Sabrina Piedimonte

Evaluating the use of machine learning in endometrial cancer: a systematic review

To review the literature on machine learning in endometrial cancer, report the most commonly used algorithms, and compare performance with traditional prediction models. This is a systematic review of the literature from January 1985 to March 2021 on the use of machine learning in endometrial cancer. An extensive search of electronic databases was conducted. Four independent reviewers screened studies initially by title then full text. Quality was assessed using the MINORS (Methodological Index for Non-Randomized Studies) criteria. P values were derived using the Pearson's Χ Among 4295 articles screened, 30 studies on machine learning in endometrial cancer were included. The most frequent applications were in patient datasets (33.3%, n=10), pre-operative diagnostics (30%, n=9), genomics (23.3%, n=7), and serum biomarkers (13.3%, n=4). The most commonly used models were neural networks (n=10, 33.3%) and support vector machine (n=6, 20%).The number of publications on machine learning in endometrial cancer increased from 1 in 2010 to 29 in 2021.Eight studies compared machine learning with traditional statistics. Among patient dataset studies, two machine learning models (20%) performed similarly to logistic regression (accuracy: 0.85 vs 0.82, p=0.16). Machine learning algorithms performed similarly to detect endometrial cancer based on MRI (accuracy: 0.87 vs 0.82, p=0.24) while outperforming traditional methods in predicting extra-uterine disease in one serum biomarker study (accuracy: 0.81 vs 0.61). For survival outcomes, one study compared machine learning with Kaplan-Meier and reported no difference in concordance index (83.8% vs 83.1%). Although machine learning is an innovative and emerging technology, performance is similar to that of traditional regression models in endometrial cancer. More studies are needed to assess its role in endometrial cancer. CRD42021269565.

Predicting recurrence and recurrence‐free survival in high‐grade endometrial cancer using machine learning

AbstractObjectiveTo develop machine‐learning models to predict recurrence and time‐to‐recurrence in high‐grade endometrial cancer (HGEC) following surgery and tailored adjuvant treatment.MethodsData were retrospectively collected across eight Canadian centers including 1237 patients. Four models were trained to predict recurrence: random forests, boosted trees, and two neural networks. Receiver operating characteristic curves were used to select the best model based on the highest area under the curve (AUC). For time to recurrence, we compared random forests and Least Absolute Shrinkage and Selection Operator (LASSO) model to Cox proportional hazards.ResultsThe random forest was the best model to predict recurrence in HGEC; the AUCs were 85.2%, 74.1%, and 71.8% in the training, validation, and test sets, respectively. The top five predictors were: stage, uterus height, specimen weight, adjuvant chemotherapy, and preoperative histology. Performance increased to 77% and 80% when stratified by Stage III and IV, respectively. For time to recurrence, there was no difference between the LASSO and Cox proportional hazards models (c‐index 71%). The random forest had a c‐index of 60.5%.ConclusionsA bootstrap random forest model may be a more accurate technique to predict recurrence in HGEC using multiple clinicopathologic factors. For time to recurrence, machine‐learning methods performed similarly to the Cox proportional hazards model

Validation of the Integrated Prediction Model algorithm for outcome of cytoreduction in advanced ovarian cancer

We previously developed the Integrated Prediction Model using a 4-step algorithm of unresectable stage IVB, patient factors, surgical resectability, and surgical complexity to predict outcome of 1 (30% vs 11%, 0.045), lower pre-operative albumin (37 vs 40, p<0.001), and higher CA-125 (970 vs 227.5, p<0.001). They also had higher patient factors (2 vs 0, p=0.013), surgical resectability (4 vs 1, p<0.001), and anticipated surgical complexity (8 vs 5, p<0.001). There was no significant difference in outcome of cytoreduction (<1 cm residual disease: 85% for primary surgery vs 87% interval surgery, p=0.12)In this validation cohort, triaging patients with patient factors ≤2, surgical resectability score ≤5, and surgical complexity score ≤9 to primary surgery had a sensitivity of 91% for optimal cytoreduction <1 cm and a specificity of 81%. The positive predictive value, negative predictive value, and accuracy were 83%, 90%, and 86%, respectively. Application of the Integrated Prediction Model would have prevented five patients from receiving suboptimal cytoreduction and triaged them to neoadjuvant chemotherapy. We validated the proposal that a triage algorithm integrating patient factors, surgical complexity, and surgical resectability in advanced ovarian cancer had high sensitivity and specificity to predict optimal cytoreduction <1 cm.

Similar Overall Survival Using Neoadjuvant Chemotherapy or Primary Debulking Surgery in Patients Aged Over 75 Years with High-Grade Ovarian Cancer

To perform a hypothesis-generating evaluation of patient outcomes following neoadjuvant chemotherapy (NACT) compared with those following primary debulking surgery (PDS) in patients over age 75 with high-grade ovarian cancer. This was a retrospective cohort study of consecutive patients aged 75 years and older, with high-grade ovarian cancer. Data were analyzed in SPSS 25.0 using descriptive statistics to characterize groups based on primary treatment modality, Kaplan-Meier survival curves to estimate overall and progression-free survival, and Cox proportional hazards to analyze confounders. Of 429 patients with stages III and IV high-grade ovarian cancer (endometrioid and serous), 71 were aged older than 75 years and met our criteria for inclusion; 58 were treated with NACT while 13 underwent primary debulking. Sixteen patients did not undergo interval debulking following NACT. There were no significant differences in demographic characteristics between the groups. Following NACT, more patients were completely debulked-36.2% versus 21% (P = 0.000)-and had a shorter length of stay (5 vs. 7 d; P = 0.018). Overall survival was similar between the NACT and PDS groups (58.7 vs. 59.7 mo; LR -0.836; P = 0.361) despite lower progression-free survival in the NACT group (25.9 vs. 47.1 mo; P = 0.042; LR 4.31). Both progression-free and overall survival were significantly higher when patients undergoing NACT achieved complete debulking (21.7 and 102.3 mo, respectively) compared with suboptimal debulking (12.03 and 14.2 mo, respectively). In this select group older patients with stage III and IV high-grade ovarian cancers, neoadjuvant chemotherapy may be considered without compromising outcomes and contributes to complete debulking.

BRCA testing in women with high-grade serous ovarian cancer: gynecologic oncologist-initiated testing compared with genetics referral

Up to 15% of patients with high-grade serous ovarian, tubal, or peritoneal carcinoma harbor a mutation in Gynecologic oncology-initiated genetic testing was implemented at a single university hospital center with input and collaboration from gynecological oncologists, nurses, and genetic counselors. All patients diagnosed with high-grade serous ovarian, tubal, or peritoneal carcinoma after August 2017 were offered gynecologic oncologist- initiated genetic testing for a panel of 13 hereditary breast and ovarian cancer susceptibility genes. Data from this group was then compared with a historic cohort of patients who received traditional genetic counseling between January 2014 and August 2017 (control group). Patients that had genetic testing through a clinical trial were excluded. The primary outcome was the uptake of genetic testing in both groups. Secondary outcomes included difference in time from diagnosis to genetic result between both cohorts. Data was analyzed using SPSS 25.0 and medians (ranges) were reported. A total of 152 women with high-grade serous ovarian, tubal, or peritoneal carcinoma were included in this study. Between January 2014 to July 2017 there were 108 patients with high-grade serous ovarian, tubal, or peritoneal carcinoma, among which 50.9% (n=54) underwent genetic testing following referral to genetics. The prevalence of Gynecologic oncologist-initiated genetic testing at the time of high-grade serous ovarian, tubal, or peritoneal carcinoma diagnosis leads to increased uptake and decreased delays in testing compared with referral for traditional genetic counseling.

Treatment outcomes and predictive factors in patients ≥70 years old with advanced ovarian cancer

AbstractObjectiveTo evaluate treatment outcomes, survival, and predictive factors in patients ≥70 with advanced epithelial ovarian cancer (AEOC).MethodsA retrospective single institution cohort study of women ≥70 with Stage III–IV AEOC between 2010 and 2018. Patients had either primary cytoreductive surgery (PCS), neoadjuvant chemotherapy (NACT) with interval cytoreductive surgery (ICS), chemotherapy alone, or no treatment. Demographics, surgical outcome, complications, and survival outcome were compared between groups.ResultsAmong 248 patients, 69 (27.7%) underwent PCS, 99 (39.9%) had ICS, 56 (22.5%) had chemotherapy alone. Twenty‐four (9.6%) remained untreated. Optimal cytoreduction (≤1 cm) was achieved in 72.4% of PCS and 77.8% of NACT/ICS (p = 0.34), without difference in grade ≥3 postoperative complications (15.9% vs. 9.1%, p = 0.37). Progression‐free survival (PFS) was 23.5 months in PCS and 15.0 months in ICS patients (hazard ratio [HR]: 1.4, p = 0.041). Patients in the surgical arms, PCS or ICS, had better 2‐year overall survival (OS) compared to chemotherapy alone (79%, 68%, 41%, respectively, HR: 3.58, p &lt; 0.001). In a subgroup analysis, patients ≥80 had improved 2‐year OS when treated with NACT compared to PCS (82% vs. 57%) and a trend toward improved PFS. Age, stage, and CA‐125 were determinants of undergoing PCS.ConclusionIn patients ≥70 with AEOC, surgery should not be deferred based on age alone. Fit, well selected patients ≥70 can benefit from PCS, while patients ≥80 might benefit from NACT over PCS.

Using a machine learning algorithm to predict outcome of primary cytoreductive surgery in advanced ovarian cancer

AbstractObjectiveTo develop a machine learning (ML) algorithm to predict outcome of primary cytoreductive surgery (PCS) in patients with advanced ovarian cancer (AOC)MethodsThis retrospective cohort study included patients with AOC undergoing PCS between January 2017 and February 2021. Using radiologic criteria, patient factors (age, CA‐125, performance status, BRCA) and surgical complexity scores, we trained a random forest model to predict the dichotomous outcome of optimal cytoreduction (&lt;1 cm) and no gross residual (RD = 0 mm) using JMP‐Pro 15 (SAS). This model is available at https://ipm-ml.ccm.sickkids.ca.ResultsOne hundred and fifty‐one patients underwent PCS and randomly assigned to train (n = 92), validate (n = 30), or test (n = 29) the model. The median age was 58 (27–83). Patients with suboptimal cytoreduction were more likely to have an Eastern Cooperative Oncology Group 3–4 (11% vs. 0.75%, p = 0.004), lower albumin (38 vs. 41, p = 0.02), and higher CA125 (1126 vs. 388, p = 0.012) than patients with optimal cytoreduction (n = 133). There were no significant differences in age, histology, stage, or BRCA status between groups. The bootstrap random forest model had AUCs of 99.8% (training), 89.6%(validation), and 89.0% (test). The top five contributors were CA125, albumin, diaphragmatic disease, age, and ascites. For RD = 0 mm, the AUCs were 94.4%, 52%, and 84%, respectively.ConclusionOur ML algorithm demonstrated high accuracy in predicting optimal cytoreduction in patients with AOC selected for PCS and may assist decision‐making.