Stability of nutrients in complex liquid and powder food matrices: learnings from shelf-life studies in foods for special medical purposes

Abstract

We investigated the effects of physical state, temperature, pH, protective atmosphere, protein hydrolysis, fats, fibre, packaging and flavour on nutrient degradation by analyzing the results of shelf-life studies of 1400 recipes of Foods for Special Medical Purposes (FSMP). For each nutrient, the adaptive LASSO method was applied to identify which factors drive degradation. Degradation percentage was estimated by repeated measure model using the list of identified factors and time as covariates. Nutrients that had little or no degradation under all tested conditions included fat, protein, individual fatty acids, minerals and the vitamins B2, B6, E, K, niacin, biotin and beta carotene. Important nutrient degradation which was, mainly driven by temperature, were only observed in vitamin A (powder products), vitamin C, B1 and D (liquid products) and pantothenic acid (acidified liquids). These nutrients used as tracers in stability studies would be sufficient to confirm the nutritional suitability of products until their end of shelf-life. Our results show that the most important factors driving nutrient degradation were liquid format, temperature and pH. The fat content, relative humidity, presence of fibre, flavours or packaging size/type did not impact on the stability of any nutrients. These results provide a sound basis to guide health authorities to define which nutrients should ideally be included in FSMP shelf-life studies in addition to providing valuable insight into which factors impact on nutrient degradation in general.

Graphical abstract

Highlights

• •Statistical method used on a very large dataset to assess degradation of nutrients.
• •Little or no degradation for macronutrients, minerals and several vitamins.
• •Liquid format, temperature and pH were the most important degradation factors.
• •Fat, humidity, fibre, flavour, packaging type did no impact nutrient stability.

1. Introduction

A shelf-life assigned to a product by the manufacturer is an assurance of safety, quality, and delivery of nutrients to the consumer. One of the driving factors that can define the shelf-life of a product is the extent of losses of its key nutrients through degradation over time. These key nutrients can be nutrients for which nutrition or health claims are made on packs and that are one of the reasons behind the purchase of the products by the consumer. They can also be nutrients that are crucial for the appropriate growth and development of the consumers such as in infant formula or for the health of the consumers such as in clinical nutrition products, the so-called Foods for Special Medical Purposes (FSMPs, also called medical foods in the US). FSMPs are highly specialised foods intended for feeding patients who have a disease, disorder, or medical condition.

Although several studies in the literature discuss the relative stability of nutrients in different food categories (see table 1), they mainly focus on vitamins and some additional nutrients like sugars and amino acids. Studies on the stability of other nutrients (such as macronutrients and minerals) are limited or unavailable. There is also no comprehensive report that evaluates the impact of different factors (temperature, pH, oxygen level …) on the stability of all nutrients (or even all vitamins only) in a same product. This study aims to fill this gap.

FSMPs constitute a good choice for such a study since these products have a long shelf-life. They allow the investigation on the impact of intrinsic and extrinsic parameters on the stability of a wide range of nutrients interacting with each other in a complex food matrix. FSMPs are indeed typically composed of the 3 macronutrients (fats, proteins and carbohydrates), contain all vitamins and minerals and can also include amino acids and other specialised ingredients such as nucleotides. As a result, and because of their intended use, FSMPs have complex nutritional labels and the delivery of these nutrients in adequate amount is deemed essential to meet the nutritional needs of the patients consuming them, especially when they are used as sole source of nutrition.

Verification of the stability of nutrients over the shelf-life of FSMPs is thus important. In the past few years, regulatory authorities have started to develop regulations or guidelines for this heavily regulated category of food that list which nutrients should be part of the stability tests to be submitted as justification for the duration of the shelf-life proposed by the manufacturers. In the absence of reliable published data, analyses of all nutrients are sometimes requested, even those that display no losses. To address this absence of data and provide a fact-based foundation for future regulatory guidelines, global FSMP manufacturers shared their stability data on FSMPs and statistically analysed them to understand the behaviour of nutrients and the factors impacting their stability over shelf-life. The aim was to identify which nutrients were stable and which ones were not and under which shelf-life conditions. The results can be used as potential recommendations when verifying the shelf-life of FSMP products. While specific to FSMPs, they are likely to apply to products with similar complex food matrices such as infant formula, follow-up formula and meal replacements for weight control as well as to provide relevant insights into the behaviour of nutrients in foods in general.

2. Materials and methods

2.1. Data collection

In total, 5 world-leading food companies (i.e. Abbott, Fresenius Kabi, Nestlé Health Science, Danone Nutricia Research and, Reckitt) provided anonymized stability data for 67 nutrients. The data was generated during shelf-life studies conducted on a range of their FSMP products. FSMPs were classified in categories defined according to the 9 below characteristics with the following 28 indicators thought to have the potential to affect nutrient stability:

• •Physical state (liquid, paste, powder),
• •Intended age (infant, non-infant),
• •Packaging type (metal can, carton, composite, flexible packaging, plastic opaque bottle),
• •Nitrogen flushing (No, Yes),
• •
  Level of protein hydrolysis:

  • -
    intact protein,
  • -
    extensively hydrolyzed protein (i.e. protein hydrolyzed into small peptides),
  • -
    partially hydrolyzed protein (i.e. protein hydrolyzed into large peptides),
  • -
    amino acid based products (i,e, recipe in which only single amino acids are added but no peptides nor proteins),
• •Fat content (with, without),
• •pH (acidified i.e. with a pH below 4.6 in agreement with (Codex Code of Hygienic Practice), non-acidified, Not applicable),
• •Storage temperature (20 °C [18°C–23 °C], 25–30 °C [23°C–33 °C], 35–40 °C [33°C–44 °C]),
• •Storage humidity (less than 60 %, 60 %–75 %, 70 %–75 %, Not applicable).

In total, data for 150 combinations of these 28 indicators were available in the database for the analysis.

All nutrients quantified at Time 0 (T0 – straight after manufacturing) and at least one time point after T0 were included in the final database, resulting in a total of 31,262 stability curves from 1463 recipes on 67 different nutrients of interest. These nutrients included the 3 macronutrients (protein, fat, total available carbohydrate, excluding dietary fibre in line with (Codex Alimentarius guideline on nutrition), total sugars, glucose and lactose, minerals (sodium, potassium, chloride, calcium, phosphorus, magnesium, iron, zinc, copper, manganese, chromium, molybdenum, selenium, iodine, fluoride), fatty acids (linoleic acid, alpha-linolenic acid, arachidonic acid, docosahexanoic acid, eicosapentanoic acid, erucic acid, saturated fatty acid, monosaturated fatty acids, polyunsaturated fatty acids), amino acids, vitamins (A, C, B6, B12, D, E, K, beta carotene, biotin, niacin, pantothenic acid, riboflavin, thiamin, folic acid), total nucleotides and some other nutrients (choline, inositol, carnitine, taurine). All stability curves have been normalised by dividing the data by the value at Time 0, such that all normalised stability curved started from 1 and decreased towards 0 in case of nutrient loss.

Data was also provided on affiliated recipes. Affiliated recipes were defined as identical recipes packed in the same packaging type and maintained under the same environmental conditions during the shelf-life tests and that only differ either by their flavour; or by the absence or presence of fibre; or by the packaging size; or by the temperature storage.

2.2. Main statistical analysis

2.2.1. Estimation of the degradation rate for each nutrient

The adaptive Least Absolute Shrinkage and Selection Operator (LASSO) method (Zou, 2006) was used to identify the indicators driving degradation for each nutrient. The slope of each nutrient stability curve was correlated with the possible 28 indicators, in addition to the double interaction and triple interaction indicators (i.e. 1757 potential effects) to identify the statistically significant effects on the degradation rate.

Following the LASSO selection for statistically significant effects, the following post selection process was used:

• •Selected effects with less than 5 % of the total degradation have been excluded from the model.
• •Selected effects present in less than 5 stability curves have been excluded from the model as potentially not reliable.
• •The three physical state indicators (liquid, powder, paste) were always maintained as separate subgroups by default as it was visible that they were the main driver of the difference in nutrient stability.
• •If an interaction of indicators was selected, the single indicators were added, e.g. if “Flushed ∗ Acidified” was selected, “Flushed” and “Acidified” were added.
• •If a middle indicator was selected (e.g. 23–33 °C, Extensively or Partially Hydrolyzed protein, …), the other indicators of the same category were also selected.

2.2.2. Degradation rate model by nutrient

The adaptive LASSO method identified the indicators driving degradation for each nutrient. For each nutrient, data was then split into several subsets of stability curves which share the same combination of indicators responsible for driving degradation. For each subset, a repeated measure model was used with the degradation percentage as the dependent variable; time (in months) as the covariate/fixed effect and the identifier of the stability curve as the random effect.

Both the linear degradation rate, i.e 1-normalised value and log-transformation of the normalised value, i.e log (normalised value) were investigated. In both cases, the value at time 0 was 0 (due to the normalisation) so no intercept was added. For each transformation, linear or log-transformation, the LASSO effect selection, the post selection process and the repeated model was used for each nutrient. For each nutrient, the most appropriate model (linear or log-transformed) was identified using the smallest mean square error (MSE), converted in the linear normalised scale. For each nutrient, the best model was then used to estimate the degradation at 12 months for liquids and at 24 months for powders for each subset.

2.2.3. Nutrient stability categorization

Through the above-mentioned models, the losses were estimated 12 months after manufacturing for liquid products and after 24 months for powder products, the most common shelf-life durations for these categories of food products. For the purpose of this study, nutrients were considered stable if their estimated degradation was less than the following thresholds established for this study: 5 % for macronutrients; 7.5 % for fatty acids and 10 % for vitamins and minerals, amino acids and nucleotides. These thresholds are much tighter than most of the global tolerances for the nutritional declaration defined in relevant food guidances and regulations. Tolerances are the accepted difference between the value declared in the nutrient declaration on the label of the product and the value obtained by analysis from the time of manufacture through the end of the product shelf life. Tolerances consider shelf-life losses, analytical and manufacturing variabilities and variabilities in nutrient content of raw materials.

2.3. Complementary statistical analysis

Concerning the affiliated recipes, the impact of flavour, presence of fibre, packaging size and/or temperature storage on the nutrient degradation rates was analysed using a three-level organizational model with a random slope nested by stability curve ID (level 2) and Affiliated Id (level 3). The affiliation ID identified a group of stability curves from the same recipe code and with the same characteristics related to physical state, age, fat, protein hydrolysis, nitrogen flushed, pH, storage humidity, packaging type and, if not used as the analysed parameter, fibre, flavour, packaging size and storage temperature. This analysis was performed for each nutrient and physical state (i.e liquid, powder, paste) for which sufficient data was provided. Only data for liquid recipes was used due to the number of available data except for the effect of storage temperature for which powdered affiliated recipes were also analysed.

To show no effect of the variable of interest, the 95 % confidence interval of the difference was compared with the stability limit (used as the equivalence margin). To show equivalence, the 95 % confidence interval must have been within the limit of the equivalence margin.

For fibre and flavour analysis, the estimate and 95 % confidence interval of the effect difference on nutrient stability for each 2 by 2 effect (e.g. no fibre versus fibre, chocolate versus vanilla flavours) at 1 year (for liquid) was used for the statistical comparison. For packaging size, the estimate and 95 % confidence interval of the effect difference on nutrient stability for an increase of 125 mL, 250 mL and 500 mL at 1 year (for liquid) was used for statistical comparison. For storage temperature, the estimate and 95 % confidence interval of an increase of 5°, 10° and 15° at 1 year (for liquid) and at 2 years (for powder) was used for statistical comparison.

3. Results

3.1. Main analysis

During shelf-life under the tested conditions (storage temperature between 20 and 37 °C for a duration up to 24 months), most nutrients in FSMPs showed no or low losses (Table 2). These nutrients are the macronutrients, all minerals, all fatty acids, most amino acids (except tryptophan, cysteine and histidine) and some of the vitamins.

Table 2.

Nutrients in bold letters showed no or minimal losses both in the main and complementary analyses. The other nutrients showed losses either in the main analysis or in the complementary analysis or both.

Macronutrients	Minerals	Fatty acids	Amino acids	Vitamins	Other nutrients
Protein	Sodium	Linoleic Acid	Isoleucine	Vitamin A	Choline
Fat	Potassium	Alpha-Linolenic Acid	Leucine	Vitamin D	Inositol
Carbohydrates, Available	Chloride	Arachidonic Acid	Valine	Vitamin E	Carnitine
Total Sugars	Calcium	Docosahexaenoic Acid	Lysine	Vitamin K	Taurine
Glucose	Phosphorus	Erucic Acid	Methionine	Beta carotene	Total nucleotides
Lactose	Magnesium	Eicosapentanoic acid	Phenylalanine	Biotin B7
	Iron	Saturated fatty acids	Threonine	Niacin B3
	Zinc	Monounsaturated fatty acids	Aspartic acid/asparaginea	Pantothenic acid B5
	Copper	Polyunsaturated fatty acids	Alanine	Riboflavin B2
	Manganese		Arginine	Thiamin B1
	Chromium		Glycine	Pyridoxine B6
	Molybdenum		Proline	Vitamin C
	Selenium		Tyrosine	Folic acid B9
	Iodine		Cystine	Cobalamin B12
	Fluoride		Histidine
			Tryptophan
			Glutamic acid/glutamineb
			Serine

^aMeasured as aspartic acid.

^bMeasured as glutamic acid.

In powder products, vitamin A was the only nutrient to display significant losses. These losses increased with the temperature (see Fig. 1). Its average percentage of degradation at two years was −11 % at 20 °C and −20 % at 25/30 °C.

Fig. 1.

Degradation rate over time – Vitamin A.

In liquid products, thirteen nutrients showed significant losses under some circumstances (see Table 3). Vitamin C, pantothenic acid, folic acid and thiamin were the 4 nutrients displaying the highest losses. The main factor responsible for degradation of vitamin C is higher temperature (see Fig. 2). Other secondary factors driving the degradation of vitamin C are low pH, absence of protective atmosphere (gas flushing) and degree of protein hydrolysis. The losses in pantothenic and folic acids are increased greatly in acidified liquid products and slightly with temperature (see Table 3 and Fig. 3, Fig. 4). The losses in thiamin increased importantly in temperature above 20 °C but were not affected by the pH of the product (see Fig. 5). The 2 main factors responsible for degradation of vitamin D is the protein degree of hydrolysis or a low pH (see Fig. 6). Vitamin A was the only nutrient to be stable in acidic liquid products but to display marked degradation in non-acidified liquid products (at 25/30 °C) (see Fig. 1). Histidine, cysteine and tryptophan were stable in non-acidified conditions but showed losses at low pH that were temperature dependent for tryptophan (see Table 3). Vitamin B12 losses increased with temperature (see Table 3). Lactose and glucose showed losses over time irrespective of the conditions and product characteristics (see Table 3). Total sugars display losses that were more important in acidified conditions than in non-acidified conditions.

Table 3.

Average percentage of degradation after one year in liquid FSMPs.

Nutrient	Non-acidified liquid FSMPs		Acidified liquid FSMPs
	20 °C	25/30 °C	20 °C	25/30 °C
Vitamins
Pantothenic acid (B5)	Stable	−10 %	−53 %	−72 %
Vitamin C(1)	−20 % (with flushing) −46 % (no flushing)	−55 % (with flushing) −65 % (no flushing)	−35 % (intact protein) −49 % (amino acid based/hydr.)	Not enough data (intact protein) −68 % (amino acid based)
Folic Acid (B9)	−11 %	−16 %	−35 %	−56 %
Thiamin (B1)	−20 %	−50 %	−20 %	−50 %
Vitamin D	Stable with intact proteins −33 % (extensively hydrolyzed proteins)		−28 % (with intact proteins)
Vitamin B12	−11 %	−17 %	−11 %	−17 %
Vitamin A	Stable	−15 %	Stable
Amino acids
Tryptophan	Stable		−21 %	−31 %
Cysteine	Stable		−20 %
Histidine	Stable		−19 %
Sugars
Glucose	−19 %
Lactose	−10 %
Total Sugars	−7 %		−14 %

Fig. 2.

Degradation rate over time – Vitamin C.

Fig. 3.

Degradation rate over time – Pantothenic acid.

Fig. 4.

Degradation rate over time – Folic Acid.

Fig. 5.

Degradation rate over time – Thiamin.

Fig. 6.

Degradation rate over time – Vitamin D.

The different factors that affect the stability of each of these 13 nutrients are summarized in Table 4. The most important factor is whether the product is a powder or a liquid. Only vitamin A shows significant degradation in powder products over a 2 years shelf-life compared to 13 nutrients (including vitamin A) in liquid products over a (shorter) shelf-life of one year. The second most important factors (in terms of number of nutrients affected and amplitude of the losses) were pH and temperature. Use of protective atmosphere (gas flushing of the packaging) only impacted the stability of vitamin C. The original degree of hydrolysis of protein as raw material impacted the stability of vitamin C and D. Packaging types, intended age of the products, fat content, relative humidity during storage did not impact the stability of any nutrients.

Table 4.

Factors that affect stability of nutrients.

Unstable nutrients	Product Physical State	Storage Temperature	Product pHa	Protective atmosphere in packaging	Hydrolyzed Protein
Vitamins
Pantothenic acid (B5)	Liquid	Yes	Low pH	No	No
Vitamin C	Liquid	Yes	Low pH	Yes	Yes
Folic Acid (B9)	Liquid	Yes	Low pH	No	No
Thiamin (B1)	Liquid	Yes	No	No	No
Vitamin D	Liquid	No	Low pH	No	Yes
Vitamin B12	Liquid	Yes	No	No	No
Vitamin A	Powder and Liquid	Yes	High pH	No	No
Amino Acids
Tryptophan	Liquid	Yes	Low pH	No	No
Cysteine, Histidine	Liquid	No	Low pH	No	No
Sugars
Lactose, Glucose, Total Sugars	Liquid	No	No	No	No

^aIn liquid FSMPs only.

Paste FSMPs were represented in our dataset by a smaller numbers of recipes compared to the powder and liquid recipes. Nutrient losses were overall similar to what was observed in liquids (Fig. 1, Fig. 2, Fig. 3, Fig. 4, Fig. 5, Fig. 6).

3.2. Complementary analyses

3.2.1. Impact of temperature

A complementary analysis, performed on a smaller dataset than the main analysis and that compared the losses in identical recipes stored at different temperatures, indicated that there was a strong correlation with the results obtained in the main analysis. The average losses obtained by this analysis (Table 5) allow to roughly compare the relative extent of the losses between nutrients. They should not be used as benchmarks of average absolute losses as other differences between recipes (e.g. pH) may impact these numbers.

Table 5.

Effect of temperature increase on losses of single nutrients over 12 month shelf-life in liquid affiliated recipes.

Nutrient	No. of affiliated recipes	P-value	5 °C increasea	10 °C increasea	15 °C increasea
Thiamin	88	P less than 0.0001	20.3 % [18.6–22.1]	40.7 % [37.1–44.2]	61.0 % [55.7–66.3]
Vitamin C	102	P less than 0.0001	14.6 % [13.0–16.2]	29.3 % [26.1–32.5]	43.9 % [39.1–48.7]
Sugars	16	P = 0.0263	7.1 % [0.9–13.2]	14.1 % [1.7–26.5]	21.2 % [2.6–39.7]
Arginine	22	P = 0.0015	6.8 % [2.7–10.9]	13.5 % [5.3–21.7]	20.3 % [8.0–32.6]
Glycine	16	P = 0.0007	6.2 % [2.7–9.7]	12.5 % [5.5–19.4]	18.7 % [8.2–29.1]
Proline	17	P = 0.0009	5.8 % [2.5–9.2]	11.6 % [4.9–18.3]	17.4 % [7.4–27.5]
Lysine	32	P less than 0.0001	5.7 % [3.1–8.4]	11.5 % [6.2–16.8]	17.2 % [9.2–25.1]
Alanine	14	P = 0.0004	5.3 % [2.5–8.2]	10.7 % [5.1–16.3]	16.0 % [7.6–24.5]
Histidine	31	P less than 0.0001	5.0 % [2.8–7.1]	9.9 % [5.6–14.3]	14.9 % [8.4–21.4]
Vitamin A	99	P less than 0.0001	4.6 % [3.4–5.8]	9.2 % [6.8–11.6]	13.8 % [10.2–17.4]
Vitamin D	88	P less than 0.0001	4.2 % [2.5–6.0]	8.5 % [5.0–11.9]	12.7 % [7.6–17.9]
Serine	17	P = 0.004	4.0 % [1.3–6.6]	7.9 % [2.6–13.3]	11.9 % [3.9–19.9]
Methionine	18	P = 0.047	3.9 % [0.1–7.8]	7.8 % [0.1–15.5]	11.7 % [0.2–23.3]
Aspartic Acid	16	P = 0.0164	3.8 % [0.7–6.9]	7.6 % [1.4–13.8]	11.4 % [2.2–20.7]
Tyrosine	20	P = 0.0343	3.5 % [0.3–6.7]	7.0 % [0.5–13.4]	10.5 % [0.8–20.2]
Carbohydrates, available	36	P = 0.0394	2.4 % [0.1–4.6]	4.8 % [0.2–9.3]	7.1 % [0.4–13.9]

^aMean and 95 % confidence intervals.

Among the 7 nutrients, the stability of which was shown to be affected by temperature in the main analysis (see Table 4), thiamin, vitamins C and A showed similar (although somewhat lower) average temperature dependent losses in the complementary analysis (see Table 5). Folic acid, pantothenic acid and vitamin B12 were not considered to be affected by temperature in the complementary analysis because their average losses were smaller than 10 % for a 15 °C increase in temperature (the cut-off below which we categorized vitamins and minerals as displaying no or minimal losses). However, they still displayed losses included between 7 and 9 % not so different from the average losses of 16–20 % observed in the main analysis. Tryptophan was not affected by temperature in the complementary analysis as expected based on the main analysis since the data was collected on non-acidified liquids.

In addition to the 7 nutrients classified as sensitive to the temperature in the main analysis, the complementary analysis classified available carbohydrates, sugars, vitamin D and the following 13 amino acids added as single ingredients (arginine, glycine, proline, lysine, alanine, methionine, serine, aspartic acid, tyrosine, isoleucine, leucine, threonine, histidine) as unstable in liquid FSMPs. The complementary analysis was performed on same recipes (stored at different temperatures) contrary to the main analysis that was performed on a dataset of heterogeneous recipes differing not only by temperature but also by other aspects such as for example pH and composition of the recipe. Consequently, much of the background noise present in the data used for the primary analysis due to differences between recipes is absent from the secondary dataset allowing to detect more subtle losses in nutrients. Therefore, it is not surprising that more nutrients are classified as unstable in the complementary analysis compared to the primary analyses. Among this list of nutrients classified as unstable in the complementary but not in the main analysis, it is indeed striking that all the losses at 12 months are relatively small (making them harder to be detected in a heterogeneous dataset) (see Table 5).

In summary, the results of the 2 analyses are congruent. Both analyses conclude that vitamin A is the only nutrient degraded by higher temperature in powder and that 6 vitamins are sensitive to temperature in non-acidified liquids: Vitamin C, thiamin, vitamin B12, folic acid, vitamin A and pantothenic acid. Proteins, fats, all fatty acids and all minerals that were part of the complementary analysis (see Table 6) were stable confirming the results of the main analysis. The complementary analysis indicates that available carbohydrates, sugars, vitamin D and 10 amino acids are slightly sensitive to temperature in liquids when stored for a year.

Table 6.

List of studied nutrients in the complementary analysis (only nutrients with sufficient data, i.e. at least 10 recipes, were considered).

Group	Nutrient	Flavours	Fibre	Packaging size	Temperature
		Liquid	Liquid	Liquid	Powder	Liquid
Macronutrients	Protein				x	x
	Fat				x	x
	Carbohydrates, Available					x
	⁃Sugars					x
	⁃Glucose					x
	⁃Lactose					x
Minerals	Sodium	x			x	x
	Potassium	x			x	x
	Chloride				x	x
	Calcium	x			x	x
	Phosphorus				x	x
	Magnesium	x			x	x
	Iron	x			x	x
	Zinc	x			x	x
	Copper	x			x	x
	Manganese	x			x	x
	Chromium				x	x
	Molybdenum					x
	Selenium				x	x
	Iodine				x	x
	Fluoride					x
Fatty acids	Linoleic Acid C18:2 n-6				x	x
	Alpha-Linolenic Acid C18:3 n-3				x	x
	Arachidonic Acid C20:4 n-6				x
	Docosahexaenoic Acid C22:6 n-3				x
	Erucic Acid C22:1 n-9					x
	Saturated fatty acids					x
	Monounsaturated fatty acids					x
	Polyunsaturated fatty acids					x
	Trans fatty acids (TFA)					x
Amino acids	Isoleucine				x	x
	Leucine				x	x
	Valine				x	x
	Lysine				x	x
	Methionine				x	x
	Phenylalanine				x	x
	Threonine				x	x
	Aspartic acid				x	x
	Alanine				x	x
	Arginine				x	x
	Glycine				x	x
	Proline				x	x
	Tyrosine				x	x
	Cystine				x
	Histidine				x	x
	Tryptophan				x	x
	Glutamic acid				x
	Serine				x	x
Vitamins	Vitamin A	x	x	x	x	x
	Vitamin D	x	x	x	x	x
	Vitamin E	x	x	x	x	x
	Vitamin K	x	x	x	x	x
	Beta carotene	x	x	x		x
	Biotin B7	x	x	x	x	x
	Niacin B3	x	x	x	x	x
	Pantothenic Acid B5	x	x	x	x	x
	Riboflavin B2	x	x	x	x	x
	Thiamin B1	x	x	x	x	x
	Pyridoxine B6	x	x	x	x	x
	Vitamin C	x	x	x	x	x
	Folic acid B9	x	x	x	x	x
	Cobalamin B12	x	x	x	x	x
Other nutrients	Choline				x	x
	Inositol				x	x
	Carnitine				x	x
	Taurine				x

3.2.2. Impact of flavours

Additional complementary analyses on the impact of different flavours on a same recipe indicate that none of the flavour had a significant effect on any of the studied nutrients (see Table 6 for the list of studied nutrients). None of the 2 by 2 effect was significant at an alpha level of 0.01 (to account for multiplicity) (See Fig. 7 as an example for the effect of different flavours on Vitamin C, the most unstable nutrient). The studied flavours were chocolate, coffee, fruit flavours, vanilla and no flavour. Since flavours does not impact the stability of nutrients, it is not relevant nor required to conduct distinct shelf-life tests on FSMP products that only differ by their flavour.

Fig. 7.

Effect of flavours (A vs B) on the degradation rate of Vitamin C at 1 year in liquid FSMP (P = 0.1659).

3.2.3. Impact of fibre

Similarly, no difference in stability of the studied nutrients (see Table 6 for the list of studied nutrients) was detected between recipes that only differ by the presence or absence of fibre. All the 95 % confidence intervals of the effect of fibre are within the stability limits defined in the material and method section for the different classes of nutrients. Identical recipes with or without fibre can be considered statistically equivalent if the stability limit defined is considered as the equivalence margin. Since presence or absence of fibre does not impact the stability of nutrients, it is not relevant nor required to conduct distinct shelf-life tests on FSMP products that only differ by the presence or absence of fibre.

3.2.4. Impact of packaging size

For the nutrients with sufficient data (i.e. with at least 10 affiliated recipes) (see Table 6 for the list of studied nutrients), the 95 % confidence interval for a packaging change of 500 mL are for all nutrients, except for Vitamin C in liquid FSMPs, within the stability limits defined in the material and methods section. Hence, the degradation rate of all studied nutrients except vitamin C can be considered as ‘equivalent’ in all different packaging sizes. Even for vitamin C, the differences in losses between different sizes of a same packaging remain small compared to losses induced by temperature and a low pH: −10 % for liquid FSMPs in packaging sizes differing by 500 mL (see Fig. 8). Therefore, for a product commercialized in different sizes of a same primary packaging, shelf-life tests conducted on a single pack size are sufficient to justify the shelf-life of the product in all the different pack sizes.

Fig. 8.

Effect of packaging size on the degradation rate of Vitamin C at 1 year in liquid FSMP (P less than 0.0001).

4. Discussion

While multiple publications evaluated the degradation of a limited selection of nutrients by a single physico-chemical parameter (such as temperature or pH) in foods (see Table 1), our study stands out both by its broad scope documenting the effects of multiple factors (physical state of the product, temperature, pH, protective atmosphere, level of protein hydrolysis, presence or absence of fats, fibre, type of packaging, flavours) on most nutrients (covering macronutrients, fatty acids, amino acids, vitamins and minerals) and by the quantity of analysed data (analyses of data collected on more than 1400 different recipes).

Table 1.

Literature overview about stability of nutrients in different food categories.

Nutrients	Food category	Conclusions	Reference
Vitamin A, C	Powdered milk	Good stability of Vitamin C, Vitamin A shows temperature dependent losses	Baéz et al. (2012)
Vitamin A, C, E, B1	Powdered enteral formula	Content of vitamin A, E and B1 decreased with temperature and time, vitamin C stayed stable	Yang et al. (2022)
Vitamins B1, B2, B6 and E	Freeze dried meals	Mean 24-month retention rates across all temperatures (1, 30 and 40 °C) and fortification levels were 94 %, 97 %, 86 % and 77 % for vitamins B1, B2, B6 and E	Coad and Bui (2020)
Vitamins A, C, E Fe, Se	infant milk-based powdered formula	All vitamins showed decreases during storage (25 and 40 °C), Fe and Se content remained unchanged	Chávez-Servín et al., (2008)
Niacin, Folic acid, Vitamins B1, B2, B6, A and E	liquid infant milk	All vitamins beside A stayed constant at 20, 30 and 37 °C over 12 months	Albalá-Hurtado et al., (2000)
Vitamins B1, E, A	liquid enteral feeding formulas	High degradation at all temperatures (4, 20 and 30 °C) for all vitamins after 9 months, highest degradation for vitamin A (all-trans-retinol and 13-cis-retinol)	Frias and Vidal-Valverde (2001)
Vit A and E	growing up formula	Decreases in vitamin content were higher in formulas stored at 37 °C compared to storage at 23 and 30 °C	Haro-Vicente et al., (2013)
Vit A, B1, C,	liquid enteral feeding formulas	Higher decrease in vitamin content at higher temperatures (4 °C, 25 °C, 30 °C and 40 °C tested)	Yang et al. (2023)
Vitamin A	enteral feeding formulas, powder	No degradation found over 12 months of storage (22–30 °C)	Duarte Fávaro et al., (2011)
Vitamin C	fortified foods	- Review -Temperature, the form of vitamin and the matrix are the factors affecting most the stability of vitamin C in foods and beverages.	Steskova et al. (2006)
All vitamins	Fortified foods and supplements	factors that affect the stability vary:	Ottaway (2008)
		-Vitamin A: light, oxygen, low pH (esters are more stable)
		-Beta-Carotene: light, oxygen, low pH
		-Vitamin E: oxygen, temperature (Tocopherol); high pH (Tocopheryl acetate)
		-Vitamin D: light, low pH
		-Vitamin K: light, high pH
		-Thiamin: temperature, high pH, sulphites, copper ion, aromatic aldehydes
		-Vitamin B2: light
		-Niacin: very stable
		-Pantothenic acid: low and high pH
		-Folic acid: low pH, light, vitamin B2, UHT treatment
		-Vitamin B6: light
		-Vitamin B12: light, oxidising and reducing agents, other vitamins
		-Biotin: good stability, only UV light
		-Vitamin C: oxygen, traces of heavy metal ions, pH, light, temperature
Vitamin C, fatty acids and tocopherols	freeze-fried human milk	Store time 3 months, at 4 °C or 40 °C. Vitamin C significantly decreased at both temperatures, while antioxidant capacity only decreased at 40 °C. Fatty acids composition and both γ-tocopherol and δ-tocopherol contents remained unaltered	Lozano et al. (2014)
Lipids	General food	- Review-Food lipids are the foods components that are most susceptible to oxidation, therefore oxidation reactions are one of the major sources of deterioration that occurs during manufacturing, storage, distribution and final preparation of foods.	Ahmed et al. (2016)
Fat	Reconstituted enteral formula	Significant effect of packaging materials and storage temperature on the shelf-life attributes of developed ready to reconstitute enteral formulae across storage. Minimal quality loss was recorded at product stored at 4 °C as compared to the product stored at 27 °C. Among the different packaging materials, the airtight glass container had better barrier properties owing to minimal quality losses in all the formulae across the storage.	Bordoloi et al. (2020)
Sugars, amino acids	General food	- Review-This paper reviews some of the research designed to lead to an increased understanding of the chemistry of the Maillard reaction, based on recent developments, and its influence on food properties like colour, flavour and nutritional value.	Martins et al. (2000)
Lysine	infant formula	Storage temperature and time promoted the production and accumulation of Maillard products and the loss of available lysine, but storage humidity did not have a significant effect on it.	Wang et al. (2022)
Lysine	infant formula	Available lysine contents decreased significantly throughout the storage time, and the contents at 37 °C were lower than at 20 °C	Ferrer et al. (2003)

According to this study, nutrients that display no or low losses under all tested conditions in FSMPs include fat, protein, individual fatty acids, minerals and the vitamins B2 (riboflavin), B3 (niacin), B6 (pyridoxine), B7 (biotin), vitamin E and K and beta carotene. These nutrients should therefore be excluded from the scope of any FSMP shelf-life tests as they are not parameters that would limit the shelf-life of the product. The stability of some of these nutrients such as minerals was expected and therefore almost never studied nor documented in the literature before. Such data is, however, valuable as it may help to convince national regulatory authorities to exclude these nutrients from shelf-life tests. In some countries, mineral analyses are currently mandatory to be included in shelf-life tests of FSMP products by some regulatory authorities. Since FSMPs include the term “medical”, guidance on the shelf-life tests of pharmaceutical products is sometimes taken as the reference for the shelf-life tests to be conducted for FSMPs. Drug stability guidelines are however inappropriate for FSMPs. Indeed, FSMPs are foods and not drugs. Their properties and composition differ widely from drugs. Drugs usually contain only one or very few active pharmaceutical ingredients (API) whereas FSMPs do not contain API but multiple nutrients (up to 40+). Despite the fundamental difference between an API and a nutrient, some FSMP guidelines recommend the inclusion of all nutrients declared in the nutritional table of the FSMP label to be included in the shelf-life tests of these products, to mimic the pharmaceutical guidelines which request API testing over shelf-life. Since all macro and micronutrients must be declared, such shelf-life tests end up including 40 or more nutrients, many of them stable and having no impact on the determination of the shelf-life duration.

Based on both the main and complementary analyses of this study, it can be concluded that a limited number of nutrients display significant losses over the product shelf-life (see Table 2). In powders, only vitamin A showed losses. In liquids, nutrients displaying significant losses included the B vitamins B1 (thiamin), B5 (pantothenic acid), B9 (folic acid) and B12, the vitamins A, C and D and most amino acids. Observed losses in total sugars, glucose, lactose, and amino acids are likely due to Maillard reducing reactions taking place during shelf-life (Martins et al., 2000; Wang et al., 2022).

These results largely confirm previous studies (see Table 1) although some differences exist. In our study, only Vitamin A shows losses in powders whereas Yang et al. (2022) in powdered enteral formula reported that the highest decrease when temperature increased was indeed observed for vitamin A but they also reported losses in vitamin B1 and E. For Vitamin E, this can be explained by different forms of added tocopherols. Where tocopherol esters have a good temperature stability, natural occurring tocopherol is temperature labile (Ottaway, 2008). For vitamin B1, the losses described by Yang et al. (2022), could be due to a variety of reasons such as inclusion of sulfites (Ottaway, 2008), different manufacturing processes or types of ingredients. Chávez-Servín et al. (Chávez-Servín et al., 2008) detected decreases in vitamins C, E and A during storage in infant milk-based powdered formula, however these trends were found by analysis of two products only when our study evaluated more than 500 recipes for these three vitamins. In liquid infant milk, Albalá-Hurtado et al. (Albalá-Hurtado et al., 2000) detected no degradation at 20, 30 and 37 °C of the B vitamins B1 (thiamin), B2 (riboflavin), B3 (niacin), B6 (pyridoxine), B9 (folic acid) and vitamin E, but vitamin A was decreasing. In our study, vitamin A losses at high temperatures in liquids were observed as well. However, in contrast to the results from Albalá-Hurtado et al. (Albalá-Hurtado et al., 2000), vitamin B1 displayed very high degradation rates at higher temperatures and some minor degradation of folic acid occurred. One explanation for these differences might be that Albalá-Hurtado et al. (Albalá-Hurtado et al., 2000) only presented the results on vitamin A and did not define what they considered as no significant losses. In any cases, it is obviously critical to ensure that the right analytical methods (specifically validated for the food matrices in which it will be used) are used in all studies aiming at quantifying nutrient content and losses over shelf-life.

In non-acidified liquid FSMPs, our study identified 3 nutrients that displayed losses much larger than the others. These 3 nutrients are by decreasing amplitude of losses: vitamin C, thiamin and vitamin D (losses only in products with hydrolyzed proteins). In acidified liquid FSMPs, 5 nutrients displayed losses much larger than the others. These nutrients are by decreasing amplitude of losses: pantothenic acid, vitamin C, folic acid, thiamin and vitamin D (losses only in products with intact proteins). In powder the only nutrient displaying losses was vitamin A. From a nutritional perspective, it is these more labile nutrients that will define and reduce the shelf-life of products. These nutrients with higher rates of degradation during shelf-life can therefore be used as tracers or markers to indicate product nutritional suitability and help define product shelf-life. If the levels of these more unstable nutrients remain adequate at the end of shelf-life, the levels of the other, more stable, nutrients will be adequate too. Typically, therefore, shelf-life tests for FSMPs should only require a quantitative analysis of a limited number of identified nutrient tracers.

Higher dosing of these labile vitamins during manufacturing is typically conducted to compensate for the losses occurring over shelf-life and ensure an adequate amount throughout the whole shelf-life. This is only possible if tolerances around the declared values are broad enough. While guaranteeing safety for the target consumers, the tolerances of labile nutrients must take into consideration the nutrient losses during shelf-life in addition to the sources of variability applicable to all nutrients (such as analytical and manufacturing variabilities, variability in raw material content).

Packaging type and size, degree of humidity, addition of fibres or use of different flavours did not affect (significantly) the stability of any nutrient. A shelf-life study performed on a single product is therefore sufficient to define the shelf-life duration of a group of products that would only differ by these parameters. The fact that humidity, packaging types and sizes were not observed in this study to have any impact on the stability of nutrients can be explained by common precautions taken when formulating and designing FSMPs. Such precautions include the systematic use of opaque packaging to protect light-sensitive nutrients such as Vitamin A and E (Ottaway, 2008). Air-tight packaging is also the norm to maintain the low water activity of powder products over long shelf-life (of 2 years or more) as well as inert gases like nitrogen that are used to flush out oxygen from the packaging in order to prevent oxidation of sensitive nutrients such as fatty acids (Wang et al., 2023).

The study has some limitations that warrant consideration. Firstly, it primarily focuses on nutrient degradation as the key factor in determining shelf life but does not address other aspects, such as sensorial and physical degradation, which could affect product use and consumer acceptability and, therefore, impact the final shelf life of a product. Secondly, different testing methods were used which could potentially have yield disparate results. However, only state-of-the-art methods (such as AOAC INTERNATIONAL, ISO methods) specifically designed for FSMPs and as a mitigating step normalised data rather than absolute values were used. Thirdly, the study does not explore whether the nutrient levels themselves impact their rate of degradation. Lastly, the scope of the study focused on shelf-life losses in sealed products.

One of the strength of the study is that the results generated over the shelf-life of FSMPs are likely to be applicable to other food products with complex matrices such as infant and follow-up formula or meal replacement for weight control packed in similar packaging materials. In addition, the broad scope of the study also provides valuable insight into which parameters affect which nutrients in foods in general.

CRediT authorship contribution statement

Laurent Ameye: Conceptualization, Investigation, Supervision, Writing – original draft, Writing – review & editing. Sophie de Brouwer: Methodology, Data curation, Formal analysis, Writing – original draft, Writing – review & editing. Donald L. Gilliland: Conceptualization, Investigation, Writing – original draft, Writing – review & editing. Julia Heckmann: Conceptualization, Investigation, Writing – original draft, Writing – review & editing. Kaushik Janakiraman: Investigation, Writing – review & editing. Bridget-Anne Kirwan: Methodology, Supervision, Writing – original draft. Xavier Lavigne: Conceptualization, Investigation, Writing – original draft, Writing – review & editing. Francisco Moya: Conceptualization, Investigation, Writing – original draft, Writing – review & editing. Vanessa Zammit: Conceptualization, Investigation, Writing – original draft, Writing – review & editing.

Declaration of competing interest

Laurent Ameye is an employee of Nestlé Health Science, Donald L Gilliland is a retired employee of Abbott Nutrition, Julia Heckmann is an employee of Fresenius Kabi Deutschland GmbH, Kaushik Janakiraman is an employee of Reckitt, Xavier Lavigne and Francisco Moya are employees of Abbott Nutrition, Vanessa Zammit is an employee of Danone Specialised Nutrition.

Handling Editor: Professor A.G. Marangoni

Data availability

The data that has been used is confidential.