Dioxins

5. How can dioxin exposure be linked to health effects?

• 5.1 Is there a known relationship between dioxin exposure and cancer?
  • 5.1.1 Human cancer
  • 5.1.2 Experimental studies
• 5.2 Can a model predict non-cancer effects?
• 5.3 How are dioxins mixtures accounted for?
  • 5.3.1 What are TEFs and TEQs?
  • 5.3.2 What are the limits of the TEF concept?

The source document for this Digest states:

DOSE-EFFECT MODELLING

The key issues concerning modelling which were discussed included: the appropriateness of the data sets for the relevant endpoints, the kind of model used, the uncertainties in the model, and the transparency of the model with regard to the base assumptions. The model requires validation, e.g. with multiple data sets, before acceptance; this is often not done. In addition, use of raw data rather than summary data will substantially improve the models.

5.1 Is there a known relationship between dioxin exposure and cancer?

• • 5.1.1 Human cancer
  • 5.1.2 Experimental studies

5.1.1 Human cancer

The source document for this Digest states:

The choice of data sets is determined to a large extent by the richness and completeness of the data. Therefore, in modelling human cancer, the most useable data sets are the industrial cohorts discussed by IARC in their 1997 monograph. In all these cohorts, exposure is back calculated from serum levels measured after the exposure had ceased. Average body burden over a lifetime was estimated assuming constant background levels of exposure before and after employment, and an assumption of continuous exposure to TCDD alone in the work place. The back calculation from the lipid adjusted serum levels observed after the end of the industrial exposure assumed a constant half-life of 7.1 years. A multiplicative linear hazard model was used to estimate a slope, using a maximum likelihood estimate. The ED01 to maintain the steady state body burden associated with a 1% excess risk over a lifetime results in a body burden of 3 -13 ng/kg, which is associated with a daily dose in the range of 2-7 pg/kg/day. If risk is related to peak exposure, rather than to continuous exposure, the estimate would be low. If the majority of exposure in the studied cohorts occurred within the earliest year instead of uniformly over the span of employment, the ED01 would increase by approximately a factor of three. It is important to note that the average exposure over time is not very different from these values (e.g., in the NIOSH cohort in the lower "exposure" group, those with less than one year of occupational exposure, resulted in an average lifetime body burden of approximately 10 ng/kg). However, the model does assume linearity within the range of the data, which is likely to provide a conservative position.

Source & ©: WHO-IPCS  Assessment of the health risk of dioxins:
re-evaluation of the Tolerable Daily Intake (TDI) page 13

5.1.2 Experimental studies

The source document for this Digest states:

Experimental cancer studies

Two approaches to modelling were used: mechanistic and curve fitting. It is important to note that the mechanistic model had many assumptions and that other assumptions may be equally plausible and can lead to other models that adequately describe the data sets.

Mechanistic Model

Molecular, cellular, and promotion data was used to predict the incidence of liver Tumours in the female Sprague Dawley rat observed in the Kociba study. The model assumed that dioxin exposure induced increased cellular proliferation and indirectly led to an increase in mutation rate due to induction of hepatic enzymes leading to oxidative stress. The hypothesis of no mutational effect was tested and could not be rejected for this model. Each part of the model was allowed to vary independently and was not constrained a priori to either a linear or nonlinear association. The ultimate best fit linear model lead to an excess 1% lifetime cancer risk associated with a steady state body burden of 2.6 ng/kg, resulting from a daily exposure of 150 pg/kg/day. Thus, while the ED01 on a daily dose base for rodents is much higher than that for humans, the steady state body burden for rodents is in the same range as that estimated for humans. This is due to the pharmacokinetic differences between the species.

Curve Fitting Model

The Armitage-Doll model was used to calculate a shape parameter to describe the results of multiple animal tumour studies in both rats and mice. The shape of the curve could be fitted by either linear or various non-linear power functions. 8 Out of 13 studies were best fitted by a linear model. However, the data may be described by a non-linear model. The ED01 based on a steady state body burden ranged from 10 ng/kg to 746 ng/kg, associated with daily doses of 1.3 ng/kg/day to 41.4 ng/kg/day. If these results are compared to the human cancer estimates, the body burdens again are similar, but the daily doses, as expected from pharmacokinetic considerations, are much higher in rats and mice. The animal estimates do not involve a large extrapolation to go from the observed data to a calculated ED01.

Source & ©: WHO-IPCS  Assessment of the health risk of dioxins:
re-evaluation of the Tolerable Daily Intake (TDI) page 14

5.2 Can a model predict non-cancer effects?

The source document for this Digest states:

Non-cancer

No models have been evaluated for non-cancer effects in humans. The animal data sets that were modelled involved those with at least four dose groups and those in which a maximal response was at least approached, if not achieved. The Hill equation was used with non-linear least squares to fit the parameters and weighted for the observed variance and the shaping function was applied to assess linearity or non-linearity. For multiple dose studies, the average body burden at steady state was calculated as in the cancer studies. For the bolus studies, the body burden was assumed to be equivalent to the administered dose or that estimated by calculation at the time of response measurement based on first order elimination kinetics.

Modelling of 45 non-cancer studies in rodents demonstrated that 21 were best fitted by a near-linear model, while 24 demonstrated non-linearity. The biochemical endpoints were mainly linear; but most of the clearly adverse endpoints were non-linear. However, the decline in sperm count following prenatal exposure was linear. When the ED01 for the biochemical endpoints was compared with the observed LOEL, the ED01 was often higher than the measured response. This may reflect the sensitive measurements that can be made for biochemical responses. For the decrease in sperm count, the ED01 was lower than the LOAEL. This may reflect measurement sensitivity, study design, and complexity of response. In some studies, estimation of the maximum response was problematic and the biological plausibility of the curve fits was unclear, underlying the need for mechanistic models for non-cancer endpoints.

The utility of the models is that they allow a common method of comparison, e.g. a 1% excess response. Use of this methodology allows comparison across responses. This benchmark methodology is less sensitive to the "ability to detect" a response based on the different study designs used to assess different endpoints. It is important to note that the ED01 for many of the non-cancer endpoints ranged for <1 to 100 ng/kg body burden. This is also true for cancer.

Regarding the importance of modelling to the human risk assessment for dioxin, the predictions of effects were compared with the actual data. At times, obvious discrepancies arise, leading to caution in the use of models. While recognizing that modelling is not suitable for human risk assessment of dioxins yet, it provides additional insights into the observational data, and adds to the transparency of the review.

Source & ©: WHO-IPCS  Assessment of the health risk of dioxins:
re-evaluation of the Tolerable Daily Intake (TDI) page 14-15

5.3 How are dioxins mixtures accounted for?

• • 5.3.1 What are TEFs and TEQs?
  • 5.3.2 What are the limits of the TEF concept?

5.3.1 What are TEFs and TEQs?

The source document for this Digest states:

APPLICABILITY OF TEF CONCEPT

The complex nature of polychlorinated dibenzo-p-dioxin (PCDD), dibenzofuran (PCDF), and biphenyl (PCB) mixtures complicates the risk evaluation for humans. For this purpose the concept of toxic equivalency factors (TEFs) has been developed and introduced to facilitate risk assessment and regulatory control of exposure to these mixtures. TEF values for individual congeners in combination with their chemical concentration can be used to calculate the total TCDD toxic equivalents concentration (TEQs) contributed by all dioxin-like congeners in the mixture using the following equation which assumes dose additivity:

The majority of studies assessing the manner in which binary and complex mixtures of dioxin-like PCDD, PCDF and PCB congeners interact to cause toxicity have demonstrated that the interaction does not deviate significantly from dose additivity. This includes investigations conducted in various classes of vertebrates (fish, birds and mammals) and on environmental relevant mixtures. TEFs for dioxin-like compounds apply only to AhR-mediated responses. The criteria for including a compound in the TEF scheme for dioxin-like compounds are that the compound must:

• Show a structural relationship to the PCDDs and PCDFs
• Bind to the Ah-receptor
• Elicit Ah receptor-mediated biochemical and toxic responses
• Be persistent and accumulate in the food chain.

To reassess the TEFs for mammals a WHO expert group recently applied a tiered approach in which results of animal toxicity studies, especially those involving (sub)chronic exposure, were given significantly more weight than results of in vitro or biochemical studies. The results of this activity are summarized in Table 3.

Source & ©: WHO-IPCS  Assessment of the health risk of dioxins:
re-evaluation of the Tolerable Daily Intake (TDI) page 15-16

5.3.2 What are the limits of the TEF concept?

The source document for this Digest states:

While additivity predominates in the majority of experimental studies, non-additive interactions of PCDDs, PCDFs and PCB mixtures have been reported at greater than environmental levels of exposure. These non-additive effects are considered to be due to multiple mechanisms of action of individual congeners and/or to pharmacokinetic interactions. For the mono- ortho PCBs especially, certain endpoints such as carcinogenicity, porphyrin accumulation, alterations in circulating thyroid hormone concentrations and neurotoxicity could arise by both Ah-receptor-mediated and non-Ah receptor-mediated mechanisms.

In addition, non-Ah receptor-mediated mechanisms of action of the mono-ortho PCBs may be shared by certain di-, tri-, and tetra-chloro ortho-substituted PCBs. This increases uncertainty in the use of TEFs for mono-ortho PCBs. While recognizing that these and other uncertainties exist in the use of the TEF concept for human risk assessment, pragmatically it remains the most feasible approach. Use of TCDD alone as the only measure of exposure to dioxin-like PCDDs, PCDFs and PCBs severely underestimates the risk to humans from exposure to these classes of compounds. Thus, the TEF approach is recommended for expressing the daily intake in humans of PCDDs, PCDFs, non-ortho PCBs and mono-ortho PCBs in units of TCDD equivalents (TEQs) for comparison to the tolerable daily intake (TDI) of TCDD.

Source & ©: WHO-IPCS  Assessment of the health risk of dioxins:
re-evaluation of the Tolerable Daily Intake (TDI) page 16-17