannot induce the battery of peroxisome proliferation-associated genes in human hepatoma cells (33,34). Conversely, the human ApoA1 gene is responsive to fibrate PPs, whereas the equivalent rat gene is not (35). Such findings isolate the human hypolipodaemic effects of PPs from the rodent cancer effects.
Although the precise mechanism of the carcinogenic action of PPs in the rodent liver remains to be determined, all of the phenomena associated with this response of rodent hepatocytes (peroxisome proliferation, ACO gene expression, induction of cell proliferation, and the suppression of apoptosis) are absent in human hepatocytes. This body of data provides a plausible mode of carcinogenic action for the rodent liver, which, when coupled with the clinical epidemiology showing an absence of a human cancer risk, provides substantial weight of evidence that the PP class of nongenotoxic rodent hepatocarcinogens does not pose a potential cancer hazard to the human liver. With respect to available human data, Roberts et al. refer to clinical studies on clofibrate and gemfibrozil in male subjects as providing evidence of "an absence of a human cancer risk" for fibrate PPs. In a previous review of these data, Ashby et al.

Ruth Roberts
(2) noted a small increase in basal cell carcinomas of the skin in gemfibrozil-treated patients, but concluded that the epidemiologic studies on hypolipidemic drugs "are of limited value only, because of the short time periods involved." The World Health Organization trial on the prevention of ischemic heart disease by clofibrate was last updated in 1982 and included 13 years of observation, 5 years during the treatment period plus 8 years of follow-up (3). That study revealed an excess of deaths for nonmalignant diseases of the liver, gall bladder, and intestines in the clofibrate-treated group compared to controls. Because the Correspondence latency period for clinical manifestation of cancer may be 20 years or more postexposure, the current data are insufficient to permit a definitive conclusion on the presence or absence of a causal association between exposure to fibrate lipid-lowering drugs and human cancer (4). It is not clear why Roberts et al. claim that "there are no remaining concerns about the human carcinogenic potential of the clinical PPs" inasmuch as a previous review from their laboratory of the same epidemiologic studies found these data to be of limited value due to short study durations (2), and there are no available studies on female cancer risk. Furthermore, in contrast to the view given by Roberts et al., the Physicians' Desk Reference (5) warns of the tumorigenicity of clofibrate [and of gemfibrozil (6)] in rodents and the possible increased risk of malignancy associated with clofibrate in the human.
For DEHP, no epidemiologic studies have been reported.
Roberts et al. also conclude that PPs would "not pose a potential cancer hazard to the human liver." However, as I noted previously (1), it might not be appropriate to expect exact site correspondence for effects of PPs in rodents and humans because of species differences in tissue expression of PPARs. For example, the demonstration of a functional PPAR in human breast cancer cell lines (7) and the finding of enhanced cell proliferation by DEHP in human breast cancer cells (8) indicate a possible breast cancer risk. Furthermore, liver is not the only target organ of tumor induction by hepatic PPs. Several PPs induce tumors of the testis and pancreas in laboratory animals, and tumor induction at these sites occurs without induction of peroxisomes in these affected organs (9).
Roberts et al. claim that available data provide "a plausible mode of carcinogenic action" for PPs, which is based on induction of hepatocyte proliferation and suppression of apoptosis. The latter effects are reported to be absent in cultured human hepatocytes. However, as noted in the Physicians' Desk Reference (5,6), changes in peroxisome morphology and numbers have been observed in humans after treatment with several members of the fibrate class, including clofibrate, when liver biopsies were compared before and after treatment in the same individual.
Several additional issues influence the plausibility of the mode of action for liver carcinogenicity claimed by Roberts  The inappropriate dismissal of positive animal cancer findings in assessments of human risk could have serious health consequences. Protection of public health requires rigorous testing and validation of mechanistic hypotheses rather than reliance on assertions of plausibility.

Ronald Melnick National Institute of Environmental
Health Sciences Research Triangle Park, NC E-mail: melnickr@niehs.nih.gov