Dietary protein consumption profiles show contrasting impacts on environmental and health indicators

Patterns of protein intake are strong characteristics of diets, and protein sources have been linked to the environmental and nutrition/health impacts of diets. However, few studies have worked on protein profiles, and most of them have focused on specific diets like vegetarian or vegan diets. Furthermore, the description of the environmental impact of diets has often been limited to greenhouse gas emissions (GHGe) and land use. This paper analyzes the alignment of environmental pressures and nutritional impacts in a diversity of representative protein profiles of a western population. Using data from a representative survey in France (INCA3, n = 1,125), we identified protein profiles using hierarchical ascendant classification on protein intake (g) from main protein sources (refined grains, whole grains, dairy, eggs, ruminant meat, poultry, pork, processed meat, fish, fruits & vegetables, pulses). We assessed their diet quality using 6 dietary scores, including assessment of long-term risk for health, and associated 14 environmental pressure indicators using the Agribalyse database completed by the SHARP database for GHGe. Five protein profiles were identified according to the high contributions of ruminant meat, pork, poultry, fish, or, conversely, as low contribution from meat. The profile including the lowest protein from meat had the lowest impact on almost all environmental indicators and had the lowest long-term risk. Conversely, the profile with high protein from ruminant-based foods had the highest pressures on most environmental indicators, including GHGe. We found that the protein profile with low contribution from meat has great potential for human health and environment preservation. Shifting a large part of the population toward this profile could be an easy first step toward building a more sustainable diet.


Abstract 23
Patterns of protein intake are strong characteristics of diets, and protein sources have been linked to 24 the environmental and nutrition/health impacts of diets. However, few studies have worked on protein 25 profiles, and most of them have focused on specific diets like vegetarian or vegan diets. Furthermore, 26 the description of the environmental impact of diets has often been limited to greenhouse gas 27 emissions (GHGe) and land use. This paper analyzes the alignment of environmental pressures and 28 nutritional impacts in a diversity of representative protein profiles of a western population. 29 Using data from a representative survey in France (INCA3, n = 1,125), we identified protein profiles 30 using hierarchical ascendant classification on protein intake (g) from main protein sources (refined 31 grains, whole grains, dairy, eggs, ruminant meat, poultry, pork, processed meat, fish, fruits & 32 vegetables, pulses). We assessed their diet quality using 6 dietary scores, including assessment of 33 long-term risk for health, and associated 14 environmental pressure indicators using the Agribalyse 34 database completed by the SHARP database for GHGe. 35 Five protein profiles were identified according to the high contributions of ruminant meat, pork, 36 poultry, fish, or, conversely, as low contribution from meat. The profile including the lowest protein 37 from meat had the lowest impact on almost all environmental indicators and had the lowest long-term 38 risk. Conversely, the profile with high protein from ruminant-based foods had the highest pressures on 39 most environmental indicators, including GHGe. 40 Graphical abstract 45 46 Environmental pressure by profile: All data are in percentage compared to the mean value of the 47 population. GHG is the emission of greenhouse gases measured in kg CO 2 eq. The ozone depletion is 48 in kg CFC-11eq. The photochemical ozone formation is in kg of Non-Methane Volatile Organic 49 Compounds eq. Particulate matter is in kg of PM 2.5 emitted. The acidification is in mol H + eq. The 50 terrestrial eutrophication is in mol N eq, the freshwater eutrophication is in kg P eq, and the marine 51 eutrophication is in kg N eq. The freshwater ecotoxicity is based on the USEtox model. Land use is in 52 kg C deficit, water use in m 3 , fossils resource use in MJ, and Metals and minerals use in kg SB eq. 53 Nutritional and health indicators by profile: All data are in percentage of difference to the mean value 54 of the population. The HiDiet was used to assess the diet impact on long-term mortality and morbidity 55 (variation between -1 and 1). The Alternative Healthy Eating Index (AHEI-2010) is a modified 56 Healthy Eating Index, assessing the adherence to Dietary Guidelines for Americans, improving target 57 food choices and macronutrient sources associated with reduced chronic disease risk (maxpoint = 58 100). The PANDiet evaluates the probability of adequate nutrient intake (maxpoint = 100). The 59 SecDiet evaluates the nutrient risk of overt deficiency (maxpoint = 1). The Literature-Based 60 Adherence Score to the Mediterranean Diet (LAMD) assesses adherence to the Mediterranean diet 61 (maxpoint = 16). sPNNS-GS2 the adherence to the French Food-based Dietary Guidelines (maxpoint 62 = 10.5). 63 64 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)  Since information on trans fatty acids was not available, only ten parameters were considered, namely  144   vegetables, fruit, whole grains, sugar-sweetened beverages and fruit juice, nuts and legumes,  145 red/processed meat, long-chain (n-3) fatty acids (EPA + DHA), polyunsaturated fatty acids (PUFA), 146 sodium, and alcohol. All AHEI-2010 components were scored from 0 (worst) to 10 (best), resulting in 147 a total AHEI-2010 score ranging from 0 (non-adherence) to 100 (perfect adherence). The scoring 148 criteria are described in Supplemental  Table 3. 152 The PANDiet evaluates the probability of adequate nutrient intake. This score is a 100-point 153 probabilistic score evaluating adequate overall nutrient intake. It combines an adequacy sub-score and 154 a moderation sub-score. The adequacy sub-score is calculated as the average probability of adequacy 155 of nutrients for which the usual intake should be above a reference value, multiplied by 100. The SecDiet evaluates nutrient deficiency risk. This score is based on the intake of twelve critical 163 nutrients for nutritional risk of overt deficiency (Salomé et al., 2021). For each nutrient, we calculated 164 the probability of having sufficient-enough intake to avoid overt nutrient deficiency. We used the 165 probability distribution of the standard normal distribution for nutrient requirements while taking into 166 account the mean intake, the day-to-day intake variability, the inter-individual variability and the 167 nutrient deficiency threshold. 168 The HiDiet was used to assess the diet impact on long-term mortality and morbidity. The HiDiet score 169 is based on the principle of the Comparative Risk Assessment but applied to the risk of one individual 170 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 7, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. ; grains, dairy, eggs, ruminant meat, poultry, pork, processed meat, fish, fruits -vegetables, pulses, and 197 others) exhibiting differences in nutritional properties and the way they are eaten. Each group's daily 198 protein intake was then calculated for each individual. Data from the SHARP database were used to complete missing values for the GHGe indicator 221 (Mertens et al., 2019). This database also includes data on land use but because the unit is different, 222 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. The hierarchical ascendant classification identified five distinct profiles (figure 1). The characteristics 242 of the population and the different profiles are described in Table 1. Profile 1 (Low meat profile) 243 represents consumers with a low protein intake from meat (9 g/day vs. 20.8 g/day in the total 244 population), who also have a relatively higher intake of protein from dairy products (13.9 g/day vs. 9.8 245 g/day in the total population). Supplemental Table 5 shows that the higher dairy protein intake mostly 246 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022.
comes from the higher consumption of milk (88 g/day vs. 80 g/day in the total population) and cheese 247 (54 g/day vs. 42 g/day in the total population). 248 Profile 2 (Poultry profile) represents people with high intakes of protein from poultry (10.8 g/day vs. 249 5.3 g/day in the total population), while profile 3 (Fish profile) represents individuals with the highest 250 fish protein intake (9.6 g/day vs. 4.4 g/day in the total population). Profile 4 (Ruminant meat profile) 251 represents individuals whose protein intake is largely contributed by ruminant meat proteins (12.4 252 g/day vs. 6.1 g/day in the total population), and profile 5 (Pork profile) represents people with high 253 intakes of protein from pork (9.6 g/day vs. 2.7 g/day in the total population). 254 The consumption of 33 food groups in grams per day is also presented in Supplemental Table 5.  Supplemental Table 6). 257 The profile with high intake of protein from ruminant has the highest environmental pressure for most 258 of these indicators. In fact, this profile has the highest values for 5 of the 14 indicators. Notably, this 259 profile also has the highest GHGe (7.7 kg CO2 eq/day vs. 6.4 kg CO2 eq/day in the total population). 260 Compared to the total population, the land use, the emission of particulate matter, the acidification, 261 and the terrestrial eutrophication are also at their highest for this profile, being 26%, 14%, 16%, and 262 19% higher, respectively. 263 The profile with high intake of protein from fish and the one with the highest pork protein intake have 264 the next most elevated GHGe (6.6 kg CO2 eq/day and 6.4 kg CO2 eq/day). For the other indicators, 265 these two profiles are different. The profile with high intake of protein from pork contributes the most 266 to freshwater ecotoxicity (164 CTUe/day vs. 151 CTUe/day in the total population), marine 267 eutrophication (26 kg N eq/day vs. 25 kg N eq/day in the total population), ionizing radiation (1.7 kg 268 U235eq/day vs. 1.5 kg U235eq/day in the total population), energy use (67 MJ/day vs. 62 MJ/day in 269 the total population), and metals and minerals use (10.3 kg Sb eq/day vs. 9.7 kg Sb eq/day in the total 270 population). For profile with high intake of protein from fish, it contributes the most to ozone 271 depletion (0.73 Freon-11/day vs. 0.59 Freon-11/day in the total population), photochemical ozone (20 272 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. ; https://doi.org/10.1101/2022.07.07.22277350 doi: medRxiv preprint kg NMVOC/day vs. 17 kg NMVOC/day in the total population), freshwater eutrophication (1.15kg P 273 eq/day vs. 1.03 kg P eq/day in the total population) and water use (7.1m3/day of water vs. 6.7 1m3/day 274 of water in the total population). 275 The profiles with the lowest environmental impacts are the poultry profile and the low meat profile. 276 These profiles emit the smallest amount of GHGe with 5.6 kg CO2 eq/day for the high consumers of 277 poultry protein and 4.6 for the low meat consumers. This last profile has the lowest impact for most 278 environmental indicators (11 from the 14 indicators) and is the second-best profile for the three others. consumers has the highest scores for the AHEI, the LAMD, and the second-best score for the 285 PNNSGS2, the SecDiet, the PANDiet, and the HiDiet. The profile with the lowest meat protein 286 consumption has the highest HiDiet score, and the one with the highest ruminant meat protein intake 287 has the lowest HiDiet score (respectively +3% and -4% compared to the total population). The profile 288 with the lowest meat protein consumption also has the lowest PANDiet (-3%) and a neutral SecDiet 289 (0%) score. The value of the health scores are presented in Supplemental Table 8. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. The low meat profile had the smallest environmental pressure. This is in line with the literature as 315 most publications pointed out that meat, especially from ruminant and to a lesser extent from 316 monogastric, has by far the most significant environmental impact, specifically for GHGe (Poore and 317 Nemecek, 2018). Indeed, low GHGe in this profile should be ascribed to the very low consumption of 318 ruminant meat. This is similar with other studies showing plant-based diets having lower GHEe 319 compare to omnivorous diet (Rabès et al., 2020). Note that dairy intake was relatively high for this 320 profile -in fact the highest of the 5 profiles -which may be considered a practical conflict between 321 the production of ruminant meat and the production of milk. This is because of the co-production 322 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. This low-meat protein profile also has drawbacks on a nutritional aspect, with the lowest PANDiet 336 values of all profiles. This is mainly explained by the low intake of nutrient-dense food sources, in 337 particular red meat, which is not offset by higher intakes of fruits and vegetables. It should be noted 338 that even if it is the most plant-based profile, the consumption of pulses, whole grains and other 339 recommended plant food groups is far from what is recommended in optimization studies (Dussiot et 340 al., 2022;Willett et al., 2019). Moreover, the nutrient security in this low-meat profile was not lower 341 than that of the general population, confirming that even if there are some slightly lower intakes of 342 some nutrients, there is no overall increase in the risk of overt deficiency. As far as long-term health is 343 concerned, this profile has the lowest estimated risk of long-term morbidity or mortality from coronary 344 heart disease, stroke, type 2 diabetes, and colorectal cancer, as assessed using the HiDiet score. Given 345 that the small increase in nutrient inadequacy does not increase the risk of overt deficiency, this profile 346 shows strong co-benefits for environmental and human health. 347 Compared to the other profiles high in animal flesh (meat or fish), the profile with high intakes of 348 protein from poultry has a lower impact on most of the environmental indicators. In fact, GHGe are 349 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. The profile with a high fish consumption has high values for all health and nutritional scores. This was 359 expected as fish consumption is positively weighted in all of the scores considered in this study. Fish 360 is a very good contributor to bioavailable nutrients and has well-known health benefits against chronic 361 disease (Bogard et al., 2019). The high intake of indispensable nutrients associated with the higher 362 level of protein for fish but also an overall dietary profile of higher quality explains the high PANDiet 363 and SecDiet score. In contrast, this high fish intake has some important drawbacks on the 364 environmental indicators. As described before, water use and photochemical ozone formation are two 365 important issues of fish consumption (Ruiz-Salmón et al., 2021). It is important to note that there is 366 significant variability between fish species, and it may be possible to improve the health and 367 environmental impact by consuming different fish species (ADEME, 2020a). However, these results 368 need to be verified with other LCA as environmental pressures from fish are not precise, being mostly 369 measured indirectly. French representative survey to date. The environmental database Agribalyse has also some 374 limitations. For example, the quantification of soil carbon storage and removal is not considered in the 375 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. ; calculation of the GHGe. Biodiversity and the impact of phytosanitary products are also not yet 376 calculated. Another drawback is that there is no information about farming practices (such as organic 377 production) for foods production that are consumed in the survey, and the environmental indicators are 378 only for mean of consumed foods differing in farming practices. However, this database includes a 379 large variety of environmental indicators covering the entire food chain from production to the plate 380 and virtually all the foods consumed by the French population since it was created for this purpose. In the present study, we showed that the profiles of protein intake of the population are varied and 387 have contrasting associations with health and environmental impacts. The protein profile marked by 388 ruminant meat had the worst scores on both health and environmental aspects. Conversely, the most 389 environment-friendly protein profile is very low in meat, and this profile also had the lowest risk for 390 long-term morbidity and mortality. As we studied real profiles identified in the general population, the 391 differences in health and environmental impacts between profiles may be useful to consider realistic 392 targets for acceptable changes in the diet. These changes may be more practical than those identified 393 by modeling studies. Our results thus support the importance of protein profiles for health and 394 environmental impacts. 395 396 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. ; . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. ; Volatile Organic Compounds eq. Particulate matter is in kg of PM 2.5 emitted. The acidification is in 656 mol H + eq. The terrestrial eutrophication is in mol N eq, the freshwater eutrophication is in kg P eq, 657 and the marine eutrophication is in kg N eq. The freshwater ecotoxicity is based on the USEtox model. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

LEGEND FOR FIGURES
The copyright holder for this preprint this version posted July 7, 2022. ; https://doi.org/10.1101/2022.07.07.22277350 doi: medRxiv preprint 673 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 7, 2022. ; https://doi.org/10.1101/2022.07.07.22277350 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted July 7, 2022.