General limitations of food composition data
AUSNUT 2011–13 has been developed specifically for use with the 2011–13 Australian Health Survey (AHS) and therefore may not be appropriate to use with other surveys or for other purposes. When using or comparing nutrient data across surveys, consideration must be given to survey methods and operations such as questionnaire wording, data processing methods, and the way foods and nutrients are reported in the survey database.
Nutrient data presented in AUSNUT 2011–13 should be regarded as approximations of the likely nutrient content of the food or beverage to which they refer. This is because the nutrient composition of foods is variable and dependent on a range of factors. There are inherent factors affecting the nutrient composition of foods related to variety, soil type and season. There are also factors associated with variability in production and processing, formulation and storage practices (Pennington, 2008).
In addition, analysis of nutrients is associated with its own uncertainty. This can be particularly significant when levels of a nutrient are low and close to the level at which they can be reliably quantified (the 'limit of reporting' or LOR).
Where the presence of a nutrient could not be quantified, assumptions were made about the concentration of this nutrient in the food and beverage in question. For most nutrients reported in AUSNUT 2011–13, where laboratories reported a nutrient concentration as being below the LOR, the concentration was assumed to be zero. In these cases, the LOR was well below what was considered a nutritionally significant concentration and the use of a zero value had no effect on estimated population nutrient intakes. In a limited number of cases, where it was judged that a food was likely to contain a particular nutrient or the laboratory has reported that trace levels are present, the food has been assigned a concentration equivalent to half the LOR determined in FSANZ's most recent analytical program. Specific issues associated with reporting of some nutrients are discussed below.
Issues associated with specific nutrients
Iodine and selenium
Iodine and selenium were assumed to be present in almost all foods reported in AUSNUT 2011–13. However, for many foods including widely consumed foods such as beer and wine, fruit and vegetables and margarine, laboratories have reported iodine and selenium concentrations as being below the LOR. As elements present in the environment, it can reasonably be expected that iodine and selenium will be present in almost all foods even though it may not be possible to quantify levels. Therefore in almost all cases where laboratories were not able to quantify iodine or selenium levels, a value equivalent to half the LOR was assigned. This approach was also used in AUSNUT 2007 for iodine.
The sodium content of home-prepared foods (in particular mixed dishes) is underestimated in AUSNUT 2011–13. This is because salt added during cooking has not been accounted for in FSANZ recipes, except where the addition of salt is needed for reasons other than flavour (e.g. bread dough).
Folate values have been substantially altered since AUSNUT 1999 through the incorporation of Australian derived analytical data, the majority of which have been determined using an improved method of analysis which tends to yield higher folate values than previously.
Values for vitamin E drew largely on analysed values, some of which were relatively old and may not reflect current commercial practices relating to the addition of alpha tocopherol as an antioxidant in oily foods. In addition, the available data do not distinguish between the different isomers of alpha tocopherol, which may have differing bioactivity. Analysed values for gamma- and delta-tocopherols were included in the estimation of vitamin E where available, but not alpha-tocotrienol as very little data was available on levels of this vitamer.
Total long chain omega-3 fatty acids
Long chain omega-3 fatty acids may occur in a range of animal foods. While concentrations are generally highest in marine foods, the total mass of these acids may still be significant in higher fat foods derived from land animals, such as eggs and cheese. Concentrations of individual long chain omega 3 fatty acids were at or just above the LOR for a number of foods reported in AUSNUT 2011–13, and therefore their quantification was associated with considerable uncertainty. As a result of improvements in the measurement of these acids in recent years, the most recent analytical data were used in this database. For reduced fat milk, the most recent data indicated that no long chain omega-3 fatty acids were present at or above the LOR. In contrast, the most recent data for yoghurt indicated a concentration of two of those acids at the LOR. For fatty acids, values less than the LOR were assigned a zero value.
Issues with specific approaches to generating nutrient data
Using label data to generate nutrient data for foods assumes that the label information is reliable. While there have been no wide-scale assessments of the reliability of NIP data, a small study conducted between 2004 and 2005 analysed 350 foods and compared the analysed values with the values presented on the product label. The study found a discrepancy between -13% to +61% for individual nutrient values (Fabiansson, 2006). A small study conducted by FSANZ between 2008 and 2009 analysed 363 foods also found 60% of analysed sodium values were within 20% of the level reported on the label (FSANZ, 2009).
Where label data has been used to assign a vitamin or mineral concentration for fortified foods, the values for added vitamins and minerals declared on labels may underestimate actual values as extra nutrient may be added to ensure the declared levels, at a minimum, are achieved throughout the life of the product. No allowance was made for any fortificant overages when using label data, so values reported in AUSNUT 2011–13 for these nutrients should be regarded as indicative.
For some special purpose foods, including protein powders, label data were sometimes insufficient to give confidence in the robustness of the nutrient profiles produced, because they sometimes contain ingredients that do not have published compositional data. Many of these profiles are based on borrowed data for whey protein powder, with some adjustments where possible to reflect label information on protein content or the presence of ingredients containing caffeine.
Although common cooking practices were used as a guide, the choice of recipes used to prepare nutrient data was somewhat subjective. Because of the extensive variation in recipe ingredients that were reported, streamlined recipes were developed for many foods that could cover a large number of variants of a basic dish. For example, for homemade curries, a large number of vegetable combinations were reported and were dealt with using a single 'not further defined' vegetable line that reflected overall vegetable usage in curries, but not necessarily the specific ones a particular respondent may have reported.