Background
Breast cancer is the most common malignant tumor in women, and its incidence is increasing globally, posing a significant threat to women’s health and life [
1]. Malnutrition can lead to decreased immunity, metabolic disorders, and decreased treatment tolerance, subsequently affecting the effectiveness of cancer treatments and patient prognosis [
2‐
7]. Thus, evaluating the nutritional status of patients with breast cancer and implementing appropriate interventions are of great significance for improving their quality of life and prolonging their survival.
Currently, nutritional assessment indicators primarily include patient-generated subjective nutrition assessment (PGSGA), global leadership initiative on malnutrition (GLIM), controlling nutritional status (CONUT), nutritional risk index (NRI), and prognostic nutritional index (PNI) [
8‐
12]. The PGSGA, which primarily relies on patients’ subjective evaluations, includes the assessment of physical function, nutritional status, and metabolic stress, and is known to accurately reflect the patient’s nutritional status. However, its evaluation results may be affected by the subjectivity of the patient’s self-evaluation. The GLIM criteria, widely acknowledged as the global guidelines for nutritional assessment, facilitate the identification of malnutrition in patients; however, they require an evaluation of body weight changes over time, thus lengthening the evaluation process [
13]. Moreover, while the CONUT, NRI, and PNI indicators are based on laboratory biochemical indicators and offer simplicity and objectivity, their effectiveness in specific populations remains to be established [
14,
15]. Additionally, due to physiological differences, like estrogen levels and body fat content, there is observed variation in the nutritional status between premenopausal and postmenopausal age women. Therefore, the applicability of these indicators for breast cancer patients warrants further investigation.
This large-scale, multicentre, population-based cohort study investigated the relationship between multiple nutritional assessment indicators and mortality in breast cancer patients. We compared the predictive effect of these indicators on patients’ overall survival (OS). Additionally, we refined the existing nutritional indicators to identify the optimal nutritional assessment indicator to predict the prognosis of breast cancer. Ultimately, this study aims to provide more accurate indicators for the nutritional assessment of breast cancer patients and better guidance to clinicians.
Methods
Study population
This study was based on the “Investigation on Nutritional Status and its Clinical Outcomes of Common Cancers” (INSCOC) project, which was registered at chictr.org.cn (registration number ChiCTR1800020329). The INSCOC project prospectively collected clinical data of patients with cancer in more than 40 hospitals in China. In this study, we initially screened 2,775 female breast cancer patients who visited the clinic between June 2012 and June 2021. Of these, we excluded 1999 patients due to incomplete clinical or survival data, resulting in 776 patients for the final data analysis. All patients were older than 18 years and had complete clinical data and follow-up information. This study was approved by the institutional review boards of all participating institutions and was conducted according to the guidelines of the Declaration of Helsinki. All enrolled patients provided informed consent for the use of their clinical data, and their personal information was anonymized. A flowchart detailing the screening of the study patients is presented in Additional file
1: Fig. S1.
Patient characteristics and outcomes
Demographic information, clinical parameters, laboratory tests, and physical measurements of all included patients at baseline were comprehensively collected. This includes age, smoking status, drinking status, comorbidities, family history of cancer, tumor pathology, tumor stage, treatment, biochemical indicators, and anthropometric indicators such as height and weight. All these data were obtained from the electronic medical record system. Patients were classified into premenopausal (< 50 years old) and postmenopausal (≥ 50 years old) categories based on their age at diagnosis. For all patients, the body mass index (BMI) was calculated using the formula: BMI = weight (kg)/height (m)2. The patients were then divided into three groups according to the BMI: normal weight (< 24 kg/m2), overweight (24.0–28.0 kg/m2), and obese (≥ 28 kg/m2). All these demographic and clinical pathological data were collected at the initial inclusion in the INSCOC project.
The primary endpoint of this study was OS, defined as the duration from diagnosis until death from any cause. Patient survival information was sourced through regular telephone contacts, outpatient visits, or hospitalizations. The follow-up process continued until either the patient's death or the point at which we could no longer contact the patient.
Malnutrition assessment
Trained staff assessed and recorded the PGSGA at baseline. In addition, we reassessed the GLIM, CONUT, NRI, and PNI based on data collected during the baseline period (Additional file
1: Table S1). The GLIM diagnostic criteria include etiological criteria (reduced food intake or assimilation, inflammation or disease burden) and phenotypic criteria (weight loss, low BMI, and reduced muscle mass). Patients are diagnosed with malnutrition when they meet at least one etiological and one phenotypic criterion. Since all the cancer patients in our study met at least one etiological criterion, our focus was primarily on the phenotypic criteria [
16]. According to the GLIM criteria, weight loss is considered significant if it is greater than 5% within 6 months. A BMI of < 20 and < 22 kg/m
2 for those aged < 70 and ≥ 70 years, respectively, was deemed low. Muscle loss is indicated by a calf circumference (CC), weight-standardized hand grip strength, or mid-arm muscle circumference (MAMC) < 15 percentile (P15) in women. The P15 values for CC, weight-standardized handgrip strength, and MAMC are 29, 0.2144, and 17.06 cm in women, respectively. The CONUT scores are determined from the albumin, lymphocyte, and total cholesterol levels [
17]. Albumin levels > 35, 30–34, 25–29, and < 25 g/L; lymphocyte counts ≥ 1.6, 1.2–1.59, 0.8–1.19, and < 0.8 *10
9/L; and total cholesterol levels ≥ 180, 140–180, 100–139 mmol/L, and < 100 mmol/L are assigned scored as 0, 2, 4, and 6 points, respectively. The scores of albumin, lymphocyte, and total cholesterol levels are then combined. A total score of ≥ 2 points indicates malnutrition. The NRI and PNI are calculated using the following formulas: NRI = 1.519 × albumin (g/L) + 41.7 × current weight/ideal body weight (IBW) [
17]. IBW = [height (m)]
2 × 22. PNI = albumin (g/L) + 5 × lymphocyte count (× 10
9) [
18].
Statistical analysis
Continuous variables were expressed as mean ± standard deviation (± SD) or median (interquartile range). Continuous variables with normal distribution were evaluated using Student’s t-test, while continuous variables with non-normal distribution were tested using the Mann–Whitney U test. Categorical variables were presented as frequencies or percentages, and χ2 tests or Fisher’s exact tests were applied. Continuous nutritional indicators were dichotomized based on optimal cut-offs, determined using maximally rank statistics. The time-dependent receiver operating characteristic curve (ROC) was utilized to evaluate the predictive power of the different nutritional assessment indices for OS. A machine learning method was employed to screen variables and to construct new and improved indicators. Restricted cubic spline (RCS) plots were used to explore the associations between the modified malnutrition indicators and OS. Kaplan–Meier curves and log-rank tests were used to compare survival between the groups. Univariate and multivariate Cox regression analyses were conducted to analyze the independent prognostic value of nutritional indicators for OS in breast cancers. In the Cox regression analyses, model a represented the univariate regression model; model b included tumor stage and BMI; and model c incorporated tumor stage, BMI, diabetes, hypertension, coronary heart disease, smoking, alcohol consumption, surgery, chemotherapy, and radiation. Statistical significance was established as a two-sided P-value < 0.05. All statistical analyses were performed using R version 4.2.1.
Discussion
In this study, we explored the correlations between the nutritional indicators PGSGA, GLIM, CONUT, NRI, and PNI with the prognosis in breast cancers to ascertain their prognostic value. Among these five indices, the time-dependent ROC showed that PNI is the most effective predictor for the prognosis of breast cancer patients. Building upon the PNI, we developed a modified index, CPNI, which incorporates total cholesterol, albumin, and lymphocytes. Kaplan–Meier survival curves and Cox regression analyses suggest that CPNI acts as an independent prognostic factor for breast cancers. Furthermore, time-dependent ROC demonstrated that the predictive capability of CPNI exceeds those of the previously mentioned nutritional indicators.
Analysis using the CPNI index revealed that postmenopausal women experience a higher incidence of malnutrition than premenopausal women, corroborating previous research findings. With aging, women undergo numerous physiological changes, especially between the premenopausal and postmenopausal periods. Factors such as a decreased metabolic rate, hormonal fluctuations, and altered eating habits may impact a woman's nutritional status. These can impede the absorption and utilization of essential nutrients like fats, proteins, and trace elements, thereby contributing to elevated malnutrition rates [
19]. Concurrently, as age progresses, tolerance to malnutrition decreases, leading to an escalating risk of malnutrition-related mortality. The relationship between malnutrition and BMI is complex. The CPNI analysis indicated an incremental rise in malnutrition across non-overweight to obese patients. While the risk of malnutrition-related mortality increased for non-overweight and overweight patients, it appeared somewhat mitigated in obese patients as compared to their overweight counterparts.
Using the PNI as a foundation, we developed CPNI by screening individual variables within PGSGA, GLIM, CONUT, and NRI, ultimately formulating a new evaluation formula based on weighted averages. The CPNI method offers advantages due to its non-invasiveness, simplicity, objectivity, and suitability for dynamic monitoring. Early detection through CPNI can accurately identify malnutrition and potential poor prognosis, thereby enabling timely clinical interventions. This could substantially enhance patient quality of life and potentially prolong survival. CPNI encompasses three key nutritional indicators: total cholesterol, albumin, and lymphocytes, each closely linked to one's nutritional status. For instance, malnutrition is often signaled by diminished serum cholesterol and albumin levels, as well as reduced lymphocyte counts. These indicators not only reflect nutritional shifts but also highlight the degree of inflammatory response. Several studies have indicated that elevated cholesterol levels might amplify the risk of certain cancers and are associated with more severe malignancy and unfavorable prognoses in tumor patients [
20,
21]. Albumin, an integral nutrient reserve, plays a crucial role in vital biological processes, such as regulating immune functions and balancing bodily fluids [
22]. Tumor progression often coincides with hypoproteinaemia, attributed to factors like decreased liver synthesis capabilities, inadequate nutrient intake, and metabolic disturbances in cancer patients [
23]. Lymphocytes, both in number and functionality, are considered important markers of the nutritional status of tumors [
24]. Malnutrition could adversely affect lymphocyte count and effectiveness, thus impairing immune surveillance against tumors. Additionally, tumor cells might produce factors that suppress lymphocytes, further evading immune monitoring. Therefore, in tumor nutritional evaluations, lymphocyte dynamics warrant significant attention. Enhancing lymphocyte count and function through nutritional intervention might strengthen the body's immune surveillance against tumors, potentially leading to improved prognosis.
This study has several limitations that deserve attention. First, the entire study population comprised Chinese patients. Considering racial differences, the extrapolated results of this study may not fully represent a more diverse global population. Second, the absence of a universally accepted gold standard for the diagnosis of malnutrition presents a challenge in comparing the diagnostic accuracy of different nutritional indicators in breast cancers. Third, although we constructed the CPNI indicator based on PNI, the predictive ability of CPNI requires external validation in future studies. Fourth, due to the limited number of underweight patients, this study combined underweight and normal-weight patients into a single non-overweight category for analysis. This approach might obscure the more nuanced relationship between malnutrition and various BMI categories. Additionally, the molecular subtype of breast cancer is closely related to patient prognosis, but the INSCOC project currently lacks molecular subtype data for breast cancer. Investigations into the NHANES, SEER, and Kailuan databases also did not yield comprehensive datasets with both hematological indicators and molecular subtype data for breast cancer. Consequently, subgroup analyses or prognostic analyses of patients with different molecular types could not be conducted, which we plan to further refine in subsequent data collection of the INSCOC project. Fifth, we stratified patients into premenopausal and postmenopausal groups based on age. While this classification approach serves as a common and convenient method to classify menopausal status in large-scale studies, we acknowledge that it may not fully capture the intricacies of the menopausal transition and the individual hormonal changes experienced by patients. However, the INSCOC project currently lacks the specific timing of patient menopause. We will incorporate this information in subsequent studies. In the future, we will also consider more refined menopausal status assessment methods to improve the accuracy of our findings. Finally, the underlying mechanisms of CPNI and breast cancer prognosis have not been fully elucidated, Future laboratory research is planned to investigate this relationship more deeply.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.