This study is based on several key hypotheses:

H1: We hypothesize that there is a positive relationship between body self-esteem and sexual self-esteem. This suggests that maintaining a positive body image correlates with higher satisfaction in sexual life.

H2: We postulate that there is a positive relationship between sexual self-esteem and marital satisfaction. This hypothesis suggests that marital satisfaction is correlated with positive sexual self-esteem.

H3: We propose that there is a correlation between marital satisfaction and body image.

H4: We hypothesize that there is a negative correlation between psychological distress and marital satisfaction. This suggests that higher levels of psychological distress are inversely proportional to marital satisfaction.

H5: We suggest that there is a positive relationship between regular physical activity and marital satisfaction. This hypothesis posits that consistent engagement in physical activity may lead to increased marital satisfaction. This proposition is supported by evidence indicating a likely connection between physical activity and marital contentment, as well as the known psychological benefits of exercise.

Our study targeted adult men and women. The research was conducted in most Tunisian cities, starting with deploying a questionnaire on the Google Drive platform. Carried out online from April 21, 2021, to November 11, 2021, the survey was also shared across various social media platforms. Among the initial 520 received responses, 443 were retained after eliminating incomplete or aberrant information-containing questionnaires. Consequently, our study cohort comprised 443 participants, with an average age of 48 years.

Before answering the questionnaire, each participant provided informed consent to participate in the study. The entire research was approved by the Ethics Committee of the Faculty of Medicine of Sousse on March 21, 2022, with reference number No. 113 (Ref: CEFMS 113/2022).

Our study population consisted of 443 participants, predominantly female (52.4%), resulting in a sex ratio of 0.91. The average age of the participants was 42.84 ± 8.09 years, ranging from 24 to 77 years old, with 57.1% of participants being over 40 years old. Nearly all participants (n = 427) were married, with marriage duration ranging from 1 to 45 years and an average of 13.22 ± 8.09 years; 74.7% had at least 2 children.

Physical Activity Level: The assessment of physical activity level was conducted using the Ricci and GAGNON questionnaire, modified by Laureyns and Séné. This questionnaire determined a score ranging from 18 to 35 points. It comprised nine questions rated from 1 to 5 points, evaluating domestic, leisure, work-related activities, as well as sedentary behavior. Participants were classified into three groups based on their scores (Less than 18: Inactive, Between 18 and 35: Active, more than 35: Very Active) (48).

Body Self-Esteem was measured using the French version of the Body Esteem Scale, a multidimensional self-questionnaire. It consisted of 23 items rated on a five-point Likert scale, distributed into three subscales: “Appearance,” “Attribution,” and “Weight.” Higher scores indicated a better body esteem and a positive appreciation of appearance. Some items were reverse scored (4, 7, 9, 11, 13, 17, 18, 19, and 21). Various validation studies (49–53). Demonstrated a satisfactory internal consistency across the three subscales. The scale was validated with adolescents and young adults. The test-retest reliability (n = 151) for the “satisfaction with weight” subscale was 0.76; 0.69 for the “satisfaction with general appearance” subscale, and 0.63 for the “desire for change and negative effects associated with general appearance” subscale. Cronbach's alphas for the three subscales from the exploratory factor analysis of the BES (satisfaction with weight; satisfaction with general appearance; desire for change and affects associated with general appearance) were 0.90, 0.83, and 0.80, respectively (53).

Sexual Self-Esteem was assessed using a subscale of the Multidimensional Sexuality Inventory (54). IMS: Snell et al. translated by Ravart et al. (52). The IMS evaluated twelve spheres of sexuality using 60 Likert-type items on a five-point scale ranging from 1 “does not characterize me at all” to 5 “characterizes me a lot.” The authors reported good internal consistency of the subscale, with a Cronbach's alpha of 0.87, and good factorial validity of the IMS (52). Summing up the items of this subscale yielded a sexual self-esteem score. A higher score indicated better sexual self-esteem.

Psychological distress was measured using the Kessler Scale (K10) developed by Kessler and colleagues (45). According to the authors' conceptualization, psychological distress is a non-specific psychological state characterized by feelings reflecting depressive or anxious mood. A French translation of this instrument used in the Canadian Community Health Survey (CCHS) was employed in this study. Instructions highlighted indicating how often the participant felt, for example, sad or depressed in the last 30 days. Five response choices were offered: never, rarely, sometimes, most of the time, and all the time, associated with scores ranging from 0 to 4. Therefore, obtained results varied between 0 and 40, with a higher score indicating greater psychological distress.

Marital Satisfaction Level was measured using the Dyadic Adjustment Scale (DAS-4), developed by Sabourin, Valois, and Lussier (55). This instrument is the abbreviated French version of the Dyadic Adjustment Scale developed by Spanier (1976) (DAS-32), shortened to 4 items (DAS-4) (55). These 4 items were: (1) thinking about separation or divorce; (2) trusting the partner; (3) being happy in the relationship; (4) sharing positive elements. The DAS-4 underwent five studies with 8,256 married or dating participants. It proved its ability to measure the unidimensional aspect of marital satisfaction and predict couple stability while avoiding social desirability bias. Apart from being quick to complete, this scale is an effective measure for predicting couple dissolution (55). Each question was evaluated using a 5 or 6-point Likert scale depending on the different items. The sum of the four items constituted the satisfaction score. A high score indicated increased marital satisfaction. A score below 13 indicated a distressed spouse. It had good internal consistency, with a Cronbach's alpha of 0.89 for women and 0.78 for men (55).

The data analysis relies on using the PLS-SEM (Partial Least Squares Structural Equation Modeling) method with the software SmartPLS 3.3.2. This approach was chosen to assess the complex relationships between observable variables and latent variables, implementing Principal Component Analysis (PCA) to describe these associations. During model estimation, the PLS approach employs least squares regression methods. The goal of PLS modeling is to maximize the explained variance of the dependent latent variable. For instance, in modeling, it's possible to obtain empirical measures that highlight the correlations between indicators and constructs, as well as interactions between latent components (56, 57). Therefore, our aim is to explore theoretically established links based on field-collected data. In this perspective, empirical measures enable us to assess how well the theory aligns with data acquired through experience. The PLS method offers us the opportunity to evaluate the predictive capability of the model and determine its overall quality (56, 57).

The SmartPLS program was utilized for model estimation (v. 3.3.2). Using the PLS approach, it enables the description of complex associations between observable variables and latent variables through a statistical method known as Principal Component Analysis (PCA).

In our study, we conducted tests on both the measurement model and the structural model. These two sub-models are used together to compute the structural equations (56, 58). Utilizing the measurement model allows us to determine the nature of the relationships between observable indicators and latent variables. As for the structural model, its role is to identify causal links between latent variables when studying them. The primary objective is to establish whether certain latent variables influence others in the model, either directly or indirectly (59).

Following the PLS methodology, start by testing measurement and structural models. Then, check the reflexive measures of the dependent variable. Examine model hypotheses next, using the dependent variable formatively. This helps evaluate impacts on components and the overall effect. The measurement model assessment focuses on reliability, composite reliability, convergent, and discriminant validity.

In a model constructed using SmartPLS, to assess construct reliability, we estimate it using three coefficients: Cronbach's alpha coefficient, composite reliability (CR) with Joreskog's rho, and Rho-A coefficient (Dijkstra-Henseler's Rho).

For acceptable internal consistency analysis, it's recommended that the Cronbach's alpha reliability coefficient should be above 0.6. The Joreskog's rho composite reliability index should be higher than 0.7 (60). The Rho-A reliability coefficient (61) falls between Cronbach's alpha and composite reliability. It's generally considered a good indicator for reliability assessment, with a recommended threshold of 0.70 (59).

Table 1 shows that the Cronbach's alpha coefficient ranges between 0.844 and 0.951. These values suggest that the indicators can consistently measure the constructs. Additionally, the values for Rho-A composite reliability were all adequate, ranging from 0.844 for marital satisfaction to 0.959 for the Distress construct (see Table 1).

Convergent validity of indicators provides information on the extent to which they measure the constructs they are supposed to measure, meaning the circumstance in which a measure of a construct is positively related to another measure of the same construct.

The average variance shared between a concept and its measures is called the AVE. Indicators linked to a specific construct have their squared loadings calculated as the overall mean of squared loadings (the total squared loadings divided by the number of indicators), defined as such (62).

The average extracted variance (AVE) and item loadings are examined when analyzing the convergent validity of a measure using the PLS technique (62).

The average variance shared between a concept and its measures should be higher than the average variance shared between the construct and other latent variables in the same model, if possible (63).

Calculating AVE is included in the analytical program of the PLS environment. In this case, the concept explains more than half of the variation of its indicators, with an AVE value equal to or greater than 0.5. Conversely, an AVE lower than 0.50 suggests that there are more errors in the indicators than the average variance explained by the constructs, which is the case for most indicators. Therefore, an AVE value greater than or equal to 0.50 is considered acceptable in general (62).

Table 1 shows that all AVE values are above the acceptable threshold of 0.5, indicating that the convergent validity of the measurement model has been validated.

The discriminant validity of a concept concerns the uniqueness of the construct, meaning if the phenomenon claimed by a construct is distinct and not reflected by other latent variables in the model (62).

The Fornel-Larcker criterion tests the second discriminant validity of a concept by comparing the square root of AVE values with correlations of latent variables (60).

In Table 2, the discriminant validity of the Fornel-Larcker criteria and cross-loadings is examined, and the results show that discriminant validity is satisfactory.

An investigation by Henseler et al. (62), based on their Monte Carlo simulation, introduced a new criterion called HTMT that can be used to assess the effectiveness of a treatment. According to Henseler and colleagues,for discriminant validity to be achieved, the HTMT score must fall within the confidence interval values of −1 and 1.

Using the Fornell-Larcker criteria and the Heterotrait-Monotrait Ratio of correlations, discriminant validity can be determined by examining the cross-loading between the study components (HTMT). Therefore, for a valid discriminant construct, its loadings on itself should be high while its loadings on other constructs are low (see Table 3) (64).

Along the diagonal of the correlation matrix, square roots of AVE coefficients are displayed. The square roots of AVEs should be higher than their greatest association with all other constructs to be considered significant (60).

The structural model focuses on latent variables and indicates the theoretically established relationship that exists between the provided data on both the input and output sides of the equation. The purpose of the analysis, based on latent variables, is to predict output training data based on the input data in the model, using the data from latent constructs. In another formulation, the structural model is used to demonstrate one or more dependent relationships that are conceptually linked to the idea of the considered model. An evaluation technique for the structural model was presented by Hair and colleagues, which included several consecutive steps that could all be completed in a single session (60).

When two or more independent variables are highly associated with each other, multicollinearity is present in the data. Multicollinearity in ordinary least squares regression inflates standard errors, renders tests of variation of an independent variable inaccurate, and prevents the researcher from determining the relative importance of one independent variable over another (65).When the variance inflation factor (VIF) coefficient is greater than 4, multicollinearity might be an issue.

We found that the PLS method generated VIF values ranging from 1.919 for item APQ3 to 4.046 for item SS1, depending on the indicator we examined. With the exception of SS1, all numbers are below 4, demonstrating that the multicollinearity problem does not exist (see Table 4).

Using the values of R-squared (R²) for endogenous variables, the predictive ability of a specific model or construct, as well as the derivation of the standard path coefficient for each connection between exogenous and endogenous variables, are assessed in PLS analysis. The significance of R² values in SmartPLS is similar to interpreting R² values obtained from multiple regression analysis. In the construct, R² values reflect the extent to which the variation of the construct can be explained by the model. The degree of variation explained by the external variable in its endogenous counterpart is shown by the R² statistic (66). It represents a depiction of the model's parameter quality (54). However, Adjusted R-squared is a modified form of R-squared in which the number of predictors in the model has been corrected. The updated R² increases only if the additional term improves the model more than expected. It decreases when a predictor enhances the model by an amount lower than what would have been anticipated by chance.

According to Chin (66), an R² of 0.67 is considered high, 0.33 is considered moderate, and 0.19 is considered low.

Psychological distress has an R² value of 0.57, indicating that the combined effect of sedentary behavior and marital satisfaction can lead to a change in psychological distress by 22%, while the adjusted R² value for psychological distress is 21.8%. Both R² and adjusted R² values show weak effects. Their complexity is determined by the model's complexity and research discipline (67).

Similarly, the combined impacts of AEC, SEL, and EVAL-CAUSE on company reputation represent 65.8% of the variance in this variable.

On the other hand, psychological distress, sexual life satisfaction, physical activity, and sedentary behavior managed to explain 49.6% of marital satisfaction. However, body appearance explained 28.7% of sexual life satisfaction (refer to Table 5).

Cohen's f² values have also been calculated to emphasize the significance of the pathway. f² is a measure of the magnitude of an effect caused by a pathway.

f² can be calculated using the following equation: f² = R² (included) - R² (excluded)/1 - R² (included).

To interpret the effect size, Cohen recommends using the following criteria: small effect size 0.02; medium effect size 0.15; and large effect size 0.35 (68).

The results are presented in Table 6.

R-squared values gauge the model's predictive accuracy, while the predictive index Q² indicates the accuracy in prediction, often referred to as “Stone-Geisser's T² value” (60, 66). A Q² value above zero signifies the predictive relevance of the relationship for that specific construct (56, 60).

Q² values can be determined using the Blindfolding process for an omission distance D that has been specified. “Blindfolding” is a resampling approach where each data point in endogenous construct indicators is omitted, and parameters are estimated using the remaining data points from the remaining data (62, 69).

Assessing the effect size of Q² is important for determining whether exogenous constructs have a significant influence on the endogenous construct. It's crucial to evaluate both the significance of the pathway being considered and how much it changes the explanatory power of the endogenous concept (70). When used in isolation, the path coefficient doesn't provide any information about the significance of the influence of exogenous latent variables on endogenous constructs.

During the effect size estimation process (f²), it's advisable to consider the predictive relevance (Q²) of the model, which was done to determine its predictive ability.

The Stone-Geisser test, a non-parametric test, can be used to determine the predictive significance of the research design (66). As a result of the outcome extracted from cross-validation, it's possible to estimate the predictability of endogenous constructs and, consequently, the overall quality of the model. According to Hair et al. (57), Q² not only analyzes model values but also estimates the values and parameters of the model in question.

For the SmartPLS model presented in this study, the Q² effect sizes demonstrate the predictive relevance of coefficients, all of which are greater than 0 (see Table 7).

According to Henseler and Sarstedt (62), the Root Mean Squared Error (RMSE) can be used as a goodness-of-fit statistic in the PLS-SEM approach. It represents the difference between the observed correlation and the expected correlation expressed as a root mean squared error. Hence, the average size of the discrepancies between observed and anticipated correlations can be used to assess model fit criteria in an absolute sense. A good fit is defined as a result less than 0.10 and preferably less than 0.08 (60).

The model generated a Root Mean Squared Error (RMSE) with a reasonable value of 0.045, as indicated in Table 8.

The execution of the PLS algorithm and bootstrapping calculations in the SmartPLS software provided the path coefficient of these relationships, indicating the strength of the relationships, along with the P-value, which can be used to determine if the relationship is statistically significant or not (refer to Table 9 and Table 10).

For the direct relationships, except for the effect of sedentary behavior on marital satisfaction, all links exhibit highly significant effects.

For the indirect effects, significant relationships have been identified between physical appearance and marital satisfaction. Similarly, sedentary behavior showed an indirect effect on marital satisfaction.

The discussion of the hypotheses formulated in our study reveals significant results that shed light on the complex relationships between various psychological and relational factors (see Supplementary Figure 1). First, hypothesis H1 suggests a correlation between body self-esteem and sexual self-esteem. Our study confirms this hypothesis by establishing a link between body self-esteem and sexual satisfaction. Individuals with a positive body image demonstrated higher sexual satisfaction, which is consistent with previous research: For example, studies on body image suggest a correlation between body perception and sexuality. Thus, the literature highlights that among young women, a negative body image is often linked to sexual problems, sexual dysfunction, and sexual dissatisfaction (71).

Various authors claim that body image has an impact on sexual activity (72–74). Our findings align with the results of Hannier et al. (75), which highlight a relationship between women's sexual self-esteem and their satisfaction with their bodies., in a survey study published in a women's health and fitness magazine, women aged 14–74 reported that body image satisfaction was associated with greater comfort with their bodies during sexual activity, higher frequency of sexual behaviors, including increased initiation of sexual activity by women, and higher orgasm frequency (76). Similarly, among a non-clinical sample of 187 college female students, Hoyt and Kogan found that women dissatisfied with their sex lives and/or dating lives experienced more body image dissatisfaction compared to women satisfied with their sex lives (77). The authors suggested that these results could be attributed to discomfort in intimate situations involving female nudity, such as undressing in front of one's sexual partner, especially among women with low body self-esteem. Furthermore, studies on body image suggest a correlation between body perception and sexuality.

However, some studies have failed to establish this correlation. For instance, in two studies involving middle-aged women, no significant association was observed between body image and sexual satisfaction, especially after adjusting for other variables such as menopausal symptoms and mood (41).

Hypothesis H2, proposing a connection between sexual self-esteem and marital satisfaction, finds support in our findings. Our results suggest that individuals with positive sexual self-esteem tend to experience higher levels of marital satisfaction, which is consistent with prior research (78–81). Longitudinal studies have consistently reported that fluctuations in sexual satisfaction are closely linked to fluctuations in marital satisfaction (34, 39). Furthermore, a multitude of studies have demonstrated a positive association between sexual satisfaction and marital satisfaction (29–35).

Numerous studies have highlighted a significant and positive association between sexual satisfaction and the quality or satisfaction of the marital relationship (33, 82–84). Many investigations exploring the link between sexual and marital satisfaction have indicated that these two variables significantly predict each other (85–87). Brezsnyak, Fielder, and Klemer suggests that “A good sexual adjustment usually, but not always, requires a fairly good total marriage relationship” (85).

However, some researchers have suggested that there may not always be a direct relationship between marital satisfaction and sexual satisfaction (88, 89). Pazak and Berg-Cross propose that sexual dissatisfaction may occur even within otherwise happy marriages. Additionally, it's noted that sexually satisfied spouses may still experience dissatisfaction within their marriages. Similarly, Colebrook Seymour, found no significant relationship between marital satisfaction and sexual satisfaction (90).

Furthermore, our study confirms hypothesis H3, which highlights a correlation between marital satisfaction and body appearance. We observed a significant indirect effect between body appearance and marital satisfaction, suggesting that this influence is mediated by other factors not explicitly measured in our study. Individuals with higher marital satisfaction also exhibited a more positive perception of their body appearance, which reinforces previous findings (91–93).

Several studies support the notion that dissatisfaction with one's body is correlated with dissatisfaction in marital relationships, and vice versa, even after adjusting for variables such as age, body mass index, self-esteem, and gender (77, 94, 95). Specifically, women's bodydissatisfaction is associated with dissatisfaction in their marital relationships for both partners. This implies that not only are women who view themselves negatively less satisfied with their marital relationship, but their partners are also less satisfied (96). However, there is only one study that does not find a significant link between these variables (97).

In summary, our findings contribute to the growing body of literature suggesting a complex interplay between body image and marital satisfaction. While our study provides evidence of a correlation between marital satisfaction and body appearance, further research is needed to elucidate the underlying mechanisms and potential mediating factors in this relationship.

The validation of hypothesis H4, which suggests a correlation between psychological distress and marital satisfaction, is reinforced by our findings, consistent with existing research. Our results align with studies indicating that higher levels of psychological distress are significantly associated with lower marital satisfaction, as documented in previous literature (45, 98). Numerous studies have demonstrated that individuals facing heightened psychological distress tend to report lower quality in their marital relationships (42–44). Furthermore, our findings correspond with a study conducted by Karen (99), which highlights a negative association between marital satisfaction and psychological distress, emphasizing the importance of marital quality as a predictor of mental health status among married African Americans. This convergence of evidence underscores the significance of considering marital dynamics in comprehending psychological well-being.

Finally, our study supports hypothesis H5 positing a relationship between sedentary behavior and marital satisfaction. We observed a significant indirect effect of sedentary behavior on marital satisfaction, suggesting that this relationship is influenced by intermediary variables not accounted for in our initial model. Thus, no previous study had explicitly highlighted this link between sports and dyadic adjustment. In reality, what our results support is rather an indirect link, supported on the one hand by numerous theories linking sports practices to better psychological well-being (11), and on the other hand by studies associating low psychological distress with better marital satisfaction (42–44).

These findings offer valuable perspectives for targeted interventions aimed at improving relationship quality. However, it is essential to acknowledge the limitations of our study, particularly its cross-sectional nature, and to encourage future research to deepen the understanding of these complex relationships.

Despite the valuable insights provided by our study, several limitations must be acknowledged. Firstly, the cross-sectional nature of our research design limits our ability to establish causal relationships between the variables studied. Future longitudinal studies are warranted to elucidate the temporal dynamics of these associations.

Secondly, our reliance on self-report measures introduces the potential for response bias and social desirability effects, which may influence the accuracy of our findings. Incorporating objective measures and observational data could enhance the validity of future investigations.

Furthermore, our study focused primarily on heterosexual couples, limiting the generalizability of our findings to diverse populations. Future research should strive to include more diverse samples to ensure the applicability of our findings across different demographic groups.

Moreover, our study did not explore the influence of cultural and societal factors on the variables under investigation. Cultural differences in attitudes towards sexuality, body image, and marital dynamics may significantly impact the observed associations. Therefore, future studies should consider cultural context in their analyses.

In conclusion, while our study contributes to a deeper understanding of the complex interplay between psychological, relational, and behavioral factors in marital satisfaction, further research is needed to address the limitations and provide more comprehensive insights into this important area of inquiry.