From:  Survival prediction in triple-negative breast cancer: a Cox model with fairness assessment using ISO/IEC TR 24027:2021 in a MENA cohort

 Summary of key studies on survival prediction in triple-negative breast cancer (TNBC), including cohort characteristics, feature sets, modeling approaches and performance metrics.

ReferenceAuthor, dateOutcomeData specificationsFeaturesMethodEvaluationResults
[5]Zarean Shahraki et al., 2023Survival prediction in different molecular subgroups in TNBC3,580 patients (1991–2021), Shahid Beheshti University’s Cancer Research Center, IranDemographic features such as age, clinical features such as tumor stage, and treatment-related features such as the type of surgeryTime-to-event deep learning-based models (N-net survival model)Cross-validation5-year survival probability risk: 84%, 10-year: 74%, 15-year: 66%
[6]Zhou and Chen, 2023Investigate prognostic factors and survival outcomes in TNBC33,654 TNBC patients, SEER database, USADemographic features such as age, clinical features such as histologic grade, and treatment-related features such as surgery of the primary siteUnivariable and multivariable Cox regressionHold-out method, internal and external validationC-index: 0.79 (internal), 0.79 (external)
Better outcomes for younger, white, married individuals with lower disease grade and stage
[7]Meng et al., 2023Predict 10-year survival outcomes in non-metastatic TNBC patients32,836 non-metastatic TNBC patients (2010–2019), SEER database, USADemographic features such as age at diagnosis, clinical features such as histologic grade, and treatment-related features such as surgeryThe least absolute shrinkage and selection operator (Lasso) regressionHold-out methodAUC: 0.86 for 1 year after diagnosis, 0.75 for 10 years after diagnosis
[8]Chen et al., 2022Determine BCSS and overall survival (OS) in female patients with metastatic TNBC1,962 metastatic TNBC female patients (2010–2017), SEER database, USADemographic features such as age at diagnosis, clinical features such as tumor stage, and the feature of metastasis in different organs, such as bone metastasisCox regressionHold-out method1-year BCSS: 56.55%, 2-year: 29.86%, 3-year: 19.71%
1-year OS: 52.55%, 2-year: 26.03%, 3-year: 16.69%
[9]Sheng et al., 2022Predict disease free survival (DFS) models using a combination of clinical features, ultrasound, and mammography636 TNBC patients (2011–2015), ChinaDemographic features such as age, clinical features such as histologic grade, and treatment-related features such as adjuvant chemotherapyMultivariate logistic regression, Cox regressionHold-out method, 0.69 (validation)C-index: 0.69 (training), 0.69 (validation)
1-year DFS AUC: 0.78 (training), 0.49 (validation)
3-year DFS AUC: 0.70 (training), 0.72 (validation)
5-year DFS AUC: 0.69 (both training and validation)
[10]Huang et al., 2022Assess the impact of chemotherapy on survival and predict OS in TNBC4,696 TNBC patients (2010–2016), SEER database, USADemographic features such as age, clinical features such as tumor status, and treatment-related features such as surgical approachNine ML models, with a focus on LightGBMTen-fold cross-validationAUC: 0.81, accuracy: 85% for OS
[11]Haiderali et al., 2021Predict event free survival (EFS), time to recurrence, and OS in TNBC patients450 TNBC patients (236 received neoadjuvant treatment, 72 received neoadjuvant therapy with surgery, 102 had only surgery, and 40 did not receive any treatment or surgery), USADemographics such as age at initial TNBC diagnosis, treatment-related features such as duration from initial TNBC diagnosis to surgery, and information about the initial diagnosis of the diseaseCox regressionWithin-group comparisons by pCR statusMortality rates: surgery (23.6%), neoadjuvant (16.5%), surgery + neoadjuvant (14.7%), no treatment (35%)
[12]Cui et al., 2021Investigate 3-year and 5-year OS and BCSS in TNBC patients4,593 TNBC patients aged 18–45 (2010–2015), SEER database, USADemographic features such as age, clinical features such as tumor size, and treatment-related features such as breast surgeryCox regressionHold-out methodAUC:
0.78 (3-year OS)
0.77 (5-year OS)
0.78 (3-year BCSS)
0.77 (5-year BCSS)
[13]Shi et al., 2019Predict the prognosis of TNBC379 TNBC patients from China (2008–2014), First Affiliated Hospital of Wenzhou Medical University, ChinaDemographic features such as age, clinical, and treatment-related features such as surgery type, blood factor features such as neutrophilsUnivariate and multivariate Cox regressionHold-out methodC-index:
0.69 (DFS)
0.740 (OS)
5-year ROC:
0.66 (DFS)
0.70 (OS)
[14]Yang et al., 2019Predict DFS and OS for operable TNBC based on Chinese breast cancer data296 invasive operable TNBC patients (2002–2014), Sun Yat-sen Memorial Hospital (108 patients, The Second Xiangya Hospital and Peking University Shenzhen Hospital, external validation), ChinaDemographic features such as age, clinical features such as tumor size, and treatment-related features such as surgery typeCox multivariate analysisBootstrapping, external validationC-index:
0.74 (DFS training)
0.78 (DFS validation)
0.79 (OS training)
0.78 (OS validation)
[15]Guo et al., 2018Predict BCSS and OS in TNBC patients21,419 TNBC patients (2010–2015), SEER 19 Cancer Registry, USADemographic features such as age, clinical features such as gradeLog-rank tests, Cox analysis, competing risk modelHold-out validationC-index:
0.78 (OS internal)
0.79 (BCSS internal)
0.77 (OS external)
0.79 (BCSS external)
1-year BC-specific mortality: 2.7%, 3-year: 12.5%, 5-year: 17.1%
[16]Lin et al., 2018Predict the prognosis of patients with TNBC404 TNBC patients from the Affiliated Union Hospital of Fujian Medical University (2006–2012) for training, 200 patients (2012–2014) for validation, ChinaClinical features such as tumor size, laboratory features such as CA125Univariate and multivariate Cox regressionExternal validationC-index of OS: 0.76 (training), 0.72 (validation)
Return to article view

AUC: area under the curve; C-index: concordance index; pCR: pathological complete response; ROC: receiver operating characteristic.