A low nickel diet calls for the reduction or elimination of some nutrient dense foods, such as nuts, seeds, and beans. Some nutrients are particularly helpful for managing SNAS, for skin health, for gut health, for pain management, and immune system support. Therefore, we wanted to explore whether a low nickel diet is nutritionally unbalanced, and if there are vulnerabilities, how they could be addressed.
There are several questions to be answered:
It's important to say outright that we have no credentials in the field of nutrition. However, we are scientists and we are fairly good with data, and so we proceed to the best of our ability until someone more qualified takes up the task. Our models have limitations and are inexact, so all of the results found here should be reviewed by a professional to verify their applicability to your situation. We hope they will provide a starting point for the discussion.
Do not use these findings blindly. The goal of this work was to see whether it CAN be done, not necessarily how to do it for you as an individual. In particular, these results do not account for the nutritional needs of pregnant or breastfeeding women, or children. See a professional for advice!
In order to answer the questions, we combined our nickel data with nutrition data from the Canadian Nutrient File. We used this data to experiment with different models of food choices. See the Appendix for details of the data and the models. If you're only interested in the results and don't care how we got there, proceed to the Summary.
To answer this question, we defined an overall "nutrient density" as the sum of the percent contribution of the food to the recommended daily amount (RDA) for the 22 nutrients that we were interested in, what we call ND22 (equation 2). Figure 1 shows a bubble chart of how ND22 varies with nickel. The colour of the bubble indicates the type of food, and the size of the bubble represents the number of nutrients for which the food contains more than 15% of the RDA. For reference, a hypothetical food that contains 5% of the RDA for every nutrient (making it a "good source" of everything) has a nutrient density value of ND22 = 110. The nutrient density appears to increase with nickel content for some foods; it is particularly evident in nuts, seeds, and beans (the red bubbles). However, there is no shortage of nutrient dense foods with less than 20 μg of nickel per serving outside of this category, which are considered to be low-to-moderate nickel by our low nickel diet. So, the answer to this question is no, a food does not have to be high in nickel to be nutrient dense.
To answer this question, we had to compare a diet where there are no restrictions on food options to a diet where the nickel per serving is limited. To see the effect of how strict the limitations are, we chose a cutoff of 10 μg nickel per serving (LND10) and a cutoff of 20 μg nickel per serving (LND20). The LND20 food options allows for the selection of more nutrient-dense foods (as shown in Figure 1) and more variety. We also had to compare different strategies for selecting foods, because if foods are selected with nutrition in mind, there may be fewer nutrients that are vulnerable to deficiency.
The technique we developed relies on the ability to divide the foods into categories and assign a number of servings per day to each category. The categories we chose are shown in Table 1, along with the number of servings we settled on. The table shows the number of foods that were removed because of the nickel restrictions.
Category | Number of servings daily | Number of foods, Free | Number of foods, LND20 | Number of foods, LND10 |
---|---|---|---|---|
Beer and wine | 2/7 (2 per week) | 3 | 3 | 3 |
Coffee and tea | 1 | 6 | 5 | 4 |
Condiments | 2 | 16 | 16 | 15 |
Dairy and alternatives | 3 | 11 | 9 | 7 |
Fats | 3 | 15 | 15 | 15 |
Grains | 3 | 39 | 31 | 22 |
Meat and alternatives | 3 | 46 | 20 | 13 |
Treats | 1 | 29 | 19 | 14 |
Vegetables and fruits | 7 | 73 | 61 | 42 |
To model a person who is equally likely to choose any food in a category (i.e., the probability of choosing a food follows a uniform random distribution), we created the "Uniform Model". Over time, each food will be chosen the same number of times. The average for a nutrient times the number of servings of a category in a day (defined in Table 1) gives the average total intake of that nutrient in a day (equations 4 and 5). Similarly, the total nickel for the day is the average nickel of the foods in a category times the number of servings of the category in a day.
To model a person choosing foods that have a higher nutrient density (ND22), we created an "ND22-biased model". In this model, the average for a category is replaced by a weighted average over all the foods in a category, where the weighting is proportional to the ND22 values (equations 4 and 6).
Because of the limitations of our models, we can't definitively say that a deficiency in a nutrient will occur, so we call them "vulnerable nutrients" to indicate that a nutrient is at risk of being deficient. For the purposes of this discussion, a "vulnerable nutrient" is one for which a model gives ≤ 105% of the RDA; we want some wiggle room to ensure that we captured all of them.
Model | Uniform | ND22-biased | ||||
---|---|---|---|---|---|---|
Options | Free | LND20 | LND10 | Free | LND20 | LND10 |
%RDA Ca | 105 | 100 | 85 | 134 | 127 | 108 |
%RDA Fe | 95 | 81 | 83 | 117 | 103 | 103 |
%RDA Mg | 147 | 113 | 91 | 186 | 145 | 115 |
%RDA P | 227 | 220 | 216 | 282 | 268 | 251 |
%RDA Zn | 142 | 153 | 159 | 185 | 196 | 185 |
%RDA Cu | 259 | 191 | 133 | 356 | 271 | 158 |
%RDA Mn | 346 | 246 | 151 | 429 | 316 | 181 |
%RDA Se | 261 | 226 | 236 | 330 | 294 | 284 |
%RDA vitamin A | 104 | 122 | 102 | 153 | 181 | 144 |
%RDA vitamin B1 | 168 | 166 | 189 | 209 | 210 | 232 |
%RDA vitamin B2 | 183 | 209 | 223 | 241 | 268 | 283 |
%RDA vitamin B3 | 140 | 176 | 195 | 178 | 221 | 225 |
%RDA vitamin B5 | 127 | 132 | 141 | 158 | 162 | 167 |
%RDA vitamin B6 | 150 | 168 | 174 | 186 | 206 | 206 |
%RDA vitamin B12 | 218 | 318 | 312 | 315 | 439 | 362 |
%RDA Folate | 116 | 104 | 100 | 152 | 141 | 140 |
%RDA Choline | 83 | 92 | 110 | 93 | 101 | 117 |
%RDA vitamin C | 246 | 276 | 209 | 361 | 421 | 299 |
%RDA vitamin D | 26 | 35 | 36 | 41 | 51 | 50 |
%RDA vitamin E | 98 | 85 | 74 | 123 | 108 | 98 |
%RDA vitamin K | 362 | 366 | 217 | 730 | 800 | 435 |
Fibre (g) | 36 | 27 | 21 | 44 | 32 | 24 |
Nickel | 356 | 140 | 92 | 419 | 158 | 103 |
Several of the vulnerable nutrients highlighted in Table 2 are known issues for a general Canadian population (calcium, iron, vitamin A, choline, vitamin D, and vitamin E):
Table 2 shows that the %RDA achieved for calcium, iron, magnesium, folate, vitamin E, and fibre all decrease as more food restrictions are applied. In the uniform model, calcium, iron, and vitamins D and E are vulnerable even without any restrictions. When a cutoff of 20 μg is used (LND20), folate becomes vulnerable. Magnesium and fibre only become vulnerable with the more strict LND10 food options. So while the target total daily nickel of less than 150 μg is met by restricting the food options, this results in several nutrient vulnerabilities.
Interestingly, the %RDA achieved for the rest of the B-vitamins, choline, and vitamin D show an opposite trend, increasing as higher nickel foods are removed from the food options. Since choline is vulnerable under LND20, it is added to the list of vulnerable nutrients. Vitamin A has an odd trend where it is higher in LND20 than in either the unrestricted or LND10 options; since it is vulnerable under LND10 in the uniform model it is also added to the list of vulnerable nutrients. So, if a low nickel diet is attempted without regard to choosing the more nutritious options, there is a greater vulnerability to deficiency for calcium, iron, magnesium, vitamin A, folate, choline, vitamin E, and fibre.
The ND22-biased model gives a higher total daily nickel than the Uniform model for all sets of restrictions. From this, we conclude that more nutrient dense food options tend to have more nickel, which is not surprising but needed proving.
Table 2 shows that when the ND22-biased model is applied to the unrestricted food options, it fixes all of the issues except choline and vitamin D, which we’re not too concerned about for the reasons above. The LND20 with ND22 bias does not achieve a sufficiently low total nickel (less than 150 μg) and is vulnerable in iron as well as choline and vitamin D. Meanwhile, the LND10 with ND22 bias has a low enough total nickel, but becomes vulnerable in vitamin E in lieu of choline, and is lower in fibre than the recommended 25 grams. We needed to find a way to bias our food choices to favour those with more of the vulnerable nutrients rather than just a general nutrient density, and less nickel.
In conclusion, the nutrients that appear to be most vulnerable in a low nickel diet are: calcium, iron, magnesium, vitamin A, folate, choline, vitamin D, vitamin E and fibre. The ones that are specific to a low nickel diet are magnesium, folate and fibre; the remainder were pre-existing issues for an unrestricted diet which low nickel generally makes worse.
We needed to bias the food choices to those with (1) more of the vulnerable nutrients rather than just a general nutrient density, and (2) less nickel. For (1), we defined another nutrient density, ND9 (equation 2), that is the same as ND22 but only includes the 9 vulnerable nutrients. For (2), we defined a metric that we call "bang for your buck", or BFYB. BFYB takes the nickel content into account by giving us a measure of the nutrient density per microgram of nickel in a serving of food (equation 7). Then we applied this "BFYB9-biased model" to the unrestricted, LND10, and LND20 diets. Table 3 compares the results of the BFYB9-biased model to those of the uniform model from Table 2.
Model | Uniform | BFYB9-biased | ||||
---|---|---|---|---|---|---|
Options | Free | LND20 | LND10 | Free | LND20 | LND10 |
%RDA Ca | 105 | 100 | 85 | 129 | 122 | 113 |
%RDA Fe | 95 | 81 | 83 | 110 | 109 | 110 |
%RDA Mg | 147 | 113 | 91 | 143 | 129 | 111 |
%RDA P | 227 | 220 | 216 | 258 | 253 | 242 |
%RDA Zn | 142 | 153 | 159 | 167 | 169 | 164 |
%RDA Cu | 259 | 191 | 133 | 245 | 220 | 158 |
%RDA Mn | 346 | 246 | 151 | 270 | 236 | 176 |
%RDA Se | 261 | 226 | 236 | 268 | 277 | 278 |
%RDA vitamin A | 104 | 122 | 102 | 220 | 235 | 209 |
%RDA vitamin B1 | 168 | 166 | 189 | 196 | 198 | 208 |
%RDA vitamin B2 | 183 | 209 | 223 | 268 | 276 | 287 |
%RDA vitamin B3 | 140 | 176 | 195 | 180 | 194 | 195 |
%RDA vitamin B5 | 127 | 132 | 141 | 163 | 169 | 172 |
%RDA vitamin B6 | 150 | 168 | 174 | 179 | 186 | 182 |
%RDA vitamin B12 | 218 | 318 | 312 | 359 | 416 | 409 |
%RDA Folate | 116 | 104 | 100 | 149 | 147 | 149 |
%RDA Choline | 83 | 92 | 110 | 117 | 128 | 139 |
%RDA vitamin C | 246 | 276 | 209 | 308 | 320 | 241 |
%RDA vitamin D | 26 | 35 | 36 | 49 | 51 | 52 |
%RDA vitamin E | 98 | 85 | 74 | 118 | 113 | 107 |
%RDA vitamin K | 362 | 366 | 217 | 638 | 651 | 411 |
Fibre (g) | 36 | 27 | 21 | 33 | 29 | 25 |
Nickel | 356 | 140 | 92 | 189 | 125 | 94 |
Table 3 shows that when the BFYB9-biased model is used with any of the food options, vitamin D is the only remaining vulnerable nutrient, and LND20 results in a total daily nickel below 150 μg. We conclude that it is possible to design a low-nickel diet that addresses the vulnerabilities.
The BFYB9-biased model on a low nickel diet actually does better than unrestricted food options with the foods chosen at random. Thus, a well-planned low nickel diet can not only meet the nutritional targets, but it can actually be more nutritious than an unplanned unrestricted diet made of whole foods.
First, let’s look at what low-nickel foods (less than 10 μg per serving) are good sources of the vulnerable nutrients. We use Health Canada’s food labelling guidelines as a basis for calling a food "high in" a nutrient (more than 15% of the RDA) or a "good source of" a nutrient (between 5% and 15%). These are summarized in Table 4.
Category | Low nickel foods providing the vulnerable nutrients |
---|---|
Beverages | Beer,2 wine |
Fats | Grapeseed oil, rice oil, sunflower oil, canola oil, soybean oil, olive oil, corn oil, avocado oil, butter, peanut oil |
Grains | Wheat and rye products,5 fortified grain products,2 wheat bran cereals,2 corn cereals, white rice, rice cakes |
Meat and alternatives | Eggs,6 fish,5 meats,3 pate |
Milk and alternatives | Milk,5 yogurt,3 cheeses,2 cottage cheese |
Fruits | Citrus fruits,5 mango,4 kiwi,4 banana,3 jackfruit,3 cherries,2 strawberries,2 breadfruit, pear, watermelon, apple, blueberries |
Vegetables | Spinach,7 leeks,7 mushrooms,3 lettuce (other than iceberg),3 broccoli,3 peppers,3 carrots,2 beets,2 cauliflower,2 cabbage,2 corn,2 rutabaga,2 onions, potato, celeriac, eggplant, celery, iceberg lettuce, zucchini, spring onion, radish |
Treats | Molasses,3 ice cream |
Now let's look at which higher-nickel foods give you nutritional "bang for your buck", foods with between 10 and 20 μg of nickel per serving that have a great nutrient profile. For Table 5, we used a cutoff of BFYB9 = 4.5, based on a hypothetical "good" food that has on average at least 5% of the RDA for each of the 9 vulnerable nutrients and 10 μg nickel.
Category | Moderate nickel foods providing bang for your buck in the vulnerable nutrients |
---|---|
Grains | Whole wheat bread, whole wheat pasta, quinoa |
Meat and alternatives | Shrimp |
Milk and alternatives | Almond beverage, rice beverage |
Fruits | Guava, cantaloupe, dates |
Vegetables | Kale, chard, bok choy, sweet potato, brussels sprouts |
Treats | Corn chips |
We also looked for foods with more than 10 μg nickel per serving that have a great nutrient profile across all 22 nutrients, using a BFYB22 defined in equation 8. We considered BFYB22 > 11 to be high bang for your buck, based on a hypothetical food with 10 μg nickel per serving and 5% of the RDA for every nutrient (ND22 ≥ 110). Brown rice, barley, multigrain bread, crab, pineapple, papaya, and popcorn satisfied this condition.
Table 6 brings all the high-nutrient foods together. The second column shows the foods that have fewer than 10 μg nickel per serving and either have a high ND22, i.e. provide at least 5% of the RDA on average over all 22 nutrients, or are listed in Table 4 as addressing a particular vulnerable nutrient. The third column contains foods with slightly more nickel that provide high overall nutrition (BFYB22) or are high in one or more of the 9 vulnerable nutrients (BFYB9, Table 5).
Category | Low nickel, high nutrition | Moderate nickel, high bang for your buck |
---|---|---|
Beverages | Beer, wine | |
Fats | Grapeseed oil, canola oil, rice oil, sunflower oil, soybean oil, olive oil, corn oil, avocado oil, butter, lard | |
Grains | White wheat and rye bakery products, fortified cereals (except rice), white wheat pasta, white rice, rice cakes | Whole wheat bread, whole wheat pasta, quinoa, brown rice, barley |
Meat and alternatives | Eggs, fish, pate, beef, pork, lamb, veal, poultry, sausage, bacon | Crab, shrimp |
Milk and alternatives | Milk, yogurt, cottage cheese, cheeses | Almond beverage, rice beverage |
Fruits | Strawberries, blueberries, cherries, citrus fruits, kiwi, mango, breadfruit, banana, jackfruit, grapes, pear, apple, watermelon | Guava, cantaloupe, dates, pineapple, papaya |
Vegetables | Spinach, lettuce, cabbage, potato, rutabaga, celeriac, beets, radish, onion, leeks, spring onion, mushrooms, broccoli, cauliflower, peppers, carrots, corn, eggplant, zucchini | Kale, chard, bok choy, sweet potato, brussels sprouts |
Treats | Ice cream, molasses, pretzels, potato chips | Corn chips, popcorn |
Not on this list are low nickel foods that are low nutrient density, but remember that this is only by the way we defined "nutrient density". There are many aspects of nutrition that aren’t being measured here.
We omitted multigrain bread from Table 6 because it is nutritionally similar to whole wheat but higher in nickel. Comparing whole wheat bread to white bread, the whole wheat bread is higher in vitamin E by 2% and fibre by 2 grams, but lower in iron by 3%, because white bread is fortified with iron. So it's a bit of a toss-up. Brown rice, on the other hand, is superior to white rice in all nutrients except folate, and has 15% more magnesium.
Soybean oil may, anecdotally, trigger a reaction in some people with SNAS [no citation available] despite being as low in nickel as the other oils.
Please note, we are not advocating alcohol consumption by listing beer and wine; please follow the health guidance of your region on that, and watch for adverse reactions since alcohol affects histamine levels in the body, and histamine may play a role in systemic nickel allergy.
We’ve shown that a healthy low nickel diet can be done, but what about when there are other restrictions due to allergies, food sensitivities, or preferences? Each has its own potential nutritional pitfalls, for example:
We can't model many of these dietary restrictions because of their wide variety of non-standardized potential implementations. We were able to look into dairy-free and gluten-free, however.
With the uniform model, a diet combining LND20 with dairy-free was vulnerable in iron, folate, choline, and vitamins D and E, and LND20 combined with gluten-free was vulnerable in the same nutrients plus calcium and fibre. However, when we applied the BFYB9-biased model, all nutritional targets were met except vitamin D, in both cases. We even found that a combined gluten-free, dairy-free, and low-nickel diet also met all the nutritional targets except for vitamin D with the BFYB9-biased model, but in order to satisfy a total daily nickel of less than 150 μg, we had to reduce the nickel cutoff to 15 μg/serving. The uniform model showed vulnerabilities in iron, folate, choline, and vitamins D and E, and fibre, but not calcium as we expected to see. On closer inspection, while folate, iron, and choline decreased when dairy-free and gluten-free restrictions were added, calcium and vitamin E increased. The relationship between restrictions and nutrition is evidently less straightforward than we anticipated.
For the dietary restrictions we can't currently model, we can't answer the question of whether there is some way to meet nutritional targets. We can only predict that adding low-nickel to the mix could make it harder for some nutrients. We recommend seeking the help of a professional if you want to combine them with low nickel.
Health Canada now recommends that everyone over the age of 50 supplement vitamin D, and suggests that anyone over the age of 2 may do the same if needs aren’t being met by dietary sources. Sun exposure may not be sufficient in Canada.
This study only looks at 22 nutrients. It leaves biotin, omega-3, and omega-6 out of the discussion due to a lack of data, and it omits molybdenum and iodine because they aren’t part of the CNF database. It doesn’t consider nutrients one might want to curtail, such as sodium or saturated fat, and it doesn’t look at beneficial substances like flavonoids, prebiotics, and peptides, for which there is no defined RDA.
We chose to focus on the RDAs of the nutrients, but future work might look at the adequate intake (AI) reference values, because intake below the RDA does not necessarily predict deficiency. This work may be overestimating the risks.
The calories, protein, carbohydrates, and fat varied depending on the food options and the model, with these values tending to increase when the model preferred higher nutrition foods. For the most part they stayed within what we felt was a reasonable range, but it could be important depending on other health goals. So it actually might be useful to include these "negatives" in the weighting.
We made some decisions about what foods to include and exclude based on our own limited experience. Without having a good handle on what an "average Canadian" eats, it’s hard for us to say whether this is a reasonable representation of a typical Canadian diet. We also have forced a whole foods diet here, which is appropriate for comparison to a low nickel diet but is likely not realistic. We had good agreement with known nutrient shortfalls, though, which makes us a bit more confident about it.
The uniform model’s baseline assumption that a person will be as likely to choose any food in a category is huge and probably not realistic. People like some foods more than others [citation needed], however, this is exceedingly difficult to work into a model as a generality. The number of servings in each category is also debatable, changing with the diner’s preference and caloric requirements. Again, the vulnerable nutrients generally agree with known shortfalls, so that makes us feel better about the model choice.
We made the assumption of a predominance of adult females under age 50 for the RDAs, which isn’t appropriate for children with a systemic nickel allergy, women over 50, or men. The RDA for iron in particular changes significantly.
Different models could be developed that include calories, protein, carbohydrates, and fat. Other databases could be used to fill in the gaps on molybdenum, iodine, omega-3, omega-6 and biotin. Other substances such as flavonoids and peptides could be added.
The adequate intake reference values could be used to find the minimal set of vulnerable nutrients and determine whether there may be an overestimate of the risk.
It would be interesting to investigate the odd trend in the %RDA of the B-vitamins, choline, and vitamin D with nickel restrictions, where they increase as higher nickel foods are removed from the food options. Do higher nickel foods tend to be low in these nutrients? The trend where %RDA vitamin A is higher in LND20 than in either the unrestricted or LND10 options is also interesting.
These models could potentially be used to build low-nickel meal plans that meet nutritional targets.
The goal was to create a single spreadsheet with both nutritional and nickel data. For the nutritional data, we retrieved the Canadian Nutrient File (CNF) database and collected the data for the following nutrients per 100 grams in an Excel 2003 workbook:
Vitamin A was collected as retinol equivalents to measure carotenoids and retinol at the same time. We would have liked to have included biotin for its effect on inflammation and nickel allergy, but there were too few measurements in the database. There was sparse coverage (only 40%) for the essential fatty acids (omega-3 and omega-6), which are important in the management of dermatitis. Although molybdenum and iodine are essential nutrients, they were not available in the Canadian data set. It may be worthwhile investigating the impact of a low nickel diet on intake of these minerals at a later date.
Each entry was manually assigned an ID number linking it to the corresponding entry in the (unweighted) nickel data worksheet, thereby also linking it to the serving size for that food, as set out by Health Canada's reference amounts. Since each food on the nickel sheet can have several corresponding entries in the nutrition data, an average of the nutrition data entries was taken to give a 1:1 relationship.
For this analysis we did not use food for which the nickel content was unreliable, or foods that we did not consider to be "generally popular and acceptable" (e.g. organ meats, foods that are difficult to prepare, foods that would demand an acquired taste). This included seafood such as octopus, squid, mussels, oysters, clams, and lobster. We felt that removing liver was justified since liver ceased to be explicitly mentioned in the Canada Food Guide in 1992 and its extreme nutrient density could bias the results unrealistically, and this approach has been taken in the literature. We chose to use the cooked versions of pasta, beans, and rice (because cooking reliably lowers the nickel in these items), but the uncooked versions of meats, fish, and vegetables (because they are not all cooked in the same way, e.g. boiling decreases nickel and nutrition, whereas frying may increase nickel, so it would be like comparing apples and oranges for both nutrition and nickel). We removed grains that are used as ingredients (e.g. flour) in favour of their edible products, herbs and spices, and food that isn’t really food such as protein powder and energy drinks. As recommended in the 2019 Canada Food Guide we removed fruit juices. Entries in the CNF for which there was an obvious input error were also omitted. While this reduced the number of foods to 250, we assume that enough variety remains that it won't distort the end results. It should be noted that the foods on our list are somewhat idealized, i.e. what processed foods are on the list are only minimally processed. This may bias the representation of a "normal" diet to one that is slightly healthier than reality.
Since nickel allergy predominantly affects females, we assume that it follows that most SNAS sufferers are female. Therefore we used the dietary reference intakes provided by Health Canada for an adult female under age 50. Where an RDA was not available we used the adequate intake (AI) value instead, but still refer to it as the RDA for simplicity. The RDA is the "intake level that's sufficient to meet the nutrient requirement of nearly all (97 to 98 percent) individuals", whereas the AI "is expected to meet or exceed the needs of most individuals" and is lower than the RDA, meaning that not meeting the RDA does not necessarily mean a deficiency will occur. Table 7 shows the reference values we used in this study. Unless marked as an AI, we used the RDA.
Nutrient | Reference value | Nutrient | Reference value |
---|---|---|---|
Vitamin A | 700 RAE | Vitamin E | 15 mg |
Vitamin B1 | 1.1 mg | Vitamin K | 90 µg (AI) |
Vitamin B2 | 1.1 mg | Calcium | 1000 mg |
Vitamin B3 | 14 mg | Iron | 18 mg |
Vitamin B5 | 5 mg (AI) | Magnesium | 320 mg |
Vitamin B6 | 1.3 mg | Phosphorus | 700 mg |
Folate | 400 µg | Zinc | 8 mg |
Vitamin B12 | 2.4 µg | Copper | 0.9 mg |
Choline | 425 mg (AI) | Manganese | 1.8 mg (AI) |
Vitamin C | 75 mg | Selenium | 55 µg |
Vitamin D | 15 µg | Fibre | 25 g (AI) |
Most notably for this discussion, the RDA for iron is significantly lower after menopause and for males, so the discussion about iron may not be relevant for those groups. We used an average height and weight from 2008 statistics for a "low active" 40-year-old female to compute the approximate calorie requirements (about 2200 kcal).
Letting represent the amount of nutrient found in food , the %RDA for a nutrient in a food is defined as:
where is the reference value for nutrient , as given in Table 7.
The Nutrient Rich Foods Index defines a metric NR, where n is the number of nutrients, as the sum of the percent daily value (%DV) that a food contributes for each nutrient . In this metric, if a food provides more than 100% of the RDA for a nutrient, its contribution to the sum is capped at 100%. We define a similar metric that uses the %RDA as we defined above rather than the %DV, which we will refer to as "nutrient density" for simplicity. For a food and nutrients selected out of the set of all available nutrients:
In our case, the ND22 metric uses the 22 nutrients in Table 7: Ca, Fe, Mg, P, Zn, Cu, Mn, Se; Vitamins A, B1, B2, B3, B5, B6, folate, B12, choline, C, D, E, K; and fibre. The other nutrients (K, Na, omega-3, omega-6, carbohydrates, protein, fat, and calories) are tracked but are not of interest at this time.
With the nutrient density computed, we had a single worksheet that contained: Food name; Average %RDA for each nutrient in the nutrition data; Nutrient density per serving (ND22); and Nickel per serving (μg). At this point we could begin investigating the relationship between nickel and nutrition.
Our model needed to be capable of customizing the set of food choices based on either nickel or dietary restrictions, and it needed to deliver the nutritional and nickel intake based on how a person makes their food choices.
The first part was simple; we manually removed food options that are not allowed. For the second part, we devised a way of modelling food selection that is based loosely on the 2007 Canada Food Guide recommendations, which divides foods into categories and recommends a certain number of servings per day from each category. We fine-tuned these using a sample meal plan provided by Alberta Health Services, and then further adjusted until the macronutrients were within reason for an average adult woman. The newer guideline was not as compatible with our method since it uses a "plate" concept rather than a number of servings. The categories and number of servings we used are given in Table 1. A person must select a food from a category the specified number of times per day.
The selection of a food in our model is probabilistic, meaning that every food has a certain probability (likelihood) that the person will select it. That likelihood could be based, for example, on the food's nutritional profile, or nickel content, or even a person's general preferences. If we let represent a variable that reflects the relative likelihood of selecting a food in category , then the probability of selecting a food in category is:
where is the set of all foods in category .
Because measures such as the %RDA and nickel are all computed in the same way, for this discussion let's refer to them collectively as "" to avoid repetition.
In the long term, the expected total amount of is given by the sum of the expected value for each category times the number of servings of category , :
where is the number of foods in category , is the amount of in food , and is the probability of a person selecting food in category .
If the person is equally likely to select any food in a category, the probability distribution is called a uniform random distribution, and is the same number for all foods in a category . Then,
This is equivalent to taking an average over all foods in a category.
If the person chooses food with overall nutrition in mind, a nutrient density using all 22 nutrients (ND22) can be used to define the probability distribution:
where is the set of all foods in category .
In these models, we wanted to bias food choices to those with higher nutrient density, and also bias against foods with higher nickel. The "bang for your buck" (BFYB) variable takes both the nutrient density and the nickel content into account, giving us the nutrient density per microgram of nickel in a serving of food :
where η is the average nickel content in a serving of the food. This number will be higher when a food provides a lot of the nutrients of interest, and when a food is lower in nickel. If nickel content is very high, even the most nutrient dense foods will have a low bang for your buck. However, without adding a lower limit to the denominator, when a food is very low in nickel BFYB would blow up no matter how low the nutrient density was, giving a false picture of what you’re getting out of it. Therefore, for foods with less than 5 μg nickel we changed the equation to only be dependent on the nutrient density. The value of 5 was chosen intuitively; future work could delve into a theoretically justified lower bound.
The BFYB22-biased model uses the overall nutrient density, ND22, in equation 7, giving us the following probability distribution:
where is the set of all foods in category .
The BFYB9-biased model uses a more focussed nutrient density, ND9, in equation 7, only including the 9 vulnerable nutrients: Ca, Fe, Mg, vitamins A, D, and E, folate, choline, and fibre. The probability distribution for this model is:
where is the set of all foods in category .
The spreadsheet is .