Beyond Scars: Development and Validation of a Four-Dimensional Body Image Score for Women with Breast and Gynecological Cancers - A Tunisian Multicenter Study

2.1 Study design and setting

This was a cross-sectional study conducted over a two-month period from March 1, 2025, to April 30, 2025. The study took place in 6 Tunisian healthcare facilities of different levels:

Level 2:

Level 3:

The study was conducted among women with breast and gynecological cancers using a questionnaire.

A self-administered paper questionnaire was used to collect data in a private room at the outpatient clinic. Participants completed questionnaires independently, with research staff available outside the room to clarify item wording if requested. Staff did not interpret items or suggest responses.

2.2 Study population

Patients meeting the following criteria were included:

Inclusion criteria

Exclusion criteria

2.3 Development of the composite score

Further methodological information on the EFA and CFA procedures described in this article are available at this link: https://doi.org/10.7910/DVN/ZGECTK.¹⁷

2.3.1 Methodology – Scoring procedure for the 4D-BISC

The development of the 4D-BISC involved several stages: a literature review, question design, creation of a scoring system (question scoring), and pre-testing.

The composite score was developed based on a comprehensive review of validated instruments in the field of body image and female sexual health, including the BESAA, the Body Image Scale (BSI), and the FSFI. The aim was to assess the impact of medical diagnosis or treatment on women’s perception of their body image.

The final version included 16 items; each rated on a 5-point Likert scale ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). Higher scores indicate a greater negative impact on body image.

2.3.2 Confirmatory Factor Analysis (CFA)

A confirmatory factor analysis revealed a four-factor structure, reflecting the following dimensions:

Each factor demonstrated acceptable to excellent internal consistency, with Cronbach’s alpha coefficients (α) ranging from 0.68 to 0.92.

2.3.3 Scoring procedure

For each participant, the following scores were computed:

To facilitate comparison across dimensions, all scores were transformed to a 0–100 scale using the following formula:

Higher transformed scores reflect greater negative body image impact.

Data processing was conducted using IBM SPSS Statistics (Version 26.0; IBM Corp, Armonk, NY, USA) and standardized syntax was applied to ensure consistency in scoring across the entire sample.

2.4 Data collection

Data collection was conducted during hospital consultations for each included woman followed for gynecological and breast cancers.

French and Arabic versions of the questionnaire were specifically developed for the needs of this study.

Data were collected through self-administered questionnaires in a private setting, with assistance available if needed. The primary goal of the questionnaire was to assess body image. Demographic (marital status, education level, socioeconomic level, ethnicity …) and clinical data (primary cancer diagnosis, time since diagnosis, current treatment modality [chemotherapy/radiation/surgery/hormonal therapy], treatment history …) were also recorded.

To preserve participant anonymity, no personally identifiable information, such as names or contact details, was collected.

2.5 Statistical methods

Descriptive statistics were used to summarize participant characteristics and score distributions. Categorical variables (e.g., cancer type, marital status) are summarized using frequencies and percentages. Continuous variables (age, scale scores) are reported as means ± standard deviations (SD) if normally distributed, or medians and interquartile ranges (IQR) if non-normal.

Internal consistency of the newly developed scale (total score and identified subscales) was assessed using Cronbach’s alpha. An alpha coefficient ≥ 0.70 was deemed acceptable for research purposes.¹⁸

Convergent validity was evaluated using Spearman’s rank-order correlations (ρ) between the new scale’s total score and validated measures: the FSFI¹⁵ and BESAA.¹⁶ Moderate-to-strong positive correlations (ρ ≥ 0.40) were hypothesized a priori based on conceptual overlap.

All analyses were performed using IBM SPSS Statistics (Version 26.0; IBM Corp, Armonk, NY, USA) for data management, descriptive statistics, correlation analysis, and non-parametric tests.

Exploratory Factor Analysis (EFA) and Cronbach’s alpha computation were conducted in JASP (Version 0.18.1; JASP Team, Amsterdam, Netherlands).

Prior to analysis, data were screened for missing values, outliers, and normality. Missing item-level data on the new scale were handled using pairwise deletion during EFA and reliability analysis. Total/subscale scores were computed only for participants with ≤20% missing items on the respective scale; otherwise, scores were treated as missing. Continuous variables were assessed for normality using Shapiro-Wilk tests (p < .05 threshold) and visual inspection of Q-Q plots.

The analysis followed several steps^19,20:

2.5.1 Assessment of suitability for exploratory factor analysis:

The adequacy of the data for factor analysis was evaluated using the Kaiser-Meyer-Olkin²¹ (KMO) measure and Bartlett’s Test of Sphericity. A KMO value above 0.80 was considered very good, while a significant Bartlett’s test (p < .05) indicated that correlations between items were sufficient to proceed with factor extraction.

2.5.2 Exploratory factor analysis:

EFA was performed on the new scale’s items using Principal Axis Factoring (PAF) to extract common variance. Factor retention was determined by:

2.5.3 Factor retention criteria:

Items with pattern matrix loadings ≥0.40 on a primary factor and cross-loadings <0.30 were retained. Items with uniqueness >0.90 or communality <0.20 were reviewed for conceptual misfit.

2.5.4 Measure of Sampling Adequacy (MSA):

The MSA was calculated for each item to assess its individual contribution to the overall factor model. Values above 0.50 were deemed acceptable.²²

2.5.5 Interpretation of factors:

Factors were labeled based on items loading ≥0.50.

Differences in scale scores across categorical clinical variables (e.g., cancer type [Grouped as: Breast vs. Gynecological vs. Colorectal], active treatment [Yes/No]) were tested using:

2.6 Ethical considerations

The study protocol was approved on 13 February 2025 by the institutional ethics committee of Charles Nicolle Hospital, Tunis, Tunisia before conducting the study with approval number FWA 00032748- IORG0011243.

Verbal informed consent was selected instead of written consent due to the specific context of the study population.

Data were collected from population considered vulnerable due to both their cancer diagnosis and the psychological impact of treatment. Requesting written consent in this setting risked creating additional stress or discouraging participation, particularly among women with lower literacy levels.

This approach was reviewed and approved by the Institutional Ethics Committee of Charles Nicolle Hospital as the study involved minimal risk, collected no identifying or sensitive information, and ensured complete anonymity of responses.

All participants received detailed information regarding the study’s objectives. They were clearly informed of their right to withdraw from the study at any point without having to justify their decision. Confidentiality of all participant data was strictly preserved throughout the study.

3.1 Exploratory factor analysis

An EFA was conducted to examine the underlying structure of the composite score following oncological treatment. All comprehensive details of the EFA/CFA conducted in this study can be found at the following link: https://doi.org/10.7910/DVN/ZGECTK.¹⁷

The data showed strong suitability for factor analysis (KMO = 0.854; Bartlett’s test χ² = 872.986, df = 153, p < .001).

Two interpretable factors were extracted using Promax rotation, accounting for 43.8% of the total variance:

Items Q6 “Since your diagnosis, has your perception of your body evolved?” and Q18 “Have you discussed your concerns about body image with your care team?” showed weak loadings and high uniqueness, suggesting they are poorly represented by the extracted factor structure.

All other items had acceptable Measures of Sampling Adequacy (MSA > 0.5), with the highest for items Q5 “Do you feel that your body is weaker or less capable than before?”, Q9 “Does your body image affect your participation in social activities”, and Q15 “Do you feel anxious or depressed because of the physical changes related to your illness or treatment?”.

Following the removal of problematic items 17 “Have you received any psychological or aesthetic support to help you accept the changes in your physical appearance?” and 18, we conducted a CFA to test the hypothesized 4-factor structure comprising 16 items. The model demonstrated acceptable to good fit across multiple indices. While the significant χ² statistic (χ² (98) = 165.087, p < .001) suggested some model-data discrepancy, this is expected with larger samples. More importantly, the CFI (0.914) and SRMR (0.069) met excellent fit thresholds, while the TLI (0.895) and RMSEA (0.082 [0.060-0.104]) approached or met acceptable levels. The GFI (0.830) indicated moderate fit, representing a notable improvement over previous specification.

Reliability analysis revealed strong internal consistency for three factors:

While Appearance Concerns showed marginally acceptable reliability (ω = 0.704, AVE = 0.377), we retained this factor due to its theoretical importance in the measurement model.

One persistent concern emerged regarding item 6, which demonstrated poor factor explanation (R² = 0.033) and non-significant loading (β = -0.188, p = 0.089). This item had shown consistent problems in prior analyses, suggesting it may not adequately capture the intended construct.

These results support the validity of the refined 4-factor structure while suggesting potential for further improvement through removal of item 6. Such modification would likely enhance the model’s psychometric properties by improving factor coherence and increasing average variance extracted, while maintaining comprehensive coverage of the theoretical domains. The current solution nevertheless represents a statistically and conceptually sound measurement model for assessing body image concerns in this population.

The analysis revealed strong and statistically significant correlations (all p < .001) between all four body image factors and the global score. Appearance concerns (F1) showed particularly strong associations with Self-esteem (F4; ρ = .639) and Social exposure (F2; ρ = .607), suggesting these dimensions share considerable variance in how individuals experience body image. The Sexual desirability factor (F3) demonstrated slightly lower but still substantial correlations with other factors (range: ρ = .465-.597), indicating it captures both related and unique aspects of body image.

The global 4D-BISC showed the strongest relationships with all individual factors (ρ = .784-.830), with the highest correlation occurring with Social exposure (F2; ρ = .830).

3.2 External validation

The correlation analysis revealed significant associations between participants’ body image perceptions and their sexual functioning across multiple domains ( Table 4).

All four body image factors - Appearance Concerns (F1), Social Exposure (F2), Sexual Desirability (F3), and Self-Esteem (F4) - demonstrated moderate to strong positive correlations with various aspects of sexual function (all p < .001). These consistent patterns suggest that greater body image concerns tend to co-occur with more reported sexual difficulties.

The strongest positive relationships emerged for sexual desire, which showed particularly robust associations with both the Sexual Desirability factor (ρ = .706) and the global 4D-BISC (ρ = .747). This indicates that women’s perceptions of their sexual attractiveness may be especially relevant for maintaining sexual interest.

Similarly, the global 4D-BISC showed strong correlations with sexual pain (ρ = .655) and overall sexual function as measured by the FSFI total score (ρ = .693), suggesting that body image disturbances may contribute to both physical and psychological aspects of sexual dysfunction.

Notably, the BESAA showed an inverse pattern of relationships. Lower body esteem scores (reflecting more negative body appraisal) were significantly associated with greater body image concerns across all factors (ρ = -.386 to -.470) and poorer sexual function in most domains (ρ = -.201 to -.295). The exception was sexual satisfaction, which did not show a significant correlation with body esteem (ρ = -.083, p = .409), possibly indicating that satisfaction depends on additional factors beyond body image alone.

3.3 Comparative study

Table 5 presents the comparative analysis among the study population.

Illiterate patients reported the highest levels of appearance-related concerns (median = 50) and social exposure anxiety (median = 60), though these education-related differences only approached statistical significance (p = 0.08 for appearance). Those with primary education showed the most negative body image profiles overall, particularly regarding sexual desirability (median = 25).

Socioeconomic status significantly impacted self-esteem (p = 0.024), with low-income patients maintaining better self-image (median = 40) compared to middle and high-income groups.

Marital status emerged as an important factor, particularly for sexuality-related body image (p = 0.018). Widowed and single patients reported the most severe concerns about sexual desirability (median = 0), while married participants showed the most positive body image profiles overall (global median = 38.8). These results suggest that relationship status may serve as either a protective or risk factor for body image disturbances.

The most pronounced differences appeared when examining cancer types. Breast cancer patients demonstrated significantly poorer body image across all domains (global p = 0.003), with particularly low scores for social exposure (median = 26.7, p = 0.006) and sexual desirability (median = 30, p = 0.002). In contrast, ovarian cancer patients reported the highest levels of sexual concerns (median = 60, p = 0.001), while endometrial cancer patients showed relatively better self-esteem (median = 46.7, p = 0.013). These cancer-specific patterns highlight the need for tailored interventions addressing the unique body image challenges associated with different cancer diagnoses.

Patients with a family history of cancer showed significantly better body image across nearly all factors (global p = 0.008), particularly regarding sexual desirability (median = 50 vs 35 for those without family history, p = 0.009). This finding suggests that prior experience with cancer in the family may foster more adaptive body image perceptions. Urban-rural differences approached significance for sexual concerns (p = 0.077), with urban residents reporting higher scores, possibly reflecting differing cultural norms around body image.

The correlations between body image factors and patient characteristics are summarized in Table 6.

Regarding age, all correlations with body image dimensions were weak and negative, but none reached statistical significance. This indicates that older patients did not significantly differ in their perception of appearance, exposure, sexuality, self-esteem, or global body image compared to younger patients.

For the number of pregnancies, correlations were negligible and non-significant across all dimensions, suggesting that parity did not influence body image perception in this sample.

Similarly, the number of children showed no meaningful correlations with body image scores, indicating no significant relationship between parental status and body image.

In contrast, time since treatment completion showed a significant positive correlation with the sexuality dimension (r = 0.207, p = 0.038), suggesting that sexual well-being improved with longer time elapsed after treatment. No significant associations were found with appearance, exposure, self-esteem, or the global body image score.

4.1 Comparative analysis of body image assessment in breast cancer patients

Our findings demonstrate both convergence and divergence with existing literature on body image assessment in breast cancer populations.

Like Baxter et al. and Gonçalves et al., found that body image is a multidimensional construct requiring comprehensive assessment.^23,24 However, while their Body Image after Breast Cancer Questionnaire (BIBCQ) identified six factors, our analysis revealed four primary dimensions (Appearance concerns, Social exposure, Sexual desirability, and Self-Esteem), suggesting potential cultural or methodological differences in factor structure.

The reliability coefficients in our study (ω = 0.704-0.918) compare favorably with those reported in multiple validations of the BIBCQ (α = 0.77-0.87) and the Body Image, Sexuality, and Breast Cancer questionnaire (BAS-BC) (α = 0.91), indicating similarly robust measurement properties. Notably, our Sexual Desirability factor showed particularly strong reliability (ω = 0.918), paralleling findings from Zhou et al.²⁵ where sexuality-related subscales demonstrated high internal consistency (α up to 0.88).

Our results showing significant body image differences by cancer type (particularly poorer outcomes for breast cancer patients) align with Guedes et al.’s report of 74.8% body image dissatisfaction post-treatment.²⁶ However, we extended these findings by identifying specific vulnerabilities in sexual body image (median = 30) and exposure concerns (median = 26.7) among breast cancer patients, providing more nuanced clinical targets.

The significant improvement in sexuality-related body image over time post-treatment (r = 0.207, p = 0.038) in our study complements Sema Koçan et al.’s systematic review findings that body image satisfaction improves over time after mastectomy.²⁷ However, our identification of age-related trends (older patients reporting fewer appearance concerns) adds new dimensions to their observation that younger women face greater body image challenges.

Unlike Derbis and Czerwik’s Polish adaptation that found limited BIBCQ correlations with sexual functioning, we observed robust associations between our Sexual Desirability factor and time since treatment.²⁸ This discrepancy may reflect cultural differences in sexuality expression or our focus on temporal rather than functional aspects.

4.2 Study strengths and limitations

While our study shares the commonly cited limitation of limited sample size, as highlighted by Zhou et al.,²⁵ it advances existing research in several meaningful ways:

Moreover, our findings address key gaps outlined in the review by Muzzatti and Annunziata.²⁹ Whereas they observed that most existing tools are narrowly tailored to breast cancer populations in English-speaking countries, our study demonstrates broader cultural and linguistic applicability across different cancer types in a non-English-speaking context.

Additionally, the four-factor structure proposed in our analysis offers a more parsimonious and clinically practical alternative to the eight different instruments reviewed in their study, thus contributing to simplified and streamlined assessments in routine oncology care.

4.3 Clinical implications

Our findings underscore the importance of targeted interventions to address sexual body image concerns, particularly among breast cancer patients in the early post-treatment phase, when vulnerability appears to be highest.

The validated dimensions identified in this study offer a theoretical and practical framework for culturally adapted assessments, helping to overcome limitations previously noted in cross-cultural validation studies.

The observed temporal patterns in body image adjustment provide critical insights into the optimal timing for intervention delivery, supporting the development of time-sensitive psychosocial support programs.

4.4 Recommendations for future research

Future research should seek to integrate our dimensional approach with existing comprehensive psychosocial evaluation frameworks, as proposed in prior studies, to facilitate the development of more personalized and context-sensitive intervention strategies.

Given the robust psychometric properties of our 4D-BISC, further studies are warranted to pursue its cross-cultural adaptation and validation. This would help address the ongoing need for diverse, culturally appropriate assessment tools in psycho-oncology, as consistently emphasized in the literature.