A dietary exposure assessment is the process of estimating how much of a food chemical a population, or population sub group, consumes. Dietary exposure to food chemicals is estimated by combining food consumption data with food chemical concentration data (see Equation A2 1). The process of doing this is called ‘dietary modelling’.
Equation A2 1: Dietary exposure calculation
Dietary exposure = Σ(food chemical concentration x food consumption)
FSANZ’s approach to dietary modelling is based on internationally accepted procedures for estimating dietary exposure to food chemicals. Different dietary modelling approaches may be used depending on the assessment, the type of food chemical, the data available and the risk assessment questions to be answered.
For the Plasticisers Survey, dietary modelling was conducted using FSANZ’s custom developed computer program Harvest, which was designed to automate dietary exposure calculations. Harvest multiplied the chemical concentration for each food consumed in the national nutrition survey with the amount of that food that each survey respondent consumed to estimate each individual’s exposure to that chemical from each food. Once this had been completed for all foods determined as containing a particular chemical, the total amount of the chemical consumed from all foods was summed for each individual. Population statistics (e.g. mean and 90th percentile exposures) for each age group were derived from the individuals’ ranked exposures. Where the results are expressed on a bodyweight basis, each individual’s exposure from all foods was divided by their own bodyweight before population summary statistics were derived.
The use of Harvest for dietary modelling brings many benefits. Harvest enables the dietary exposure assessments to be conducted using actual diets, as recorded in national nutrition surveys, in place of the ‘average’ diets which were used prior to the 19th ATDS. The use of specific food consumption data greatly improves the reliability and accuracy of the dietary exposure estimates and takes account of the different eating patterns of consumers.
Once dietary exposure to the chemical from the total diet had been estimated, this is compared to relevant HBGVs to assess the potential risk to human health.
Further detailed information on conducting dietary exposure assessments at FSANZ is provided in Principles and Practices of Dietary Exposure Assessment for Food Regulatory Purposes (FSANZ 2009), available at:
Number of respondents in each of the population groups assessed
A range of population groups were assessed including infants, children and adults. Table A2.1 shows the number of individuals in each age group assessed.
Table A2.1: Number of respondents and mean bodyweight for each age group assessed
No. of respondents
(2 day average)
Mean bodyweight (kg)
17 years & above*
* derived using the 2011–12 Australian National Nutrition and Physical Activity Survey
Dietary exposure assessments were conducted for infants and children as separate groups as they generally have higher exposures because they consume more food on a kilogram body weight basis compared to adults.
Food consumption data
The 2011–12 Australian National Nutrition and Physical Activity Survey (NNPAS) undertaken by the Australian Bureau of Statistics is the most recent food consumption data for Australia. This survey includes dietary patterns of a sample of 12,153 Australians aged from 2 years and above. The survey used a 24-hour recall method for all respondents, with 64% of respondents also completing a second 24-hour recall on a second, non-consecutive day. The data were collected from May 2011 to June 2012 (with no enumeration between August and September 2011 due to the Census). Day 1 and Day 2 24-hour recall data for respondents were used for this assessment. There were 7735 respondents with two days of data, and these were averaged for estimating dietary exposure for this assessment. A separate set of sample weights are used for the 7735 respondents with two days of data to ensure that when using this subset, they are representative of the Australian population. Consumption and respondent data from the survey were incorporated into the Harvest program from the Confidentialised Unit Record Files (CURF) data set (ABS 2014).
As no food consumption data from the NNPAS were available for children under two years of age, a model diet was constructed to enable dietary exposure to be assessed for infants.
Construction of the model diet for 9 month old infants
By the age of 9 months, most infants will be consuming a mixed diet and will be exposed to food chemicals from a range of foods in addition to human breast milk and/or infant formula. To enable food chemical exposures for 9 month old infants to be estimated a model diet was constructed. The model diet was based on recommended energy intakes, mean bodyweight, the proportion of milk and solid foods in the diet for a 9 month old infant and 2011-12 NNPAS data on foods consumed by a two year old child. The recommended energy intake for a 9 month old boy (FAO 2004) at the 50th percentile weight (WHO 2006) (2936 kJ/day) was used as the basis for the model diet. Boys’ weights were used as boys tend to be heavier than girls at the same age and therefore have higher energy and food requirements. The bodyweight of a 50th percentile 9 month old boy was 8.9 kg.
It was assumed that 50% of energy intake was derived from infant formula and 50% from solids and other fluids (Hitchcock et al. 1986). The patterns of consumption of a two year old child from the 2011-12 NNPAS survey were scaled down and used to determine the 50% solid and other fluids portion of the 9 month old infant’s diet. As two year olds consume many foods that are not appropriate for infants, these foods need to be taken out of the infant diet and the energy intake from those foods attributed to the remaining foods. Certain foods such as nuts and coffee were removed from the diet since nuts are not recommended for infants because of choking risk (NHMRC 2012) and coffee is unsuitable for infant consumption (ACT Community Care 2000). Bran is not recommended in the diet of infants due to the potential interference with the absorption of minerals (The Children's Hospital at Westmead 2008) and to the immaturity of the infant gut (H.J Heinz 2010a). For this reason, consumption of breakfast cereals is usually assumed to be in the form of either infant cereal or single grain breakfast cereals. As no infant specific cereal was sampled for this survey, consumption of mixed grain cereals could not be substituted to it. Furthermore, as the packing type was similar and the survey’s focus was on migrating chemicals from packaging, mixed grain breakfast cereals were not excluded from the model infant diet. Since cow’s milk is not recommended as the main milk source for children aged less than 12 months of age (NHMRC 2012; H.J Heinz 2010b), all milk consumption was assumed to be in the form of infant formula.
As the model diet is based on mean food consumption amounts only for all nutrition survey respondents, a distribution of food consumption was not available and hence, a distribution of food chemical exposures was not able to be produced. Therefore, the 90th percentile dietary exposures were estimated using the calculation shown in Equation A2 2 below. Exposures were then compared to the HBGVs where relevant.
Equation A2 2: 90th percentile dietary exposure calculation for the 9 month old infant model diet
* (WHO 1985)
Respondents versus consumers
Estimates of dietary exposure can be calculated for all survey respondents or only for those who reported consuming a food containing the chemical on the day of the survey (‘consumers’). This study reports exposure estimates for ‘consumers’ for the population aged 2 years and above. The model diet for infants is on an ‘all respondent’ basis therefore dietary exposures are for respondents.
The plasticisers investigated in this study are distributed across a wide range of foods and are frequently consumed by all members of the population.
The number and proportion of consumers in each age group is provided in Table A4.1.
Food chemical concentrations for dietary exposure estimates
As a number of composite samples were analysed for each food in the survey, the median concentrations of each of the plasticisers for each of the foods sampled was derived and used in the dietary exposure assessment, with the exception of hamburgers and instant noodles. In these cases, the concentrations of the two different types of hamburgers (Brand A only and other fast food chain hamburgers) were analysed and reported separately because a small survey limited to hamburgers previously undertaken by FSANZ had identified one brand as having higher levels of phthalates than other brands. Concentrations for the two instant noodle types analysed (packaged in cups/bowls and plastic wrapping) were also reported separately.
For the purposes of dietary exposure estimates a median concentration was derived for all hamburgers and for all noodles, as food consumption data for each of the types of these foods could not be sufficiently differentiated by brand or packaging type (see food mapping section for further details). In some case, where more than 50% results were non-detects for a given food, the median concentration for modelling purposes was a non-detect, whereas the mean concentration may have a numerical value.
The median concentration in an analysed food was also carried over to all of the mixed foods in which it was used as an ingredient based on FSANZs recipe dataset for mixed foods.
Treatment of analytical values below the Limit of Reporting (LOR)
Some analytical results for some samples were ‘not detected’, or in other words, were below the LOR for the analytical method. In order to take account of these samples in the dietary exposure assessment, a numerical concentration value must be assigned to these samples. Assumptions were made about the concentration of the plasticisers in food samples where the analytical results were below the LOR or where there were no detects. In the case of contaminants that occur naturally in the environment, it is not reasonable to assume that the contaminant is not present in the food when the analytical concentrations are less than the LOR. In the case of the plasticisers assessed, the LOR was reported as equal to the LOD. Actual concentrations below the LOR could in reality be anywhere between zero and the LOR. To allow for this uncertainty, the results for dietary exposure to plasticisers were presented as a range. The lower end of the range is calculated based on the assumption that results below the LOR are equal to zero. The upper end of the range, representing a very conservative ‘worst-case’ estimate, is calculated on the assumption that results below the LOR are equal to the LOR.