Abstract
Background
Our objective is to develop a predictive model utilizing the ferritin and transferrin ratio (FTR) and clinical factors to forecast overall survival (OS) in breast cancer (BC) patients.
Methods
We conducted a retrospective analysis of clinical data from 2858 BC patients diagnosed between 2013 and 2021. Subsequently, the cohort of 2858 BC patients underwent random assignment into distinct subsets: a training cohort comprising 2002 patients and a validation cohort comprising 856 patients, maintaining a proportional ratio of 7:3. Employing multivariable Cox regression analysis within the training cohort, we derived a prognostic nomogram. The predictive performance was assessed using calibration curves, C-index, and decision curve analysis.
Results
The final prognostic model included the TNM stage, subtype, hemoglobin levels, and the ferritin-transferrin ratio. The nomogram achieved a C-index of .794 (95% CI: .777-.810). The nomogram demonstrated superior predictive accuracy for OS at 3, 5, and 7 years for BC, with area under the time-dependent curves of .812, .782, and .773, respectively. These values notably outperformed those of the conventional TNM stage. Decision curve analysis reaffirmed the greater net benefit of our nomogram compared to the TNM stage. These findings were subsequently validated in the independent validation cohort.
Conclusion
The FTR-based prognostic model may predict a patient’s OS better than the TNM stage in a clinical setting. The nomogram can provide an early, affordable, and reliable tool for survival prediction, as well as aid clinicians in treatment option-making and prognosis evaluation. However, further multi-center prospective trials are required to confirm the reliability of the existing nomogram.
Plain Language Summary
Background
Our objective is to develop a predictive model utilizing the ferritin and transferrin ratio (FTR) and clinical factors to forecast overall survival (OS) in breast cancer (BC) patients.
Methods
We conducted a retrospective analysis of clinical data from 2858 BC patients diagnosed between 2013 and 2021. Subsequently, the cohort of 2858 BC patients underwent random assignment into distinct subsets: a training cohort comprising 2002 patients and a validation cohort comprising 856 patients, maintaining a proportional ratio of 7:3. Employing multivariable Cox regression analysis within the training cohort, we derived a prognostic nomogram. The predictive performance was assessed using calibration curves, C-index, and decision curve analysis.
Results
The final prognostic model included the TNM stage, subtype, hemoglobin levels, and the ferritin-transferrin ratio. The nomogram achieved a C-index of .794 (95% CI: .777-.810). The nomogram demonstrated superior predictive accuracy for OS at 3, 5, and 7 years for BC, with area under the time-dependent curves of .812, .782, and .773, respectively. These values notably outperformed those of the conventional TNM stage. Decision curve analysis reaffirmed the greater net benefit of our nomogram compared to the TNM stage. These findings were subsequently validated in the independent validation cohort.
Conclusion
The FTR-based prognostic model may predict a patient’s OS better than the TNM stage in a clinical setting. The nomogram can provide an early, affordable, and reliable tool for survival prediction, as well as aid clinicians in treatment option-making and prognosis evaluation. However, further multi-center prospective trials are required to confirm the reliability of the existing nomogram.
Introduction
Breast cancer (BC) is a malignancy that originates in breast tissue, typically affecting females and stands as a prominent cause of cancer-related mortality in women.1,2 By 2024, it is projected that there will be approximately 313,510 new cases of BC and 42,780 deaths worldwide. BC has emerged as a global public health challenge, with its incidence continuing to rise at an annual rate of approximately .6%. 3 Current treatment patterns, including surgery, radiotherapy, chemotherapy, targeted therapy, and hormone therapy, are applied to cure BC patients.4,5 For early-stage BC, surgical resection typically serves as the preferred initial treatment, whereas advanced-stage cases often necessitate systemic therapy, particularly chemotherapy.6,7 Currently, clinical decisions, prognosis assessments, and predictions in BC predominantly hinge on the anatomically-based tumor-node-metastasis (TNM) staging system. 8 Indeed, BC is a complex disease that displays a large degree of heterogeneity. 9 Even among patients with identical TNM staging and similar treatment approaches, disparities in clinical outcomes persist.10,11 This indicates that there is a certain degree of inaccuracy in predicting prognosis using the TNM staging system. To improve the accuracy of prognosis prediction, efforts have been made from multiple perspectives. These include examining genomics and immuno-oncological markers within tumors and MRI features.12-15 Although these methods have successfully identified patients with poor prognoses, their widespread adoption in clinical practice is challenging due to their high cost or technical limitations. Therefore, there is a dire need to explore inexpensive and reliable clinical biomarkers to more accurately predict the prognosis of BC patients and devise personalized treatment plans. This endeavor will contribute to enhancing both the survival rates and the quality of life for patients.
Recent research has indicated that nutritional and inflammation markers are closely associated with the prognosis of cancer patients.16,17 Inflammation around BC is considered to be the starting point in the development of BC.18,19 The inflammatory state can have a significant impact on iron metabolism, leading to an increase in antimicrobial iron levels while decreasing serum iron levels. 20 This occurs because inflammation can disrupt the regulation of serum ferritin (SF) and transferrin (TRF), both of which are synthesized by the liver and play pivotal roles in iron absorption and distribution. 21 During inflammation, SF is typically considered an acute-phase inflammatory marker, and its production can significantly rise.22,23 TRF, responsible for transporting and distributing iron in the body, may also be subject to regulation under inflammatory conditions, resulting in alterations in iron absorption and distribution.24,25 In the study, ferritin and transferrin ratio (FTR) was utilized as a potential prognostic factor for BC patients. In the assessment of anemia, SF and TRF have been thoroughly investigated in the past for their capacity to disclose underlying inflammatory conditions.26,27 Nevertheless, this study is the initial one to show it as a ratio in BC. The study created a practical nomogram that integrates blood biomarkers and clinical-pathological risk factors to forecast BC survival rates. This model was then validated in an independent cohort for reliability. Furthermore, the research conducted tests to assess how the nomogram’s predictive abilities compared with the existing staging system. These findings enhance our understanding of prognostic factors in BC patients and introduce a potential blood biomarker that could enhance prognosis assessments.
Materials and Methods
Study Population
Based on the available data, we conducted a retrospective study involving 2858 patients diagnosed with histologically confirmed BC who received treatment at the Affiliated Tumor Hospital of our university between 2013 and 2021. The inclusion criteria were as follows: (1) patients with BC diagnosed by clinicians and pathologists; (2) no preoperative therapy for initial BC; (3) absence of acute infections or other inflammatory conditions in the two weeks preceding surgery; (4) availability of complete follow-up information and clinical data; (5) availability of peripheral blood hematological markers before treatment. Exclusion criteria included: (1) a history of cancer treatment antitumor therapy (e.g., chemotherapy, radiotherapy) (n = 889); (2) lack of complete and definite pathological diagnosis and medical history information (n = 96); (3) presence of other malignant tumors except for BC (n = 63); (4) diagnosis of autoimmune diseases or chronic inflammatory conditions (n = 79); (5) relapse or de novo BC(n = 35). Subsequently, 2858 BC patients were randomly assigned to a training cohort (2002 patients, approximately 70% of the data) and a validation cohort (856 patients, the remaining 30% of the data) by R software in a 7:3 ratio. To develop the nomogram models, the training group was utilized, and the validation cohort was employed to assess the model’s generalizability. Figure 1 illustrates the detailed process for selecting patients. Flowchart depicting the patients identified in this investigation.
Data Collection
In this study, we collected a comprehensive set of clinicopathological, and laboratory data from 2858 BC patients. Clinicopathological data included the patient’s age, histopathological type, TNM stage, grade, subtype, and clinical TNM stage based on the most recent AJCC staging system (8th edition). 8 Pre-treatment inflammation and nutritional biomarkers included the levels of SF, TRF, hemoglobin (HGB), and albumin (ALB). To facilitate analysis, we also transformed certain clinicopathological characteristics into categorical variables. Furthermore, we calculated several inflammation-related ratios, such as neutrophil–lymphocyte ratio (NLR), platelet–lymphocyte ratio (PLR), and FTR, based on their known associations with the outcomes of interest.
Follow-Up
We conducted follow-up assessments using a combination of phone interviews and an outpatient surveillance system. The median follow-up time was 54 months (range: 52 to 55 months). The main outcome measure we focused on was overall survival (OS), which was defined as the time between the date of surgery and death from any cause or last follow-up, whichever came first. The follow-up period for our study extended until December 2022, or until the date of a patient’s death if it transpired earlier.
Ethical Statement
This study was strictly conducted according to the provisions of the Declaration of Helsinki of 1995 and was approved by the Medical Ethics Committee of Guangxi Medical University Cancer Hospital, Nanning, China. The ethical approval statement includes the name (the Medical Ethics Committee of Guangxi Medical University Cancer Hospital) and the location of the review board (Wuxiang Campus, Guangxi Medical University Cancer Hospital, No. 50 Liangyu Avenue, Liangqing District, Nanning City, Guangxi, China), the approval number (LW2023174), and the date of approval to this manuscript (January 4, 2023). The reporting of this study conforms to TRIPOD guidelines. 28 Written informed consent for their data to be used was obtained from all of the patients. To safeguard patient confidentiality, computer-generated ID numbers were employed to anonymize the identities of the individuals enrolled in this study. On admission, all patients provided written consent for their anonymized medical data to be analyzed and published for research purposes.
Statistical Analysis
Categorical variables were described as numbers (percentages). Categorical variables were tested using the Chi-square test for equal proportion. OS was defined as the duration between the date of surgery and the occurrence of death from any cause or the date of the last follow-up. The cumulative OS was estimated using Kaplan-Meier analysis. Initially, the variable of clinicopathologic candidate predictors was reduced using a univariate Cox regression, with a significance threshold of P < .05, in the training cohort. Important OS factors were selected using a multivariate Cox proportional hazards regression model. A nomogram was created using the training cohort of 2002 breast cancer patients, employing significant predictors identified through multivariable Cox regression analysis and constructed through the utilization of R’s rms package.
The nomogram’s performance was assessed in both the training and validation cohorts, focusing on its ability to discriminate and calibrate. The discrimination performance for predicting OS was quantitatively evaluated using Harrell’s concordance index (C-index). Furthermore, the agreement between the predicted survival probability and the average real survival probability, or the calibration curve, has been plotted to assess the nomogram’s calibration.
In addition, the analysis utilized receiver operating characteristic (ROC) curves to assess and compare the nomogram’s capacity to discriminate and predict 3-year, 5-year, and 7-year OS rates, as measured by the area under the curve (AUC) value, with the TNM stage. To assess the practical value of the nomogram, a decision curve analysis (DCA) was performed. This analysis served as a comprehensive approach to evaluate and compare the nomogram with the TNM stage. The net benefits were calculated for various threshold probabilities in both the training cohort and the validation cohort. Using the X-tile program’s predetermined cut-off values for total points, patients were categorized into low-risk, intermediate-risk, and high-risk groups for OS in both the training and validation cohorts. A Kaplan-Meier curve was generated to demonstrate discrimination among the three groups, using a log-rank test. The ROC curves were generated using the “timeROC” package. The construction and calibration plots of the nomogram were executed using the “rms” package. The DCA analysis was conducted using the “R version 4.2.1”. The statistical analysis was performed using SPSS Statistics 23.0 and R software (R version 4.1.3). A two-tailed P-value less than .05 was deemed statistically significant.
Results
Characteristics of Patients and Disease
Patient Demographics and Clinical Characteristics.
Abbreviations: TNM: tumor-node-metastasis staging; IDC: Invasive ductal carcinoma; ILC: Invasive lobular carcinoma.
A High Concentration of FTR is Related to the Poor Prognosis of BC
To evaluate the prognostic value of FTR in BC, we compared the survival differences among patients with different concentrations of FTR by drawing the Kaplan-Meier survival curve. ROC curves were employed to determine the optimal cutoff points for FTR using MedCalc software. The data was divided into two groups based on a cut-off value of FTR, and a Kaplan-Meier survival curve was generated (Figure 2). The analysis showed that the group with high FTR had considerably shorter OS (P < .001, Figure 2A) in the training group. In the validation cohort, there was a significant association between the high group and poor OS (P = .032, Figure 2B). Kaplan-Meier curve of FTR. (A) Kaplan-Meier curve of FTR in the training cohort. (B) Kaplan-Meier curve of FTR in the validation cohort.
Construction and Validation of the Nomogram
Univariate and Multivariable Cox Regression Analysis for the Prediction of OS in the Training Cohort.
Abbreviations: TNM: tumor-node-metastasis staging; IDC: Invasive ductal carcinoma; ILC: Invasive lobular carcinoma.
Nomogram for predicting the 3-year, 5-year, and 7-year OS for BC patients in the training cohort.
The calibration curves for predicting patient OS at three years, five years, and seven years in the training cohort(A) and three years, five years, and seven years in the validation cohort(B).
Comparison of Predictive Accuracy and Clinical Usability Between Nomogram Models and TNM Staging Systems
To assess the predictive capacity of the nomogram, we conducted a comparison between the nomogram and the TNM stage model in both the training cohort and the validation cohort. The nomogram’s C-index outperformed that of the TNM stage, with a value of .794 (95% CI: .777-.810) in the training cohort and .833 (95% CI: .813-.853) in the validation cohort. The ROC analysis confirmed that the nomogram outperformed the TNM stage alone in predicting 3-year, 5-year, and 7-year OS. In the training cohort, the nomogram had a higher area under the curve (AUC) values compared to the TNM stage alone: .812 vs .759 for 3-year OS, .782 vs .732 for 5-year OS and .773vs .732 for 7-year OS. Similarly, in the validation cohort, the nomogram had higher AUC values: .881 vs .850 for 3-year OS, .785 vs .753 for 5-year OS (Figure 5), and .780 vs .762 for 7-year OS. ROC curves compare the prediction accuracy of the nomogram with the TNM stage in predicting 5-year OS (A) in the training cohort, and (B) in the validation cohort.
Ultimately, DCA was employed to assess the clinical utility of the nomogram in comparison to the conventional TNM stage. Figure 6(A) and (B) of the DCA visually demonstrated that the nomogram outperformed the typical TNM stage for predicting OS over a larger range of threshold probabilities, respectively, for the training and validation cohorts. Decision curve analysis for the nomogram and TNM stage in prediction OS of patients at 5-year point (A) in the training cohort, and (B) in the validation cohort.
Performance of the Nomograms in Risk Stratification of OS
The prognostic scores of each independent variable were determined using the established nomogram. The ideal cut-off values were obtained using X-tile, based on the total scores. Individuals diagnosed with BC were categorized into three distinct risk groups: low-risk group (total points: 0 to 77), middle-risk group (total points 78 to 118), and high-risk group (total points: 119 to 199). The Kaplan-Meier OS curves demonstrated significant differentiation between the three risk categories, while the TNM staging system showed limited capacity to identify patients at high risk in both the training (Figure 7A) and validation cohorts (Figure 7B). Kaplan–Meier curves of three groups based on total points (A) in the training cohort, and (B) in the validation cohort.
Discussion
In this study, a nomogram prognostic model was developed, and for the very first time, a remarkable correlation between increased serum FTR levels and an adverse prognosis in individuals diagnosed with BC. We developed a nomogram to forecast the prognosis of patients with BC. The nomogram’s validation demonstrated its significant ability to distinguish and accurately calibrate. This nomogram, based on routinely available hematological risk factors and clinical characteristics predicts the survival probability for individual BC patients.
Previous studies have created nomograms to forecast survival in BC. However, our research has significant advantages over these earlier papers. For example, Soo-Yeon Kim et al. developed a prognostic model using imaging data and clinical-pathologic variables to predict the outcome of BC patients. 13 Zhili et al. and Ju Wang et al. developed a nomogram using molecular profiling and transcriptomics data.29,30 These methods have been restricted due to the requirement for intricate apparatus, or expensive, or intrusive procedures. Notably, the convenient and reasonably priced indicator is used in our nomogram model. The acquisition of FTR parameters is noninvasive, inexpensive, and simple. It is crucial that our nomogram model is easily adaptable and can be used in basic medical facilities. Furthermore, the nomogram offers the shortest turnaround time, delivering results in as little as .5 to 6 hours, compared to the molecular profiling and imaging features system, which may take up to 10 days. The utilization of the FTR-based prediction model in this study not only improves the financial benefits for patients but also assists clinicians in decision‐making treatment. Systemic inflammation has been demonstrated to play a role in the onset and advancement of various forms of malignancies.31,32 To assess the impact of inflammation and nutrition on cancer prognosis, we examined many prognostic markers, including systemic inflammation and nutrition measures such as NLR, PLR, HGB, ALB, and FTR. Through multivariate analysis, it was determined that only HGB and FTR were independent risk variables for prognosis in BC patients. Incorporating FTR into our model, along with other indicators, yielded a good prognosis model for OS. This enhances the development of more precise prognostic forecasts. Furthermore, the patients were categorized into three separate risk groups for OS according to the cumulative points obtained from the nomogram. The Kaplan-Meier survival curves for OS were distinctly separated in both the training cohorts and the validation cohort. These data suggest a positive correlation between greater total point scores and an increased risk of mortality. Consequently, patients in the three groups may be subjected to distinct treatment options and active supervision.
Based on the multivariate analysis results, we have constructed and internally validated a nomogram to predict survival probability in patients with BC. We identified four independent predictors TNM stage, subtype, HGB, and FTR which were embedded into the current nomogram. We divided patients into low-, middle-, and high-risk groups according to their nomogram total points. The Kaplan-Meier OS curves exceeded the traditional staging technique in discriminating between the three risk groups, showing substantial variations in OS. The nomogram exceeded the traditional staging approach in identifying the high-risk population. In the current study, the nomogram incorporates several elements, including blood markers and clinicopathological characteristics, into a quantitative model. It has been demonstrated to outperform certain traditional staging systems, like the American Joint Committee on Cancer (AJCC), 8 in predicting prognosis and making clinical decisions. Traditionally, the TNM staging system has been the primary method for determining the prognosis of patients with BC. Typically, the stages of this system are closely linked to the OS. Nevertheless, patients at the same stage exhibited varying prognoses. This prognosis heterogeneity could be explained by that subtype, inflammatory and nutritional indicators, and other factors are not considered in the TNM staging system. Thus, we conducted a comparison between the nomogram, which incorporates a greater number of factors, and the usual TNM staging scheme. The ability of discrimination and risk stratification of the model of the nomogram indicated that the nomogram had better predictive capability than the TNM staging system alone. The strength of the current nomogram is that it incorporates clinicopathological and some routinely available clinical data variables, which are critically important for predicting OS but cannot be adopted by the TNM stage system. Furthermore, DCA demonstrated that our nomogram excelled over the standard staging method in terms of predicting survival with enhanced therapeutic benefit and utility.
Our analysis revealed that the TNM staging system had the most significant impact on predicting OS in BC patients. The TNM staging system is largely acknowledged as the primary prognostic factor for BC patients. 33 Several studies have indicated a correlation between anemia and a negative tumor oxygenation status. Consequently, the level of HGB seems to have an impact on cancer survival, which aligns with our findings.34,35 FTR reflected the diverse effects of SF and TRF on tumor progression, which was a significant predictor of prognosis in BC.21,25 Fatemeh Zali et al. demonstrated a correlation between the FTR and the prognosis of COVID-19 patients, mirroring the findings of our research. 36 Ishaan Vohral et al. introduced the ratio of ferritin to transferrin as an indicator for assessing the prognosis of hepatocellular carcinoma, consistent with our research. 37 Serum ferritin and transferrin, as important indicators of iron metabolism, play crucial roles in assessing anemia. 38 Ferritin functions as a protein responsible for the storage of iron, offering insights into the body’s iron reserves.39,40 Conversely, transferrin, a protein involved in iron ion transportation, reflects cellular iron requirements. 40 The collaborative action of these proteins is pivotal in preserving iron ion equilibrium within the body, essential for cellular viability and metabolic processes. 41 Consequently, cancer cells frequently manifest irregularities in iron metabolism in cancer patients. Cancer often presents with a concurrent chronic inflammatory response, characterized by the release of various cytokines such as tumor necrosis factor (TNF), interleukin-6 (IL-6), and interferon-gamma (IFN-γ). These cytokines influence the process of iron metabolism by promoting iron accumulation and inhibiting transferrin production. 19 This leads to an elevation in the TFR index, thereby affecting iron metabolism. These markers serve as surrogate indicators of the inflammatory milieu provoked by BC, a process further potentiated by the tumor’s intrinsic characteristics. Monitoring changes in ferritin and transferrin levels offers valuable information on aberrations in iron metabolism among cancer patients, aiding in treatment decision-making. Elevated ferritin levels and decreased transferrin levels may indicate increased tumor iron requirements, impacting patient survival. Moreover, assessing the effects of chemotherapy or radiotherapy on ferritin and transferrin levels can inform treatment effectiveness. Physicians can adjust treatment plans based on these biomarker fluctuations, optimizing iron utilization and selecting alternative therapies. However, the prognostic value of serum FTR in BC has not been reported. Our study is the first to reveal a correlation between elevated serum FTR levels and poor prognosis in BC patients, suggesting that FTR, as a biomarker, holds promise for clinical application in evaluating treatment efficacy and providing prognostic insights for cancer patients. FTR plays a crucial role in tumor progression by exerting significant influence on the tumor microenvironment, enhancing migration and invasion capabilities, and promoting proliferation.21,27,42,43 Although FTR has been associated with human tumors, its significance in BC has not been fully investigated. Therefore, further research is warranted to expand the existing literature on the role of the FTR in predicting the prognosis of BC.
There were several limitations in the development of this nomogram. First, we must acknowledge that this study is retrospective, as it gathers data from medical records, which inherently carries the risk of selection bias, ultimately weakening the reliability of the results. Second, our study specifically concentrated on the nomogram’s ability to predict OS in BC patients and did not gather any data regarding disease-free survival (DFS). It did not assess its ability to accurately predict DFS in BC patients. DFS considers the significant proportion of BC patients who experience disease recurrence. Including OS and DFS may offer a more comprehensive prognostic perspective. In future research, we will continue to collect DFS data and investigate the influence of prediction models on DFS. Third, this study is limited to a single center, which means the results may not apply to other populations. Despite excellent internal validation, the presented nomogram was not yet suitable for universal application before the predictive models were validated with external datasets. Thus, it is necessary to conduct external and multicenter prospective cohorts with large sample numbers to authenticate the clinical use of our nomogram. In future studies, we plan to expand our collaboration by establishing partnerships with other health care institutions or industry organizations to acquire additional data or samples. Increasing the sample size will enhance the dependability and statistical power of the validation process. Moreover, we aim to explore diverse data sources, encompassing various regions, ethnic groups, and disease types, to enhance the generalizability and applicability of the model.
Conclusion
Our study indicates that preoperative FTR may serve as an independent prognostic factor in patients with BC. Consequently, we developed and validated a novel nomogram based on the FTR. This nomogram can provide an accurate, early, cost-effective, and user-friendly method for predicting OS in BC patients. Its performance surpasses that of the existing staging system alone, both in terms of prediction capability and clinical applicability. The nomogram is a valuable tool for identifying patients with different levels of survival risk, aiding clinicians to make informed decisions and helping patients consider their treatment options. However, further validation through multi-center prospective studies is necessary to confirm its reliability.
Footnotes
Acknowledgments
The authors gratefully acknowledge each editor and reviewer for their efforts in this study.
Author Contributions
Shuqing Huang, Hao Lai, Qiumei Lin collected the data. Shuqing Huang, Xiaolan Pan and Caibiao Wei analyzed the data and interpreted the results. Yuling Qin, Min Fang, Fengfei Liu and Wencheng Huang collected and performed the data curation. Shuqing Huang, Hao Lai, Wencheng Huang, and Caibiao Wei drafted and reviewed the manuscript. All authors read and approved the final manuscript.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by grants from National Science Foundation of Guangxi (2022GXNSFAA035510), National Science Foundation of China (81760530), and Postdoctoral Science Foundation of China (2021M693803).
Ethical Statement
Data Availability Statement
The datasets generated and/or analysed during the current study are not publicly available due to that the data also forms part of other ongoing studies but are available from the corresponding author on reasonable request.