Machine Learning-Based Mortality Prediction Model for Critically Ill Cancer Patients Admitted to the Intensive Care Unit (CanICU).

BACKGROUND: Although cancer patients are increasingly admitted to the intensive care unit (ICU) for cancer- or treatment-related complications, improved mortality prediction remains a big challenge. This study describes a new ML-based mortality prediction model for critically ill cancer patients admitted to ICU. PATIENTS AND METHODS: We developed CanICU, a machine learning-based 28-day mortality prediction model for adult cancer patients admitted to ICU from Medical Information Mart for Intensive Care (MIMIC) database in the USA (n = 766), Yonsei Cancer Center (YCC, n = 3571), and Samsung Medical Center in Korea (SMC, n = 2563) from 2 January 2008 to 31 December 2017. The accuracy of CanICU was measured using sensitivity, specificity, and area under the receiver operating curve (AUROC). RESULTS: A total of 6900 patients were included, with a 28-day mortality of 10.2%/12.7%/36.6% and a 1-year mortality of 30.0%/36.6%/58.5% in the YCC, SMC, and MIMIC-III cohort. Nine clinical and laboratory factors were used to construct the classifier using a random forest machine-learning algorithm. CanICU had 96% sensitivity/73% specificity with the area under the receiver operating characteristic (AUROC) of 0.94 for 28-day, showing better performance than current prognostic models, including the Acute Physiology and Chronic Health Evaluation (APACHE) or Sequential Organ Failure Assessment (SOFA) score. Application of CanICU in two external data sets across the countries yielded 79-89% sensitivity, 58-59% specificity, and 0.75-0.78 AUROC for 28-day mortality. The CanICU score was also correlated with one-year mortality with 88-93% specificity. CONCLUSION: CanICU offers improved performance for predicting mortality in critically ill cancer patients admitted to ICU. A user-friendly online implementation is available and should be valuable for better mortality risk stratification to allocate ICU care for cancer patients.

Effect of High-Versus Low-Frequency of Abdominopelvic Computed Tomography Follow-Up Testing on Overall Survival in Patients With Stage II Or III Colon Cancer.

BACKGROUND: Intensive surveillance of colon cancer by using the abdominopelvic computed tomography (AP-CT) is common in real world practice; however, it is still unclear whether high-frequency surveillance using AP-CT in patients with these risk factors is superior to that in the low-frequency surveillance. PATIENTS AND METHODS: We retrospectively reviewed 1803 patients with stage II-III colon cancer receiving curative surgery between January 1, 2005 to December 31, 2015. We evaluated the average scan-to-scan intervals of postoperative AP-CT testing and assigned patients with an interval of 5 to 8 and 9 to 13 months to the high-frequency (HF) and low-frequency (LF) groups, respectively. The cutoff value of preoperative and postoperative CEA levels was 5 ng/mL. We also applied propensity score matching (PSM) and inverse probability of treatment weighting to adjust clinicopathologic differences between the 2 groups. RESULTS: We matched 1:1 for each surveillance group yielding a cohort of 776 matched patients. After PSM, Baseline demographics were overall well balanced between 2 groups. Stage III (OR, 2.00; 95% Confidence interval [CI], 1.21-3.30) and postoperative CEA elevation (OR, 2.30; 95% CI, 1.08-4.92) were independent risk factors of recurrence in multivariate analyses. Patient in the HF group had more surgery plus chemo- or radiotherapy as postrecurrence treatment than patient in the LF group (46.2% vs. 23.1%, P = .017). This trend was retained after PSM, although it is not significant (44.4% vs. 23.1%, P = .060). However, survival outcomes of high-frequency AP-CT surveillance were not superior to those of low-frequency surveillance in all subgroups, including stage III (HR 0.99, 95% CI 0.40-2.47) and postoperative CEA elevation (HR 1.36, 95% CI 0.45-4.11). CONCLUSION: High-frequency AP-CT testing is associated with a higher proportion of surgery plus chemo- or radiotherapy as postrecurrence treatment, without improvement in 5-year overall survival.

Detection of circulating KRAS mutant DNA in extracellular vesicles using droplet digital PCR in patients with colon cancer.

Background: Extracellular vesicles secreted by tumor cells contain double-stranded DNA called extracellular vesicle DNA (evDNA). EvDNA is genomic DNA that reflects cancer driver mutations. However, the significance of evDNA analysis in the diagnosis and surveillance of colon cancer remains unclear. This study aimed to investigate the clinical utility of extracellular vesicles and evDNA isolated from the plasma of colon cancer patients harboring KRAS G12D and G13D mutations. Methods: Cell-free DNA (cfDNA) and evDNA were collected from the plasma of 30 patients with colon cancer. KRAS mutation status (G12D and G13D) was detected using a droplet digital polymerase chain reaction assay (ddPCR). Sensitivity and specificity were evaluated in patients with wild-type KRAS tumors. Mutation status was correlated with carcinoembryonic antigen (CEA) levels and overall survival (OS). Results: Thirty cfDNA and evDNA pairs showed a KRAS fractional abundance (FA) ranging from 0 to 45.26% and 0 to 83.81%, respectively. When compared with eight wild-type KRAS samples, cfDNA exhibited 70% sensitivity and 100% specificity, whereas evDNA achieved 76.67% sensitivity and 100% specificity. The concentration of evDNA was significantly lower than that of cfDNA, but it obtained a higher FA than cfDNA, while showing a positive correlation with CEA. Conclusions: Our findings demonstrate the feasibility of evDNA as a complementary tool to aid current methods of patient evaluation in the diagnosis and surveillance of colon cancer.