Search

394 Background: Previous work by our group has demonstrated that leveraging Machine Learning on diagnostic codes from Electronic Health Records (EHRs), can identify individuals at high-risk for Pancreatic Duct Adenocarcinoma (PDAC), as early as 1 year before current cancer diagnosis. We aim to improve the performance of our existing PDAC risk stratification model, by using an independent, multi-center dataset, and adding lab test features. Methods: EHR data from TriNetX, a federated global health research network, was utilized to develop Logistic Regression (LR) models. Diagnoses and lab test data from 32 different Health Care Organizations in the United States from 2015-2020 was used. PDAC patients ages 60-80 years, were identified using ICD codes, and cross-checked with tumor registry and pathology data to decrease false positives. Only patients with one or more clinical encounter/s, at least 6 months prior to cancer diagnosis, were included. Prediction time cutoffs of 180, 270, and 360 days before PDAC diagnosis were used. Preliminary basic data analysis was initially performed to explore potential lab test features that could be used to improve model performance. The discriminatory capabilities of the LR models were compared using Area Under the Receiver Operating Characteristic Curve (AUC) and 95% Confidence Interval using empirical bootstrap over test data were computed. We used L2-regularized LR, and performed evaluation using cross-validation. We report cross-validation performance. In contrast to prior published work that used predefined feature sets for model development, we incorporated a wide range of indicators, and relied on regularization to address potential overfitting risk. Results: The LR models were trained and evaluated on diagnoses and labs for 25,644 patients (cases= 1352; age-sex paired controls). Lab test administration per patient (i.e., for a given patient, what lab tests were administered and how frequently), was found to be the most valuable feature for improving discrimination. For almost every type of lab test, the average number of administrations per patient was higher for PDAC patients than controls. The top lab tests with highest discriminatory coefficients included glucose, potassium, hematocrit, hemoglobin, sodium, chloride and creatinine. With a 365-day lead time, the diagnoses-based LR obtained a test AUC of 0.58, the lab-test based LR obtained a test AUC of 0.72. The combined diagnoses and lab-test model (“concatenated LR model”) outperformed both of these models, obtaining a test AUC of 0.73. Conclusions: Our findings demonstrate that LR models based on concatenated lab test and diagnoses feature sets (“concatenated LR models”), can outperform both diagnoses-based LR models and lab-test-based LR models, and can be utilized in early prediction of PDAC development.

243 Background: Circulating Cell-free Genome Atlas (CCGA; NCT02889978) is a multi-center, case-control, observational study with longitudinal follow-up to develop a cfDNA assay in which classifiers were trained on whole-genome bisulfite sequencing (WGBS) and targeted methylation (TM) sequencing data for detection of multiple cancer types. Previously, we showed that the fraction of ctDNA fragments (TF) was a stronger predictor of cancer detection than clinical stage and an equivalent predictor for survival. Given that CRC tumors can be described via surface area (TSA) and microscopic tumor extent (microinvasion), CRC was used as a model to examine the biophysical determinants of TF. Methods: Detection of multiple cancers with WGBS at 98% and TM at > 99% specificity, and methods for determining TF, were previously reported. A model to predict the presence of detectable cfDNA fragments for CRC adenocarcinomas of stages I, II, and III included TSA and microinvasion beyond the subserosa. Predictors were combined assuming a linear increase of cfDNA shedding with tumor size, with scaling factors depending on microinvasion. Model parameters were determined for 27 participants (7, 11, 9 for stages I, II, III, resp.) with WGBS and applied to 40 participants (12, 15, 13 for I, II, III, resp.) with TM assay and information on tumor size and microinvasion. Results: CRC detection at stages I/II/III was 33/46, 61/73, 57/74% for WGBS/TM. TF predicted detection with AUC = 97.6. The model predicted TF as TSA multiplied by 3.81*10−6 / mm2 for tumors that invaded beyond the subserosa (p < 0.001). This was 4.4x higher than estimates for tumors below the subserosa. The model trained on the WGBS assay predicted CRC detection in the TM assay with an AUC of 0.844. Conclusions: This model used TSA (number of tumor cells) and microinvasion (bloodstream access) to predict the fraction of CRC ctDNA fragments in blood without needing to account for stage. Tumors not penetrating the subserosa had low ctDNA shedding that likely limited detection. These findings may generalize to other cancer types, providing principles to predict ctDNA shedding and thus cancer detectability based on microinvasion and surface area. Clinical trial information: NCT02889978.