Dear Susanne,

 

I have tried using this approach to calculate missing values in multi-ingredient foods with varying results.

 

For the procedures described by Westrich et al., the keywords are “linear programming” (see also http://en.wikipedia.org/wiki/Linear_programming).

This a common mathematical, numerical analysis approach which is used to solve/optimize a set of simultaneous linear relationships/equations. It is used in diet/nutrient planning.

Most statistical software contains the procedures to carry out these calculations – as far as I remember USDA’s software is using a SAS procedure.

It is also available as an Add-In to Microsoft Excel.

 

Search Google (and Google Scholar) for “linear programming” or better “linear programming excel” (and maybe even “linear programming excel nutrients”) and you will get a better idea of what is available.  

 

The simple form of this procedure is solving a system of n equations with n variables, which can be done by Gaussian elimination, e.g. using the Gauss-Jordan  algorithm or more complex with the Simplex method.

Each equation resembles the summation of the contributions from each ingredient to the total content of a certain nutrient. In a simple food (tow ingredients), you can construct the following equation for e.g. fat (one of the labeled nutrients)

 

fat content ingr1 (g/100 g) * amount of ingr1 (g)/100 (g)  +  fat content ingr2 (g/100 g)) * amount of ingr2 (g)/ 100 (g) = total fat (g/100 g)

 

Similarly, you can construct similar equations for the other labeled nutrients.

The variables are the unknown amounts of ingr1, ingr2, etc.

The equations are completely the same as the ones you are using for calculating nutrients from a recipe. The only difference is that you do not know the amounts, the linear programming is the approach to find the amounts.

 

The solution (if any ;-) to the set of linear equations are the amounts of ingr1, ingr2, ingr3, etc. – from these amounts you can then (back-) calculate the recipe based on the ingredient list for all nutrients provided that you have the information on all nutrients in all ingredients, including water (or all nutrients in all significant ingredients).

This means that for a 5 ingredient food, you will need five (independent) equations for five labelled nutrients. For a six ingredient food, 6 equations for six nutrients, etc.

 

One of the great challenges in using this approach to calculate missing nutrient information in a food (product) is having proper and good information on all nutrients in question in all (significant) ingredients, i.e. also in the industrial ingredients used in the food. This is often very difficult.

However, if the producer has calculated the nutrition labelling information based on your food composition data, you should be able to get good results in determining the amounts of the ingredients.

 

All the best,

Anders

 

 

 

 

 

 

 

 

From: Food Composition Discussion Group [mailto:[log in to unmask]] On Behalf Of Susanne Westenbrink
Sent: 06 May 2013 14:29
To: [log in to unmask]
Subject: calculation of micro nutrient content of brand data

 

Dear all,
In the Dutch food composition database we need to include more and more brand data in order to cover the needs of our users. Both researchers and dietitians ask for details on mineral and vitamin content of these foods. Unfortunatelly this information is generally not available from the labels nor from the producers. We try to fill in these missing data as good as possible by borrowing from similar foods, estimations or calculations.  Such calculations can be based on the ingredients as known from the labels. By trying several senario's we try to get the composition for the macronutrients as close as possible to the values on the label and use the micronutrients as calculated at the same time.

This is a time consuming approach that gives an estimation of the composition, and it is the best we have at the moment. To work more efficient we are looking for way to find a more automated procedure.
A literature search so far revealed the approach used for the NDSR database as desribed by Westrich (1994 and 1998) who developed sofware to calculate the composition of brand foods, based on the ingredients on the label. This software is at the moment not available to use with other e.g. European food composition databases. I did not find any more papers on this issue.

My question to you are:
Are you aware of any other publications on this topic?
Are you aware of software that is available for this purpose?
Do you have specific experience with calculating composition based on the ingredients on labels (with or without using specific software)?
 
I suppose filling in missing values of brand foods by recipe calculations is done by other food database compilers as well. I would be glad to hear and learn from you.
Thanks in advance,
with kind regards,
Susanne


Susanne Westenbrink
Rijksinstituut voor Volksgezondheid en Milieu (RIVM) / National Institute for Public Health and the Environment
Centrum voor Voeding, Preventie en Zorg /
PO Box 1
3720 BA Bilthoven
The Netherlands
tel:  +31 (0) 30 274 2448
fax: +31 (0) 30 274 4466
e-mail: [log in to unmask]
for more information on  the Dutch food composition database see
http://www.rivm.nl/nevo
http://nevo-online.rivm.nl




Proclaimer RIVM http://www.rivm.nl/Proclaimer


To unsubscribe from the InFoods-Food-Comp-L list, click the following link:
https://listserv.fao.org/cgi-bin/wa?SUBED1=InFoods-Food-Comp-L&A=1



To unsubscribe from the InFoods-Food-Comp-L list, click the following link:
https://listserv.fao.org/cgi-bin/wa?SUBED1=InFoods-Food-Comp-L&A=1