Gender differences in dietary patterns and physical activity: an insight with principal component analysis (PCA)

Abstract

Background

Gender differences in dietary patterns and physical activity are known to influence metabolic health, but research exploring these differences using principal component analysis (PCA) is limited. This study aims to identify distinct patterns of eating behaviour, body composition and physical activity between men and women, in order to develop tailored interventions.

Methods

A cross-sectional study was conducted on 2,509 adults attending a metabolic health centre. Data on eating habits, physical activity and body composition were collected by means of questionnaires and bioimpedance analysis. PCA was used to identify patterns of eating behaviour and physical activity. Statistical analyses were performed to explore gender-specific differences.

Results

Based on the PCA, five distinct groups of participants were identified: Balanced Eaters, Focused on Home Cooking, Routine Eaters, Restaurant Lovers and Varied Eaters. Significant gender differences in food preferences were observed, with men consuming more meat and women more vegetables. Men also reported greater participation in strength and endurance sports, while women showed a more structured eating routine.

Conclusions

This study, using principal component analysis (PCA), revealed gender-specific patterns in diet, physical activity and body composition. PCA identified five distinct behavioural groups, revealing that men tended to consume more meat and engage in strength training, while women adhered to structured, vegetable-rich diets. The application of PCA provided more insight than traditional analysis, highlighting the complexity of gender-specific behaviour. These results emphasize the need for tailored interventions, focusing on increasing vegetable intake in men and encouraging strength training in women. Future research should exploit PCA to explore behavioural patterns longitudinally for more refined and personalised health strategies.

Clinical trials registered

This study is registered on ClinicalTrials.gov (NCT06654674).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-024-05965-3.

Introduction

The prevalence of metabolic disorders, such as obesity, type 2 diabetes, and cardiovascular diseases, has increased dramatically in recent decades, becoming a major public health concern worldwide [1]. These conditions are often linked to unhealthy eating patterns, sedentary lifestyles, and poor sleep quality [2, 3]. Understanding the intricate relationships between diet, body composition and physical activity is crucial for developing effective prevention and intervention strategies.

Dysfunctional eating behaviours, including irregular meal patterns, emotional eating, and late-night snacking, have been associated with negative health outcomes, such as increased body fat, insulin resistance and metabolic syndrome [4, 5]. These behaviours are not only influenced by individual factors, but also by broader socio-cultural contexts, including gender, social role and expectations [6, 7]. Traditionally, most research studies on eating behaviours have focused on describing and trying to explain pathological eating patterns, often overlooking adaptive eating patterns. However, more recently, researchers have emphasised the importance of paying greater attention to the latter, arguing that these patterns represent not simply the absence of pathological eating, but play a significant role in promoting well-being and health [8]. It has been established that men and women exhibit different eating behaviours and respond differently to dietary interventions, highlighting the need for gender-specific analyses in nutritional studies [9].

Physical activity, particularly regular engagement in sport, plays a significant role in modulating body composition and metabolic health. Regular exercise is associated with numerous health benefits, including improved cardiovascular fitness, better weight management and increased insulin sensitivity [10]. However, the interaction between physical activity, eating habits and sleep quality remains complex and is often influenced by gender and lifestyle factors [11]. Previous studies have shown that the impact of physical activity on health outcomes can vary significantly between men and women, depending on the type, frequency and intensity of exercise [12, 13].

In recent years, principal component analysis (PCA) has emerged as a valuable tool in nutrition and metabolic research, enabling the identification of patterns in complex data sets, particularly in dietary behaviour. PCA reduces the dimensionality of large data sets, uncovering underlying structures that may not be immediately apparent through traditional statistical methods [14]. By classifying individuals according to common characteristics, such as eating habits, socioeconomic factors and anthropometric data, PCA can help to identify distinct groups with unique health risks and behavioural patterns [15]. Such an approach is particularly useful for exploring gender-specific differences, as previous studies have shown that men and women often exhibit different responses to dietary and lifestyle factors [16]. In this study, we applied PCA to examine how dietary behaviours, food preferences and physical activity interact with body composition in men and women, with the aim of uncovering gender-specific trends that can help to define more tailored dietary and lifestyle interventions [17, 18].

The main objective of this study was to investigate the associations between dietary behaviour, physical activity and body composition in a large sample of adults attending a nutritional and metabolic health centre. Thanks to the exclusive application of PCA, this study explores gender differences in dietary patterns and physical activity in an Italian context, offering insights not extensively addressed in the previous literature.

Methods

Participant recruitment

This cross-sectional study was conducted among adult patients attending a medical centre focused on nutrition and metabolic health in Rome, Italy. Participants were recruited between May 2023 and June 2024, enrolling 3,302 individuals. Participants were Caucasian adults attending a nutrition and metabolic health center for either preventive care or management of metabolic conditions such as overweight, obesity, or type 2 diabetes. A preliminary power analysis was conducted to ensure the robustness of the study and the reliability of the results. A desired power of 0.80 was set with an alpha level of 0.05, and the mean expected effect sizes in the associations between dysfunctional eating behaviour, nutritional status, and sleep patterns, a minimum sample size of 3,000 participants was determined. After applying exclusion criteria — participants under 18 years of age, without consent, showing incomplete or inconsistent data, with diagnosed psychiatric disorders, pregnancy, with alcohol dependence, with incomplete body composition measurements — a final sample of 2,509 participants was included in the analysis. The study received ethical approval from the IRCCS San Raffaele Ethics Committee (protocol number RP 23/13), adhering to the Declaration of Helsinki. In this study, the protocols and procedures were registered at ClinicalTrials.gov under registration number NCT06654674.

Body composition assessment

Patients were assessed in the morning, after an overnight fasting, with empty bladder and adhering to the following instructions: no alcohol consumption or intense physical activity within the previous 12 h, no significant calorie restriction within the last 48 h, and no use of diuretic medications during the previous week. Testing was performed at least 3 h after waking up. Initially, weight and height were measured, with patients wearing only underwear. The body mass index (BMI) was calculated as weight (kg) to the square of height (m²) ratio. Waist circumference (WC) was measured at the midpoint between the iliac crest and the last rib using a common tape measurement. Finally, body composition was analysed using a Tanita BC-420 MA bioimpedance analysis (BIA) device, a validated tool that, with the use of no electrodes, provides accurate measurements of fat mass (FM), lean mass (FFM) and hydration status (TBW), with an accuracy of 100 g.

Diet and lifestyle assessment

Prior to the medical visit, patients completed an online questionnaire on their eating habits, followed by a discussion of the results with the medical staff. The questionnaire was adapted from previously validated food preference surveys, including the Food Liking Questionnaire validated in a French-Canadian population [19], and other tools tailored for specific populations, such as the food frequency questionnaire validated for adults living in Sicily [20]. A pilot test was conducted with a small group to ensure clarity and relevance, which led to minor adjustments in the wording. The questionnaire was self-administered, accessible via a link and could be completed in approximately 30 min on any Internet-enabled device. Although no formal validation was conducted, the survey was in line with established food preference questionnaires [19]. All responses were recorded anonymously and organised into four sections, in line with previous research. The questions covered topics such as meal frequency (1–6 meals per day), meal skipping (rarely to frequently), eating speed (slowly to quickly), cooking habits (from frequent home-cooking to preferring eating out), snacking frequency (rarely to very frequently), and routine stability (from well-defined to highly flexible). Responses were categorised based on frequency. The second part explored food preferences across various categories, including dairy products, meat and meat alternatives, vegetables, fruits, cereals, and dark chocolate. It also examined consumption habits for meals, water, alcohol, and sugary drinks, along with a 24-hour recall of food intake. Sleep patterns were assessed using two questions included in the questionnaire. The first question, “How do you sleep at night?” aimed to evaluate self-reported sleep quality, with response options ranging from “Good” to “I wake up several times during the night.” The second question, “Do you wake up to eat at night?” assessed nighttime eating behavior with response options including “Never,” “Rarely (once a month),” and “Often (> 1/week).” These variables were analyzed across gender and PCA groups to explore their potential associations with dietary patterns and physical activity behaviors.

Physical activity assessment

Physical activity was assessed by evaluating structured sports activities, such as endurance sports, strength training and team sports. These categories were chosen for their measurable characteristics, including intensity, frequency and regularity, which allow for a reliable comparison between participants. Physical activity data were divided into categories such as endurance sports, strength training and team sports. The classification of sports activities according to the main categories is shown in Supplementary Table 1 S. Data were collected by means of a self-report questionnaire, in which participants indicated their participation in these sports and the average weekly duration. Although this approach provides valuable information on structured physical activity, we are aware that it does not capture unstructured physical activities, such as walking, housework or commuting. These types of activities, although often less intense and less regular, also contribute significantly to overall physical activity levels and health. Future studies should aim to include a broader range of physical activity types to provide a more complete understanding of participants’ physical activity profiles.

Principal component analysis (PCA)

To identify patterns of eating behaviour, we applied PCA, a statistical method designed to reduce data complexity and identify underlying patterns in large datasets. PCA has been widely used in recent studies to analyse the influence of eating behaviour on body composition and physical activity, particularly to highlight gender differences in these contexts [21, 22]. PCA was chosen for this study as it identifies uncorrelated components that represent the maximum variance in the data, making it ideal for summarising dietary and physical activity behaviour. Unlike factor analysis (FA), PCA does not assume underlying latent constructs, aligning with our goal of identifying observable patterns rather than exploring theoretical variables. Factor analysis, while useful for hypothesis-driven research, assumes an underlying model structure and focuses on latent variables, which was not the goal of this study. In our study, PCA was conducted using dietary, socioeconomic and anthropometric variables to classify participants into distinct groups reflecting different dietary behaviours and potential health impacts. The methodology included standardisation of variables, calculation of principal components with eigenvalues greater than 1, and application of varimax rotation to improve interpretability. The number of selected components was determined by examining the scree plot and the cumulative variance explained. The results of the PCA analysis, including the main characteristics of the identified groups and their associations with dietary patterns, body composition and physical activity, are reported in the Results section and detailed in Supplementary Table 2 S. This categorisation provides valuable input for the development of customised and gender-specific nutritional interventions.

Statistical analysis

Statistical analysis was conducted to assess the differences between the five groups identified by PCA according to key variables such as gender distribution, age, smoking habits and body composition parameters (weight, BMI, fat mass percentage, etc.). Appropriate statistical tests were used to compare continuous and categorical variables. For continuous variables such as weight, BMI and fat mass percentage, the one-way ANOVA was used to compare the averages between the groups. If assumptions of normality or homogeneity of variances were not fulfilled, the Kruskal-Wallis test was applied as a non-parametric alternative. For categorical variables, such as gender distribution, smoking habits and eating frequency, the chi-square test was used to assess associations between groups. Where necessary, Bonferroni corrections were applied to adjust for multiple comparisons and reduce the risk of type I errors. To control for multiple comparisons, the Bonferroni correction was applied to adjust p-values, ensuring a stricter significance threshold for identifying differences across groups. No additional correction methods, such as the Holm or Benjamini-Hochberg procedures, were utilized, as the Bonferroni approach provided the necessary control for Type I errors in this study’s context. The analysis was conducted using SPSS v. 28 software (IBM Corp., Armonk, NY, USA). Descriptive statistics, including means, standard deviations and frequencies, were calculated to describe sample characteristics and delineate dysfunctional eating behaviour. Stratification by gender was used to explore gender-specific differences in eating behaviour and body composition.

Results

Demographic and anthropometric characteristics

The table below (Table 1), shows the results of the gender comparison across several anthropometric and socioeconomic variables. Notably, significant gender differences were observed in almost all variables analyzed. Males had higher average weight, BMI, and lean body mass, while females showed higher percentages of fat mass and body fat percentage. In terms of socioeconomic status, a higher percentage of females reported an annual income less than €20,000 compared to males. Age distribution also varied significantly between genders, with females being slightly older on average than males.

Table 1.

Statistical analysis of gender differences in anthropometric and socioeconomic variables

			Total (n)	M	F	p-value
			2508	1021	1487	< 0.001
	Smokers	yes	599 (23.9%)	238 (23.3%)	361 (24.3%)	0.577
Age	Age	yrs	40.65 ± 13.23	39.37 ± 13.03	41.53 ± 13.3	< 0.001
	18–24	n (%)	281 (11.2%)	135 (13.22%)	146 (9.82%)	0.0002
	25–44		1304 (51.99%)	561 (54.95%)	743 (49.97%)
	45–59		677 (26.99%)	246 (24.09%)	431 (28.98%)
	60–74		222 (8.85%)	71 (6.95%)	151 (10.15%)
	75–90		24 (0.96%)	8 (0.78%)	16 (1.08%)
	Weight	kg	78.67 ± 18.01	88.4 ± 18.04	71.99 ± 14.63	< 0.001
BMI	BMI	Kg/m2	27.69 ± 5.44	28.46 ± 5.48	27.17 ± 5.35	< 0.001
	< 18.5	n (%)	34 (1.4)	10 (1.0)	24 (1.6)	< 0.001
	18.5–24.9		803 (32.0)	268 (26.2)	535 (36.0)
	25.0–29.9		936 (37.4)	390 (38.3)	546 (36.8)
	30.0–34.9		481 (19.2)	237 (23.2)	244 (16.4)
	35.0–39.9		188 (7.5)	87 (8.5)	101 (6.8)
	40.0–44.9		43 (1.7)	16 (1.6)	27 (1.8)
	≥ 45		19 (0.8)	11 (1.1)	8 (0.5)
	FM	kg	24.32 ± 11.19	22.57 ± 11.59	25.51 ± 10.75	< 0.001
	FM	%	30.17 ± 9.76	24.41 ± 9.15	34.11 ± 8.06	< 0.001
	AC	cm	95.94 ± 14.22	100.25 ± 14.52	92.94 ± 13.2	< 0.001
	Lean Body Mass	kg	51.63 ± 11.23	62.58 ± 8.36	44.16 ± 5.18	< 0.001
	Body Water	kg	38.33 ± 8.42	46.36 ± 6.25	32.87 ± 4.35	< 0.001
	BMR	Kcal	1636.8 ± 341.4	1948 ± 277.8	1422.5 ± 176.4	< 0.001
INCOME PER YEAR	<€20.000	N (%)	404 (16.11%)	138 (13.52%)	266 (17.89%)	0.0175
	€20.000 - €40.000		1694 (67.54%)	704 (68.95%)	990 (66.58%)
	€40.000- €60.000		332 (13.24%)	149 (14.59%)	183 (12.31%)
	> €60.0 00		71 (2.83%)	28 (2.74%)	43 (2.89%)

Descriptive statistics and p-values for anthropometric and socioeconomic variables by gender (M = males, F = females). Includes total participants (n), number/percentage of smokers, mean age ± standard deviation (SD) with age group distribution, weight (kg), Body Mass Index (BMI, kg/m²) with BMI categories, fat mass (FM, kg and %), abdominal circumference (AC, cm), lean body mass (kg), body water (kg), and Basal Metabolic Rate (BMR, kcal). Annual income distribution is shown, with 7 participants (0.28%) opting not to declare (M: 2, F: 5). p-values < 0.05 indicate significant gender differences

Significant differences in food preferences between men and women are displayed in Supplementary Fig. 1S. In the comparative analysis between men and women regarding protein sources consumption, significant differences emerged among several categories (Fig. 1). The consumption of meat, fish, dairy products and legumes was significantly different between genders (p < 0.05). Men showed a higher frequency of meat and fish consumption, whereas women reported more frequent consumption of dairy products and legumes. The differences in consumption of processed meat, eggs and soy did not reach statistical significance (p > 0.05).

Fig. 1.

Gender differences in protein sources consumption. Comparison of protein source consumption frequencies between men and women. Protein sources include meat, processed meat, fish, eggs, dairy products, legumes, and soy. The values shown represent the average consumption frequencies for each protein source by gender. The chi-square test was used to assess statistically significant differences. p-values are displayed above the corresponding bars: meat (p = 0.02), processed meat (p = 0.08), fish (p = 0.01), eggs (p = 0.15), dairy products (p = 0.05), legumes (p = 0.01), soy (p = 0.11), and total number of protein sources (p = 0.03)

Our analysis reveals significant gender differences in the intake of several foods (Fig. 2). Males were significantly more likely to consume meat (p = 6.83e-08), red meat (p = 1.15e-05) and processed meat (p = 1.49e-06), whereas females preferred cooked vegetables (p = 5.26e-12), raw vegetables (p = 0.0007) and whole grain foods (p = 1.81e-08). Significant differences were also highlighted in the consumption of vegetable drinks (p = 0.0238) and eggs (p = 0.0053). Other foods, such as cow’s milk, fish and fruit, showed no statistically significant differences between the sexes (p > 0.05).

Fig. 2.

Gender differences in eating habits: A comparative analysis on food intake patterns. Radar chart showing differences in food intake between males and females in several food categories. Foods highlighted in red indicate food sources with higher average consumption among women, while items highlighted in blue indicate food sources with higher average consumption among men. The p-values were calculated using an independent t-test to compare the proportion of ‘Yes’ responses between genders for each food. Significant differences (p < 0.05) were observed for several foods, including Vegetable drinks (p = 0.0238), Meat (p = 6.83e-08), Red meat (p = 1.15e-05), Processed meat (p = 1.49e-06), Eggs (p = 0.0053), Cooked vegetables (p = 5.26e-12), Raw vegetables (p = 0.0007) and Whole grains (p = 1.81e-08)

Differences in Food Patterns and PCA Groups

The application of Principal Component Analysis (PCA) allowed for the identification of five distinct dietary behaviour groups: Balanced Eaters, Home-Cooked Focused, Routine Eaters, Restaurant Lovers, and Varied Eaters. Each group represents unique patterns of food preferences, dietary routines, and potential implications for health outcomes. Table 2 summarizes the key characteristics, potential health impacts, and typical foods or meals associated with each group.

Table 2.

Characteristics, health implications and dietary patterns of PCA groups

Group	Characteristics	Potential Health Impact	Typical Foods/Meals
Balanced Eaters	Stable routine, prefer meals at home, eat slowly, cook over the weekend.	Likely positive due to a balanced diet.	Home-cooked balanced meals (e.g., grilled chicken with vegetables).
Home-Cooked Focused	Mainly consuming home-cooked meals, especially over weekends, consuming convivial meals .	Positive if meals are nutritionally balanced.	Traditional home-cooked meals (e.g., pasta, casseroles).
Routine Eaters	Follow a regular routine with few variations, prefer meals at home.	Stable, possibly positive due to consistency.	Simple, consistent meals (e.g., sandwiches, salads).
Restaurant Lovers	Prefer to eat out, especially over the weekends.	Potentially negative due to eating out often with an unbalanced diet.	Restaurant meals (e.g., pizza, burgers).
Varied Eaters	Great variability in eating habits, frequent snacking and changes.	Mixed impact; risks due to inconsistency.	Snacks, varied meal types (e.g., fast food, home-cooked, takeout).

The table summarises the main characteristics of participants in five distinct groups identified through principal component analysis (PCA): Balanced Eaters, Focused on Home Cooking, Routine Eaters, Restaurant Lovers and Varied Eaters. Each group is characterised by unique eating behaviours and physical activity, potential health impacts based on their eating patterns and typical foods or meals they may prefer. The table provides a concise overview of the differences between these groups, offering insights into their potential health outcomes and food preferences. The classification of sports activities referenced in this table is detailed in Supplementary Table 1 S

Table 3 shows the distribution of several anthropometric and socioeconomic variables in all five PCA groups. The gender distribution displays a consistent pattern between the groups, with women outnumbering men in each group, but showing no significant differences (p = 0.5539). Age is the only variable that showed a statistically significant difference between the groups (p = 0.0425), and group 2 (‘Focused on home cooking’) showed a slightly higher average age than the other groups. Other variables, such as smoking habits, weight, BMI and body composition metrics, showed no significant differences between the groups.

Table 3.

Comparative analysis of anthropometric and socioeconomic variables between PCA groups

PCA	Balanced Eaters PCA 1	Home-Cooked Focused PCA 2	Routine Eaters PCA 3	Restaurant Lovers PCA 4	Varied Eaters PCA 5	p-value
Patients	558	367	450	495	636
Gender	M: 220 (39.4%), F: 338 (60.6%)	M: 158 (43%), F: 209 (57%)	M: 193 (42.9%), F: 257 (57.1%)	M: 191 (38.6%), F: 304 (61.4%)	M: 259 (40.7%), F: 377 (59.3%)	0.5539
Smoker	23.8%	22.3%	24.3%	24.8%	24.57%	0.9204
Age	40.2 ± 13.2	42.4 ± 13	39.9 ± 13.1	40.3 ± 13.1	40.8 ± 13.5	0.0425
Frequency of Bowel Movements (per week)	6.1 ± 1.6	6.2 ± 1.5	6.2 ± 1.6	6.1 ± 1.5	6.06 ± 1.63	0.9357
Occupation	Clerk (42.9%), Student (9%), Housewife (3.6%)	Clerk (42%), Student (8.7%), Retired (5.5%)	Clerk (39.1%), Student (11.2%), Housewife (3.8%)	Clerk (42.7%), Student (8.3%), Retired (3.6%)	Clerk (43.1%), Student (11.8%), Retired (4.4%)	0.5412
Category	Sales and Services (45.4%), Student (8.8%), Professional Services (7.5%)	Sales and Services (44.4%), Student (8.7%), Healthcare and Wellness (7.6%)	Sales and Services (41.1%), Student (11.2%), Professional Services (6.5%)	Sales and Services (47.2%), Student (8.3%), Professional Services (7.5%)	Sales and Services (45.9%), Student (11.8%), Professional Services (5.8%)	0.7082
Weight (kg)	78.9 ± 18.7	79.5 ± 19.2	79.6 ± 18.5	77.7 ± 17.3	78.2 ± 16.9	0.6462
BMI	27.8 ± 5.4	28 ± 5.9	27.7 ± 5.6	27.6 ± 5.4	27.5 ± 5.1	0.8345
Fat Mass (%)	24.5 ± 10.9	24.9 ± 12.3	24.5 ± 11.8	24.1 ± 11.2	23.9 ± 10.3	0.8891
AC (cm)	96.3 ± 15	96.9 ± 14.7	95.8 ± 13.9	95.7 ± 14.0	95.3 ± 13.6	0.6990
FFM (kg)	51.7 ± 11.8	51.8 ± 11.5	52.3 ± 11.5	50.9 ± 10.5	51.6 ± 11	0.5662
TBW (L)	38.3 ± 8.9	38.5 ± 8.6	38.8 ± 8.7	37.8 ± 7.9	38.3 ± 8.1	0.5781
BMR (kcal/day)	1640.9 ± 360.4	1640.5 ± 353.9	1657.8 ± 350.4	1616.1 ± 318.4	1632.4 ± 327.6	0.6682

This table represents a comparative analysis of anthropometric, socioeconomic, and behavioural variables among the five groups identified through principal component analysis (PCA). The table includes gender distribution, smoking status, age, bowel movement frequency (Bowel Movements/wk), occupation, income category, and various body composition metrics such as weight, body mass index (BMI), fat mass, abdominal circumference (AC), fat-free mass (FFM), total body water (TBW), and basal metabolic rate (BMR). P-values are provided to assess the statistical significance of differences between groups

The heatmap in Fig. 3 shows significant differences between the sexes in food consumption in all five PCA groups for various food categories. In PCA group 1, men consumed more red meat, processed meat and cooked vegetables, while women consumed more raw vegetables. Significant differences were also observed for fish. In the PCA 2 group, women consumed more cooked and raw vegetables. In the PCA 3 group, men consumed more meat, red meat and processed meat, while women consumed more cooked and raw vegetables. In the PCA 4 group, significant differences were found for meat, processed meat and cooked and raw vegetables, with men consuming more meat and women more vegetables. In the PCA 5 group, men preferred meat and red meat, while women preferred vegetables and fish.

Fig. 3.

Gender differences in food consumption in the five PCA groups for the food categories in the test. Heatmap representing the correlations between PCA components and variables, including dietary, socioeconomic, and anthropometric factors. Warmer colors indicate stronger positive correlations, while cooler colors indicate negative correlations. The figure illustrates key associations, such as strong positive correlations between structured eating patterns and physical activity levels. Annotated values represent PCA loadings, with corresponding significance levels (p-values) for each food category: PCA 1: Red Meat (p = 0.0040), Processed Meat (p = 0.0105), Cooked Vegetables (p = 0.0010), Raw Vegetables (p = 0.0411), Fish (p = 0.0041). PCA 2: Cooked Vegetables (p = 0.0051). PCA 3: Meat (p = 0.0119), Red Meat (p = 0.0038), Processed Meat (p = 0.0016), Cooked Vegetables (p = 0.0283), Raw Vegetables (p = 0.0246). PCA 4: Meat (p < 0.0001), Processed Meat (p = 0.0110), Cooked Vegetables (p = 0.0157), Raw Vegetables (p = 0.0018), Fish (p = 0.0012). PCA 5: Meat (p = 0.0008), Red Meat (p = 0.0129), Cooked Vegetables (p = 0.0005), Raw Vegetables (p = 0.0473), Fish (p = 0.0002)

The analysis of gender differences in the eating behaviours of the five PCA groups revealed several statistically significant differences (Fig. 4). More precisely, gender differences were found in the frequency of daily meals, with significant differences observed in PCA group 1 (p = 0.0142). Similarly, the ‘skipping meals’ habit showed significant differences between genders in PCA groups 1 (p = 0.0447) and 5 (p = 0.0152). Further differences were observed in eating pattern-related behaviours, such as ‘eating quickly’, where significant differences were found in all PCA groups, particularly in PCA group 3 (p < 0.001). The behaviour of ‘eating out at meals’ also showed gender disparities, with significant differences in PCA groups 1 (p = 0.0008), 4 (p = 0.0043) and 5 (p = 0.0001). Other eating behaviours such as ‘snacking between meals’ and ‘waking up to eat at night’ showed statistically significant gender differences in PCA group 3 (p = 0.0078) and PCA group 1 (p = 0.0111), respectively.

Fig. 4.

Gender differences in eating behaviours within PCA groups. This heatmap represents the p-values of gender differences in various eating behaviours within the five groups identified by the PCA analysis. Coloured cells indicate where the differences between males and females are statistically significant (p < 0.05), with the PCA groups indicated along the x-axis (1 to 5) and the behavioural variables on the y-axis

The specific food preferences between men and women in the five PCA groups are given in Supplementary Table 3 S.

Analysis of sleep patterns revealed notable differences between genders and across PCA groups (Tables 4S and 5S). Women more frequently reported good sleep quality compared to men (51.7% vs. 48.3%). However, women also indicated a higher prevalence of waking up multiple times during the night (40.8% vs. 34.6%). Regarding nighttime eating behavior, women were more likely to report “Never” waking up to eat compared to men, but similar proportions of men and women indicated waking up often (> 1/week). Among PCA groups, “Varied Eaters” (PCA Group 5) displayed the poorest sleep quality, with the highest proportion of individuals reporting waking up multiple times during the night. This group also showed a greater frequency of waking up to eat at night compared to other groups. Conversely, “Balanced Eaters” (PCA Group 1) exhibited the best overall sleep quality, with the majority reporting “Good” sleep. These findings suggest that disrupted sleep patterns may correlate with specific eating behaviors and PCA-defined dietary groups, highlighting the need for further investigation into the interplay between sleep, eating habits, and physical activity.

Table 4.

Total and gender sports participation in the five PCA groups

PCA, Gender	Total	Yes %	No %	Gender p-value	Endurance Sports %	Skill Sports %	Strength Training %	Team Sports %
(1, ‘F’)	338	44.7	55.3	0.0163	40.9	4.0	51.0	4.0
(1, ‘M’)	220	55.5	44.5		38.0	14.0	42.1	5.8
(2, ‘F’)	208	50.5	49.5	0.0641	43.8	8.6	41.9	5.7
(2, ‘M’)	158	60.8	39.2		31.6	8.4	51.6	8.4
(3, ‘F’)	257	50.6	49.4	0.0715	44.1	10.2	38.6	7.1
(3, ‘M’)	193	59.6	40.4		29.2	18.6	45.1	7.1
(4, ‘F’)	304	46.7	53.3	0.001	38.6	7.9	47.1	6.4
(4, ‘M’)	191	62.3	37.7		37.6	9.4	41.9	11.1
(5, ‘F’)	377	51.7	48.3	0.0198	43.5	9.8	42.0	4.7
(5, ‘M’)	259	61.4	38.6		30.1	10.9	48.7	10.3

The table shows sports participation in the five PCA groups, broken down by gender. Absolute numbers and percentages for each group are presented, along with p-values for gender comparisons. The p-value for the comparison of sports participation between males and females across the five PCA groups is p < 0.001. The p-value for the comparison of sports participation among men across PCA groups is 0.633, and among women is 0.321. The p-value for overall gender differences across PCA groups is 0.954. The p-value for comparing sport participation across all PCA groups is 0.639. For gender differences within each PCA group, PCA 1 has a p-value of 0.023, PCA 2 has a p-value of 0.322, PCA 3 has a p-value of 0.068, PCA 4 has a p-value of 0.528, and PCA 5 has a p-value of 0.031. For differences across PCA groups in sport participation by gender, males have a p-value of 0.352, and females have a p-value of 0.616.For each sport category across PCA groups, the p-value for Strength Training, Team Sports, Endurance Sports, and Skill Sports in both males and females is p < 0.001

Table 5.

Principal component analysis (PCA) Groups: key findings and recommended actions

	Key Findings	Recommended Actions
PCA groups	The groups were defined as “Balanced Eaters,” “Home-Cooked Focused,” “Routine Eaters,” “Restaurant Lovers,” and “Varied Eaters,” each with unique dietary and physical activity habits.	Tailor nutritional and physical activity interventions based on the specific characteristics of these groups.
Gender differences in PCA groups	Women displayed healthier dietary habits in the “Balanced Eaters” and “Home-Cooked Focused” groups, while men consumed more meat and were more engaged in strength and endurance sports.	Develop dietary plans encouraging men to incorporate more vegetables and fiber and promote higher sports participation among women.
“Restaurant Lovers” and “Varied Eaters” at higher risk	These groups showed more flexible eating habits and a tendency to eat out more, often associated with less healthy choices.	Educate individuals on making more mindful food choices when eating out and encourage both sexes to establish more regular eating routines.
Importance of meal regularity in groups	In the “Balanced Eaters” and “Routine Eaters” groups, regularity and structured meals were associated with better metabolic profiles.	Promote the importance of following stable meal routines, particularly among men who tend to introduce more variability in meal timing and composition.
Targeted interventions on sports and physical activity	Sports participation varied significantly across the PCA groups, with men more active in strength and endurance sports.	Implement targeted sports programs to increase women’s participation, especially in groups with lower engagement.

The table summaries the main findings of the principal component analysis (PCA) on dietary patterns, physical activity and body composition. The table highlights the five identified groups and provides recommended actions for targeted interventions to address gender-specific differences and improve health outcomes

Sports participation in PCA groups

The analysis of sports participation in the five PCA groups revealed significant gender differences (Table 4). Men reported higher participation rates (55.5-62.3%) than women (44.7-51.7%), with the largest gaps in the PCA 4 and PCA 5 groups. No group had a higher female participation. Overall, there were no significant differences in sports participation between genders in the PCA groups (p = 0.954), but significant differences between genders were found in the PCA 1 (p = 0.023) and PCA 5 (p = 0.031) groups, with borderline significance in the PCA 3 group (p = 0.068). The sport categories showed significant gender differences in all PCA groups for strength training, team sports, endurance sports and skill sports in both men and women.

Discussion

Gender dynamics influence dietary behaviour, body composition and participation in physical activity, thus representing a crucial aspect for understanding the determinants of health and for designing tailored interventions [23]. It has been established that women tend to prefer diets rich in fibre and vegetables, whereas men are more likely to consume animal-derived proteins and red meat [24, 25]. These behavioural differences are well documented in the literature, with women showing to pay greater attention to health and dietary balance than men. This divergence may be influenced by socio-cultural, biological and psychological factors, which vary depending on the context and specific populations studied [26]. In accordance with these observations several studies recognized greater adherence to dietary recommendations in women compared to men, the latter showing more flexibility and less interest in health as a determinant in daily food choices [27, 28]. Interestingly, Wardle et al. found that women are more likely to avoid high-fat foods and consume greater amounts of fruit and vegetables compared to men, regardless of their country of origin, thus suggesting that the causal mechanisms behind gender differences in food choices go far beyond socio-cultural factors and geographical location [29]. On the other hand, social differences, such as class and type of work, influence men and women differently in the management of eating habits. Indeed, employment status affects only women’s food behavior, while marital status impacts both men and women, with married individuals adhering more closely to dietary guidelines. Moreover, as mentioned above, only women, particularly those with young children, show greater alignment with dietary guidelines [30].

By categorizing dietary and activity patterns through PCA, this study highlights actionable gender-specific health interventions, enhancing the current understanding of personalized nutritional strategies. In this study, we aimed to investigate the associations between eating behaviour, physical activity and body composition in a large sample of adults, attending a nutritional and metabolic health centre. To this purpose we took advantage of PCA, a technique that reduces the complexity of the data, making it possible to identify groups of individuals with similar behavioural characteristics [23]. This approach has been found particularly useful for exploring gender differences, as it allows individuals to be grouped according to anthropometric, socioeconomic and behavioural variables, revealing trends that might not emerge through traditional statistical methods [31, 32]. In fact, PCA allowed us to identify distinctive patterns in eating behaviour and sports participation between men and women in our sample. In particular, we used five distinct groups, each characterised by unique dietary and physical activity behaviours (Table 1). Each of the PCA groups exhibited distinct food preferences and physical activity patterns. However, differences in body composition were primarily observed between genders rather than across PCA groups. This suggests that gender plays a pivotal role in shaping eating habits and physical activity, with potential implications for body composition. While our focus on structured sports activities offers insight into gender differences in high-intensity exercise engagement, unstructured activities such as walking or cycling may provide equally significant benefits for health and well-being, particularly for populations less inclined toward competitive or organized sports. Indeed, in accordance with previous results by Millis and colleagues [33]. Balanced Eaters and Focused on Home Cooking groups were more likely to have regular and structured eating patterns, with a preference for home-cooked meals. These groups display a relatively balanced approach to eating, which may contribute to more favourable results in terms of body composition. On the contrary, the groups of restaurant lovers and varied eaters showed more flexible and inconsistent eating patterns, with frequent dining out and snacking, potentially predisposing these individuals to less favourable metabolic profiles [34].

We previously demonstrated that gender plays a key role in shaping dietary preferences, with women favoring healthier food choices and men exhibiting distinct meal behaviors [35]. In this analysis, gender differences emerged in the PCA groups revealed nuances far beyond simple stereotypes about red meat and vegetables consumption, highlighting the relevance of other aspects, including food preferences, dietary habits and sports participation. By analyzing these variables within our groups, we gained critical insights into how dietary and physical activity behaviors are differently managed across genders, also pointing out the complex interplay between biological sex, dietary patterns, and exercise habits. In addition, these behaviors appear to be influenced by broader socio-cultural frameworks, which shape individual lifestyle choices. In detail, in the Balanced Eaters group, both men and women display a focus on regular/balanced eating, but what distinguishes them is how they experience food. Men tend to prefer foods with an energy function, often influenced by established habits that respond more to practical needs than to personal taste. Women, on the other hand, seem to approach food with a more holistic approach, also considering the aspect of sensory pleasure and overall well-being that a balanced meal can offer. Notably, these results are in accordance with our previous studies investigating gender differences in dietary preferences and eating behaviors [35]. We also found that different fat mass distribution and body composition observed between men and women contribute to food preferences [36], thus promoting a vicious cycle potentially responsible for the failure to respond to conventional nutritional therapies in subjects with overweight/obesity. In this context, findings emerging from the Balanced Eaters group’s analysis reveal stability of behaviour which in turn could represent a gender tailored intervention strategy to prevent food-related metabolic diseases.

Home cooking is associated with healthier eating habits and represents an important prevention strategy to tackle obesity development in adults [33, 37]. Results from NHANES study revealed that meal preparation is predominantly managed by women in the United States [29, 38]. Similarly, in our investigation, the Home-Cooked Focused group displayed a more evident social dynamic, where women appear more involved in meal preparation, often associated with a sense of care and community. Men in this group, while appreciating the quality of home-cooked meals, tend to be less involved in the process. Such evidence confirms that personalised nutrition strategies should take into account that each individual is embedded in social and cultural contexts [39]. Nevertheless, both sexes recognise the nutritional and social value of home-cooked meals. It is essential to highlight that the abandonment of home cooking, often replaced by ready-made meals or meals eaten outside the home, may lead to an increase in the intake of processed and high-calorie foods, contributing to a worsening of public health [40]. For this reason, promoting greater male participation in meal preparation could not only better distribute the household load, but also reinforce the importance of home-prepared food for family well-being, particularly in couple and parenting dynamics. Interventions that encourage both sexes to be more active in meal preparation could reduce the overuse of ready-made meals and fast food, improving overall health and social bonding around mealtimes [41].

Gender differences also emerge in the Routine Eaters group, distinguished by a greater regularity in eating habits. Eating routines are a critical concern as habitual eating patterns have a direct impact on both nutritional intake and overall health. Indeed, regular food consumption can influence dietary quality, metabolic homesotasis, and contribute to long-term health outcomes, such as the risk of chronic diseases [42, 43]. In our analysis, women from the Routine Eaters group follow a more structured and stability-oriented routine, whereas men are more inclined to introduce variations, both in the timing and choice of meals. This difference may reflect men’s flexibility in their responses to daily commitments, but also less attention to the metabolic consequences of these variations. Interventions that promote greater adherence to eating routines could be useful for men in this group to ensure better weight and energy control in the long term. These considerations are in accordance with results from chrono-nutrition studies demonstrating that disruptions in circadian rhythms, such as misalignment of eating and sleeping patterns, can interfere with metabolic functions regulated by energy balance and appetite hormones, thus contributing to increased fat accumulation, impaired glucose tolerance, and increased risk of obesity and type 2 diabetes [44]. In the Restaurant Lovers group, strong differences in food tastes and in the management of meals out of home are observed. Men tend to see restaurants as an opportunity to consume indulgent, higher-calorie meals, whereas women, while appreciating the social aspect of eating out, show a greater ability to maintain a nutritional balance, even in less controlled settings. This reflects not only a different management of tastes and habits, but also a greater female awareness in balancing the pleasure of food along with maintaining a healthy diet [45]. An intervention for men in this group could focus on providing tools to make more conscious and healthy choices even in less regulated social contexts such as restaurants, without giving up the pleasure of food [46].

Finally, in the Varied Eaters group, variability in eating habits represents a complex challenge. It has been established that irregular meal patterns are associated with increased cardio-metabolic risk [47, 48]. In our analysis, men in the Varied Eaters group show a certain flexibility in organising meals, which could be attributed to a more hectic and disorganised lifestyle. On the other hand, women tend to compensate for a lack of regularity with frequent snacking, which, although maintaining a certain energy intake, could lead to nutritional imbalances in the long run. Interventions for this group should aim to stabilise eating habits, helping both sexes to find a balance that does not compromise the nutritional quality of their diets, while maintaining the flexibility needed to manage daily life routines [36, 49].

The analysis of sports participation in our study revealed significant gender differences between the five groups in the PCA. In particular, the Restaurant Lovers and Varied Eaters groups showed the widest gender gap in participation, with men reporting greater involvement in physical activities, such as strength training and endurance sports than women. Such observation is in accordance with previous studies suggesting that men and women display different levels of physical activity, thus highlighting the importance of tailored interventions taking into account gender gap related to sport participation [50–52]. Encouraging structured sports programmes tailored to the specific needs and constraints of women in these groups could prove effective in bridging these gaps [53]. In contrast, in the Balanced Eaters and Routine Eaters groups, sports participation was more evenly distributed between genders, reflecting a more controlled and structured lifestyle. These groups may already benefit from healthier behaviours, as they show that when eating habits are more regular and balanced, physical activity levels also tend to follow. This finding emphasises the importance of adopting a gender-sensitive approach in public health interventions, integrating nutrition and physical activity promotion to ensure holistic health improvements, especially in groups with lower sports participation [54]. In summary, while gender-specific patterns in sports participation varied across the PCA groups, there is clear evidence that addressing these differences, particularly in groups like Restaurant Lovers and Varied Eaters, could lead to better health outcomes by improving both dietary habits and physical activity levels among women. Significant gender and group differences in sleep patterns were also observed, with ‘varied eaters’ showing worse sleep quality and ‘balanced eaters’ reporting better sleep. These results suggest that structured eating patterns may promote better sleep, while irregular habits may disrupt it. Future studies should explore the role of chronotypes and their interaction with dietary behaviour and physical activity.

A number of limitations must be taken into account when interpreting the results of the present study. Firstly, the cross-sectional nature of the research design does not allow causal relationships to be established between dietary patterns, physical activity and body composition. Another limitation is the exclusion of general physical activity, such as unstructured or incidental activities (e.g. walking, housework), which could provide a more complete understanding of participants’ physical activity levels. Furthermore, the use of self-completed questionnaires to collect data on dietary habits and physical activity may be subject to participants’ social desirability bias or memory errors. Although PCA is a powerful tool for identifying patterns in complex data, the components identified may be influenced by the variables chosen for analysis, limiting the generalisability of the results to other populations. Furthermore, the sample was drawn exclusively from a single metabolic health centre located in Italy. This focus may introduce cultural biases, as dietary patterns and physical activity behaviour are influenced by cultural, social and environmental factors. Therefore, the results may not be fully applicable to populations in other cultural or geographical contexts. Finally, our results are based on a specific sample of individuals attending a metabolic health centre, which may reduce the transferability of the results to larger, heterogeneous populations. Future longitudinal studies, including more diverse populations and objective data collection, will be useful to confirm our results and improve understanding of gender differences in dietary and physical activity behaviour.

Conclusions

This study shows significant gender differences in the eating and physical activity patterns of adults attending a metabolic health centre. Women demonstrated healthier eating patterns, with a preference for vegetables and structured routines, while men were more likely to consume meat and showed more flexible eating habits. Men also reported greater participation in strength and endurance sports, while women were less active in these areas, particularly in the ‘restaurant lovers’ and ‘various eaters’ groups. Our results underline the need for gender-specific strategies in the development of diet and lifestyle interventions. For example, encouraging women to participate in structured strength training and promoting regular physical activity of any kind, including walking, cycling and dancing, could improve overall health and well-being. Similarly, dietary interventions for men should focus on increasing vegetable and fibre intake.

Tailor-made approaches that address these gender differences can improve cardiometabolic health, both in the general population and in patients undergoing rehabilitation programmes. By closing existing gaps in dietary and physical activity habits, such interventions could improve quality of life and long-term health outcomes (Table 5).

Future research should focus on longitudinal studies to assess the stability and evolution of behavioural patterns identified through PCA. These studies could confirm the consistency of these patterns over time and explore their predictive value for health outcomes. Furthermore, incorporating objective measures of food intake, physical activity and body composition in future research would reduce potential biases and improve the robustness of the results. These approaches would provide a more solid basis for clinical applications, enabling the development of personalised and gender-specific nutritional interventions.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

Alessandra Feraco contributed to the conceptualisation and writing of the original draft. Stefania Gorini managed the data, edited the visualisation and participated in the revision and editing of the manuscript. Elisabetta Camajani handled data visualisation and writing of the original draft. Tiziana Filardi contributed to the formal analysis and revision of the manuscript. Sercan Karav participated in the methodology, supervision and revision of the text. Edda Cava performed supervision and revision tasks. Rocky Strollo contributed to the formal analysis and revision of the text. Elvira Padua supervised and participated in the revision and editing of the sport parts. Massimiliano Caprio worked on the methodology, supervision and revision of the manuscript. Andrea Armani was responsible for data management and writing the original draft. Mauro Lombardo contributed to the conceptualisation, supervision, writing of the original draft and final revision of the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by institutional fundings from IRCCS San Raffaele Pisana (Ricerca Corrente).

Data availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request. All data will be shared in a de-identified format to protect participant confidentiality.

Declarations

Ethics approval and consent to participate

The participants provided their written informed consent to participate in this study. The studies were conducted in accordance with local legislation and institutional requirements. The studies involving humans were approved by the IRCCS San Raffaele Ethics Committee (registration number RP 23/13).

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.