Design and study population

We analyzed the dietary data collected through the SFFQ of Ensanut 2012 (SFFQ-2012). Ensanut 2012 is a probabilistic survey stratified by cluster, and representative at the national, regional and urban/rural levels. For determining sample size a design effect of 1.836 was considered, allowing for the estimation of the following expected prevalences for the different population age groups: 2% for adults (>20 years), 3% for adolescents (12-19 years) and 4% for pre-school (1-4 years) and school (5-11 years) children, with a 95% confidence interval and response rates of 0.81 in pre-school as well as school children and 0.78 in adolescents as well as adults in general. The Ensanut 2012 obtained information from 50 528 households, and its SFFQ was randomly applied in a population subsample was made up of one out of every six subjects per age group: pre-school children (1-4 y), school children (5-11y), adolescents (12-19 y) and adults (>20 y). Survey sampling and methodology procedures have been described elsewhere.^5,6

The Ensanut 2012 protocol and survey tools were approved by the Ethics Committee of the National Institute of Public Health in Mexico (INSP by its Spanish acronym, from Instituto Nacional de Salud Pública). Informed consent was obtained from all study subjects.

SFFQ-2012: food items and portion sizes

The SFFQ-2012 was based on the previous SFFQ used by Ensanut 2006 (SFFQ-2006).⁷ Designed to characterize the diets of the Mexican population by age group, the SFFQ-2006 collected dietary data over the last seven days to obtain current dietary data for the population, and was composed of selected foods based on their contribution to total consumption of energy and the 11 nutrient intake from 24-hour recall (24-HR) of the National Nutrition Survey 1999 (ENN 99, by the Spanish acronym for Encuesta Nacional de Nutrición, 1999). The foods that contributed ≥90% of the total energy and nutrients intake were included in the SFFQ-2006 list. The questionnaire was also complemented with food items identified as inhibitors or facilitators of iron and zinc absorption. The portion sizes used in the 2006 and 2012 SFFQs were divided into (1) standard portion size (the principal portion size specified in the food list) based on the median food intake by population age group from the 24-HR of the ENN 99; and (2) alternative (standard) portions sizes ranging from very small to extra-large. All the portion sizes for each food item were expressed in standardized home-measurements.

On the other hand, the SFFQ-2012 incorporated 39 additional food items given their relation to the high prevalence of non-communicable diseases across all age groups. The new food items were selected by a group of experts in nutrition from the Center for Nutrition and Health Research in the National Institute of Public Health in Mexico. Commonly consumed foods and dishes high in calories, sodium, fats and sugars were added (Table I). With 140 food items classified into 14 general groups (Table II), the final SFFQ-2012 was validated for estimating energy and nutrient intake (further details are published elsewhere).⁸

Table I Stages of work and criteria for inclusion of new food items in SFFQ-2012 

Table II Food groups listed in the Ensanut 2012 SFFQ 

SFFQ application

The SFFQ-2012 was administered by trained health personnel (i.e., nutritionists, nurses and physicians) using laptop computers (Hewlett Packard 435) with specifically designed Ensanut 2012 software (Visual Fox Pro program, v.7) for data entry. Interviewers asked study subjects to recall all foods (and portions) consumed in the seven days prior to the interview. The SFFQ standard portion sizes specified by interviewers were based on the average weight value assigned to each food item per age group. The subjects who reported not having consumed the standard portion size were asked to select an alternative portion size, ranging from very small to extra-large. They were then asked to specify the number of days and the number of times per day they consumed the food item in question during the seven days prior to the interview, as well as the number of portions consumed on each occasion. For estimations, the number of days was multiplied by the number of times per day that the food item was consumed in the last seven days. Exclusively in the case of children aged from 1 to 11, the questionnaire was administered to their mothers or caretakers; for adolescents and adults (≥12 years of age) it was applied directly. Figure 1 shows SFFQ-2012 software and its use.

Figure 1 Software to capture the SFFQ-2012 from Ensanut 

Food composition database

We constructed a database to estimate the energy and nutrient intake data collected through the SFFQ-2012. It was based on a more general food composition database (GDFC) containing 1 600 foods which had been compiled and updated that same year by the Center for Nutrition and Health Research of the INSP.⁹ To construct our database, we followed a series of steps and criteria: Step 1. We identified and selected the GDFC foods that represented each food item on the SFFQ-2012 food list. For example, in the case of the "banana" food item, we selected all GDFC banana types: dominico, manzano, macho, red and tabasco. For the mature and fresh cheese item, we selected all GDFC cheeses and classified them within the fresh and mature categories according to the corresponding Official Mexican Regulation.¹⁰Step 2. We complemented the food items in Step 1 with those reported under the Ensanut 2012 24-HR, taking into account information from all the population age groups (pre-school and school children, adolescents, adults and older adults). Step 3. For 40% of the SFFQ list, we estimated the composition of each food item, and computed the average energy, fiber and nutrient content (per 100 g) taking into account the matching GDFC foods. For the remaining 60%, we weighted the average energy and nutrient content based on the Ensanut 2012 24-HR data. Step 4. In the case of the newly incorporated dish items, we estimated the energy and nutrient content (per 100 g of preparation) according to their raw ingredients.

Nutrient estimates

Using the food composition database and the weights of the size portions described above, we estimated the average daily intake of energy, fiber, macronutrients and micronutrients of the population (Figure 2). Net grams of food consumed were determined considering the density factor for beverages and the edible portion factor for fruits, vegetables and meats. Daily dietary intake per person was calculated using Access (Microsoft Office, 2003), Visual Basic Code and SQL consults, and Stata v.13 software. SFFQ-2012 asked about consumption of food supplements and complements. However, we constructed a qualitative variable that allowed us to identify who had consumed food supplements/ complements in the last seven days. Figure 2 shows the process followed for computing the average daily intake obtained through the SFFQ-2012.

Figure 2 Process for estimation of average daily intake per individual 

Finally, we generated DI data for 133 nutrients: energy, water, fiber, five macronutrients, 12 minerals, 26 vitamins, 66 types of fatty acids, 19 amino acids and other nutrients such as phytic acid and tannins. We also estimated the proportion of the population that reported consuming food complements/supplements: <11% of adults, 16.7% of older adults, 6.3% of adolescents, 8.8% of pre-school and 8.3% of school children.

Cleaning DI data

We prioritized the cleaning of DI data as regards to energy, fiber and the 13 nutrients identified as most relevant in the area of public health according to evidence from Mexico. These consisted in five macronutrients (carbohydrates, proteins, total fats, saturated fats and polyunsaturated fats), five vitamins (vitamin A, vitamin C, folates, vitamin D and vitamin B12), and three minerals (iron, zinc and calcium). Cleaning was performed in two stages. We first cleaned data in grams per food item consumed per subject of study. At this level it is important to clarify that there were no individuals with missing data. Those who had consumed one or more foods above three standard deviations (>3 SD) were excluded from analysis (Figure 3).¹¹ We also evaluated the biological plausibility of food intake and the contribution percentage of each food vis-à-vis total DI. Additionally, the data identified as high was corroborated. Extremely low values were considered valid.

* Excluded for having one or more implausible values in energy, fiber and nutrients: energy=49, macronutrients=8, vitamins=62 and minerals=23

^‡ Excluded for having one or more implausible values in energy, fiber and nutrients: energy=49, macronutrients=6, vitamins=1 and minerals=3

^§ Excluded for having one or more implausible values in energy, fiber and nutrients: energy=121, macronutrients=9, and vitamins=22

^# Excluded for having one or more implausible values in energy, fiber and nutrients: energy=99, macronutrients=12, and vitamins=37

^& Excluded for having one or more implausible values in energy, fiber and nutrients: energy=42, macronutrients=6, and vitamins=1

Figure 3 Sample size analyzed 

In a second stage, we cleaned daily intake regarding energy, macronutrients, micronutrients and fiber. To clean data at the upper extreme values of energy intake, we estimated the ratio of energy intake/estimated energy requirement (EER). As reference, we used two sets of equations from the Institute of Medicine (IOM).¹² With regard to body mass maintenance, we used specific equations for the population with obesity, overweight and normal nutritional status by age group. For children <3 years of age, only one type of equations was applied. For study subjects without weight and height information, values were imputed according to data drawn from the same survey population of the same age and sex. Only data for 5.4% of subjects from the total SFFQ-2012 population were imputed. The physical activity level of each subject was considered according to several studies regarding data of the ENN-99.¹³ We assigned the physical activity factor as follows: a light physical activity factor was assigned to pre-school and school children as well as adolescent and adult males, whereas a sedentary activity value was assigned to adolescent and adult females. In relation to nutritional status, physical activity factors of 1.13 and 1.16 were assigned to non-obese boys and girls ages 3 to 18, respectively, whereas factors of 1.12 and 1.18 were assigned to obese boys and girls of the same age, respectively. For adults' ≥19 years, factors of 1.11 and 1.0 were assigned to men and women, respectively. Physical activity levels were not assigned to children <3 years.¹²

To clean data at the lower extreme values of energy intake, we excluded subjects with energy intake/basal metabolic rate (BMR) ratios below 0.5. We estimated the BMR for adults (≥19 years of age) according to the Mifflin-St Jeor equations¹⁴ for overweight and obese populations by gender. For subjects <19 years, we used the Food and Agriculture Organization (FAO) equations¹⁵ according to age and gender.

Cleaning of nutrient data was performed by age group using the following ratios: intake/estimated average requirement (EAR) for micronutrients^16-18 and proteins;¹² intake/adequate intake (AI) for fiber;¹² and intake/range of acceptable macronutrients distribution range (AMDR).¹² For carbohydrates and total fats, average values of 55 and 30% of AMDR, respectively, were considered to be adequate.

We cleaned iron data for all age groups using the specific recommendations for the Mexican population,¹⁹ as they take into account the low bioavailability of iron in the diet (<10%). Intake of phytates -the main inhibitor of iron absorption- is high in Mexico. Additionally, the Mexican diet is low in heme iron (<5% of total iron), which has a higher absorption rate than non-heme iron.²⁰ In light of the foregoing, iron intake recommendations for the Mexican population are higher than those issued, for instance, for the US population by its respective IOM, with the latter positing an iron bioavailability of 18% and a heme intake of ≈ 10% of total iron.¹⁶ For adults >60, we used the WHO nutrient recommendations²¹ given their comparability with Ensanut 2006 intake results, as well as their similarity to the IOM recommendations, except as regards folates (400 μg/d recommended by the WHO and 320 μg/d recommended by the IOM).

All respondents with ratios above +3 SD were excluded. All DI cleaning was done using Stata v.13.1 (StataCorp. 2011, College Station, TX: Stata Press).

Finally, we obtained DI for 7 214 study subjects from SFFQ-2012 with plausible data for a total 7 810 subjects. Between 4.8% and 9.6% of population from the different age groups were excluded for lack of plausible information (Figure 3). Final estimated daily intake did not include energy or nutrients from food complements/ supplements consumed by <11% of adults, 16.7% of older adults, 6.3% of adolescents, 8.3% of school children and 8.8% of pre-school children.