Association of dieting self-efficacy, weight management nutrition knowledge, physical activity, and dietary pattern among Turkish adults: a cross-sectional study

Abstract

Background

Unhealthy diets and obesity are major contributors to noncommunicable diseases worldwide. Dieting self-efficacy and nutrition knowledge are believed to influence eating behavior; however, their combined effect on dietary pattern is not well understood. This study aimed to examine the relationship between dieting self-efficacy, weight management-related nutrition knowledge, and dietary pattern among adults.

Methods

This cross-sectional study recruited 1457 adults (aged 19–64) from Erzurum, Türkiye. Participants completed validated measures assessing dieting self-efficacy (DIET-SE), weight management nutrition knowledge (WMNKQ), and dietary intake using a brief food frequency questionnaire. A dietary pattern score was derived using principal component analysis (PCA). Physical activity was assessed using the International Physical Activity Questionnaire–Short Form (IPAQ-SF). Partial Spearman correlations (adjusting for age and sex), multiple linear regressions, and path analysis were conducted to examine associations and mediation effects.

Results

The sample had a median age of 24 years, and 10.2% were obese. DIET-SE and WMNKQ scales showed high internal reliability (Cronbach’s α ≥ 0.72). Nutrition knowledge was positively correlated with self-efficacy (ρ = 0.12, p < 0.01); however, both were inversely associated with dietary pattern (ρ = -0.22 and − 0.15, respectively; p < 0.01). Regression analyses confirmed that higher DIET-SE and WMNKQ scores independently predicted poorer dietary pattern (β = -0.21 and − 0.13; p < 0.001). Female sex, higher education, and income were positively associated with WMNKQ scores, while age negatively predicted dietary pattern. Path analysis supported a partial mediation model: WMNKQ positively predicted DIET-SE (β = 0.13), which was negatively associated with dietary pattern (β = -0.20).

Conclusions

The results demonstrated that participants with higher weight-management knowledge and dieting self-efficacy had lower overall dietary pattern. These findings highlight that knowledge alone is not enough to ensure healthy eating, and that self-efficacy must be properly directed.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-24848-w.

Background

According to the World Health Organization (WHO), noncommunicable diseases (NCDs) were responsible for the deaths of at least 43 million people worldwide, equivalent to 75% of non-pandemic-related deaths globally in 2021 [1]. Among the most prevalent NCDs are obesity, cardiovascular diseases, and type 2 diabetes [2], all of which are closely linked to modifiable risk factors, particularly unhealthy dietary habits and sedentary lifestyle [3]. In this context, promoting healthy eating behaviors and improving nutrition-related competencies have become urgent public health priorities.

Overweight and obesity are key risk factors for chronic NCDs [4]. Referred to as “globesity,” the escalating global increase in obesity has become a significant public health issue for societies and healthcare systems [5]. According to the World Obesity Atlas 2025 report on March 4, 2025 (the latest data of v7) 71% of the Turkish adult population currently has a high body mass index (BMI), and 34% are living with obesity [6]. By way of comparison, European Union member states report a wide range of adult obesity rates, approximately 32% in Hungary, 28% in Poland, 21% in Germany, and 16% in Spain [6]. This underscores the particularly high burden of obesity in Türkiye.

Obesity is not only a risk factor for multiple NCDs but also imposes significant economic and psychosocial costs [7]. These concerns have prompted growing interest in behavioral and educational strategies that can empower individuals to make sustainable dietary choices [8]. In this context, two key constructs that have received increasing attention are dietary self-efficacy [9] and weight management-related nutritional knowledge [10]. Emerging evidence indicates that dieting self-efficacy is one of the most important factors in the successful re-establishment of healthy eating patterns [11]. Bouwman et al. (2020) also found that brief self-efficacy intervention can directly increase vegetable intake and indirectly improve compliance to a diet plan to eat healthier [12]. Recent studies on weight loss processes suggest that individuals with high dieting self-efficacy are more likely to believe in their ability to manage dietary obstacles and maintain behavioral changes over time [13, 14]. Higher self-efficacy is associated with improved dietary pattern, reduced calorie intake, and better weight outcomes across diverse populations [15–17]. Moreover, food choices and food literacy are increasingly recognized as pivotal factors influencing not only overall dietary patterns but also body weight regulation, weight gain prevention, and successful weight loss maintenance [18]. Recent findings highlight that food literacy is not equally distributed across demographic groups, with gender emerging as a key determinant [19]. Similarly, weight management nutrition knowledge encompasses an individual’s understanding of core dietary principles, including energy balance, portion control, food labeling, nutrient density, and dietary guidelines [20]. Recent studies have shown that adequate nutritional knowledge is positively associated with healthy eating behaviors and inversely associated with BMI [21–23].

Although dieting self-efficacy and weight management nutrition knowledge have each been shown to influence eating behaviors and body weight independently, few studies have explored how these two factors interact among the adult population. A clearer understanding of this relationship is essential, as it could lead to the development of more comprehensive and behaviorally informed dietary interventions. Some individuals may have nutritional principles but struggle to apply this knowledge in daily life due to low self-confidence in their ability to maintain healthy habits. Some may feel motivated and capable of changing their diet but lack the factual knowledge required to make informed choices. Recognizing and addressing both the psychological (self-efficacy) and cognitive (nutrition knowledge) components of dietary behavior may offer a more effective path toward sustainable health improvement. As the prevalence of obesity and diet-related NCDs continues to rise, there is an increasing need for studies on person-centered approaches to dietary behavior change. This study aimed to investigate the relationship between dietary pattern, dieting self-efficacy, and weight management nutrition knowledge in adults.

Methods

Study design and participants

This cross-sectional study was conducted between Jun 2025 and July 2025 in Erzurum, Türkiye, to examine the associations among dieting self-efficacy, weight management nutrition knowledge, dietary pattern, and physical activity among community-dwelling adults aged 19–64 years. A convenience sampling method was used to recruit participants. The eligibility criteria included age between 19 and 64 years and residence in the study area. A total of 1457 individuals participated in the study and completed questionnaire-based assessments. The exclusion criteria included pregnancy, self-reported chronic conditions preventing participation, and missing baseline variables for planned analyses.

Ethical approval was obtained from the Ethics Committee of the Faculty of Health Sciences, Atatürk University (Approval No: [2025/05/06], date: May 27, 2025), and the study protocol was implemented in accordance with the Declaration of Helsinki. All participants provided written informed consent prior to the study.

Sample size

An a priori power analysis was performed using G*Power (version 3.1.9.7). For a linear multiple regression with five predictors, the analysis indicated a required minimum sample size of 830 participants to detect a small effect (f² = 0.02) with 90% power at a significance level of α = 0.05. The final sample (N = 1457) exceeded this requirement, thereby providing at least 90% power under the specified assumptions [24].

Data collection instruments

Sociodemographic and anthropometric data

A structured questionnaire was administered via face-to-face interviews. Participants self-reported their age, sex, education level, income status, and chronic disease status. Weight and height were also self-reported, and body mass index (BMI) was calculated as weight (kg) divided by height in meters squared (kg/m²). BMI was categorized into four groups according to the WHO classification: underweight (< 18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), and obese (≥ 30.0 kg/m²) [25].

Dieting self-efficacy scale (DIET-SE)

Dieting self-efficacy was assessed using the 11-item Dieting Self-Efficacy Scale (DIET-SE) developed by Stich et al. (2009) [26] and adapted into Turkish by Hamurcu et al. (2023) [27] The scale evaluates individuals’ confidence in maintaining healthy eating behaviors under challenging circumstances. It consists of three subdimensions: (i) high-calorie food, (ii) social and internal factors, and (iii) negative emotional events. Items are rated on a 5-point Likert scale ranging from 0 (“not at all confident”) to 4 (“very confident”), yielding a total score of 0–44. Higher total and subscale scores indicate a stronger perceived dieting self-efficacy. The Cronbach’s alpha coefficient for the Turkish version was α = 0.900.

Weight management nutrition knowledge questionnaire (WMNKQ)

Nutrition knowledge was assessed using the Weight Management Nutrition Knowledge Questionnaire (WMNKQ), originally developed by Mikhail et al. (2020) [20], and adapted into Turkish by Onbaşı and Akçil Ok (2024) [28]. The instrument consists of 43 multiple-choice questions designed to evaluate essential domains of nutrition knowledge related to effective weight management. It comprises five conceptual subdimensions: (i) energy density, (ii) portion size, (iii) variety of food affecting food intake, (iv) reliable nutrition information sources, and (v) alcohol intake and sugar-sweetened beverages. Each correct response received 1 point, with total scores ranging from 0 to 43. Higher scores indicate greater weight management-related nutritional knowledge. The Turkish version demonstrated acceptable internal consistency, with a reported Cronbach’s α of 0.75.

Dietary assessment

Dietary intake was evaluated using a brief food frequency questionnaire (FFQ) that assessed the habitual consumption of 14 commonly consumed food items in the previous month. To identify the underlying dietary patterns, Principal Component Analysis (PCA) was conducted on the standardized intake frequencies. Sampling adequacy was verified using the Kaiser–Meyer–Olkin test (KMO = 0.837), and Bartlett’s test of sphericity indicated suitability for factor analysis (χ² = 6932.22, p < 0.001). Components with eigenvalues greater than 1.0 were retained, and food items with absolute factor loadings ≥|0.30| were considered a moderate correlation (See Supplementary Table 1). While our 14-item FFQ was designed to capture key food groups relevant to obesity and non-communicable disease risk, we acknowledge that it does not comprehensively assess the entire diet. Short FFQs are widely used and validated for ranking individuals’ dietary patterns rather than quantifying exact nutrient intakes [29–34]. Nonetheless, caution is warranted in interpreting the resulting pattern as a proxy for the overall dietary pattern. While less precise than longer tools, this approach is appropriate for deriving a relative diet pattern score [35], and its limitations are acknowledged in the manuscript.

The first principal component had an eigenvalue of 3.999 and explained 28.6% of total variance.

The first principal component, characterized by positive loadings for fruits, vegetables, whole grains, and low-fat protein sources, and negative loadings for processed snacks, sugary beverages, and high-fat foods, was interpreted as a general dietary pattern. To facilitate interpretability, the scores for this component were reversed, such that higher values reflected healthier dietary profiles.

International physical activity questionnaire – short form (IPAQ-SF)

Physical activity was assessed using the Turkish-validated version of the International Physical Activity Questionnaire–Short Form (IPAQ-SF) [36], which measures self-reported frequency and duration of walking, moderate, and vigorous-intensity activities performed over the past seven days. Total physical activity was expressed in MET-min/week and computed using the IPAQ scoring protocol [37], whereby standardized MET coefficients of 3.3, 4.0, and 8.0 were applied for walking, moderate, and vigorous activity, respectively, based on reported minutes and days. Participants who responded “I don’t know” or left any item blank were excluded from the MET analysis according to IPAQ data-cleaning rules. Of the 1457 participants, 521 (35.8%) were excluded due to incomplete or inconsistent IPAQ responses, leaving a final analytic sample of 936 individuals with valid MET data.

Physical activity levels were categorized based on total weekly MET-minutes using the standard IPAQ classification thresholds: low (< 600 MET-min/week), moderate (600–2999 MET-min/week), and high (≥ 3000 MET-min/week). Sitting time (hours/day) was also recorded as a single-item self-report and included in the descriptive statistics but was not used in the calculation of the total MET scores.

Statistical analysis

All statistical analyses were conducted using IBM SPSS Statistics for Windows, Version 27.0 (IBM Corp., Armonk, NY, USA) and AMOS Version 22.0 (IBM Corp., Armonk, NY, USA). The distribution of continuous variables was assessed using the Shapiro–Wilk test. As the assumption of normality was violated for all continuous variables (p < 0.05), nonparametric statistical methods were performed for the analysis. Multiple linear regression was also conducted to model the predictors of dietary pattern. Although the continuous variables deviated from normality, regression assumptions were evaluated on the residuals and were adequately met—residual normality (Q–Q plots), linearity, and homoscedasticity. Given the large sample size, the regression estimates remain robust under the Central Limit Theorem [38].

The internal consistency reliability of the DIET-SE and WMNKQ instruments was evaluated using both Cronbach’s alpha and McDonald’s omega (ω) coefficients. While Cronbach’s α is widely used, McDonald’s ω is often preferred as it does not assume tau-equivalence and may provide a more accurate reliability estimate for scales with heterogeneous items. The Omega macro for SPSS [39] was used to compute the ω values with 10,000 bootstrap samples. Items showing corrected item–total correlations below 0.10 or negative values were excluded from the ω calculations in accordance with Dunn et al. (2014) [40].

Group comparisons were conducted using the Mann–Whitney U test (for two-group comparisons) or the Kruskal–Wallis H test (for more than two groups) because of non-normal distributions. Pearson’s Chi-square test was used for categorical comparisons. Effect sizes were reported as rank-biserial correlations (for continuous variables) or Cramér’s V (for categorical variables). Effect sizes were interpreted based on conventional thresholds (e.g., r = 0.10 small, 0.30 medium, 0.50 large; Cramér’s V = 0.10 small, 0.30 medium, 0.50 large) as recommended by Cohen (1988) [41].

To evaluate potential selection bias arising from missing IPAQ data, we compared baseline characteristics between participants included in the IPAQ analysis (n = 936) and those excluded due to missing/invalid IPAQ data (n = 521). Continuous variables were compared using the Mann–Whitney U test with rank-biserial correlation (r) as the effect size; categorical variables were compared using Pearson’s chi-square test with Cramér’s V. Effect sizes were interpreted using conventional benchmarks (negligible, small, medium, large). Full results are shown in Supplementary Table 2.

Spearman’s partial rank correlations (ρ) among the PCA-derived dietary pattern score, DIET-SE, WMNKQ, MET values, and other derived indices were estimated by rank-transforming the variables and then computing Pearson partial correlations on the ranks, controlling for age (years) and sex (female = 1). Two-tailed significance was assessed at p < 0.05 and p < 0.01. A correlation heatmap was generated to visualize the directions and strengths of the associations.

Multiple linear regression models involving MET scores were conducted on a subset of participants (N = 936) with complete IPAQ-SF data; those with missing or inconsistent physical activity responses were excluded. All continuous predictor variables were z-standardized before the analysis. Each model included two blocks: Block 1 consisted of behavioral predictors (DIET-SE, WMNKQ, and MET scores), and Block 2 included demographic covariates (age, sex, education level, income level, and chronic disease status). Multicollinearity was evaluated using variance inflation factors (VIF), and all VIF values were below 1.2. The model assumptions were verified before interpreting the results. Standardized beta coefficients (β), 95% confidence intervals (CI), R², ΔR², and p-values were reported for all the models.

To assess whether the associations of WMNKQ and DIET-SE with the dietary pattern score differed by BMI, we performed pooled linear regression models with standardized predictors. In a second block, multiplicative interaction terms (WMNKQ×BMI and DIET-SE×BMI) were introduced, and changes in explained variance (ΔR²) were reported to evaluate interaction effects. Additionally, we fitted stratified models within the BMI categories (underweight, normal weight, overweight, obese) to estimate stratum-specific standardized β coefficients and two-sided p-values. All predictors (WMNKQ, DIET-SE, BMI) were z-standardized and mean-centered before creating interaction terms to reduce collinearity. All analyses were exploratory in nature; therefore, no adjustments were made for multiple comparisons.

Finally, path analysis was performed using AMOS to examine the potential mediating effect of dieting self-efficacy on the relationship between nutrition knowledge and dietary pattern. Because the mediation model with three observed variables was saturated (df = 0), global fit indices (e.g., χ², CFI, RMSEA) were not informative. Therefore, the interpretation focused on the magnitude and significance of the standardized path coefficients (β) and the bootstrap CI for the indirect effect [42]. Statistical significance was set at p < 0.05.

Results

General characteristics of the study population

The participant characteristics are summarized in Table 1. Male participants exhibited significantly higher median values for age, weight, and BMI than female participants (all p < 0.001), with a medium effect size for weight (r = −0.427). Significant sex differences were also found in the BMI category, education level, and household income (all p < 0.05), with small effect sizes (Cramér’s V < 0.21). No sex differences were observed in chronic disease status (p = 0.534). In the subsample with valid IPAQ-SF data (n = 936), physical activity levels differed by sex (p = 0.015), with a higher proportion of males classified as highly active.

Table 1.

Socio-demographic and anthropometric characteristics of the participants by sex

Variable	Male (N = 488)	Female (N = 969)	Total (N = 1457)	Test statistic	Effect size	p-value
Age (years), Median (IQR)	25.0 (16.0)	24.0 (8.0)	24.0 (11.0)	Z = −4.203	r = −0.110	< 0.001
Weight (kg), Median (IQR)	78.0 (19.0)	61.0 (16.0)	65.0 (22.0)	Z = −16.279	r = −0.427	< 0.001
BMI (kg/m²), Median (IQR)	25.2 (5.1)	22.8 (5.4)	23.5 (5.9)	Z = −8.468	r = −0.222	< 0.001
BMI category, n (%)				χ²(3) = 60.980	V = 0.205	< 0.001
Underweight	16 (3.3)	70 (7.2)	86 (5.9)
Normal	218 (44.7)	592 (61.1)	810 (55.6)
Overweight	193 (39.5)	219 (22.6)	412 (28.3)
Obese	61 (12.5)	88 (9.1)	149 (10.2)
Education level, n (%)				χ²(3) = 10.649	V = 0.085	0.014
Primary/Secondary	13 (2.7)	37 (3.8)	50 (3.4)
High School	112 (22.9)	196 (20.2)	308 (21.1)
University (Associate/Bachelor’s)	303 (62.1)	659 (68.1)	962 (66.0)
Postgraduate (Master’s/Doctorate)	60 (12.3)	77 (7.9)	137 (9.4)
Household income level, n (%)				χ²(2) = 8.546	V = 0.077	0.014
Less than expenses	72 (14.8)	157 (16.2)	229 (15.7)
Equal to expenses	261 (53.4)	574 (59.2)	835 (57.3)
More than expenses	155 (31.8)	238 (24.6)	393 (27.0)
Chronic disease diagnosis, n (%)				χ²(1) = 0.387	V = 0.016	0.534
Yes	70 (14.3)	151 (15.6)	221 (15.2)
No	418 (85.7)	818 (84.4)	1236 (84.8)

	Male (N = 329)	Female (N = 607)	Total (N = 936)	Test statistic	Effect size	p -value
*Physical activity level, n (%)				χ²(2) = 8.463	V = 0.095	0.015
Low	98 (29.8)	207 (34.1)	305 (32.6)
Moderate	157 (47.7)	309 (50.9)	466 (49.8)
High	74 (22.5)	91 (15.0)	165 (17.6)

^* N = 936 indicates the number of participants with complete IPAQ-SF MET-min/week data points. BMI was calculated using self-reported weight and height. Z-scores were obtained using the Mann–Whitney U test for continuous variables that were not normally distributed. Pearson’s Chi-square test was used for categorical comparisons. Effect size expressed as Cramér’s V for categorical variables and rank-biserial r for continuous variables. r values were calculated as Z/√N.

Supplementary Table 2 summarises the comparison of participants included vs. excluded from the IPAQ analysis. Age, sex, weight, BMI, diet self-efficacy, household income, and chronic disease did not differ meaningfully between groups (all |r| or V < 0.10). Education level differed slightly (V = 0.117, p < 0.001; a higher proportion with university/postgraduate education in the included group), and WMNKQ scores were modestly higher among included participants (r = 0.184, p < 0.001). Dietary pattern showed a statistically significant but negligible difference (r = 0.074, p = 0.005).

Internal consistency reliability of the scales

Table 2 presents the descriptive statistics and reliability. The DIET-SE total score demonstrated excellent internal consistency (α = 0.903 [0.896–0.910]; ω = 0.904 [0.893–0.911]). The subscales showed acceptable to borderline good reliability (α = 0.774–0.800; ω = 0.776–0.797). The WMNKQ showed adequate reliability (α = 0.718 [0.697–0.738]; ω = 0.757; CI unavailable). For the IPAQ SF, continuous variables were non-normal; MET min/week was available for 64.2% of participants (median 1085.5, IQR 1881.0).

Table 2.

Descriptive statistics and internal consistency reliability of the study scales (N = 1457)

Scale/Subscale	Items (n)	Score Range	Median (IQR)	Min–Max	Missing (%)	Cronbach’s [95% CI]	McDonald’s ω [95% CI]
DIET-SE	11	0–44	17.0 (14.0)	0–44	-	0.903 [0.896–0.910]	0.904 [0.893–0.911]
High-Caloric Food	4	0–16	6.0 (5.0)	0–16	-	0.774 [0.755–0.793]	0.776 [0.753–0.798]
Social and Internal Factors	4	0–16	6.0 (6.0)	0–16	-	0.800 [0.782–0.816]	0.797 [0.776–0.817]
Negative Emotional Events	3	0–12	5.0 (5.0)	0–12	-	0.786 [0.767–0.805]	0.788 [0.765–0.809]
WMNKQ	43	0–43	20.0 (8.0)	5–36	-	0.718 [0.697–0.738]	0.757 -
IPAQ-SF (MET-min/week)	-	-	1085.5 (1881.0)	99-7092	35.8	-	-

All continuous variables violated the assumption of normality (Shapiro–Wilk test, p < 0.05); therefore, median and interquartile range (IQR) values are reported. MET values were computed only for participants with complete IPAQ-SF data (N = 936). Internal consistency reliability was assessed using Cronbach’s α and McDonald’s ω coefficients. Cronbach’s α values (and their 95% CI) were derived from the intraclass correlation coefficient (Two-Way Mixed, Consistency, Average-Measures model), which is mathematically equivalent to Cronbach’s α. Omega values (and their 95% CI) were estimated using 10,000 bootstrap samples. A bootstrap CI could not be generated for the WMNKQ owing to computational instability; hence, only the point estimate (ω = 0.757) is reported. Omega coefficients for the subscales were reported only when the item count and internal consistency supported a meaningful estimation. Items with corrected item–total correlations below 0.10 or negative correlations were excluded from the omega calculations, in line with the statistical guidelines for omega estimation

DIET-SE Dieting Self-Efficacy Scale, WMNKQ Weight Management Nutrition Knowledge Questionnaire, IPAQ-SF International Physical Activity Questionnaire–Short Form

Partial correlation analyses controlling for age and sex are shown in Fig. 1. Total dieting self-efficacy (DIET-SE) was negatively correlated with dietary pattern (ρ = −0.22, p < 0.01) and positively correlated with WMNKQ (ρ = 0.12, p < 0.01). WMNKQ total score showed significant positive correlations with all five subdomains of nutrition knowledge, ranging from weak to very strong in magnitude (ρ = 0.26–0.85, p < 0.01). No significant association was observed between the WMNKQ and high-calorie dietary self-efficacy barriers (ρ = 0.04, p > 0.05). Correlations between IPAQ-SF MET-min/week and the examined behavioral variables were consistently negligible in size (|ρ| ≤ 0.08), regardless of their statistical significance.

Fig. 1.

Heatmap of partial correlations among the main study variables, adjusted for age and sex. The row and column labels indicate the variables included in the partial correlation analysis (adjusted for age and sex). “Dietary Pattern” = Dietary Pattern Score derived from PCA; “Energy Density” = Energy Density Score; “Portion Size” = Portion Size Score; “Food Variety” = Food Variety Score; “Reliable Info” = Reliable Nutrition Information Score; “Alcohol/Sugar Drinks” = Alcohol and Sugar-Sweetened Beverage Score; “Total WMNKQ” = Total Weight Management Nutrition Knowledge Score; “SelfEfficacy – Caloric/Social/Emotional/Total” = Subscale and total scores from the Dieting Self‐Efficacy Scale; “IPAQ MET/min/week” = Total MET-min/week from IPAQ-SF. Cells display Spearman’s ρ for each pairwise association, with red and blue indicating positive and negative correlations, respectively. Statistical significance is denoted by *p < 0.05 and **p < 0.01

Multiple linear regression models: dietary pattern, dieting self-efficacy, and nutrition knowledge

The results of the hierarchical multiple linear regression analysis are summarized in Table 3. In Model 1, dieting self-efficacy (DIET-SE) (β = −0.208, p < 0.001) and nutrition knowledge (WMNKQ) (β = −0.131, p < 0.001) were significant predictors of dietary pattern. Additionally, younger age was associated with better dietary pattern (β = −0.202, p < 0.001). The full model explained R² = 0.115 of variance. Adding demographic covariates in Block 2 resulted in an additional ΔR² = 0.045 of the explained variance. The overall model was significant [F(8, 925) = 15.094; p < 0.001]. Note: F-change is not provided in this table.

Table 3.

Multiple linear regression models predicting dietary pattern, dieting self-efficacy, and nutrition knowledge among adults (N = 1457)

Predictors	Model 1:Dietary Pattern (PCA)		Model 2:DIET-SE		Model 3:WMNKQ
	β[95% CI]	p-value	β[95% CI]	p-value	β[95% CI]	p-value
Block 1 (Behavioural)
Dieting Self-Efficacy (DIET-SE)	−0.208[−0.259, −0.139]	< 0.001	—	—	0.085 [0.117, 0.817]	0.009
Nutrition Knowledge (WMNKQ)	−0.131 [−0.191, −0.070]	< 0.001	0.097 [0.222, 1.545]	0.003	—	—
Dietary Pattern (PCA)	—	—	–0.206 [–2.857, –1.490]	< 0.001	−0.135 [−0.403, −0.061]	< 0.001
IPAQ-SF (MET-min/week)	−0.041 [−0.099, 0.020]	0.195	0.047 [–0.164, 1.122]	0.144	−0.038 [−0.559, 0.138]	0.236
Block 2 (Demographic)
Age (years)	−0.202 [−0.026, −0.013]	< 0.001	0.021 [–0.059, 0.081]	0.756	0.190 [−0.012, 0.063]	0.190
Sex (Female=1)	−0.065 [−0.181, 0.071]	0.392	0.071 [0.163, 2.878]	0.028	0.118 [0.639, 2.102]	< 0.001
Education (University+)	0.124 [−0.040, 0.249]	0.156	–0.021 [–2.090, 1.044]	0.513	0.162 [1.337, 3.011]	< 0.001
Income (>expenses)	0.010 [−0.109, 0.154]	0.739	0.047 [–0.363, 2.478]	0.144	0.085 [0.277, 1.810]	0.008
Chronic disease (Yes=1)	0.020 [−0.108, 0.241]	0.454	0.016 [–1.440, 2.338]	0.641	0.023 [−0.663, 1.382]	0.491
Model fit statistics
R²	0.115		0.069		0.087
ΔR² (block 2)	0.045		0.007		0.048
Adjusted R²	0.108		0.061		0.079
F (df1, df2)	15.094 (8, 925)		8.606 (8, 925)		11.005 (8, 925)
Model p-value	< 0.001		< 0.001		< 0.001
Max VIF	1.148		1.193		1.111

β standardized regression coefficient; 95% CIs are shown for the corresponding unstandardized coefficients (B), MET Metabolic Equivalent of Task, VIF variance inflation factor

Block 1 included behavioral predictors (DIET-SE, WMNKQ, and IPAQ-SF [MET-min/week]). Block 2 includes demographic covariates (age, sex, education, income, and chronic disease). Continuous variables were z-standardized before the analysis. Model 1 used the PCA-derived dietary pattern score as the dependent variable, Model 2 used the total dieting self-efficacy score, and Model 3 used the total weight management nutrition knowledge score. ΔR² denotes the incremental variance explained after adding Block 2 covariates. The incremental contribution of Block 2 was statistically significant in all models (p<0.05, F-change test). All regression assumptions (normality, homoscedasticity, and multicollinearity) were met

In Model 2, dietary pattern was a strong negative predictor of DIET-SE (β = −0.206, p < 0.001), and the WMNKQ showed a small but significant positive contribution (β = 0.097, p = 0.003). The inclusion of demographic variables resulted in a modest increase in explained variance (ΔR² = 0.007), among which, female sex was a significant positive predictor (β = 0.071, p = 0.028). The total model was significant [R² = 0.069; F(8, 925) = 8.606, p < 0.001].

In Model 3, both DIET-SE (β = 0.085, p = 0.009) and dietary pattern (β = −0.135, p < 0.001) significantly predicted nutritional knowledge (WMNKQ). Among the demographic covariates, female sex (β = 0.118, p < 0.001), higher education (β = 0.162, p < 0.001), and higher income (β = 0.085, p = 0.008) were positively associated with the WMNKQ. Age (p = 0.190) and chronic disease status (p = 0.491) were not significant predictors. The model explained R² = 0.087, with ΔR² = 0.048 added for demographics. The overall model fit was significant [F(8, 925) = 11.005, p < 0.001]. Across all models, multicollinearity was not a concern (maximum VIFs: 1.148, 1.193, and 1.111; all < 1.2).

BMI did not significantly moderate the associations of WMNKQ and DIET-SE with the dietary pattern score in the pooled model, as indicated by non-significant interaction terms (WMNKQ×BMI: p = 0.677; DIET-SE×BMI: p = 0.768) and no incremental variance explained (ΔR²=0.000, p = 0.891). In stratified analyses, inverse associations between DIET-SE and the dietary pattern score remained significant across normal (β=−0.222, p < 0.001), overweight (β=−0.178, p < 0.001), and obese (β=−0.162, p = 0.048) categories; the underweight stratum also showed inverse associations (WMNKQ β=−0.258, p = 0.015; DIET-SE β=−0.244, p = 0.022), but should be interpreted cautiously given the smaller n = 86. Full stratified results are provided in Supplementary Table 3.

The path analysis results (Fig. 2) supported a partially mediated model in which DIET-SE served as a significant intermediary between WMNKQ and dietary pattern. WMNKQ was positively associated with DIET-SE (β = 0.134, p < 0.001), and DIET-SE negatively predicted dietary pattern (β = −0.199, p < 0.001). The direct path from WMNKQ to dietary pattern remained significant (β = −0.127, p < 0.001), indicating a partial mediation. The indirect effect was statistically supported by bootstrap CI (5,000 resamples), which did not include zero. As the model was saturated (df = 0), global fit indices were not applicable, and the interpretation focused on the significance and direction of the standardized path coefficients.

Fig. 2.

Mediation model linking WMNKQ to dietary pattern (PCA-derived score) through DIET-SE. Standardized regression coefficients (β) and p-values are shown. The model is saturated (df = 0); therefore, global fit indices are not reported. ***p < 0.001

Discussion

This cross-sectional study examined the interrelationships between dieting self-efficacy, weight management nutrition knowledge, dietary pattern, and physical activity in a large adult sample, and explored associations with sociodemographic factors, including age, sex, education, income, and chronic disease.

Both the DIET-SE and WMNKQ instruments demonstrated high internal consistency, with both Cronbach’s alpha and McDonald’s omega coefficients supporting their reliability. The strong reliability of these scales adds confidence that the associations found are not due to measurement errors. High reliability is expected for well-validated questionnaires and aligns with similar instruments used in nutrition studies [20].

Statistically significant sex-based differences were observed in age, body weight, and BMI, with men reporting higher values for all three metrics. Education level, income, and BMI category also varied by sex, but the prevalence of chronic diseases did not. These differences were similar to previous studies, which showed that females tend to have greater nutrition knowledge and often better dietary pattern adherence than males [43–46]. In addition, females have been shown to display greater adherence to recommended dietary guidelines than males [47]. Similarly, we observed that female sex, higher education, and income were positively associated with nutrition knowledge, whereas physical activity did not emerge as a significant predictor. The lack of association between physical activity and any of the psychological or dietary variables in our study is noteworthy. Physical activity did not significantly predict nutritional knowledge, self-efficacy, or dietary pattern. This indicates that exercise behavior was largely independent of dietary and cognitive factors. Age was negatively related to dietary pattern, and female sex was positively related to DIET-SE. However, the lack of sex balance in the sample should be acknowledged as a limitation, since it may introduce sample bias and constrain the generalizability of the findings [48].

Our multiple regression analyses revealed that, although nutrition knowledge positively predicted DIET-SE, both constructs were inversely associated with dietary pattern. This intriguing result seems contrary to the majority of prior studies, which often find that nutrition knowledge is positively associated with healthy eating habits [49–51]. Although a strong link has been demonstrated between poor knowledge and poor eating behaviors [52], knowledge alone does not indicate healthy dietary behavior [53, 54]. Moreover, sufficient knowledge is not necessarily associated with healthy dietary behavior because of psychological and environmental factors that might play a role [55].

A recent study among Japanese university students suggests that while nutrition knowledge is positively associated with nutritional adequacy, its effect on actual eating behaviors may be limited. Although participants with higher knowledge demonstrated better alignment with dietary reference intakes, no significant relationship was found between total nutrient inadequacies and overall knowledge scores [56]. However, nutritional knowledge does not always translate into healthy behaviors due to resource limitations, family priorities, and other social or environmental restrictions [53, 57]. Similarly, nutrition students and health professionals with extensive knowledge have been found to maintain some inadequate dietary habits [58].

Knowledge assessed in surveys is often declarative (knowing “what” is healthy) rather than procedural (knowing “how” to eat healthily in daily life) [48]. In general, the phenomenon of not acting as intended is called the intention-behavior gap [59]. Our results support the notion that higher WMNKQ scores did not translate into better dietary pattern, and those with the most knowledge reported poorer dietary pattern. A recent study has shown that a lack of nutritional knowledge leads to deviations from recommended dietary practices [60]. One possible explanation is that individuals who score high on the WMNKQ might be those actively struggling with weight issues; they may have knowledge (of nutrients, calories, diets, etc.) without successfully translating it into sustained healthy eating.

Our correlation analyses demonstrated that individuals with higher knowledge and confidence reported lower-pattern dietary patterns. Interestingly, higher knowledge and confidence were associated with more energy-dense eating patterns and larger portion sizes, whereas self-efficacy showed a negative correlation with BMI. Physical activity was not associated with dietary behavior or psychological determinants. Our finding that higher knowledge was associated with larger portion sizes and more energy-dense choices suggests evidence of a relationship between knowing and doing. Nutrition knowledge should be supported by practical strategies [61], because even knowledgeable individuals may consume an energy-dense, nutritionally poor diet.

These findings suggest that the observed associations between knowledge/self-efficacy metrics and the dietary pattern are largely independent of BMI. Correlational analyses showed only a weak association between BMI and DIET-SE (ρ=−0.15) and a negligible association with WMNKQ (ρ=−0.01), indicating that BMI is unlikely to confound these associations in our models; we did not model BMI as a mediator, so mediation by BMI was not evaluated. Moreover, the absence of significant interactions, combined with consistently negative regression coefficients across BMI strata (See Supplementary Table 3), reinforces the interpretation that these associations are robust across body-weight categories. While statistically reliable, the effect sizes were small (β ≈ 0.11–0.26; R²=0.03–0.16).

The role of DIET-SE in predicting dietary pattern is also complex. Self-efficacy is a cornerstone of many behavior change theories and typically facilitates healthier choices [62]. However, we observed that higher DIET-SE scores were correlated with lower dietary pattern. It is important to consider that the DIET-SE scale is designed to measure confidence in resisting dietary temptations and maintaining adherence to restrictive eating behaviors—often in the context of high-risk or emotionally challenging situations (e.g., social pressure, negative mood, or food cravings). This construct aligns high confidence in one’s ability to “diet” or restrict (dieting self efficiancy) could plausibly lead to cycles of restriction and overcompensation, resulting in a lower-quality overall dietary pattern [63]. This distinction offers a plausible explanation for the paradoxical associations observed in our study. Participants with high dieting self-efficacy may be highly focused on avoiding specific ‘unhealthy’ foods, such as sugary snacks or fast food, which aligns with the situations described in the DIET-SE items. However, this may not necessarily translate into greater consumption of nutrient-dense, diverse foods. In fact, an excessive focus on restriction—without a concurrent emphasis on variety and adequacy—may lead to imbalanced eating patterns or cycles of restriction and overcompensation [64]. Thus, the observed negative relationship between DIET-SE scores and the dietary pattern index may reflect a restrictive, rather than holistic, approach to dietary control.

Another explanation might be that individuals with very high dieting self-efficacy might be those with a history of repeated dieting and weight cycling. Most recently, Phillips et al. (2025) showed that poor dietary quality scores in pregnant women, indicating higher nutrition self-efficacy, did not translate to higher dietary quality scores [65]. It is also possible that our measure of “dieting self-efficacy” is specific to weight management contexts (e.g., resisting overeating or following a calorie-restricted plan) and does not equate to self-efficacy in maintaining a balanced nutritious diet. This aligns with studies on metacognition in nutrition, which show that how people evaluate their knowledge and abilities matters; overestimating one’s nutritional competence can lead to choices that diverge from dietary recommendations [60]. Importantly, the dietary pattern score in this study reflects a tendency to avoid energy-dense, nutrient-poor foods such as fast food, processed snacks, and sugary beverages. This does not necessarily equate to a nutritionally adequate or diverse diet. Thus, individuals with high dieting self-efficacy may be selectively avoiding certain ‘unhealthy’ foods without engaging in balanced or nutrient-rich dietary practices. This nuance may help explain the association observed between higher self-efficacy and lower dietary pattern scores.

Our path analysis also supported a partial mediation model in which DIET-SE significantly mediated the relationship between nutritional knowledge and dietary pattern. WMNKQ had a positive effect on DIET-SE, which in turn predicted lower dietary pattern, while the direct effect of WMNKQ on dietary pattern remained significant but negative.

Path analysis extended this by demonstrating partial mediation. WMNKQ had a positive effect on DIET-SE (knowledge boosts confidence), and DIET-SE was linked to lower dietary pattern, mediating part of the relationship. Knowledge also had a remaining direct negative effect on dietary pattern, independent of self-efficacy. Future studies should explore this mechanism in depth; it may be that knowledge needs to be paired with behavioral skills training and realistic goal setting to translate into better dietary pattern. The direct path from knowledge to lower dietary pattern suggests that some knowledgeable individuals knowingly or unknowingly make suboptimal dietary choices, perhaps reflecting that weight-centric knowledge does not equal understanding of balanced nutrition. Studies have emphasized that knowing facts (“declarative knowledge”) has limited impact on diet unless individuals also have “procedural knowledge” [66–68]. Addressing the gap between knowledge and practice is crucial. Our findings are compatible with a more comprehensive approach to nutrition education [57]. This suggests that while knowledge is important, it may not translate into healthier eating habits unless individuals feel confident and motivated to act on that knowledge [69]. Studies recommend that practitioners in weight management programs assess both knowledge and self-efficacy at the start and employ behavior-specific strategies to improve them [70].

Limitations

Despite these findings, several limitations of the cross-sectional design must be acknowledged. First, because the data were collected at a single time point, we could not establish causality or the direction of the association between knowledge and self-efficacy. Behavioral and demographic variables together explained 11% of the variance in dietary pattern, which is typical of cross-sectional behavioral nutrition studies. Longitudinal relationships cannot be inferred, and the observed link is only associative. Second, both variables were measured using questionnaires, which introduced the possibility of response bias. BMI was derived from self-reported height and weight values, which may be subject to social desirability and recall biases. Moreover, BMI alone may not be the most robust indicator of adiposity or metabolic risk, as it does not account for muscle mass, fat distribution, or other relevant body composition variables. Participants with health-conscious dispositions might have rated their knowledge and confidence highly, increasing the correlation. Additionally, the nutrition knowledge test may not capture all aspects of practical knowledge for weight management, and the self-efficacy scale might not account for situational challenges (such as social or economic constraints) that affect confidence. Third, the external validity of our findings is constrained by the single-city sampling frame and predominance of younger adults. Moreover, although 35.8% of participants were excluded from IPAQ analyses, the comparative assessment (See Supplementary Table 2) showed predominantly negligible-to-small differences between included and excluded participants; only education (V = 0.117) and WMNKQ (r = 0.184) differed modestly. These findings suggest that any selection bias due to missing IPAQ data is likely limited, though IPAQ-related results should be interpreted with appropriate caution. Another limitation is that our adult sample may not fully represent a broader population. For example, if the sample had a higher education or socioeconomic status on average, the general level of nutrition knowledge and self-efficacy might differ from that of other groups. Moreover, the dietary pattern score was derived from a relatively brief 14-item FFQ, which, while capturing key obesogenic foods, does not fully reflect total dietary variety or nutrient adequacy. The resulting score likely captures a ‘low-processed/low-sugar’ pattern rather than comprehensive diet pattern. Consequently, it may underestimate healthy eating behaviors such as the intake of legumes, nuts, or unsaturated fats. These factors limit the generalizability of the findings and suggest caution in the interpretation of the results. Finally, the use of a cross-sectional survey precluded the observation of how knowledge and self-efficacy might change together over time or in response to interventions. These limitations highlight the need for further studies focusing on understanding the knowledge–self-efficacy relationship.

Conclusions

In conclusion, this cross-sectional study revealed a paradoxical relationship between knowledge, confidence, and healthy eating. While it might be assumed that knowing more about nutrition and feeling confident in weight management would lead to better diets, our data showed the opposite. Participants with the highest weight-management knowledge and dieting self-efficacy had lower overall dietary pattern. These findings highlight that knowledge alone is insufficient to ensure healthy eating. Socio-demographic factors (female sex, higher education, and income) were associated with greater nutrition knowledge and self-efficacy, underscoring the gap between understanding “what” to do and actually “doing.” Given the observed associations between sociodemographic factors and both nutrition knowledge and dieting self-efficacy, the findings suggest that broad, one-size-fits-all interventions may be insufficient. Instead, tailored educational programs that address specific population needs, particularly by sex and socioeconomic status, are essential to effectively promote healthier dietary behaviors. From a public health perspective, our study emphasizes the need for interventions that go beyond education; it is crucial to translate knowledge and motivation into actionable skills and sustainable habits. Improving dietary quality at the population level will likely require coupling nutrition education with practical support and addressing external barriers so that informed, confident individuals can turn knowledge into practice. We suggest that future studies should focus on bridging the knowledge-action gap.

Abbreviations

AMOS

Analysis of Moment Structures

BMI

Body Mass Index

CFI

Comparative Fit Index

CI

Confidence Interval

DIET-SE

Dieting Self-Efficacy

df

Degrees of freedom

FFQ

Food Frequency Questionnaire

FDR

False Discovery Rate

IPAQ

International Physical Activity Questionnaire

IPAQ-SF

International Physical Activity Questionnaire–Short Form

IQR

Interquartile Range

KMO

Kaiser–Meyer–Olkin measure of sampling adequacy

MET-min/week

Metabolic Equivalent of Task minutes per week

PCA

Principal Component Analysis

RMSEA

Root Mean Square Error of Approximation

R²

Coefficient of determination

SE

Standard Error

VIF

Variance Inflation Factor

ρ(rho)

Spearman’s rank correlation coefficient

Authors’ contributions

IMA was responsible for formal analysis, investigation, study methodology, supervision of the research process, and the writing of both the original draft and the revised manuscript. ED was responsible for formal analysis, investigation, study methodology, data visualization, and the writing of both the original draft and the revised manuscript. All authors read and approved the final manuscript.

Funding

None.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval was obtained from the Ethics Committee of the Faculty of Health Sciences, Atatürk University (Approval No: [2025/05/06], date: May 27, 2025), and the study protocol was implemented in accordance with the Declaration of Helsinki. All participants provided written informed consent prior to the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.