The association between dietary pattern and visceral adiposity index, triglyceride-glucose index, inflammation, and body composition among Iranian overweight and obese women

The aim of the present study was to investigate the association between dietary patterns, derived through latent class analysis (LCA), with visceral adiposity index (VAI), Triglyceride-Glucose Index (TyG), inflammation biomarkers, and body composition in overweight and obese Iranian women. For this cross-sectional study, dietary exposure was assessed using a validated 147-item semi-quantitative food frequency questionnaire (FFQ). Dietary patterns were derived through LCA. Binary logistic was performed to test the associations of dietary patterns with VAI, TyG, inflammation biomarkers, and body composition. Health centers in Tehran, Iran. 376 obese and overweight women, aged > 18 years. Two dietary patterns were identified using LCA modeling: healthy and unhealthy. Women in the unhealthy class were characterized by higher consumption of fast food, sweetened beverages, grains, unhealthy oils, butter and margarine, and snacks. Compared with the healthy class, the unhealthy class was associated with an increased risk of higher fasting blood sugar (FBS) (OR = 6.07; 95% CI: 1.33–27.74, P value = 0.02), c-reactive protein (CRP) (OR = 1.72; 95% CI: 1.05–2.80; P value = 0.02), and lower fat free mass index (FFMI) (OR = 0.56; 95% CI: 0.35–0.88, P value = 0.01), after adjusting for confounders. We found that adherence to an unhealthy dietary pattern was associated with decreased FFMI and increased FBS and CRP using LCA, but not with the rest of the variables. Further studies should be conducted to confirm the veracity of these findings.


Anthropometric measurements and body composition.
A calibrated digital scale was used to evaluate the body weight of each subject, to the nearest 100 g, while they were barefoot and wearing light clothing.The height of the participants was measured using a non-elastic tape.To the nearest 0.5 cm, while they were asked to stand next to the wall and unshod.For calculating BMI (in square meters), the weight (in kilograms) was divided by the square of the height.To measure waist circumference elastic measuring tape was placed at the narrowest part of the waist, with a precision of 0.5 cm.Hip circumference evaluation was performed using strapless tape at the most prominent circumference.Each measurement was performed by one person to reduce possible measurement errors.A bioelectrical impedance analyzer (BIA) (Inbody 770 Co., Seoul, Korea) was used to measure all participant's body composition, and operated based on manufacturer's guidelines 12 .The subjects stood on the balance scale foot pad while holding the BIA handles with no shoes and any metal items or extra Dietary intake assessment.A validated 147-item semi-quantitative food frequency questionnaire (FFQ), that had previously been validated for reliability and validity, was used to assess dietary exposure 17 .In the presence of a dietitian, participants recorded their consumption frequency in grams and milliliters based on their usual diet.Utilizing the NUTRITIONIST 4 (First Data Bank, San Bruno, CA) food analyzer, dietary intake was analyzed for energy intake, macronutrients, and micronutrients 22 .
Dietary patterns derived by LCA.LCA would seek groups of subjects (classes) whom follow a common dietary pattern that is distinguishable from other groups.To determine dietary patterns, all reported food items were divided into 25 different food groups.In order to perform the latent class analysis, we first categorized people according to their energy intake according to Dietary Guidelines for Americans 2020-2025 (DGA).Following this, as LCA applies to categorized variables, individuals in each category of energy intake were split into two groups: those who consumed less than daily recommendations and those who consumed more than daily recommendations.The cut-off points for consumption categories for some food groups (legumes, meat, fish, grain, dark green, red, and orange vegetables, starchy vegetables, other vegetables, fruits, nuts, and soy) were established in accordance with the Dietary Guidelines for Americans (DGA) 23 .For other food categories for which there was no cut-point based on DGA, individuals were split into two groups based on median intake.As the amount of daily intake in different food groups can be completely different (e.g., considering spice consumption in grams compared to those of fruits), we first standardized all the variables.Then, we applied different LCA models to the data based on the different classes.Models were compared to each other using four criteria, including AIC, BIC, G 2 , and χ 2 .Based on the results, an LCA with two classes was selected as the final model.So, having the estimated classes identified, we tracked the exact pattern of consumption of a general person in each class.The average daily intake of 25 food groups was computed separately for each class.Eventually, the presence of a lower average daily intake in some food groups and a higher average in others resulted in the inference that the LCA1 class was categorized as "unhealthy", whilst the LCA2 class, with a higher average daily intake compared to one of the other classes, was classified as a "healthy" dietary pattern (Fig. 1).

Definitions.
The VAI and TyG index were calculated using the following 18,24 : Statistical analysis.Dietary patterns were obtained using the poLCA (version 1.6.0.1) package from R software (R-4.2.1).To determine the normality of data distribution, the Kolmogorov-Smirnov test was used; quantitative data were reported as means and standard deviation (SD), and categorical data were reported as numbers with percentages.General linear models [i.e., Analysis of variance (ANOVA) and analysis of covariance (ANCOVA)] were built to compare the body composition, VAI, TyG, and inflammatory profile between subjects.Binary logistic regression was used to determine whether an unhealthy dietary pattern was associated with cardiometabolic risk factors.
In adjusted model 1, age, BMI, and energy intake were controlled.In adjusted model 2, age, BMI, energy intake, education, job, and physical activity were controlled.An odds ratio (OR) with a 95% Confidence Interval (CI) was calculated.Statistical analysis was performed using SPSS v.26 software (SPSS Inc., IL, USA).Statistical significance was accepted at P < 0.05, while P = 0.05 was considered marginally significant in the present study.
Ethical standards disclosure.The present study was carried out in accordance to the ethical standards laid down in the 1964 Declaration of Helsinki.This investigation was also approved by the Ethics Committee of Tehran University of Medical Sciences, Tehran, Iran (with ethics number: IR.TUMS.MEDICINE.REC.1400.1515).All of the study participants signed a written consent form related to this study.Each individual was informed completely regarding the study protocol and provided a written and informed consent form before taking part in the study.literate family members of illiterate participants provided informed consent for the study and this method is approved by the Ethics Committee of Tehran University of Medical Sciences, Tehran, Iran.' Dietary patterns derived by LCA.Latent class models were fitted for 2 to 6 classes, and finally, two dietary pattern classes were chosen.Bayesian Information Criteria, among other model diagnostics, indicated that two classes were optimum(two classes, BIC = 10,964.66;three classes, BIC = 11,019.34;four classes, BIC = 11,094.86;five classes, BIC = 11,191.17;six classes, BIC = 11,287.96).The proportion of each class, after dividing into two categories, was also balanced.Therefore, based on the balance of statistical fit and parsimony, we concluded that the 2-class model is appropriate.Figure 1 shows the conditional probabilities of participants taking each food group in each class.We named the two chosen classes as "healthy dietary pattern" and "unhealthy dietary pattern" Each line show the food consumption of participant from each dietary pattern.Continuous line indicates healthy dietary pattern and dotted line shows unhealthy dietary pattern.

Study population characteristics.
Table 1 presents the mean consumption of food groups in each class.The dietary pattern had a higher percentage of individuals with a higher consumption of all kinds of vegetables, fruit, low-fat dairy, meat, fish, olive oil, vegetable oil, legumes, tea and coffee, nuts, and soy.This pattern was thus labelled a "healthy dietary pattern".On the other hand, another dietary pattern presented the higher frequency of people with a higher consumption of fast food, sweetened beverages, grains, unhealthy oil, butter and margarine and snacks.This pattern was labelled as an "Unhealthy dietary pattern".Overall, by conditional probability, 48.5% and 51.2% of women were characterized by unhealthy dietary pattern and healthy dietary pattern respectively.
The energy intake of women characterized by an unhealthy dietary pattern and a healthy dietary pattern was 2576.25 kcal/day and 2676.67 kcal/day, respectively.Sociodemographic characteristics, visceral adiposity index (VAI), triglyceride-glucose index (TyG), biochemical variables, and body composition according to Dietary Patterns.
The mean and SD of sociodemographic characteristics, visceral adiposity index (VAI), triglyceride-glucose index (TyG), biochemical variables, and body composition of subjects according to their dietary patterns are shown in Table 2.The "unhealthy dietary pattern" presented a lower age mean compared to the "healthy dietary pattern" (p = 0.008).In addition, physical activity was significantly lower among those with an unhealthy dietary pattern (p = 0.003).After adjusting for confounders such as age, BMI, physical activity, education, job, and energy intake, there was no significant difference between the two dietary patterns in other variables.3.
A healthy dietary pattern was used as a reference.Binary logistic analysis showed that the unhealthy dietary pattern class was associated with an increased risk of higher FBS and lower FFMI in the crude model (P < 0.05).After adjustment with confounders in model 1 (adjusting for age, energy intake, and BMI), participants with unhealthy dietary pattern had lower odds of FFMI (OR = 0.99; 95% CI: 0.35-2.81;P value = 0.040), and higher odds of FBS and LDL (OR = 0.99; 95% CI: 0.35-2.81;P value = 0.040).In the fully adjusted model (adjusting for age, energy intake, BMI, education, job, and physical activity), we found that an unhealthy dietary pattern was strongly associated with an increased risk of higher FBS (OR = 6.07; 95% CI: 1.33-27.74,P value = 0.02), lower FFMI (OR = 0.56; 95% CI: 0.35-0.88,P value = 0.01), and significant associated with an increased risk of higher CRP (OR = 1.72 95% CI: 1.05-2.80;P value = 0.02).No significant relationship between unhealthy dietary pattern and other outcomes was observed.

Discussion
In the current study, for the first time, we applied LCA modelling to investigate the association between dietary patterns (DPs) with VAI, TyG index, inflammation, and body composition in an overweight and obese Iranian female population.We found that following an unhealthy dietary pattern, characterized by high intakes of fast food, sweetened beverages, grain, unhealthy oil, butter, margarine, and snacks, was associated with lower FFMI and a higher risk of increased FBS and CRP compared to a healthy dietary pattern, with a high load of vegetables, fruit, low-fat dairy, meat, fish, olive oil, vegetable oil, legumes, tea and coffee, nuts, and soy.Moreover, we did not detect any significant association with the rest of the variables in the fully adjusted model.
Diet acknowledgedly p lays an important role in the development of insulin resistance, in particular glucose serum levels 29,30 , and the potential contributions of diet to body composition, inflammation, VAI, and TyG index have also been reported 11,21,31,32 .A study among medical university students in China that examined the impact of different dietary pattern on body composition showed that the Western pattern influenced FMI/FFMI ratio positively, while the "vegetable and fruit" pattern influenced FMI/FFMI ratio negatively 33 .Similarly, another study in 6-year-old children proposed that high intake of fruit, vegetables, grains, and vegetable oils can result in elevated www.nature.com/scientificreports/FFMI 34 .Findings from a meta-analysis that assessed the effects of saturated fatty acid (SFA), polyunsaturated fatty acid (PUFA), monounsaturated fatty acids (MUFA), and carbohydrates on glucose-insulin homeostasis revealed that only the substitution of energy intake with PUFA was related to lower FBS 35 .This significant association www.nature.com/scientificreports/may be related to the potential health benefit of PUFA in suppressing oxidative stress and insulin resistance 36 .
Another study in participants with the highest category of whole-grain consumption demonstrated lower levels of CRP and FBS, compared with participants in the lowest category of whole-grain consumption 8 .One publication, among adult Americans, indicated positive associations between VAI and glucose/insulin homeostasis markers with a higher dietary proportion of carbohydrate and sugar, total fat, and SFA, and negative associations with a diet consisting of fiber, vitamins, and minerals.An inverse association was also found between a diet rich in PUFA and MUFA with FBS 37,38 .In support of our findings, a five-year prospective study failed to find an association contributed to PUFAs, MUFAs, and SFAs with changes in VAT 39 .There was also no significant relationship between healthy dietary pattern with visceral fat level and TyG index 40 .However, in a study conducted on Iranian adults, where individuals followed a diet rich in MUFA concomitant to decreased dietary intake of total protein or PUFA, a positive association with VAI changes was observed 41 .More so, a cross-sectional study revealed a significant direct association between fat intake and visceral adipose tissue (VAT) in overweight young adults 40 .
Another study among older Americans reported a negative association between the Dietary Approaches to Stop Hypertension (DASH) diet index and VAI 42 .In addition, in a prospective cohort study, an inverse association between anti-inflammatory diet and TyG index was shown 43 .
In comparison to following a healthy dietary pattern, 6-13-year-old children who followed a Western pattern had higher levels of glucose and LDL and lower levels of HDL 44 .An investigation among Brazilian adults exhibited an inverse association between following a healthy dietary pattern with obesity-related markers (BMI, WC, and WHR), FBS, TG/HDL, and LDL/HDL values, and a positive association with HDL.In contrast, adherence to the "Traditional" pattern was positively associated with BMI, WC, and WHR and had a significant inverse association with body fat, LDL, HDL, and TyG 45 .Notwithstanding an inverse association between adherence to a traditional dietary pattern and excess body fat, no relationship was detected between the healthy, snack, and unhealthy dietary patterns with obesity or body adiposity 46 .
The putative mechanism accounting for the obtained results of following a healthy dietary pattern is that the higher fiber content of vegetables, legumes, and nuts, can lead to lower nutrient absorption or energy reduction, and therefore affects total fat mass and visceral fat accumulation.Furthermore, higher fiber consumption displayed tendencies toward improved insulin resistance and reduced visceral fat adiposity.It has been demonstrated that low glycemic index carbohydrates from vegetables, legumes, and nuts can also decrease insulin resistance 41,47,48 .On the other hand, lower systemic inflammation may be related to the anti-inflammatory and anti-oxidative properties of olive oil, nuts, soy foods, fiber, and legumes 40,49 .Consequently, the mechanism behind the effects of unhealthy dietary pattern on the aforementioned markers might be attributed to an imbalance in diet adherence with low levels of certain vital nutrients that are prone to inflammation, increased FBS, and decreased FFMI 43 .For instance, higher sugar beverage consumption with high amounts of rapidly absorbable carbohydrates can lead to enhanced FBS and insulin resistance 50 .Comparisons of our findings with previous reports are difficult to make due to the different approaches applied to specifying dietary patterns.Nevertheless, these results highlight the importance of following a healthy dietary pattern to ameliorate the above-mentioned markers.
LCA is a robust, data-driven, analytic approach that considers the source of heterogeneity from participants of a studied population instead of diet measurement variables.This person-centered novel model may be more effective at identifying dietary patterns when compared to previous models 51,52 .
To the author's knowledge, there are no comparable studies that have been conducted in developing countries that have jointly investigated the association between dietary patterns and VAI, TyG index, inflammation, and body composition using LCA.Given that participants use a combination of foods rather than a single food, evaluating dietary patterns to examine the mentioned associations provides accurate information and should be considered as an additional strength.However, there are several limitations to this study that need to be noted.Due to the cross-sectional nature of the study, we could not draw any causal inferences.In addition, the study relied on self-reported dietary data, and reporting inaccuracies are expected in this regard.The present study, Table 3. Binary logistic regression analysis of the association between dietary patterns and visceral adiposity index (VAI), triglyceride-glucose index (TyG), biochemical variables, and body composition in obese and overweight women (n = 376).Binary logistic regression was used.A healthy dietary pattern considers a reference group.Data are presented as odds ratio (OR) and (95% confidence interval).P-values < 0.05 were considered significant.P-values = 0.05 were considered marginally significant.P value with unadjusted (crude).Adjusted model 1: adjusted for age, energy intake, BMI.Adjusted model 2: adjusted for age, energy intake, BMI, physical activity, education, job.UDP unhealthy dietary pattern, HDP healthy dietary pattern, VAI visceral adiposity index, TyG triglyceride-glucose index, BMI body mass index, WC waist circumference, WHR waist to hip ratio, BFM body fat mass, FFMI fat-free mass index, FFM fat-free mass, FMI fat mass index, HDL_C high-density lipoprotein cholesterol, LDL_C low-density lipoprotein cholesterol, TG triglyceride, hs CRP high-sensitivity C-reactive protein.Significant values are in bold.

Figure 1 .
Figure 1.Item probabilities in the two latent classes (straight line: healthy dietary pattern, dotted line: unhealthy dietary pattern).

Table 1 .
Average intake of food groups for healthy and unhealthy dietary pattern (UDP) in obese and overweight women (n = 376).Values are represented as means (SD).Average consumption is reported in grams.UDP unhealthy dietary pattern, HDP healthy dietary pattern, SD standard deviation.

Table 2 .
Mean and SD of sociodemographic characteristics, visceral adiposity index (VAI), triglycerideglucose index (TyG), biochemical variables, and body composition in obese and overweight women (n = 376).Values are represented as means (SD).Categorical variables: N and %.ANCOVA (P value*) was performed to adjust potential confounding factors (age, BMI, energy intake, physical activity, education, job).p-values < 0.05 were considered significant.UDP unhealthy dietary pattern, HDP healthy dietary pattern, IPAQ International Physical Activity Questionnaire, VAI visceral adiposity index, TyG triglyceride-glucose index, BMI body mass index, WC waist circumference, WHR waist to hip ratio, BFM body fat mass, FFMI fat-free mass index, FFM fat-free mass, FMI fat mass index, HDL_C high-density lipoprotein cholesterol, LDL_C low-density lipoprotein cholesterol, TG triglyceride, hs CRP high-sensitivity C-reactive protein.a BMI is considered a collinear variable for anthropometrics and body composition variables.Significant values are in bold.