Predictive model for totally implanted venous access ports‑related long‑term complications in patients with lung cancer
- Authors:
- Published online on: May 15, 2024 https://doi.org/10.3892/ol.2024.14459
- Article Number: 326
-
Copyright: © Jia et al. This is an open access article distributed under the terms of Creative Commons Attribution License.
Abstract
Introduction
Lung cancer is the most prevalent malignancy worldwide, the incidence of which has increased annually among the elderly for the past 40 years according to the surveillance, epidemiology and end results (SEER) database (1). Treatment modalities for lung cancer include surgery, radiotherapy, chemotherapy, targeted therapy, antiangiogenic therapy and immunotherapy. Among these treatment modalities, chemotherapy is the cornerstone of adjuvant or palliative therapy. Routinely, the drugs for chemotherapy are administered through a central venous access, such as central venous catheters (CVCs), peripherally inserted catheters (PICCs) or totally implanted venous access ports (TIVAPs). PICCs are often used for short-term treatment (up to 3 months) in the perioperative or intensive care setting, while CVCs and TIVAPs are used for medium to long-term treatments (months to years) such as total parenteral nutrition and chemotherapy. Potential complications of central venous accesses usually include short-term (≤30 days of insertion) and long-term complications (>30 days post insertion). Short-term complications may present with symptoms of hemorrhage, hemothorax, pneumothorax, air embolism, cardiac arrhythmias or nerve palsy. Long-term complications mainly include catheter migration, catheter-related thrombosis and infection (2–4). These three types of central venous access reduce repeated venipuncture and avoid focal venous injury and tissue necrosis caused by repeated administration of anticancer therapies. Furthermore, TIVAPs have lower reported rates of catheter-related bloodstream infections (CRBSIs) than the other two types of central access (5). TIVAPs are also more optimal for bathing and swimming, which are restricted with external vascular access, and may appeal to patients concerned about the psychological implications of the presence of visible non-implanted catheters. A meta-analysis by Yeow et al (6) reported that TIVAPs were superior to CVCs and PICCs in terms of complication rate and quality of life without compromising cost-effectiveness. However, insertion or removal of TIVAPs requires a minor surgical operation, and long-term complications are major reasons for removal, which include pocket infection, CRBSI, catheter-related thrombosis and catheter migration (7–9). According to the literature, the incidences of catheter-related infection, thrombosis and migration were 3–10%, 1.06–11.4% and 0.05–3.5%, respectively (10–12). Since management of complications is typically time-consuming and costly, a risk prediction model of related events may be of great value. However, such models are not well established at present. The main focus of the present study is to explore the risk factors for long-term complications after TIVAP placement and construct a predictive model.
Machine learning, with its powerful and efficient computational capabilities, can assist in the diagnosis of diseases through well-trained models (13). Thus far, machine learning has been widely used in foundation and clinical medicine, new drug development and public health (14–16). Machine learning-based abnormality detection overcomes the data imbalance problems encountered in the real healthcare world (17). To the best of the authors' knowledge, the present study is the first to develop a machine learning-based risk prediction model for long-term complications associated with TIVAP implantation in patients with lung cancer.
Materials and methods
Patients and variables
The present retrospective, low-risk study was approved by The Ethics Committee of The First Affiliated Hospital with Nanjing Medical University (Nanjing, China; approval no. 2022-SR-518) and informed patient consent was waived. Clinical data between January, 2016 and December, 2018 were obtained from the inpatient recording system. The patient inclusion criteria were as follows: i) Aged ≥60 years (according to the World Health Organization criteria for the age classification of older individuals in developing countries); ii) pathologically diagnosed with lung cancer and requiring chemotherapy; and iii) had TIVAP implanted by a physician and a nurse in the operation room and aided by ultrasound guided venipuncture and intracavitary electrocardiogram (IC-ECG) guided tip localization (18,19). Patients with large amount of missing data were excluded from the study. There were 666 males and 236 females, with a median age of 67.23±0.52 (range 60–90) years. The primary end point in the present study was long-term complications and all complications were diagnosed by the current gold standard (3,20,21).
By searching the relevant literature (2,22–32), the data collected in the present study were as follows: i) Demographic characteristics, including age, sex, body mass index (BMI), smoking history, thrombus history, history of catheter placement, comorbidities, pleural effusion, cough, pathological type based on WHO standard (33) and tumor stage based on the 8th Edition of the TNM Classification of the International Association for the Study of Lung Cancer (34); ii) laboratory indicators, including white blood cell (WBC) counts, platelet (PLT) counts, hemoglobin (HB), D-dimer, activated partial thromboplastin time (APTT), fibrinogen, albumin (ALB), total bilirubin and creatinine (Cr); iii) medication for lung cancer, including platinum, pemetrexed, bevacizumab, docetaxel, paclitaxel, radiotherapy, leukocyte stimulant and PLT stimulant; and iv) data related to the TIVAPs, including implantation site (right or left side of the body), catheter length and operation time
Model development
The occurrence and detection of abnormalities is the focus of disease prediction. Anomaly detection, also known as outlier detection, was used to build the predictive model in the present study (35). Anomaly detection has a wide range of applications in various scenarios, such as earth sciences, traffic monitoring, early diagnosis of diseases and disease outbreak detection (36–38). To improve the accuracy of the developed model and to identify relevant risk factors that have not yet been recognized, all data were incorporated into the model.
The modeling process, in which machine learning algorithms suitable for supervised learning tasks [including Isolation Forest (iForest), Local Outlier Factor (LOF) and one-class Support Vector Machines (one-class SVM)] were used, was divided into steps. First, the collected data were pre-processed, including missing value processing, feature selection and standardization. The tools used for this step were Pandas version 1.5.2 (https://github.com/pandas-dev/pandas), Numpy version 1.26.0 (https://github.com/numpy/numpy) and Seaborn version 0.12.2 (https://github.com/mwaskom/seaborn). The dataset was then divided into training and test sets, ensuring that model training was performed on a representative sample of data, while retaining an independent dataset for evaluation. Second, the model parameters were adjusted (contamination=28/902) according to the data distribution after initializing the model. Training was then performed on the training set, from which patterns and relationships between input features and implanted outcomes were learned. During the training process, the algorithm parameters were iteratively adjusted to minimize the prediction error using GridSearchCV in the hyperparameter tuning. The tool used for this step was scikit-learn (sklearn) version 1.2.2 (https://github.com/scikit-learn/scikit-learn). Finally, a model evaluation was performed and the receiver operating characteristic curves (ROC) were plotted. The tool for plotting ROC was matplotlib version 3.7.1 (https://github.com/matplotlib/matplotlib). A total of five common metrics were introduced, including accuracy, precision, recall, F1 score and area under the curve (AUC), to evaluate the performance of the model with the test set. In general, the higher the accuracy, precision, and recall of the model and the closer the F1 score is to 1, the more optimal the performance of the model. In addition, Matthew's correlation coefficient (MCC) was introduced, which provided a more accurate assessment of performance with the unbalanced data sets to inform clinical decision-making and risk management. All the models were analyzed using Python version 3.10.5 (https://www.python.org/downloads/release/python-3105/).
Statistical analysis
Continuous variables are presented as the mean ± SD, while categorical variables are presented as numbers (n) and frequencies (%). Comparisons were conducted using χ2 test or Fisher's exact test with scipy version 1.10.1 (https://github.com/scipy/scipy). P<0.05 was considered to indicate a statistically significant difference.
Results
Patient clinical characteristics in the training and test sets and the occurrence of complications
A total of 902 patients were included in the present study (Fig. 1). As shown in Table I, the training and test sets consisted of 631 and 271 patients, respectively. The median age of the training and test set was 67.35±0.42 and 67.10±0.62 years respectively. Males accounted for 74.8% of the training set and 71.6% of the test set. A total of 28 patients (3.1%) developed complications. There were no significant differences between the two sets except in the number of patients administered docetaxel or leukocyte stimulant (P<0.05; Table I), which suggested that the feature distribution of the two datasets was similar after data pre-processing, ensuring that the model had good generalization ability.
Feature selection; correlation analysis with heatmaps
Seaborn was used to construct correlation-based heatmaps (Fig. 2), to perform full factor analysis and to determine any correlations with the occurrence of complications. According to the heatmap, the factors that may be associated with complications include history of thrombosis, comorbidities, pleural fluid, adenocarcinoma, tumor stage, APTT, BMI, site and time of implantation, WBC, HB, D-dimer, ALB, Cr, antineoplastic agents and leukocyte stimulants. Among these factors, those with a correlation coefficient of ≥0.05 were BMI, HB, implantation time, docetaxel and leukocyte stimulants.
One-class SVM model performs the best among the three models
The performance of the one-class SVM model (MCC, 0.078; AUC, 0.62; accuracy, 66.0%) was significantly superior than the iForest (MCC, 0.015; AUC, 0.48; accuracy, 94.0%) and LOF (MCC, −0.017; AUC, 0.51; accuracy, 96.0%; Fig. 3 and Table II) models. The classification reports for patients in category 0 (without complications) and category 1 (with complications) provided more detailed performance metrics (Table II). Overall, the one-class SVM model had a positive MCC, a relatively balanced recall and performed well in the task.
Discussion
Different algorithms have different performances for specific datasets. In the present study, three common anomaly detection algorithms were applied to the same training set and their performances were compared. These three algorithms were chosen due to the following: i) iForest excels at identifying outliers by constructing random trees and isolating anomalies in shorter paths. iForest does not assume any potential data distributions and can effectively handle large datasets (39); ii) LOF is sensitive to local context, which is important in the healthcare setting as subtle changes may indicate abnormalities. Moreover, LOF is less affected by noise and different densities, which is consistent with the inherent variability of medical data (40); and iii) one-class SVM can implement boundary learning, it builds a hyperplane around most data points to isolate a few classes (anomalies). Additionally, one-class SVM is flexible to capture the complex relationships between risk factors by tweaking the kernel function and performs well when dealing with unbalanced datasets (41,42). In summary, iForest, LOF and one-class SVM were selected in the present study due to their effectiveness in detecting anomalies, robustness to noise, processing high-dimensional data and capturing complex relationships.
The early identification of high-risk groups for long-term complications is important to improve the quality of life of patients with cancer and to reduce the waste of medical resources. To the best of our knowledge, the present study was the first to present a model that has been built on top of an anomaly detection algorithm. The results of the present study indicated that the one-class SVM model had the highest performance with an MCC, AUC and accuracy of 0.078, 0.62 and 66.0%, respectively. Reducing the occurrence of complications has always been the focus of healthcare professionals. To date, a number of factors such as catheter material, age, BMI, severe coughing, time interval from first use to placement, site of placement, hypoalbuminemia and leukopenia, have been identified in patients with cancer suffering from TIVAP-related complications (26,31,43–46).
Due to data sparseness, the present study was not able to evaluate the contribution of each variable to classification accuracy. However, the correlation-based heat map suggested that BMI, HB, implantation time, docetaxel and leukocyte stimulants may be closely related to the occurrence of long-term complications following port implantation. These findings were consistent with the real-world observations such that the occurrence and progression of an outcome event are often influenced by a combination of several factors. A previous study assessing the risk of venous thromboembolism in patients with cancer included BMI, HB and WBC count in the risk score (47), suggesting that the relationship between BMI, HB, leukocyte stimulants and catheter related thrombosis should be further studied using prediction models. Adverse reactions to antitumor drugs should also be noted. For instance, docetaxel may cause bone marrow suppression, manifesting as neutropenia, thrombocytopenia or anemia (48). This will undoubtedly increase the incidence of TIVAP-related infections. Hypoalbuminemia was found to be an independent risk factor for infections (26). However, the relationship between albumin correction and improved prognosis was not definitively identified; it can be further explored in larger datasets in the future. In the present study, the duration of implantation was shown to be associated with the development of long-term complications. However, a multicenter prospective French cohort study (ONCOCIP) showed that an average surgery duration of 25 min was not a risk factor (49). Therefore, this factor needs to be verified by further research.
In conclusion, a machine learning-based prediction model for the long-term complications associated with TIVAPs in patients with lung cancer was developed in the present study. The model will help to identify individuals at high risk of complications, which can improve their quality of life and prevent unnecessary waste of medical resources. However, the present study did have several limitations. First, since it was a single-center retrospective study, the generalizability of the conclusions was limited. Second, all of the participants were older (aged ≥60 years) and had a small number of complications. Third, no predictive model was developed for specific complications, which was mainly due to the low complication rate in the dataset. Therefore, predictive models for specific complications among individuals of different ages with other diseases should be developed in the future by building larger sample sets or conducting multicenter collaborative studies.
Acknowledgements
Not applicable.
Funding
The present study was supported by the China Postdoctoral Science Foundation (grant no. 2023M731584), the Young Scholars Fostering Fund of the First Affiliated Hospital of Nanjing Medical University (grant no. PY2021015), the Study on the Effect and Health Economics Evaluation of Home Cardiac Rehabilitation for Heart Failure Based on Digital Medicine (grant no. M2022032) and the Individual Institutional Breakthrough Theory: Theoretical Construction and Testing Based on Local Practice (grant no. 72072083).
Availability of data and materials
The data generated in the present study are included in the figures and/or tables of this article.
Authors' contributions
JJ, XTF, WHZ, YYZ and BQF contributed to the study design. XTF, ZYX and MW contributed to the data acquisition and table organization;. XTF contributed to data analysis. XTF contributed to writing the original draft. JJ contributed to review and editing. YYZ contributed to funding acquisition. JJ, XTF, YYZ and BQF confirm the authenticity of all the raw data. All authors read and approved the final version of the manuscript.
Ethics approval and consent to participate
The research protocol was approved by The Ethics Committee of the First Affiliated Hospital with Nanjing Medical University (Nanjing, China; approval no. 2022-SR-518) and patient informed consent was waived.
Patient consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
References
|
Abdel-Rahman O: Changing epidemiology of elderly small cell lung cancer patients over the last 40 years; a SEER database analysis. Clin Respir J. 12:1093–1099. 2018. View Article : Google Scholar : PubMed/NCBI | |
|
Gallieni M, Pittiruti M and Biffi R: Vascular access in oncology patients. CA Cancer J Clin. 58:323–346. 2008. View Article : Google Scholar : PubMed/NCBI | |
|
Silberzweig JE, Sacks D, Khorsandi AS and Bakal CW; Society of Interventional Radiology Technology Assessment Committee, : Reporting standards for central venous access. J Vasc Interv Radiol. 14((9 Pt 2)): S443–S452. 2003. View Article : Google Scholar : PubMed/NCBI | |
|
Nayeemuddin M, Pherwani AD and Asquith JR: Imaging and management of complications of central venous catheters. Clin Radiol. 68:529–544. 2013. View Article : Google Scholar : PubMed/NCBI | |
|
Maki DG, Kluger DM and Crnich CJ: The risk of bloodstream infection in adults with different intravascular devices: A systematic review of 200 published prospective studies. Mayo Clin Proc. 81:1159–1171. 2006. View Article : Google Scholar : PubMed/NCBI | |
|
Yeow M, Soh S, Yap R, Tay D, Low YF, Goh SSN, Yeo CS and Lo ZJ: A systematic review and network meta-analysis of randomized controlled trials on choice of central venous access device for delivery of chemotherapy. J Vasc Surg Venous Lymphat Disord. 10:1184–1191.e8. 2022. View Article : Google Scholar : PubMed/NCBI | |
|
Walser EM: Venous access ports: Indications, implantation technique, follow-up, and complications. Cardiovasc Intervent Radiol. 35:751–764. 2012. View Article : Google Scholar : PubMed/NCBI | |
|
Tabatabaie O, Kasumova GG, Eskander MF, Critchlow JF, Tawa NE and Tseng JF: Totally implantable venous access devices: A review of complications and management strategies. Am J Clin Oncol. 40:94–105. 2017. View Article : Google Scholar : PubMed/NCBI | |
|
Voog E, Campion L, du Rusquec P, Bourgeois H, Domont J, Denis F, Emmanuel E, Dupuis O, Ganem G, Lafont C, et al: Totally implantable venous access ports: A prospective long-term study of early and late complications in adult patients with cancer. Support Care Cancer. 26:81–89. 2018. View Article : Google Scholar : PubMed/NCBI | |
|
Lebeaux D, Fernández-Hidalgo N, Chauhan A, Lee S, Ghigo JM, Almirante B and Beloin C: Management of infections related to totally implantable venous-access ports: Challenges and perspectives. Lancet Infect Dis. 14:146–159. 2014. View Article : Google Scholar : PubMed/NCBI | |
|
Wildgruber M, Borgmeyer S, Haller B, Jansen H, Gaa J, Kiechle M, Meier R, Ettl J and Berger H: Short-term and long-term outcome of radiological-guided insertion of central venous access port devices implanted at the forearm: A retrospective monocenter analysis in 1704 patients. Eur Radiol. 25:606–616. 2015. View Article : Google Scholar : PubMed/NCBI | |
|
Rouzrokh M, Shamsian BS, Tabari A, Mahmoodi M, Kouranlo J, Manafzadeh G, Arzanian MT, Fallah F, Anoush M and Gorji FA: Totally implantable subpectoral vs. subcutaneous port systems in children with malignant diseases. Arch Iran Med. 12:389–394. 2009.PubMed/NCBI | |
|
Peiffer-Smadja N, Rawson TM, Ahmad R, Buchard A, Georgiou P, Lescure FX, Birgand G and Holmes AH: Machine learning for clinical decision support in infectious diseases: A narrative review of current applications. Clin Microbiol Infect. 26:584–595. 2020. View Article : Google Scholar : PubMed/NCBI | |
|
Reel PS, Reel S, Pearson E, Trucco E and Jefferson E: Using machine learning approaches for multi-omics data analysis: A review. Biotechnol Adv. 49:1077392021. View Article : Google Scholar : PubMed/NCBI | |
|
Vamathevan J, Clark D, Czodrowski P, Dunham I, Ferran E, Lee G, Li B, Madabhushi A, Shah P, Spitzer M and Zhao S: Applications of machine learning in drug discovery and development. Nat Rev Drug Discov. 18:463–477. 2019. View Article : Google Scholar : PubMed/NCBI | |
|
Mooney SJ and Pejaver V: Big data in public health: Terminology, machine learning, and privacy. Annu Rev Public Health. 39:95–112. 2018. View Article : Google Scholar : PubMed/NCBI | |
|
Alloqmani A, Abushark YB and Khan AI: Anomaly detection of breast cancer using deep learning. Arab J Sci Eng. 12:1–26. 2023.PubMed/NCBI | |
|
Pittiruti M, Pelagatti F and Pinelli F: Intracavitary electrocardiography for tip location during central venous catheterization: A narrative review of 70 years of clinical studies. J Vasc Access. 22:778–785. 2021. View Article : Google Scholar : PubMed/NCBI | |
|
Liu Z, Zheng X, Zhen Y, Liu X, Lin F, Ye Z and Liu P: Efficacy, safety, and cost-effectiveness of intracavitary electrocardiography-guided catheter tip placement for totally implantable venous access port. Ann Vasc Surg. 83:168–175. 2022. View Article : Google Scholar : PubMed/NCBI | |
|
Bell T and O'Grady NP: Prevention of central line-associated bloodstream infections. Infect Dis Clin North Am. 31:551–559. 2017. View Article : Google Scholar : PubMed/NCBI | |
|
Baskin JL, Pui CH, Reiss U, Wilimas JA, Metzger ML, Ribeiro RC and Howard SC: Management of occlusion and thrombosis associated with long-term indwelling central venous catheters. Lancet. 374:159–169. 2009. View Article : Google Scholar : PubMed/NCBI | |
|
Chang YF, Lo AC, Tsai CH, Lee PY, Sun S, Chang TH, Chen CC, Chang YS and Chen JR: Higher complication risk of totally implantable venous access port systems in patients with advanced cancer-a single institution retrospective analysis. Palliat Med. 27:185–191. 2013. View Article : Google Scholar : PubMed/NCBI | |
|
Goltz JP, Schmid JS, Ritter CO, Knödler P, Petritsch B, Kirchner J, Hahn D and Kickuth R: Identification of risk factors for catheter-related thrombosis in patients with totally implantable venous access ports in the forearm. J Vasc Access. 13:79–85. 2012. View Article : Google Scholar : PubMed/NCBI | |
|
Ma LI, Liu Y, Wang J, Chang Y, Yu L and Geng C: Totally implantable venous access port systems and associated complications: A single-institution retrospective analysis of 2,996 breast cancer patients. Mol Clin Oncol. 4:456–460. 2016. View Article : Google Scholar : PubMed/NCBI | |
|
Vidal M, Genillon JP, Forestier E, Trouiller S, Pereira B, Mrozek N, Aumeran C and Lesens O: Outcome of totally implantable venous-access port-related infections. Med Mal Infect. 46:32–38. 2016. View Article : Google Scholar : PubMed/NCBI | |
|
Skummer P, Kobayashi K, DeRaddo JS, Blackburn T, Schoeneck M, Patel J and Jawed M: Risk factors for early port infections in adult oncologic patients. J Vasc Interv Radiol. 31:1427–1436. 2020. View Article : Google Scholar : PubMed/NCBI | |
|
Wang YC, Lin PL, Chou WH, Lin CP and Huang CH: Long-term outcomes of totally implantable venous access devices. Support Care Cancer. 25:2049–2054. 2017. View Article : Google Scholar : PubMed/NCBI | |
|
Tsuruta S, Goto Y, Miyake H, Nagai H, Yoshioka Y, Yuasa N and Takamizawa J: Late complications associated with totally implantable venous access port implantation via the internal jugular vein. Support Care Cancer. 28:2761–2768. 2020. View Article : Google Scholar : PubMed/NCBI | |
|
Lin WY, Lin CP, Hsu CH, Lee YH, Lin YT, Hsu MC and Shao YY: Right or left? Side selection for a totally implantable vascular access device: A randomised observational study. Br J Cancer. 26:932–937. 2017. View Article : Google Scholar | |
|
Tian L, Li W, Su Y, Gao H, Yang Q, Lin P, Wang L, Zeng J and Li Y: Risk factors for central venous access device-related thrombosis in hospitalized children: A systematic review and meta-analysis. Thromb Haemost. 121:625–640. 2021. View Article : Google Scholar : PubMed/NCBI | |
|
Intagliata E, Basile F and Vecchio R: Totally implantable catheter migration and its percutaneous retrieval: Case report and review of the literature. G Chir. 37:n211–n215. 2017.PubMed/NCBI | |
|
Chen Y, Tsang YS, Chou X, Hu J and Xia Q: A lung cancer patient with deep vein thrombosis: A case report and literature review. BMC Cancer. 19:2852019. View Article : Google Scholar : PubMed/NCBI | |
|
Travis WD, Brambilla E, Nicholson AG, Yatabe Y, Austin JHM, Beasley MB, Chirieac LR, Dacic S, Duhig E, Flieder DB, et al: The 2015 world health organization classification of lung tumors: Impact of genetic, clinical and radiologic advances since the 2004 classification. J Thorac Oncol. 10:1243–1260. 2015. View Article : Google Scholar : PubMed/NCBI | |
|
Goldstraw P, Chansky K, Crowley J, Rami-Porta R, Asamura H, Eberhardt WE, Nicholson AG, Groome P, Mitchell A, Bolejack V, et al: The IASLC lung cancer staging project: Proposals for revision of the TNM stage groupings in the forthcoming (Eighth) edition of the TNM classification for lung cancer. J Thorac Oncol. 11:39–51. 2016. View Article : Google Scholar : PubMed/NCBI | |
|
Foorthuis R: On the nature and types of anomalies: A review of deviations in data. Int J Data Sci Anal. 12:297–331. 2021. View Article : Google Scholar : PubMed/NCBI | |
|
Arnaout R, Curran L, Zhao Y, Levine JC, Chinn E and Moon-Grady AJ: An ensemble of neural networks provides expert-level prenatal detection of complex congenital heart disease. Nat Med. 27:882–891. 2021. View Article : Google Scholar : PubMed/NCBI | |
|
Kim B, Kwon K, Oh C and Park H: Unsupervised anomaly detection in MR images using multicontrast information. Med Phys. 48:7346–7359. 2021. View Article : Google Scholar : PubMed/NCBI | |
|
Karadayi Y, Aydin MN and Ogrenci AS: Unsupervised anomaly detection in multivariate spatio-temporal data using deep learning: Early detection of COVID-19 outbreak in Italy. IEEE Access. 8:164155–164177. 2020. View Article : Google Scholar : PubMed/NCBI | |
|
Guo Y, Jiang X, Tao L, Meng L, Dai C, Long X, Wan F, Zhang Y, van Dijk J, Aarts RM, et al: Epileptic seizure detection by cascading isolation forest-based anomaly screening and easyensemble. IEEE Trans Neural Syst Rehabil Eng. 30:915–924. 2022. View Article : Google Scholar : PubMed/NCBI | |
|
Lin CH, Hsu KC, Johnson KR, Luby M and Fann YC: Applying density-based outlier identifications using multiple datasets for validation of stroke clinical outcomes. Int J Med Inform. 132:1039882019. View Article : Google Scholar : PubMed/NCBI | |
|
Mei S and Zhu H: A novel one-class SVM based negative data sampling method for reconstructing proteome-wide HTLV-human protein interaction networks. Sci Rep. 5:80342015. View Article : Google Scholar : PubMed/NCBI | |
|
Schölkopf B, Platt JC, Shawe-Taylor J, Smola AJ and Williamson RC: Estimating the support of a high-dimensional distribution. Neural Comput. 13:1443–1471. 2001. View Article : Google Scholar : PubMed/NCBI | |
|
Wildgruber M, Lueg C, Borgmeyer S, Karimov I, Braun U, Kiechle M, Meier R, Koehler M, Ettl J and Berger H: Polyurethane versus silicone catheters for central venous port devices implanted at the forearm. Eur J Cancer. 59:113–124. 2016. View Article : Google Scholar : PubMed/NCBI | |
|
Xu M, Deng L, Zhu Y, Li Y, Wang F, Li H and Zhou Y: Risk factors of catheter-related infection in unplanned extubation of totally implantable venous-accessportsin tumor patients. Emerg Med Int. 2022:42353162022. View Article : Google Scholar : PubMed/NCBI | |
|
Narducci F, Jean-Laurent M, Boulanger L, El Bedoui S, Mallet Y, Houpeau JL, Hamdani A, Penel N and Fournier C: Totally implantable venous access port systems and risk factors for complications: A one-year prospective study in a cancer centre. Eur J Surg Oncol. 37:913–918. 2011. View Article : Google Scholar : PubMed/NCBI | |
|
Liu Y, Li LL, Xu L, Feng DD, Cao Y, Mao XY, Zheng J, Jin F and Chen B: Comparison between arm port and chest port for optimal vascular access port in patients with breast cancer: A systematic review and meta-analysis. Biomed Res Int. 2020:90829242020.PubMed/NCBI | |
|
Khorana AA, Kuderer NM, Culakova E, Lyman GH and Francis CW: Development and validation of a predictive model for chemotherapy-associated thrombosis. Blood. 111:4902–4907. 2008. View Article : Google Scholar : PubMed/NCBI | |
|
Wabont G, Bergeron S, Gautier S and Barus R: Sex differences in serious adverse drug reactions in patients receiving immunotherapy, targeted therapy, or chemotherapy: A disproportionality analysis of the VigiBase(R). Eur J Clin Pharmacol. 78:1355–1356. 2022. View Article : Google Scholar : PubMed/NCBI | |
|
Decousus H, Bourmaud A, Fournel P, Bertoletti L, Labruyere C, Presles E, Merah A, Laporte S, Stefani L, Piano FD, et al: Cancer-associated thrombosis in patients with implanted ports: A prospective multicenter French cohort study (ONCOCIP). Blood. 132:707–716. 2018. View Article : Google Scholar : PubMed/NCBI |
