This cross-sectional study was conducted at the Clínica Escola de Nutrição “Irmã Anna de São José Camargo Barros” of the Centro Universitário Nossa Senhora do Patrocínio (CEUNSP, abbreviation in Portuguese), Cruzeiro do Sul Educacional, located in Itu, São Paulo, from August 2018 to February 2019.

The study protocol received approval from the Research Ethics Committee of the Max Planck Faculty (opinion number: 2.148.238), and it was conducted in accordance with Brazilian resolution 466/2012 and the 1975 Declaration of Helsinki, updated in 2013. All participants signed a free and informed consent form.

The non-probabilistic, convenience sample included adults between 20 and 59 years of age who did not follow any specific diet and were in their first nutrition consultation at the Clínica Escola de Nutrição. As exclusion criteria, the following were considered: smoking; alcohol use; use of hypoglycemic or hypolipidemic drugs; medical diagnosis of hypothyroidism or use of medication to treat hypothyroidism; individuals reporting energy intake < 800 or ≥ 4200 kcal/day; and individuals with elevated HDL-cholesterol levels (> 90 mg/dL in men and > 75 mg/dL in women), given that Wilkins et al.,²³ did not observe additional reductions in the risk of coronary heart disease in individuals with HDL-cholesterol levels above these values.

Socio-demographic data (sex and date of birth), health (medication use and personal history of diseases) and lifestyle (smoking, drinking, and physical activity) were investigated through interviews. For the assessment of lifestyle, the following indicators were considered: a) Smoking: current tobacco consumption (yes or no); b) Alcohol use: current consumption of alcoholic beverages (yes or no); and c) Physical activity: practice of physical activity (yes or no).

Weight, height, and WC were measured in duplicate, and the average was used in the analyses. For weight, a platform-type electronic scale (Welmy®, Santa Bárbara d’Oeste, São Paulo, Brazil) was used, with a capacity of 150 kg and sensitivity of 100 g. The participants wore light clothes and removed their shoes. They were instructed to position themselves in the center of the scale and remain erect, with arms extended beside the body, with their feet together.²⁴ For height, a stadiometer (Welmy®, Santa Bárbara d'Oeste, São Paulo, Brazil) was used, with a scale in millimeters, fixed on a support, so that it formed a right angle between the floor and the wall. Participants were instructed to position themselves with their feet and heels together, knees extended, in an erect posture, with arms extended next to their body,²⁴ any they positioned their heads in the Frankfurt plane (the imaginary line from the external auditory canal to the inferior orbital margin).²⁴ The reading was performed at the centimeter closest to the horizontal rod of the vertical bar of the height scale placed against the participant’s head. To calculate BMI, the means of the two measures of weight and height were used. For the classification of nutritional status, the following cutoff points of the World Health Organization were used:²⁵ underweight if BMI < 18.5; healthy weight if BMI between 18.5 and 24.9; overweight if BMI between 25.0 and 29.9; and obese if BMI ≥ 30.0. WC was measured at the midpoint between the lower edge of the last rib and the hip bone (iliac crest)²⁴ in non-obese individuals and at the level of the umbilicus, when it was not possible to identify the natural waist in obese individuals.²⁶

Sample collection and processing were carried out at the CEUNSP Biomedical Laboratory. Blood samples were collected after a 12-hour fast. Values of fasting blood glucose (glucose monoreagent-K082), total cholesterol (monoreagent-K083) and triglycerides (monoreagent-K117) were quantified by the enzymatic colorimetric method and HDL-cholesterol (enzymatic-K015) by the enzymatic trinder method. The kits used were from the Bioclin brand (Quibasa-Bioclin, Belo Horizonte, Minas Gerais, Brazil). LDL-cholesterol values were estimated using the formula by Friedewald et al.,²⁷ given that the samples had triglyceride results < 400 mg/dL.

Food consumption was assessed using a 24-hour recall performed during the first consultation. During the interview, participants reported all the foods and drinks they had consumed the day before, as well as the amounts consumed. Food portion sizes were collected in household measurements and later converted into grams using tables of household measurements of foods and preparations.^{28 , 29} The analysis of nutritional composition (total energy consumption in kilocalories [total energy in kcal], percentage of proteins in total energy, percentage of fat in total energy, and percentage of carbohydrates in total energy) from the 24-hour recall was determined using the Brazilian Table of Food Composition,³⁰ the Tables of Nutritional Composition of Food Consumed in Brazil,³¹ and the Food Composition Table: Support for Nutritional Decision.³²

To assess phytochemical intake, we used the PI proposed by McCarty,¹⁹ calculated as follows: PS = daily energy derived from phytochemical-rich foods (kcal) ÷ total daily energy intake (kcal) × 100. The food groups included in the phytochemical-rich categories were fruits, vegetables (excluding potatoes but including other tubers), whole grains, legumes, seeds, nuts, natural fruit and vegetable juices, and soy products.¹⁹

For data analysis, the statistical program IBM SPSS Statistics for Windows, version 20.0 (IBM Corporation, Armonk, New York) was used, and p-values < 0.05 were considered statistically significant. To verify the distribution of continuous variables and residuals, the Shapiro-Wilk test was used. Continuous variables were presented as mean ± standard deviation or median (interquartile ranges) according to data normality, and categorical variables were expressed as percentages. Spearman’s correlation coefficient was used to investigate the relationship between PI and CVD risk factors. Subsequently, multiple linear regression analyses were performed,³³ including only the variables that presented a statistically significant correlation with the PI in the bivariate analysis (Spearman’s correlation coefficient), with the objective of investigating whether the relationship between the variables remained statistically significant after the inclusion of potentially confounding variables. The following independent variables were included in the regression models: PI, sex (where 0 = female; 1 = male), age (years), physical activity (where 0 = no; 1 = yes), and total energy (kcal) or BMI (kg/m²) or WC (cm). The independent variables of total energy, BMI, and WC were not included in the same regression model because preliminary analyses showed a multicollinearity problem. The results of multiple linear regression were expressed as unstandardized regression coefficient (B) and 95% confidence interval (95% CI). The B values provide information about the relationship between the dependent variable and each independent variable.³⁴

Of the 141 individuals included in the study, the majority were female (90.1%) and did not practice physical activity ( Table 1 ).

Table 1

Variables	n = 141
Age (years)	36 (22)*
Sex
Female	90.1†
Male	9.9†
Physical activity
No	62.4†
Yes	37.6†
Weight (kg)	74.2 (17.9)*
Height (m)	1.60 (0.11)*
BMI (kg/m2)	29.0 (5.8)*
Low weight	0.7†
Healthy weight	19.9†
Overweight	36.2†
Obese	43.3†
WC (cm)	90.1 ± 13.4‡
Total cholesterol (mg/dL)	177 ± 36‡
LDL-cholesterol (mg/dL)	109 ± 34‡
HDL-cholesterol (mg/dL)	47 ± 10‡
Triglycerides (mg/dL)	95 (60)*
Fasting blood glucose (mg/dL)	90 (19)*

Source: Authors. HDL-cholesterol: high-density lipoprotein cholesterol; LDL-cholesterol: low-density lipoprotein cholesterol; WC: waist circumference; BMI: body mass index. *Data presented as median (interquartile ranges). †Data presented as percentage. ‡Data presented as mean ± standard deviation.

Table 2 displays the participants’ food consumption. The main sources of phytochemical-rich foods consumed were fruits, natural fruit juices, vegetables, and legumes, and the least consumed were whole grains, seeds, soy products and nuts.

Table 2

Variables	n = 141
PI	9.80 (11.45)*
Total energy (kcal)	2312 ± 487†
Carbohydrates (% of total energy)	56.5 ± 3.7†
Proteins (% of total energy)	14.7 ± 3.8†
Lipids (% of total energy)	28.8 ± 2.4†
Fruits (g)	106.3 ± 82.4†
Natural fruit juices (ml)	36.1 ± 87.9†
Vegetables (g)	28.4 ± 23.2†
Natural vegetable juices (ml)	0 ± 0†
Legumes (g)	16.5 ± 22.3†
Whole grains (g)	10.0 ± 13.3†
Seeds (g)	5.3 ± 7.3†
Nuts (g)	1.1 ± 2.4†
Soy products (g)	4.8 ± 8.2†

Source: Authors. PI: phytochemical index. *Data presented as median (interquartile ranges). †Data presented as mean ± standard deviation.

Table 3

Variables	PI (n = 141)
	rs*	p-value
BMI	-0.43	< 0.001
WC	-0.36	< 0.001
Total cholesterol	-0.07	0.43
LDL-cholesterol	-0.09	0.30
HDL-cholesterol	0.25	0.003
Triglycerides	-0.15	0.07
Fasting blood glucose	-0.16	0.06

Source: Authors. HDL-cholesterol: high-density lipoprotein cholesterol; LDL-cholesterol: low-density lipoprotein cholesterol; PI: phytochemical index; CVD: Cardiovascular Disease; WC: waist circumference; BMI: body mass index. *Spearman’s correlation coefficient.

Table 4 displays the multiple linear regression analyses, including only the dependent variables that presented a statistically significant correlation with the PI in the bivariate analysis. PI was inversely associated with BMI after adjustment for total energy, sex, age, and physical activity. There was no significant association between PI and WC after adjustment for BMI, sex, age, and physical activity. After adjusting for sex, age, physical activity, and total energy, PI was positively associated with HDL-cholesterol (Model 1). When the independent variable total energy was replaced by BMI (Model 2), PI remained positively associated with HDL-cholesterol. When the independent variable BMI was replaced by WC (Model 3), PI remained positively associated with HDL-cholesterol ( Table 4 ).

Table 4

Dependent variables	Model	Independent variables	B (SE)	95% CI	β	p-value
BMI (kg/m2)	1	Constant	7.49 (2.06)	3.42 – 11.57	NA	< 0.001
		PI	-0.08 (0.03)	-0.15 – -0.01	-0.14	0.02
		Total energy (kcal)	0.008 (0.001)	0.007 – 0.01	0.77	< 0.001
		Sex*	-7.00 (1.05)	-9.07 – -4.93	-0.40	< 0.001
		Age (years)	0.11 (0.02)	0.06 – 0.15	0.24	< 0.001
		Physical activity†	0.41 (0.59)	-0.75 – 1.57	0.04	0.49
WC (cm)	1	Constant	25.10 (4.36)	16.47 – 33.73	NA	< 0.001
		PI	-0.03 (0.07)	-0.17 – 0.12	-0.02	0.73
		BMI (kg/m2)	1.94 (0.13)	1.69 – 2.19	0.76	< 0.001
		Sex*	11.18 (2.05)	7.14 – 15.23	0.25	< 0.001
		Age (years)	0.22 (0.05)	0.12 – 0.33	0.20	< 0.001
		Physical activity†	-1.73 (1.27)	-4.24 – 0.77	-0.06	0.17
HDL-cholesterol (mg/dL)	1	Constant	54.57 (6.08)	42.55 – 66.58	NA	< 0.001
		PI	0.21 (0.10)	0.02 – 0.41	0.19	0.03
		Total energy (kcal)	-0.004 (0.002)	-0.008 – -0.0002	-0.20	0.04
		Sex*	-1.64 (3.09)	-7.74 – 4.47	-0.05	0.60
		Age (years)	-0.03 (0.07)	-0.17 – 0.11	-0.03	0.70
		Physical activity†	2.55 (1.73)	-0.87 – 5.98	0.12	0.14
	2	Constant	53.71 (5.99)	41.87 – 65.55	NA	< 0.001
		PI	0.21 (0.10)	0.008 – 0.40	0.18	0.04
		BMI (kg/m2)	-0.35 (0.18)	-0.692 – 0.002	-0.18	0.05
		Sex*	-4.94 (2.81)	-10.49 – 0.61	-0.14	0.08
		Age (years)	0.01 (0.07)	-0.13 – 0.16	0.02	0.84
		Physical activity†	2.63 (1.74)	-0.80 – 6.07	0.12	0.13
	3	Constant	57.55 (6.60)	44.50 – 70.61	NA	< 0.001
		PI	0.20 (0.10)	0.01 – 0.40	0.18	0.04
		WC (cm)	-0.17 (0.07)	-0.31 – -0.03	-0.22	0.02
		Sex*	-3.01 (2.84)	-8.62 – 2.60	-0.09	0.29
		Age (years)	0.05 (0.07)	-0.10 – 0.20	0.06	0.49
		Physical activity†	2.33 (1.72)	-1.08 – 5.73	0.11	0.18

Source: Authors. B: unstandardized regression coefficients; CI: confidence interval; HDL-cholesterol: high-density lipoprotein cholesterol; NA: not applicable; SE: standard error; β: standardized regression coefficients; PI: ihytochemical index; CVD: Cardiovascular Disease; WC: waist circumference; BMI: body mass index. *Sex: 0 = female; 1 = male. †Physical activity: 0 = no; 1 = yes