Development and validation of a nomogram for predicting chemotherapy-induced liver injury in breast cancer patients

Introduction

Breast cancer remains the most prevalent malignancy among women globally, with a significant disease burden (1). Systemic chemotherapy, utilized in neoadjuvant, adjuvant, and palliative settings, is a cornerstone of treatment (2). However, chemotherapeutic agents are a well‑recognized cause of druginduced liver injury (DILI), which can manifest as asymptomatic transaminase elevation or, in severe cases, lead to treatment discontinuation, surgical delays, and compromised patient outcomes (3-5). The incidence and impact of chemotherapyinduced liver injury (CILI) in breast cancer are increasingly acknowledged, yet clinical practice lacks a validated, individualized risk prediction tool (6).

Current research on CILI in oncology often focuses on molecular mechanisms or the hepatotoxicity of specific agents (7,8), with limited effort dedicated to developing integrated predictive models that incorporate readily available clinical and treatmentrelated factors (9). Nomograms, which provide intuitive visual risk quantification, have gained traction in oncology for prognosis prediction (10,11), but their application for predicting treatmentrelated toxicity like CILI is less explored, particularly in breast cancer (12).

This study aimed to develop and internally validate a clinically practical nomogram that integrates key demographic and treatment variables to predict the risk of CILI in breast cancer patients receiving chemotherapy. Such a tool could facilitate personalized risk assessment, inform clinical decisionmaking regarding monitoring intensity and supportive care, and potentially improve treatment safety and completion rates. We present this article in accordance with the TRIPOD reporting checklist (13) (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0782/rc).

Methods

Study design and participants

A retrospective cohort study was conducted at The Third Affiliated Hospital of Wenzhou Medical University. Patients with a histologically confirmed diagnosis of breast cancer who initiated systemic chemotherapy between May 2022 and May 2025 were considered for inclusion. Exclusion criteria were: (I) pre‑existing chronic liver disease of Child‑Pugh class B or C, or active viral hepatitis; (II) significant baseline liver function abnormalities [alanine aminotransferase (ALT)/aspartate aminotransferase (AST) >3* upper limit of normal (ULN) or total bilirubin (TBil) >2* ULN] prior to chemotherapy; (III) incomplete clinical or laboratory follow‑up data during treatment; (IV) concurrent use of other non‑chemotherapy medications with a high known risk of DILI.

Data collection and variables

Data were extracted from the hospital’s electronic medical records using a standardized form. Collected variables included:

• Demographics: age at diagnosis, menopausal status.
• Clinical comorbidities: history of hypertension, diabetes mellitus, and fatty liver disease (diagnosed by ultrasonography).
• Tumor characteristics: histopathological type, axillary lymph node involvement status.
• Treatment modalities: intent of chemotherapy (neoadjuvant, adjuvant, palliative), specific chemotherapy regimens, and type of breast surgery performed.
• Laboratory parameters: serum albumin, TBil, ALT, and AST were measured at baseline and serially during chemotherapy cycles.

Outcome definition

The primary outcome was the occurrence of CILI during chemotherapy. This was defined as an increase in ALT or AST levels exceeding the normal upper limit.

Statistical analysis

To evaluate the predictive ability of continuous variables for CILI, receiver operating characteristic (ROC) curve analysis was performed. The optimal cut‑off points were determined using the Youden index (maximizing sensitivity + specificity −1). Categorical variables were presented as counts (%) and compared using chi‑square or Fisher’s exact test.

To identify independent predictors for CILI, the least absolute shrinkage and selection operator (LASSO) regression method was employed for variable selection, penalizing the coefficients to reduce overfitting and select the most relevant features from the candidate variables. Predictors with non‑zero coefficients after LASSO selection were then entered into a multivariable logistic regression model. Based on the final logistic model, a nomogram was constructed to visualize the prediction model.

The model’s discriminative ability was evaluated using the concordance index (C‑index). Calibration, which assesses the agreement between predicted probabilities and observed outcomes, was evaluated using a calibration plot with the Hosmer‑Lemeshow test and calculation of the mean absolute error (MAE). The clinical utility of the nomogram was assessed using decision curve analysis (DCA) and a clinical impact curve (CIC), quantifying the net benefit across a range of probability thresholds. Internal validation was performed using the bootstrap resampling method (1,000 repetitions) to obtain a bias‑corrected C‑index and assess potential over‑optimism.

All statistical analyses were performed using R software (version 4.2.0). A two‑sided P value <0.05 was considered statistically significant.

Ethical statement

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was approved by the Ethics Committee of Ruian People’s Hospital (The Third Affiliated Hospital of Wenzhou Medical University) (No. YJ2026044). As this study is a retrospective analysis of previously collected clinical data, and all patient data were de‑identified to protect privacy without any additional medical interventions, the application for exemption from signed informed consent was approved.

Results

ROC analysis for continuous variables

Before variable selection, we performed ROC curve analyses for the continuous variables to determine optimal cut-off points for potential dichotomization. The optimal cut-off points were determined using the Youden index. The analysis revealed the following optimal thresholds: age at 49.5 years [area under the curve (AUC) =0.68] (Figure 1), serum albumin at 43.5 g/L (AUC =0.65) (Figure 2), and TBil at 8.6 µmol/L (AUC =0.62) (Figure 3).

Figure 1 ROC curve for age in predicting chemotherapy-induced liver injury. The optimal cut-off point (49.5 years) determined by the Youden index is marked with a red dot. ROC, receiver operating characteristic.

Figure 2 ROC curve for serum albumin in predicting chemotherapy-induced liver injury. The optimal cut-off point (43.5 g/L) determined by the Youden index is marked with a red dot. ROC, receiver operating characteristic.

Figure 3 ROC curve for total bilirubin in predicting chemotherapy-induced liver injury. The optimal cut-off point (8.6 µmol/L) determined by the Youden index is marked with a red dot. ROC, receiver operating characteristic; TBil, total bilirubin.

Patient Characteristics

A total of 422 eligible patients were included in the final analysis. The median age was 53 years (range, 28–78 years). Among the cohort, 34 patients (8.1%) received neoadjuvant chemotherapy, 368 (87.2%) received adjuvant chemotherapy, and 20 (4.7%) received palliative chemotherapy. The overall incidence of CILI during the study period was 23.7% (100/422) (Table 1).

Predictor selection and nomogram construction

The logistic regression analysis identified two variables as significant independent predictors for CILI: patient age (per year increase) and the intent of chemotherapy (neoadjuvant vs. adjuvant). Palliative chemotherapy was not retained as a separate significant predictor compared to adjuvant in the final model. In the subsequent multivariable logistic regression model, age [odds ratio (OR) =1.05 per year, 95% confidence interval (CI): 1.02–1.08] and neoadjuvant chemotherapy (OR =2.20, 95% CI: 1.45–3.35) remained significantly associated with higher CILI risk. Based on these coefficients, a nomogram was constructed (Figure 4), where points are assigned for age and chemotherapy intent, and the total points correspond to a predicted probability of CILI.

Figure 4 Nomogram for predicting chemotherapy‑induced liver injury.

Model performance and validation

The nomogram demonstrated good discriminative ability with a C‑index of 0.75 (95% CI: 0.70–0.80). The calibration plot (Figure 5) showed good consistency between the predicted probabilities and the actual observed frequencies of CILI, with a MAE of 0.024.

Figure 5 Calibration curve of the nomogram. The x‑axis represents the predicted probability of CILI by the nomogram, and the y‑axis represents the actual observed frequency of CILI. CILI, chemotherapy‑induced liver injury.

The DCA (Figure 6) indicated that using the nomogram for clinical decision‑making provided a higher net benefit compared to the strategies of “treating all” or “treating none” across a wide range of clinically reasonable risk thresholds.

Figure 6 DCA of the nomogram. The y‑axis shows the net benefit. The x‑axis is the threshold probability at which a clinician would consider intervention (e.g., enhanced monitoring). DCA, decision curve analysis.

The CIC (Figure 7) further illustrated that the nomogram could effectively stratify patients, identifying a group at high risk where a substantial proportion would actually experience the CILI event.

Figure 7 Clinical impact curve of the nomogram. The red curve (Number high risk) shows the number of patients per 1,000 who would be classified as high‑risk at each probability threshold if the nomogram were used. The blue curve (Number high risk with event) shows the number of those high‑risk patients who would actually experience the CILI event. CILI, chemotherapy‑induced liver injury.

Internal validation

After 1,000 bootstrap resamples for internal validation, the optimism‑corrected C‑index was 0.73, indicating good internal validity and minimal overfitting of the model.

Discussion

In this study, we developed and internally validated a parsimonious yet effective nomogram for predicting the risk of CILI in breast cancer patients. The model, incorporating only age and chemotherapy intent (neoadjuvant vs. adjuvant), demonstrated robust discrimination, good calibration, and meaningful clinical utility. This tool addresses a gap in personalized toxicity risk assessment for a common and potentially treatment‑limiting complication.

The association between advanced age and increased risk of CILI is biologically plausible and aligns with contemporary understanding of pharmaco‑physiological changes in aging. Older individuals often experience alterations in hepatic blood flow, reduced phase I metabolism (e.g., cytochrome P450 activity), and diminished functional reserve, which can potentiate drugrelated hepatotoxicity (14). Furthermore, polypharmacy and a higher prevalence of comorbid conditions common in older adults may exacerbate this vulnerability (15). Our findings reinforce the need for vigilant monitoring of liver function in older breast cancer patients undergoing chemotherapy.

The significantly elevated risk associated with neoadjuvant chemotherapy, compared to adjuvant therapy, is a critical and clinically relevant finding. This may be attributed to several interconnected factors characteristic of the neoadjuvant setting. Neoadjuvant regimens are frequently more dose‑intense or utilize specific cytotoxic combinations (e.g., anthracyclinetaxane sequences) aimed at achieving maximal tumor response for surgical downstaging (2,16). Several factors may contribute to the higher risk observed with neoadjuvant chemotherapy, including potentially more intensive regimens and different patient characteristics. However, due to the retrospective nature of this study, mechanistic explanations cannot be inferred (17). Last but not least, differences in baseline patient characteristics or supportive care protocols between treatment settings could contribute. This distinction underscores that toxicity risk profiles are not uniform across treatment intents, a nuance that must be incorporated into pretreatment counseling and monitoring plans (18).

The nomogram bypasses the need for complex calculations or additional costly tests, facilitating its potential integration into busy clinical workflows (19). The visual format is intuitive for clinicians, aiding rapid risk estimation during patient consultations or multidisciplinary team meetings. This approach aligns with the growing emphasis on developing practical prediction tools in supportive oncology (20). However, the main limitations of our nomogram are that it only includes age and chemotherapy intention, as CILI is influenced by many other factors. Our model may sometimes be more regarded as an exploratory tool.

The C-index of 0.75 indicates moderate discriminative ability, comparable to several established prognostic models in oncology (21). More importantly, the excellent calibration and the positive net benefit demonstrated by DCA across a wide threshold probability range suggest the model is not only statistically sound but also clinically actionable (22,23). It can help clinicians and patients weigh the benefits of interventions [e.g., more frequent monitoring, prophylactic hepatoprotectants like ursodeoxycholic acid where evidence supports (24)] against their costs and burdens, moving towards personalized supportive care strategies (25).

This study has several limitations. First, its single‑center, retrospective nature may affect the generalizability of the findings. Patient demographics, specific chemotherapy protocols, and institutional monitoring standards can vary (26). Therefore, external validation in independent, preferably multicenter cohorts is an essential next step to confirm the model’s transportability. Second, due to data availability constraints inherent to retrospective designs, we could not incorporate potentially important biomarkers such as pharmacogenomic variants (e.g., in CYP or UGT1A1 genes) or serial pharmacokinetic data, which are known modifiers of DILI risk (27). Future prospective studies integrating such “omics” data hold promise for further refining prediction accuracy (28). Third, our outcome was defined as any ALT/AST elevation above the upper normal limit. This biochemical definition may capture transient or clinically insignificant abnormalities and does not reflect other forms of hepatotoxicity (e.g., steatosis, cholestasis) that may occur without transaminase elevation. Consequently, our event rate (23.7%) may be inflated, and the model’s specificity may be limited. Future studies should use more stringent criteria, such as Common Terminology Criteria for Adverse Events (CTCAE) grade ≥2, liver fibrosis or more detailed imaging assessment definitions (29). Moreover, factors such as whether the patient is obese, their history of alcohol consumption, and the classification of chemotherapy drugs, which may have an impact on the prediction results, were not included.

In future research, if this model is improved through external validation and relevant prospective clinical trials are conducted, we hope it can be applied to the following aspects: for a patient identified as high‑risk by the model, clinicians could consider: (I) enhanced surveillance with more frequent liver function tests during treatment cycles; (II) pre‑emptive discussion of potential dose modifications or alternative regimens with lower hepatotoxicity potential; (III) evidencebased use of hepatoprotective agents in select cases; and (IV) targeted patient education regarding symptoms of liver dysfunction (30,31).

Future research directions should prioritize external validation studies. Prospective implementation research is needed to evaluate whether using this nomogram in routine care actually improves patientrelevant outcomes, such as reducing the incidence of severe liver injury, minimizing unplanned treatment interruptions, or increasing relative dose intensity. Integrating the nomogram into electronic health record systems to generate automated risk alerts could streamline its adoption (32). Furthermore, as breast cancer treatment rapidly evolves, extending the model to assess hepatic safety profiles of newer agents (e.g., antibodydrug conjugates, immune checkpoint inhibitors) and their combinations will be crucial (33,34).

Conclusions

We developed and internally validated a simple and visually understandable nomogram. Based on age and the willingness to undergo chemotherapy, this tool has the potential to predict the risk of liver damage in breast cancer patients after chemotherapy. After undergoing independent external validation and a prospective clinical trial, it may be used for personalized treatment plan formulation, enabling active monitoring, and facilitating shared decision-making, ultimately enhancing the safety and therapeutic efficacy of chemotherapy.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0782/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0782/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0782/prf

Funding: This work was supported by Wenzhou Municipal Science and Technology Bureau (No. Y20240955).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-0782/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Ruian People’s Hospital (The Third Affiliated Hospital of Wenzhou Medical University) (No. YJ2026044). As this study is a retrospective analysis of previously collected clinical data, and all patient data were de‑identified to protect privacy without any additional medical interventions, the ethics committee granted a waiver of informed consent.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Gradishar WJ, Moran MS, Abraham J, et al. NCCN Guidelines® Insights: Breast Cancer, Version 5.2025. J Natl Compr Canc Netw 2025;23:426-36. [Crossref] [PubMed]
2. Gennari A, André F, Barrios CH, et al. ESMO Clinical Practice Guideline for the diagnosis, staging and treatment of patients with metastatic breast cancer. Ann Oncol 2021;32:1475-95. [Crossref] [PubMed]
3. Andrade RJ, Chalasani N, Björnsson ES, et al. Drug-induced liver injury. Nat Rev Dis Primers 2019;5:58. [Crossref] [PubMed]
4. EASL Clinical Practice Guidelines. Drug-induced liver injury. J Hepatol 2019;70:1222-61.
5. Shen T, Liu Y, Shang J, et al. Incidence and Etiology of Drug-Induced Liver Injury in Mainland China. Gastroenterology 2019;156:2230-2241.e11. [Crossref] [PubMed]
6. Lyman GH, Greenlee H, Bohlke K, et al. Integrative Therapies During and After Breast Cancer Treatment: ASCO Endorsement of the SIO Clinical Practice Guideline. J Clin Oncol 2018;36:2647-55. [Crossref] [PubMed]
7. Chalasani N, Bonkovsky HL, Fontana R, et al. Features and Outcomes of 899 Patients With Drug-Induced Liver Injury: The DILIN Prospective Study. Gastroenterology 2015;148:1340-52.e7. [Crossref] [PubMed]
8. Robles-Diaz M, Lucena MI, Kaplowitz N, et al. Use of Hy's law and a new composite algorithm to predict acute liver failure in patients with drug-induced liver injury. Gastroenterology 2014;147:109-118.e5. [Crossref] [PubMed]
9. Du W, Chen W, Zhang D, et al. Establishment and validation of a predictive nomogram for voriconazole-associated liver injury in lung transplant patients. Int J Clin Pharm 2025;47:1416-26. [Crossref] [PubMed]
10. Utsumi T, Ishitsuka N, Noro T, et al. Nomograms in urologic oncology: a review of development, validation, and future directions. Int J Urol 2025;32:205-17.
11. Balachandran VP, Gonen M, Smith JJ, et al. Nomograms in oncology: more than meets the eye. Lancet Oncol 2015;16:e173-80. [Crossref] [PubMed]
12. Zhang M, Yang X, Jia J, et al. A competing-risk nomogram for predicting gastric cancer-specific survival in patients over 70 years: A SEER-based study. Cancer Epidemiol 2024;93:102696. [Crossref] [PubMed]
13. Collins GS, Moons KGM, Dhiman P, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ 2024;385:e078378. [Crossref] [PubMed]
14. Le Couteur DG, Ngu MC, Hunt NJ, et al. Liver, ageing and disease. Nat Rev Gastroenterol Hepatol 2025;22:680-95. [Crossref] [PubMed]
15. Ngcobo NN. Influence of Ageing on the Pharmacodynamics and Pharmacokinetics of Chronically Administered Medicines in Geriatric Patients: A Review. Clin Pharmacokinet 2025;64:335-67. [Crossref] [PubMed]
16. Spring LM, Fell G, Arfe A, et al. Pathologic Complete Response after Neoadjuvant Chemotherapy and Impact on Breast Cancer Recurrence and Survival: A Comprehensive Meta-analysis. Clin Cancer Res 2020;26:2838-48. [Crossref] [PubMed]
17. Yeniceri M, Senoymak MC, Bas S, et al. The effects of agmatine treatment on cisplatin-induced hepatotoxicity: an experimental rat study. Sao Paulo Med J 2026;144:e20253392. [Crossref] [PubMed]
18. Burstein HJ, Curigliano G, Thürlimann B, et al. Customizing local and systemic therapies for women with early breast cancer: the St. Gallen International Consensus Guidelines for treatment of early breast cancer 2021. Ann Oncol 2021;32:1216-35.
19. Benitez-Aurioles J, Wynants L, Peek N, et al. The continuous net benefit: assessing the clinical utility of prediction models when informing a continuum of decisions. Diagn Progn Res 2026;10:8. [Crossref] [PubMed]
20. Hesketh PJ, Kris MG, Basch E, et al. Antiemetics: ASCO Guideline Update. J Clin Oncol 2020;38:2782-97. [Crossref] [PubMed]
21. Tang Y, Xu L, Zhang G, et al. Survival analysis and prognostic nomogram for patients with cholangiocarcinoma after radical resection in Asia. Eur J Surg Oncol 2024;50:108659. [Crossref] [PubMed]
22. Van Calster B, Wynants L, Verbeek JFM, et al. Reporting and Interpreting Decision Curve Analysis: A Guide for Investigators. Eur Urol 2018;74:796-804. [Crossref] [PubMed]
23. Kerr KF, Brown MD, Zhu K, et al. Assessing the Clinical Impact of Risk Prediction Models With Decision Curves: Guidance for Correct Interpretation and Appropriate Use. J Clin Oncol 2016;34:2534-40. [Crossref] [PubMed]
24. Lindor KD, Bowlus CL, Boyer J, et al. Primary Biliary Cholangitis: 2018 Practice Guidance from the American Association for the Study of Liver Diseases. Hepatology 2019;69:394-419. [Crossref] [PubMed]
25. Runowicz CD, Leach CR, Henry NL, et al. American Cancer Society/American Society of Clinical Oncology Breast Cancer Survivorship Care Guideline. CA Cancer J Clin 2016;66:43-73. [Crossref] [PubMed]
26. Gobbini E, Ezzalfani M, Dieras V, et al. Time trends of overall survival among metastatic breast cancer patients in the real-life ESME cohort. Eur J Cancer 2018;96:17-24. [Crossref] [PubMed]
27. Cirulli ET, Nicoletti P, Abramson K, et al. A Missense Variant in PTPN22 is a Risk Factor for Drug-induced Liver Injury. Gastroenterology 2019;156:1707-1716.e2. [Crossref] [PubMed]
28. Nafchi HM, Solatzadeh H, Hajimaghsoudi E, et al. Personalizing cancer therapy: the role of pharmacogenetics in overcoming drug resistance and toxicity. Mol Biol Rep 2025;52:785. [Crossref] [PubMed]
29. Kleiner DE. The pathology of drug-induced liver injury. Semin Liver Dis 2009;29:364-72. [Crossref] [PubMed]
30. Korde LA, Somerfield MR, Carey LA, et al. Neoadjuvant Chemotherapy, Endocrine Therapy, and Targeted Therapy for Breast Cancer: ASCO Guideline. J Clin Oncol 2021;39:1485-505. [Crossref] [PubMed]
31. Cardoso F, Kyriakides S, Ohno S, et al. Early breast cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up†. Ann Oncol 2019;30:1194-220. [Crossref] [PubMed]
32. Böhm-Hustede AK, Lubasch JS, Hoogestraat AT, et al. Barriers and facilitators to the implementation and adoption of computerised clinical decision support systems: an overview of reviews. Syst Rev 2026;15:166. [Crossref] [PubMed]
33. Emens LA. Breast Cancer Immunotherapy: Facts and Hopes. Clin Cancer Res 2018;24:511-20. [Crossref] [PubMed]
34. Modi S, Saura C, Yamashita T, et al. Trastuzumab Deruxtecan in Previously Treated HER2-Positive Breast Cancer. N Engl J Med 2020;382:610-21. [Crossref] [PubMed]