Diet selection
To estimate the environmental impact, data was collected from different types of pet food for healthy adults. All pet foods were categorized as dry (extruded pet food with 12% or less moisture), wet (canned or pouch), and homemade diets (produced using the same ingredients as man food). Homemade pet foods were subcategorized as “commercial homemade” (produced and sold by pet food companies) or “website homemade” (recipes recommended by websites to be cooked at home by owners). To estimate the environmental impact of commercial pet foods, all commercial dry and wet diets found on the websites of the three major retailers of the pet food sector in Brazil were selected. Commercial homemade diets were selected after a search using the Google search tool using the Portuguese terms for “buy” and “homemade diet”, followed by the terms “dog”, “canine”, “cat” or “feline”. Website recipes published in Portuguese were selected using the Google search tool, and search terms were “homemade diet recipe” and “homemade food recipe” followed by the terms “dog” and “cat”. For both commercial and website homemade diets, the results obtained up to the 10th page of the search tool for each term were considered.
All diets included were advertised as complete and balanced for healthy adults, and exclusion criteria were diets for puppies or kittens, senior diets, therapeutic diets, and treats. Website homemade diet recipes were excluded if the quantity of one or more ingredients was not specified and the same recipes on different websites were also not included.
Ingredient inclusion percentage
Information regarding the ingredients (except premixes, additives, and preservatives) and guaranteed analysis from labels of all commercial diets were collected. For the recipes of homemade diets acquired from websites, the ingredients and their amounts were considered as described by the website’s authors.
The ingredient inclusion percentages for each commercial diet were estimated using a diet formulation software34, aiming at the dry matter macronutrient concentration. The guaranteed analysis information of macronutrients (crude protein, crude fat, crude fiber, and ash) was converted to a dry matter basis according to the moisture declared on the label. This information was then inserted into the nutrient composition part of each diet in the software.
For nutrients with minimum guaranteed levels (crude fat and crude protein), values for maximum inclusion in the software were considered as up to 10% of the minimum value. For nutrients with maximum guaranteed levels (crude fiber and ash), only maximum levels were inserted in the software.
The ingredient database for commercial wet and dry diets was obtained preferably from the Brazilian Association of the Pet Food Industry (ABINPET)35, but when not described in this publication, other sources were used36,37. For the homemade diets (commercial and website), the ingredient database was obtained from the USDA’s FoodData Central37 or, when not presented at FoodData Central, the Brazilian Table of Food Composition (TACO)38 was used.
After the percentages of inclusion of ingredients were estimated in a dry matter basis, they were converted to percentage of inclusion in original matter basis (as fed), considering the ingredients’ moisture35,36,37.
For the website homemade diet recipes, the amount in original matter basis was already stated, and inclusion percentage was calculated according with total amount of the recipe and the amount of each ingredient.
Macronutrient profile
The quantities of protein, fat and nitrogen-free extract (NFE) of the diets were calculated according to label information provided by the manufacturers. The information regarding the metabolizable energy and the minimum amounts of crude protein and crude fat according to the guaranteed analysis information were obtained, and with this information the amount of nutrient per 1000 kcal of the diet was estimated for the dry, wet, and commercial homemade diets. For these three types of diets, the NFE was calculated according to the NRC24 equation:
$$ \% {\text{ NFE}}\, = \,{1}00{-}(\% {\text{ crude\, protein}}\, + \,\% {\text{ crude\, fat}}\, + \,\% {\text{ crude \, fiber}}\, + \,\% {\text{ ash}}\, + \,\% {\text{ moisture}}). $$
For the website homemade diets, the information was obtained by the composition of the recipe, as they did not contain labels. According with the recipe, the metabolizable energy, protein, fat, and NFE were estimated based on the composition of nutrients37.
The metabolizable energy of each diet as informed by the manufacturers on the labels was considered for commercial dry, wet and homemade diets. The Atwater method was used to calculate the energy of website homemade diets24, considering 4 kcal per gram of protein and NFE and 9 kcal per gram of fat39.
Nutrient and energy source estimate
To better understand the source of nutrients of diets for dogs and cats, the percentage of protein, fat and metabolizable energy provided by vegetable or animal ingredients was calculated for each diet. The percentage was calculated according to the contribution of the nutrient provided by each ingredient in the diet, and if this ingredient was of animal or vegetable origin. For the calculation of the energy source percentage, the energy provided by each ingredient type was considered.
Environmental impact estimate
The environmental impact variables evaluated were greenhouse gas (GHG) emission (as carbon dioxide equivalent emission—CO2eq), land use, acidifying emission (as sulphur dioxide equivalent emission—SO2eq), eutrophying emissions (as phosphate equivalent emissions—PO43−eq), freshwater withdrawals, and stress-weighted water use per 1000 kcal of diet, according with the metabolizable energy of the diet and the percentage of inclusion of each ingredient in the diet, as the equation below:
$$Impact\, variable\, per\, 1000 \,\mathrm{kcal}=\frac{Amount \,of\, ingredient/1000 \,\mathrm{kcal} \,of \,diet \,\times \,Variable/1000 \,\mathrm{kcal} \,of \,ingredient}{1000\, \mathrm{kcal}}$$
To obtain these results, the diet composition was first converted from a dry matter basis to a 1000 kcal basis using the following equation, applied to all ingrediets present in the diet:
$$Amount\, of \,ingredient\, per\, 1000 \,\mathrm{kcal}= \frac{Amount \,of\, ingredient\, per\, kg\, of \,diet\, \times \,1000}{Metabolizable\, energy\, of \,the \,diet\, (\mathrm{kcal}/\mathrm{kg})}$$
The comparison per 1000 kcal was used to put all diets on a basis of dietary intake, as a dog or cat requires the same energy intake regardless of the diet chosen and is a reliable unit to compare dietary composition and nutrient intake.
The data used to estimate the variables of environmental impact was based on the data from Poore and Nemecek7 for nutrition functional units as 1000 kcal. When data was provided per 100 g protein or per kg of product, it was converted to 1000 kcal based on data from ABINPET35, Butolo36, TACO38, and USDA37 (Table S4). The ingredients were classified in one of the 43 groups listed by Poore and Nemecek7, for example, all types of beef meat were calculated as bovine meat.
Furthermore, the relationship between the dietary nutrient composition and the variables that were used to evaluate the environmental impact was assessed.
Statistical analysis
The statistical analysis was performed using R Core Team40. Adherence to normality was tested with the Shapiro–Wilk test, and as only the variables crude protein concentration and NFE of website homemade diets, and SO2eq of dry diets were considered to adhere to normality, non-parametric tests were performed. For the analysis of macronutrient profile and the estimated environmental impact, the Kruskal–Wallis test was used to compare variables. When at least one median was considered different, multiple comparisons between groups were performed. The comparison between energy provided by ingredients of vegetable and animal origin was performed with the Wilcoxon test, considering the variables as two dependent samples. Values of p < 0.05 were considered significant.
The principal component analysis (PCA) was used to evaluate the relation between the diet characteristics and the variables of environmental impact. As the units of the variables were different, they were scaled considering mean = 0 and variance = 1. The first (PC1) and second principal components (PC2) were responsible for 68.2% of the data variance for dogs (Fig. S3). For cats, PC1 and PC2 are responsible for 71.1% of the data variance (Fig. S4).
