Foods that we eat and their relationship to health

  • Food data

    All food databases store nutrient concentrations per 100 G weight of the food. For foods that are typically used by volume measures one must also know the density of the food, ie. how many G are in 100 mL in order to be able to figure out the weight of the food actually consumed. Food databases usually include conversion factors which allow the calculation of density. For instance if a food factor is given for a cup and we assume a cup to be 250 mL then we can calculate the density. Density of a food also depends on how it is served. A cupped of diced vs a cup of crushed pineapple. Each one of these would have a different density.

    In scientific studies with multiple coders of food data it is important to ensure consistency. When a subject mentions pasta as a consumed food, which, of multiple pasta codes must the coder use? Is more information required? A user manual which documents the decisions made during coding is essential. Coder training in the use of the manual is also essential.

    The more accurate and consistent is the data coding and entry the more valuable will be the resulting data and the more likely will the study be able to show nutrient effects.


  • Nutrient data calculations

    Food consumption research calculations are quite simple. Take the food eaten, say eggs, take a quantity, say 2 eggs and multiply the nutrient concentration needed, say protein, by the quantity (2) and you have the protein contribution by eggs. Simple… not so.

    Nutrient databases express nutrient concentrations per 100 G. How many 100 Gs in one egg? Fortunately most nutrient databases also have food factors, in this case, small egg, average egg, large egg, etc. Each one of those has a factor. For instance if your egg is 60 G then your factor should be .6. The factor has to be .6 because that is what you multiply 100 G by to get 60 G. Hence all the of nutrient concentrations have to be multiplied by this same factor to yield the proper nutrient contribution. The protein for eggs in this case would be 2 (eggs) x .6 (factor) x [protein]. The square brackets around the nutrient denote concentration per 100 G. Simple? Of course it is.


  • Week-weighted Averages in CANDAT

    When expressing and comparing results of multi-day (3 day commonly) food recalls one tends to use the average daily intake as a measure of a subject’s intake. Of course, 7 day recalls will give more accurate estimates of this daily intake and the accuracy will increase if all the days of a week are used. The assumption here is that subjects will eat differently on different days with the greatest variability been between week days and weekend days.

    Rather than simply computing an average daily intake one may get more accurate results by weighting the days of the week and the weekend days differently.

    A weight of 5 for each week day and a weight of 2 for weekend days should allow us to calculate week-weighted averages for any number of recalls. Of course, for recalls without weekends this would just be a day of the week average and vice-versa.

    The week-weighted average could be calculated as follows:

    xw = (∑ wi × xi) ÷ ∑ w

    where xw is the weighted average with  xi as the daily intakes, and i is 1 – 5 (codes for days of the week) or 0,6 codes for weekend days corresponding to Sunday and Saturday respectively. w1-5 would then be 5 and w0,6 would be 2.

    with a variance of

    Var(xw) = Var(x) × ((∑ wi2) ÷ (wi)2)

    where Var(x) would be the variance over all the days of the food intakes for the subject.