4.1. Combination of Animal and Human Toxicity Data Demonstrate the Relevance of Analogous Manifestation of Toxicity
The carcinogenicity of PA contaminated food has never been investigated in long-term animal studies. Chronic animal toxicity of single PA or extracts containing PA results in hepatic vein occlusions, mutagenic, teratogenic, and carcinogenic effects [6
]. In rodents, the signal tumor of chronic PA poisoning is hepatic hemangiosarcoma [64
]. It is concluded that inter-species qualitative comparability of metabolism and toxicity is feasible but huge quantitative differences are obvious. It is an open question whether the differences in metabolic cell response between rodents and humans should disqualify animal data for quantitative risk assessment in humans [53
Risk assessment extrapolating animal findings to humans quantitatively implies major uncertainties. The reliability of animal data to predict outcome in humans has been critically reviewed by different working groups [66
]. In general, comparisons between animal and human data show that findings in animals were not reliably replicated in humans. Heywood reported that animal experiments poorly predict adverse effects of pharmaceuticals and concluded “the best guess for the correlation of adverse reactions in man and animal toxicity data is somewhere between 5 and 25%” [66
]. Olson et al. [67
] analyzed the concordance of the toxicity of pharmaceuticals in humans and animals in a multinational pharmaceutical company survey compiling data from 12 companies with 150 compounds in clinical development. Non-rodent experiments predicted 63% of human organ toxicity, studies in rodents 43%, respectively. These discrepancies are thought to be due to differences in physiology and metabolism.
The value of animal use in the field of regulatory toxicology relies on a codified set of highly standardized acute, repeated dosing, and long-term animal studies, many of them developed in the 1960s. Their relevance has been scrutinized as more modern concepts became available to predict human outcome after exposure to xenobiotics [68
]. However, if rare tumor types found in humans correspond to tumors in animal tested with the same suspected carcinogenic agent this is a reasonable empirical argument for causality. Identical organotropism and comparable tumor pathology have been demonstrated for about 50% of the chemicals, which are accepted by the World Health Organization (WHO) as carcinogenic to humans [71
Hemangiosarcoma of unknown origin (“spontaneous tumors”) occur frequently in mice, are unusual in rats [46
], and are well known in cats and special breeds of dogs [72
]. Comparative gene expression profiling studies in Golden Retrievers suggest genetic background to mold occurrence, phenotype and biological behavior of sporadic hemangiosarcomas, at least in dogs [73
]. These findings reveal that the risk of developing hemangiosarcoma could depend upon genetic disposition.
Carcinogens associated with liver carcinomas and hemangiosarcomas of the liver in humans include Alpha-emitters, vinyl chloride, and arsenic. Intravenous injection of thorium dioxide induced hepatocellular carcinomas in hamsters and liver carcinomas, intrahepatic bile-duct carcinomas, and hemangiosarcomas in rats [71
]. Hepatic angiosarcomas can be reproducibly induced in rodents by exposure to vinyl chloride in the air [74
Rodent liver angiosarcomas develop following exposure to DNA reactive, genotoxic chemicals as well as following chronic exposure to non-DNA reactive, non-genotoxic xenobiotics. Cohen et al. [53
] have composed different modes of action for different mechanisms leading to hepatic angiosarcoma in rodents. Dysregulated angiogenesis can lead to local hypoxia followed by an overexpression of hypoxia-inducible factor 1-alpha (HIF) and vascular endothelial growth factor (VEGF), macrophage activation and thus locally increased interleukin-6 (IL-6) concentration. Both, VEGF and IL-6, can stimulate endothelial cell proliferation. The European Medicines Agency (EMA) statement [4
] describes stimulation of cell proliferation following local hypoxia as a mode of action for hepatic angiosarcoma in rodents following exposure to PA. There are several mechanisms by which carcinogens cause cancer [75
]. The mechanisms include DNA-methylation, histone methylation and acetylation, micro-RNA expression, receptor binding (aryl hydrocarbon, nuclear, peroxisome proliferator, and hormonal receptors), cytotoxicity, hormonal imbalance, chronic inflammation, oxidative stress, inhibition of apoptosis, disturbances in cell to cell communication, and induction of cell proliferation [76
]. Dose thresholds have been shown experimentally for tumor induction by non-genotoxic tumor development. There is no proof that genotoxic compounds lead to carcinogenicity only through gene mutations. The hypothesis for genotoxic compounds assumes that one DNA mutation is sufficient to initiate a cancer cell. This hypothesis has not yet incorporated the better understanding of molecular mechanisms of DNA repair and apoptosis [77
]. There appear to be significant tissue-specific and species-specific differences between the responses of endothelial cells to xenobiotics. It remains unclear whether these differences are qualitative and/or quantitative in nature. To ultimately address the question of interspecies comparisons a better understanding of biological similarities and differences between rodents and humans is needed.
Corresponding findings in (experimental) animals and humans strengthen the plausibility of a causative relationship and thereby move up the importance of the experimental findings in the hierarchy of evidence [62
]. As shown in the paragraph on the etiology of PA induced toxicity experimental data, livestock poisoning and accidental human intoxications are comparable in acute and subacute toxicity. After PA poisoning many patients recover almost completely if the alkaloid intake is discontinued; cases of liver fibrosis and cirrhosis have been reported but no case of human liver malignancies has been linked to PA ingestion [13
In regulatory toxicology, we assume threshold doses for most toxic effects. PA have the potential to produce several forms of toxicity. Acute intoxications are a result of overload of metabolic detoxification pathways and can result in secondary consequences such as hepatic occlusive disease. These effects assume a dose threshold below which no signs of toxicity occur. Due to metabolic detoxification of low doses no manifestation of acute and subacute liver toxicity is expected in the general population of developed countries with high agricultural standards [2
]. This still leaves the question unanswered: what is the risk of possible low-dose, long-term exposure to PA by food contaminants?
Different approaches have been suggested to translate animal doses to human exposure risks. These calculations try to bridge speciesspecific differences, e.g., in genetic diversity, life expectancy, and basic metabolic pathways [78
]. FDA has published a guidance document describing the use of standard specific factors that allow conversion of animal doses in (mg/kg) to human doses in (mg/kg) using the body surface area as the common denominator [79
European regulators favor the margin of exposure approach to translate doses used in animal experiments to human exposure. Safety factors (in general of 10,000) are introduced to compensate for the knowledge gap in translating the benchmark dose lower confidence limit 10% (i.e., 95% confidence limit of the lowest dose showing a specific toxic effect in 10% of the exposed animals) in animals to the human situation [43
EFSA has used the margin of exposure approach to translate the benchmark lower dose confidence limit for a 10% excess cancer risk of 70 µg/kg body weight per day for induction of liver hemangiosarcomas by lasiocarpine in male rats to a reference point for comparison with estimated dietary human exposure [43
]. The quality of estimates in translating experimental findings to the human situation very much determines how efficiently the toxicity of PA in humans is controlled and prevented.
Since in carcinogenesis experiments PA induce hepatocellular carcinomas and hemangiosarcomas of the liver in the rat, it is worthwhile to look for a match with analogous malignancies in humans.
4.2. Epidemiological Evidence
In 2012, 782,000 new cases of liver cancer were diagnosed worldwide [80
]. In the United States, 39,230 new cases of primary liver cancer (hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (HCV)) are estimated for 2016 [81
]. Other sources report about 21,000 men and 8000 women to newly develop primary liver cancer annually, while 27,170 people (18,280 men and 8890 women) are estimated to die of primary liver cancer in 2016 [81
]. Thus, for calculation purposes, 25,000 deaths/year and twice as many men affected than women seem reasonable figures.
The overall rates of HCC tripled between 1975 and 2005. This increase is explained by the rise in hepatitis B and C virus carriers [83
]. Major risk factors for HCC are chronic infection with hepatitis C and hepatitis B virus, excessive alcohol consumption, obesity, diabetes, metabolic syndrome, all can lead to liver cirrhosis. Smoking and eating foods with aflatoxin burden have also been associated with HCC [84
]. HCV has been associated with several diseases of the biliary tract or liver, such as primary sclerosing cholangitis, Caroli disease, cholelithiasis, cholangitis, liver fluke, and inflammatory bowel disease.
No epidemiological data exist to associate chronic low dose PA exposure with human disorders [2
]. The total number of primary liver cancer minus known attributable risks leaves only a low to very low risk for unknown causes. Theoretically, a fraction thereof could be associated with PA intake. The risk estimate attributable to hepatic infections accounts for well over 80% of all primary liver cancers. Thus, less than 20% are left for all other well established factors and for cancer of yet unexplained etiology, including a theoretical minority portion associated to chronic PA exposure) [82
By subtracting the cases attributable to hepatitis (25,000 − >20,000 = <5000), a theoretical number of <5000 new cases/year can be calculated for the US population. Within these <5000 cases must be those due to known risk factors besides hepatitis B and C. The possible impact of PA contamination stays speculative, but is negligible compared to the impact of the sum of tumors probably attributable to already well-recognized risks. Thus far, no clinical association has been described between human cancer and exposure to PA. Based on the extensive reports on the outcome of human exposure available in the literature, Prakash, Pereira, Reilly, and Seawright [91
] concluded that, while humans face the risk of veno-occlusive disease and childhood cirrhosis by PA poisoning, PAs are not carcinogenic to humans. 2500 annual cases linked to PA uptake seem clearly above a reasonable educated guess.
Human hepatic hemangiosarcoma (HHA) is a rare tumor [92
], one source states HHA to account for only 2% of primary liver malignancies. Other authors estimated that only about 10 to 25 such cases occur each year in the United States [94
]. Zochetti reported its frequency to be 2.5 cases every 10,000,000 persons [96
]. For calculative purposes, an annual figure of 100 HHA for the US population seems a reasonable upper end estimate of the order of magnitude.
Established etiologic factors for HHA are thorium dioxide in angiography [97
], exposure to vinyl chloride monomer at the work place [98
], and ingested inorganic arsenic [99
]. Anabolic-androgenic steroids have been associated, too [100
]. No evidence of a relationship between environmental exposure to vinyl chloride monomer and angiosarcoma of the liver has been built [93
]. An association with low dose, long-term PA exposure is without any epidemiological evidence [91
]. The risk to experience HHA due to PA contaminated biologicals is negligible. If it does exist, it must be minimal compared to other largely accepted risks of daily living.
The overall comparison of experimental evidence, case reports and epidemiological evidence highlights that qualitative acute and subacute toxicity data correspond. High PA doses lead to a comparable specific liver pathology across species. However, regarding long term, low dose exposure no link has been found between the liver carcinogenicity seen in rat animal models and the human situation. Since liver hemangiosarcomas are signal tumors for human carcinogens, the missing accordance between rats and humans in PA tumorigenesis does not support the hypothesis that low dose PA contamination of food implies a considerable human cancer risk. In contrast, a solid data base exists demonstrating health protecting effects of T&HI.
While the PA-tumor link is unproven in humans, there exists a solid and consistent data base for the health-protecting effects of T&HI.