Construction of a nomogram-based risk prediction model for depressive symptoms in middle-aged and young breast cancer patients

Abstract

BACKGROUND

Breast cancer (BC) is the second most common malignancy globally. Young and middle-aged patients face more pressures from diagnosis, treatment, costs, and psychological issues like self-image concerns, social barriers, and professional challenges. Compared to other age groups, they have higher recurrence rates, lower survival rates, and increased risk of depression. Research is lacking on factors influencing depressive symptoms and predictive models for this age group.

AIM

To analyze factors influencing depressive symptoms in young/middle-aged BC patients and construct a depression risk predictive model.

METHODS

A total of 360 patients undergoing BC treatment at two tertiary hospitals in Jiangsu Province, China from November 2023 to April 2024 were included in the study. Participants were surveyed using a general information questionnaire, the patient health questionnaire depression scale, the visual analog scale for pain, the revised family support scale, and the long form of the international physical activity questionnaire. Univariate and multivariate analyses were conducted to identify the factors affecting depression in middle-aged and young BC patients, and a predictive model for depression risk was developed based on these findings.

RESULTS

Among the 360 middle-aged and young BC patients, the incidence rate of depressive symptoms was 38.61% (139/360). Multivariate analysis revealed that tumor grade, patient’s monthly income, pain score, family support score, and physical activity score were factors influencing depression in this patient group (P < 0.05). The risk prediction model constructed based on these factors yielded an area under the receiver operating characteristic curve of 0.852, with a maximum Youden index of 0.973, sensitivity of 86.80%, specificity of 89.50%, and a diagnostic odds ratio of 0.552. The Hosmer-Lemeshow test for goodness of fit indicated an adequate model fit (χ² = 0.360, P = 0.981).

CONCLUSION

The constructed predictive model demonstrates good predictive performance and can serve as a reference for medical professionals to early identify high-risk patients and implement corresponding preventive measures to decrease the incidence of depressive symptoms in this population.

Key Words: Breast cancer; Middle-aged and young adults; Depression; Risk factors; Predictive model; Survey research

Core Tip: Breast cancer is the second most common malignancy globally. Young and middle-aged patients have higher recurrence rates, lower survival rates, and increased depression risks. This study analyzed factors influencing depressive symptoms among 360 patients treated at two hospitals in Jiangsu Province, China from November 2023 to April 2024. Key factors included tumor grade, patient income, pain score, family support, and physical activity. The predictive model showed an receiver operating characteristic curve of 0.852, aiding early identification and prevention of depression in this patient group.

INTRODUCTION

Breast cancer (BC) stands out as the leading type of cancer diagnosed among women. According to comprehensive global studies on cancer[1], in 2022 there were approximately 20 million newly reported cancer cases around the globe, with female BC accounting for about 2.3 million of these instances, positioning it as the second most prevalent cancer globally. Regarding mortality, BC had a death rate of 6.9% that year, placing it fourth when compared to other cancer types in terms of fatality rates[2]. In China, the incidence of BC has also shown an increasing trend[3]. The peak age of onset in Chinese BC patients is 45-54 years, about 10 years younger than that in developed Western countries. Patients under 50 years old accounted for a significant 57.4% of all BC cases, a rate much higher than that in developed countries[4,5]. Studies[6,7] have shown that middle-aged and young BC patients often face the stress and burdens from a series of events such as diagnosis, radiotherapy and chemotherapy, and high treatment costs. In addition, they have to cope with various psychological problems due to the impact on self-image, social barrier, sexual relationship, and career engagement. Consequently, compared to other age groups, this age group has higher recurrence rates and poorer survival outcomes[8,9], and is more prone to negative emotions like anxiety and depression[10-12].

It was reported that the detection rate of depressive symptoms in middle-aged and young BC patients was as high as 65%[13]. Maass et al[14] investigated the prevalence of depressive symptoms in young and middle-aged BC patients and healthy individuals, and the results suggested that the prevalence of depressive symptoms in young and middle-aged BC patients was 10.6%, which was higher than that of healthy individuals (4.9%), especially the occurrence of severe depression. According to Burgess et al[15], the prevalence of depressive state in young and middle-aged patients at the time of diagnosis of BC was 33%, rising to 48% 1 year after diagnosis. The presence of depressive symptoms in BC has not only been shown to diminish immune function, potentially increasing the risk of tumor recurrence, metastasis, and exacerbation[16,17], but also to elevate the mortality rate within five years[18], thereby negatively affecting treatment and rehabilitation. Consequently, creating an early detection and screening system for depressive symptoms in BC patients is crucial.

Currently, researchers have employed two methods for early screening of depressive symptoms in middle-aged and young BC patients after surgery and those diagnosed at an early stage: Scale assessments[19-21] and monitoring of peripheral blood cells[22-24]. However, scale assessments are time-consuming and prone to inaccurate interpretations, while peripheral blood cell monitoring is invasive and laborious, and is not an ideal primary method for early screening. The rise of information technology and big data has ushered in a new era for clinical risk prediction modeling, which now serves as an innovative tool for the early detection and alerting of various diseases[25]. Nomogram-based models integrate numerous risk factors as variables, visually representing their scores through graphical interfaces. Such models enable healthcare professionals to assess the likelihood of outcomes associated with these risk factors based on calculated scores, significantly aiding patient education and proactive nursing interventions[25]. This methodology has been extensively utilized in clinical research concerning conditions like esophageal cancer[26,27], stroke[28,29], and colorectal cancer[27,30], but it has not yet been implemented for the early screening of depressive symptoms in middle-aged and young BC patients. The primary objective of this study was to explore the risk factors contributing to depression in this population, develop a nomogram model tailored to these factors, and establish a foundation for the early identification and prevention of depressive symptoms among these patients.

MATERIALS AND METHODS

Study design and setting

This cross-sectional study was conducted from November 2023 to April 2024 at two tertiary hospitals in Nanjing, Jiangsu Province. Data were collected using convenience sampling in the breast departments.

Participants

A total of 360 middle-aged and young patients with BC were recruited from southern China. The inclusion criteria were as follows: (1) BC diagnosis conforming to the standards outlined in the Chinese Guidelines for BC Screening and Early Diagnosis[31] and confirmed by physical examination, imaging, biopsy, or other methods; (2) Aged ≥ 18 and < 60 years (taking into account the classification of age groups[32,33] and the epidemiological characteristics of BC incidence[34], this study defined young and middle-aged BC patients as those aged between 18 and 60 years. This definition is informed by both international standards and the specific context of BC in China); (3) Mentally competent with no communication barriers; and (4) Being aware of their condition and willing to cooperate with the researchers. The exclusion criteria were: (1) Having a history of severe heart, brain, lung, or other organic functional disorders, or a history of depression; (2) Concurrent with other cancers besides BC; and (3) Participation in other studies during the same period of this investigation.

Survey and measurements

Demographic data including gender, age, occupation, marital status, education level, economic status, tumor grade, tumor size, presence of cancer cell metastasis, smoking history, alcohol consumption history, and family history were surveyed.

Patient health questionnaire-9

The patient health questionnaire-9 (PHQ-9) uses a 4-point rating scale ranging from 0 to 3[35]. The total score is 27 points, with 0 points representing none at all, 1 point indicating several days, 2 points suggesting more than half of the days, and 3 points denoting almost every day. Interpretation of results was as follows: 0-4 points indicate no depression, 5-9 suggest mild depression, 10-14 denote moderate depression, 15-19 indicate moderately severe depression, and 20-27 represent severe depression. In this study, the Cronbach's α coefficient for this scale was 0.85. This study used PHQ-9 ≥ 5 points as the criterion for classifying the presence of depressive symptoms.

Visual analog score

The visual analog score for pain represents the degree of pain on a scale of 0 to 10, with 0 indicating no pain and 10 representing the worst imaginable pain. Patients select one number from these 11 options to indicate their level of pain.

Perceived social support from family scale

The perceived social support from family scale (PSS-Fa) was designed by Procidano and Heller in the United States[36], and the revised scale consists of 15 straightforward questions with answer options being "yes" and "no". Responses of "yes" are scored as 1 point, while "no" receives 0 point, with some items inversely scored. Prior to calculating the total score, scores for inversely scored items are reversed. The total possible score is 15, with higher scores indicating greater family support. Family support levels are categorized into three tiers: Low (0-5 points), medium (6-10 points), and high (11-15 points). The scale exhibits good reliability and validity, with a Cronbach's α coefficient of 0.77.

International physical activity questionnaire

The international physical activity questionnaire (IPAQ), formulated by the IPAQ Working Group in 2001[37], comprehensively assesses the level of physical activity with validated reliability and validity. Comprising seven sections and 27 items, the questionnaire measures the subject's physical activities over the past seven days, encompassing occupational activities, household chores, transportation, recreational exercise, sedentary behaviors, and sleep time. It demonstrates acceptable reliability (Spearman’s rho = 0.8) and validity (P = 0.30).

Statistical analysis

Data were analyzed using SPSS 26.0 software and the rms package in R 3.6.3. Apart from PHQ-9, PSS-Fa, and IPAQ scores, missing data were present for other indicators, which were addressed using multiple imputations. Normally distributed continuous variables are expressed as the mean ± SD. Independent sample t tests were used for comparisons between two groups and ANOVA for multiple group comparisons. Skewed continuous data are presented as median (interquartile range), and the Wilcoxon rank-sum test was used for two-group comparisons. Categorical data are presented as frequencies and percentages (%), and χ² tests were employed for comparisons between two or more groups. Variables with statistical significance in univariate analysis were included in logistic regression analysis. The predictive capability of the model was assessed by the area under the receiver operating characteristic (ROC) curve and the Hosmer-Lemeshow test. A nomogram was constructed based on the multivariable model, with calibration plots utilized for internal validation of the model. This study employed two-tailed tests, and P < 0.05 was considered statistically significant.

RESULTS

A total of 363 questionnaires were handed out, and 3 patients did not complete the survey. The age of the 360 middle-aged and young BC patients enrolled ranged from 27 to 59 [median: 51 (interquartile range: 44, 55)] years. Among them, 139 patients exhibited depressive symptoms, with an incidence rate of 38.61% (139/360). Details are provided in Figure 1. Univariate analysis revealed that variables such as marital status, insurance type, surgical method, smoking history, drinking history, family history of cancer, lymph node metastasis, and treatment modality did not exhibit statistically significant differences between the depression group and the non-depression group (P > 0.05). However, statistically significant differences were observed in age, body mass index (BMI), place of residence, education level, menopausal status, adverse reactions, duration of illness, tumor grade, distant metastasis, monthly income, pain score, physical activity score, and family support score (P < 0.05). Detailed information is presented in Table 1.

Figure 1  Flowchart of subject enrollment.

Table 1 Baseline characteristics of patients with and without depression, n (%).

Characteristic	Non-depressed group (n = 221)	Depression group (n = 139)	P value
Age (mean ± SD), years	47.34 ± 8.914	51.66 ± 6.159	< 0.001b
BMI (mean ± SD)	23.68 ± 3.19	24.60 ± 3.44	0.011a
Marital status			0.431
Married	211 (95.48)	135 (97.12)
Unmarried/divorced/widowed	10 (4.52)	4 (2.88)
Type of health insurance			0.052
Urban employee health insurance	155 (70.14)	81 (58.27)
Urban and rural resident health insurance	47 (21.27)	45 (32.37)
Uninsured	19 (8.60)	13 (9.35)
Place of residence			< 0.001b
Rural	43 (19.46)	54 (38.85)
Urban	178 (80.54)	85 (61.15)
Surgical approach			0.286
No surgery	59 (26.70)	29 (20.86)
Total mastectomy	109 (49.32)	80 (57.55)
Breast conserving surgery	53 (23.98)	30 (21.58)
Smoking history			0.574
Yes	3 (1.36)	1 (0.72)
No	218 (98.64)	138 (99.28)
Drinking history			0.562
Yes	1 (0.45)	2 (1.44)
No	220 (99.55)	137 (98.56)
Family history of tumors			0.602
Yes	10 (4.52)	8 (5.76)
No	211 (95.48)	131 (94.24)
Lymph node metastasis			0.181
Yes	139 (62.90)	97 (69.78)
No	82 (37.10)	42 (30.22)
Treatment method			0.269
Chemotherapy	144 (65.16)	82 (58.99)
Chemotherapy/targeted therapy	64 (28.96)	43 (30.94)
Radiotherapy/endocrine therapy/other	13 (5.88)	14 (10.07)
Education level			< 0.001b
High school or above	131 (59.28)	23 (16.55)
Junior high school or below	90 (40.72)	116 (83.45)
Menopausal status			< 0.001b
Yes	136 (61.54)	118 (84.89)
No	85 (38.46)	21 (15.11)
Adverse reactions			< 0.001b
Yes	84 (38.01)	127 (91.37)
No	137 (61.99)	12 (8.63)
Duration of illness [M (P25, P75)]	18.52 (3.00, 7.00)	24.81 (3.00, 24.00)	0.033a
Tumor grade			< 0.001b
I	150 (67.87)	12 (8.63)
II	58 (26.24)	66 (47.48)
III	13 (5.88)	61 (43.88)
Distant metastasis			< 0.001b
Yes	40 (18.10)	53 (38.13)
No	181 (81.90)	86 (61.87)
Patient's monthly income, RMB yuan			< 0.001b
< 3000	39 (17.65)	128 (92.09)
≥ 3000 but < 5000	82 (37.10)	8 (5.76)
≥ 5000	100 (45.25)	3 (2.16)
Pain score [M (P25, P75)]	0.06 (0.00, 0.00)	3.86 (2.00, 5.00)	< 0.001b
Level of physical activities			< 0.001b
Active	131 (59.28)	7 (5.04)
Suitable	83 (37.56)	11 (7.91)
Insufficient	7 (3.17)	121 (87.05)
Family support level			< 0.001b
Low level	5 (2.26)	35 (25.18)
Medium level	10 (4.52)	100 (71.94)
High level	206 (93.21)	4 (2.88)

^aP < 0.05.

^bP < 0.001.

BMI: Body mass index.

Multivariable logistic regression analysis results

The dependent variable was the occurrence of depression. The independent variables included age, BMI, place of residence, education level, menopausal status, adverse reactions, duration of the illness, tumor grade, distant metastasis, monthly income, pain score, physical activity score, and family support score. Multivariable backward logistic regression analysis showed that tumor grade, patient's monthly income, pain score, physical activity score, and family support score were significant predictors of depression (Table 2).

Table 2 Multivariable logistic regression analysis.

Predictor		OR	95%CI	SE	P value
Tumor grade	I	0.12	0.000-1.083	2.297	0.044a
	II	0.782	0.018-34.662	1.934	0.048a
Patient’s monthly income, RMB yuan	≥ 3000 but < 5000	0.003	0.000-0.176	2.029	0.005b
	≥ 5000	0.018	0.000-0.958	2.038	0.048a
Pain score		6.350	1.388-29.063	0.776	0.017a
Level of physical activities	Active	< 0.001	0.000-2.359	4.374	0.028a
	Suitable	0.051	0.003-0.822	1.422	0.036a
Family support level	Medium level	0.345	0.005-24.676	2.179	0.625
	High level	0.005	0.000-0.481	2.381	0.023a

^aP < 0.05.

^bP < 0.01.

Development and application of a depression prediction model

Statistically significant variables from the logistic regression analysis were used as predictors to construct a nomogram-based prediction model using the rms package in R 3.6.3 software, as illustrated in Figure 2. In clinical application, by integrating patient data, a vertical line was drawn from the score points of each indicator towards the "Total Score" standardized rating axis. The cumulative scores from all indicators were then added up at the corresponding point on the total score axis, and another vertical line was drawn downwards along the "Risk Probability" axis, with the resulting score representing the probable likelihood of depressive symptoms occurring in middle-aged and young BC patients. The area under the ROC curve for this model was calculated at 0.852 (95% confidence interval: 0.762-0.866), demonstrating a sensitivity of 86.80% and specificity of 89.50%. The model achieved its highest Youden index at 0.973, with a diagnostic odds ratio of 0.552, highlighting its strong discriminative ability. Additionally, the Hosmer-Leme goodness-of-fit test indicated an adequate model fit (χ² = 0.360, P = 0.981). More detailed information can be found in Figure 3.

Figure 2  Construction of a nomogram-based risk prediction model for depression.

Figure 3  Receiver operating characteristic analysis of the model for prediction of depression.

The overall trend of the calibration curve closely approximates the ideal curve, highlighting the model's good fit and predictive efficacy, as displayed in Figure 4.

Figure 4  Calibration curve of the nomogram-based prediction model.

DISCUSSION

The results of this study showed a 38.61% (139/360) incidence rate of depressive symptoms in BC patients, which is higher than that found in the cross-sectional study by Liao[38](23.3%), but lower than that reported by Lan et al[39] (76.3%) and Zhao et al[13] (65%). It is, however, comparable to the findings of Wu et al[40] (39%) and Dewan et al[41] (46%). The discrepancy might be attributed to racial differences, regional disparities, variations in depression assessment scales, and different sample sizes. Complications of depressive symptoms in middle-aged and young BC patients can potentially undermine their immunity, leading to tumor recurrence, metastasis, and deterioration[16], and even increase the mortality rate within five years[18]. Developing a risk prediction framework facilitates the early detection of individuals vulnerable to developing depressive symptoms, enabling more focused and prioritized preventive strategies. This study developed a nomogram-based predictive model distinguished by its user-friendly nature, swift application, and robust predictive accuracy, achieving an area under the ROC curve of 0.852. The Hosmer-Leme goodness-of-fit test's findings, showing a P value exceeding 0.05, indicate that this model offers satisfactory discrimination and clinical utility in assessing risks associated with depressive symptoms, effectively addressing the needs for initial screening among populations at higher risk. Furthermore, the five variables utilized within this nomogram are conveniently obtainable, allowing healthcare professionals to efficiently assign scores and compute cumulative totals. Utilizing the optimal threshold of 0.552 for distinguishing between high-risk and low-risk categories not only eases the computational load on practitioners but also streamlines the evaluation process.

Our study showed that higher tumor grade was a risk factor for depressive symptoms in middle-aged and young BC patients, which is consistent with the findings by Wu et al[40], Zhuo et al[42], Liao[38], and Patsou et al[43], among others, both domestically and internationally. The rationale behind this is that as tumor grade increases, patients tend to experience more severe clinical symptoms, face greater challenges in treatment, and undergo more invasive surgeries, which impose a heavier physical toll, consequently augmenting their psychological burden. This underscores the need for healthcare providers to promptly conduct depression evaluations for such patients, and initiate psychological interventions early, thereby preventing the onset of depressive symptoms.

Low monthly income is associated with the occurrence of depressive symptoms in middle-aged and young BC patients, aligning with previous research findings[40,43-45]. A qualitative analysis has revealed that, even with the improving survival rates among middle-aged and young BC patients thanks to advances in contemporary medical technology, the considerable expenses from follow-up surgeries, radiation therapy, chemotherapy, and other treatments contribute significantly to financial strain on these individuals. This economic pressure serves as a critical risk factor for depression within this population[46]. Consequently, it is advised that healthcare providers effectively identify those at high risk and offer regular psychological support to mitigate their mental health challenges.

The presence of pain is a risk factor influencing the likelihood of depressive symptoms in middle-aged and young BC patients, consistent with findings from studies by Kyranou et al[47], Zhuo et al[42], and others. Additionally, research has pointed out that over 37% of middle-aged and young BC patients experience depression due to pain, highlighting the substantial impact of pain on their psychological well-being[47]. Collectively, middle-aged and young BC patients who suffer from pain that impairs their daily activities and adds to the burden of additional medical visits are more susceptible to depressive symptoms. This highlights the importance to enhance pain management by healthcare providers for these patients to achieve pain relief while minimizing analgesic dosage when possible.

The role of family is crucial in alleviating patients' negative emotions. This research indicates that individuals lacking sufficient family support are at a greater risk of experiencing depressive symptoms than those who receive substantial familial support. These findings align with the studies conducted by Manczak et al[48], Wong et al[49], and Wu et al[40]. A supportive family environment that offers respect and care can help the BC patient to maintain positive emotions such that the depressive symptoms can be alleviated[48]. However, influenced by the tacit cultural genes of Chinese tradition, many Chinese families grapple with an "awkwardness in expressing emotions", leading cancer patients to hesitate in sharing their thoughts or feelings with family, thereby exacerbating depressive symptoms[50]. Moreover, depressive symptoms can stem from emotional incompatibility between spouses, lack of mutual understanding, unequal family roles, disturbance in sexual life, and conflicts among family members[49]. Therefore, healthcare professionals should encourage patients to express their emotions and feelings to their families, while concurrently providing health education to family members about the importance of understanding and supporting the patient, thereby preventing the onset of depressive symptoms.

Physical activity has a significant association with the onset of depressive symptoms among middle-aged and young BC patients, acting as one of the key predictive factors for depression within this population. Research has indicated[43] that engaging in more than 2.5 hours of physical activity per week is associated with a reduction of depressive symptoms. This suggests that for middle-aged and young BC patients, regular engagement in physical activities can act as a protective factor against depression. Thus, individuals who have insufficient physical activities due to illness burdens and reluctance to participate in outdoor events are more vulnerable to depressive symptoms. Moreover, studies by Ribeiro et al[51] have shown that leisure activities or active commuting is negatively correlated with depression, whereas occupational physical activity does not have a significant relationship with the occurrence of depressive symptoms, potentially due to occupation-related anxiety and inadequate job performance due to discomfort and pain in the upper limb. Depressive mood is a psychological side effect associated with the diagnosis and treatment of BC in middle-aged and young patients, linked to poor adherence and reduced survival rates[43]. Thus, it is important for healthcare professionals to recommend or provide feasible, personalized, and progressive exercise programs to help middle-aged and young BC patients maintain a healthy lifestyle and prevent the onset of depressive symptoms.

The present study has several limitations. First, it was conducted solely in two hospitals in Nanjing, Jiangsu Province, which may limit the generalization of the conclusion. Second, although it is well-documented that perimenopausal women face an elevated risk of depression due to hormonal fluctuations, we did not incorporate this factor into the present study. Our team is now committed to exploring the impact of perimenopausal status on depression by a larger scale study. Additionally, the model requires further verification to confirm its effectiveness and applicability to a wider population. In the future, the constructed model can be applied in a clinical setting to validate its accuracy.

CONCLUSION

This study's results indicate that risk factors for depressive symptoms in middle-aged and young BC patients include higher tumor grades (II and III), lower monthly incomes (< 3000 RMB yuan), experience of pain, insufficient family support, and lack of physical activity. Utilizing these factors, the developed nomogram model shows strong predictive capability, offering a foundation for the early detection and prevention of depressive symptoms in this group of patients.

ACKNOWLEDGEMENTS

We thank all the participants involved in the study.