JCO Clinical Cancer Informatics

Historic Clinical Trial External Control Arm Provides Actionable GEN-1 Efficacy Estimate Before a Randomized Trial

PURPOSE To inform continued development of the novel immune agent GEN-1, we compared ovarian cancer patients' end points from a neoadjuvant single-arm phase IB study with those of similar historic clinical trial (HCT) patients who received standard neoadjuvant chemotherapy. METHODS Applying OVATION-1 trial (ClinicalTrials.gov identifier: NCT02480374 ) inclusion and exclusion criteria to Medidata HCT data, we identified historical trial patients for comparison. Integrating patient-level Medidata historic trial data (N = 41) from distinct neoadjuvant ovarian phase I-III trials with patient-level OVATION-1 data (N = 18), we selected Medidata patients with similar baseline characteristics as OVATION-1 patients using propensity score methods to create an external control arm (ECA). RESULTS Fifteen OVATION-1 patients (15 of 18, 83%) were matched to 15 (37%, 15 of 41) Medidata historical trial control patients. Matching attenuated preexisting differences in attributes between the groups. The median progression-free survival time was not reached by the OVATION-1 group and was 15.8 months (interquartile range, 11.40 months to nonestimable) for the ECA. The hazard of progression was 0.53 (95% CI, 0.16 to 1.73), favoring GEN-1 patients. Compared with ECA patients, OVATION-1 patients had more nausea, fatigue, chills, and infusion-related reactions. CONCLUSION Comparing results of a single-arm early-phase trial to those of a rigorously matched HCT ECA yielded insights regarding comparative efficacy prior to a randomized controlled trial. The effect size estimate itself informed both the decision to continue development and the randomized phase II trial (ClinicalTrials.gov identifier: NCT03393884 ) sample size. The work illustrates the potential of HCT data to inform drug development.

Analytical Validation of a Deep Neural Network Algorithm for the Detection of Ovarian Cancer

PURPOSE Early detection of ovarian cancer, the deadliest gynecologic cancer, is crucial for reducing mortality. Current noninvasive risk assessment measures include protein biomarkers in combination with other clinical factors, which vary in their accuracy. Machine learning can be applied to optimizing the combination of these features, leading to more accurate assessment of malignancy. However, the low prevalence of the disease can make rigorous validation of these tests challenging and can result in unbalanced performance. METHODS MIA3G is a deep feedforward neural network for ovarian cancer risk assessment, using seven protein biomarkers along with age and menopausal status as input features. The algorithm was developed on a heterogenous data set of 1,067 serum specimens from women with adnexal masses (prevalence = 31.8%). It was subsequently validated on a cohort almost twice that size (N = 2,000). RESULTS In the analytical validation data set (prevalence = 4.9%), MIA3G demonstrated a sensitivity of 89.8% and a specificity of 84.02%. The positive predictive value was 22.45%, and the negative predictive value was 99.38%. When stratified by cancer type and stage, MIA3G achieved sensitivities of 94.94% for epithelial ovarian cancer, 76.92% for early-stage cancer, and 98.04% for late-stage cancer. CONCLUSION The balanced performance of MIA3G leads to a high sensitivity and high specificity, a combination that may be clinically useful for providers in evaluating the appropriate management strategy for their patients. Limitations of this work include the largely retrospective nature of the data set and the unequal, albeit random, assignment of histologic subtypes between the training and validation data sets. Future directions may include the addition of new biomarkers or other modalities to strengthen the performance of the algorithm.

Leveraging an Informatics Approach to Identify an Unmet Clinical Need for BRCA1/2 Testing Among Patients With Ovarian Cancer

PURPOSE Although BRCA1/ 2 testing in ovarian cancer improves outcomes, it is vastly underutilized. Scalable approaches are urgently needed to improve genomically guided care. METHODS We developed a Natural Language Processing (NLP) pipeline to extract electronic medical record information to identify recipients of BRCA testing. We applied the NLP pipeline to assess testing status in 308 patients with ovarian cancer receiving care at a National Cancer Institute Comprehensive Cancer Center (main campus [MC] and five affiliated clinical network sites [CNS]) from 2017 to 2019. We compared characteristics between (1) patients who had/had not received testing and (2) testing utilization by site. RESULTS We found high uptake of BRCA testing (approximately 78%) from 2017 to 2019 with no significant differences between the MC and CNS. We observed an increase in testing over time (67%-85%), higher uptake of testing among younger patients (mean age tested = 61 years v untested = 65 years, P = .01), and higher testing among Hispanic (84%) compared with White, Non-Hispanic (78%), and Asian (75%) patients ( P = .006). Documentation of referral for an internal genetics consultation for BRCA pathogenic variant carriers was higher at the MC compared with the CNS (94% v 31%). CONCLUSION We were able to successfully use a novel NLP pipeline to assess use of BRCA testing among patients with ovarian cancer. Despite relatively high levels of BRCA testing at our institution, 22% of patients had no documentation of genetic testing and documentation of referral to genetics among BRCA carriers in the CNS was low. Given success of the NLP pipeline, such an informatics-based approach holds promise as a scalable solution to identify gaps in genetic testing to ensure optimal treatment interventions in a timely manner.

Algorithm to Identify Incident Epithelial Ovarian Cancer Cases Using Claims Data

PURPOSE To create an algorithm to identify incident epithelial ovarian cancer cases in claims-based data sets and evaluate performance of the algorithm using SEER-Medicare claims data. METHODS We created a five-step algorithm on the basis of clinical expertise to identify incident epithelial ovarian cancer cases using claims data for (1) ovarian cancer diagnosis, (2) receipt of platinum-based chemotherapy, (3) no claim for platinum-based chemotherapy but claim for tumor debulking surgery, (4) removed cases with nonplatinum chemotherapy, and (5) removed patients with prior claims with personal history of ovarian cancer code to exclude prevalent cases. We evaluated algorithm performance using SEER-Medicare claims data by creating four cohorts: incident epithelial ovarian cancer, a 5% random sample of cancer-free Medicare beneficiaries, a 5% random sample of incident nonovarian cancer, and prevalent ovarian cancer cases. RESULTS Using SEER tumor registry data as the gold standard, our algorithm correctly classified 89.9% of incident epithelial ovarian cancer cases (cohort n = 572) and almost 100% of cancer-free controls (n = 97,127), nonovarian cancer (n = 714), and prevalent ovarian cancer cases (n = 3,712). The overall algorithm sensitivity was 89.9%, the positive predictive value was 93.8%, and the specificity and negative predictive value were > 99.9%. Patients were more likely to be correctly classified as incident ovarian cancer if they had stage III or IV disease compared with early stage I or II disease (93.5% v 83.7%, P < .01), and grade 1-4 compared with unknown grade tumors (93.8% v 81.4%, P < .01). CONCLUSION Our algorithm correctly identified most incident epithelial ovarian cancer cases, especially those with advanced disease. This algorithm will facilitate research in other claims-based data sets where cancer registry data are unavailable.

Gynecologic Survivorship Tool: Development, Implementation, and Symptom Outcomes

PURPOSE To describe the development and implementation of a new digital health clinical tool (Gynecologic Survivorship Tool [GST]) for symptom management of women surgically treated for gynecologic cancer; to assess its feasibility; and to conduct a retrospective review of the data. MATERIALS AND METHODS The GST was developed on the basis of a comprehensive review of the literature, multidisciplinary expert opinion, and feedback from women with a history of gynecologic cancer. It is composed of 17 questions addressing six main categories—gynecologic health (abnormal bleeding/pain), lymphedema, vaginal/vulvar dryness, sexual health, menopause (hot flushes/sleep difficulties), and bowel/urinary issues. An electronic version using the Memorial Sloan Kettering Cancer Center Engage platform was piloted in two clinics for patients with endometrial or cervical cancer. Health information was generated into clinical summaries and identified concerns for actionable response. Associations of symptom and survey time point were assessed by longitudinal models using generalized estimating equations. RESULTS From January 1, 2019, to February 29, 2020, 3,357 GST assessments were assigned to 1,405 patients, with a 71% completion rate (90% within 5 minutes). Sixty-eight percent were performed at home via a patient portal, 32% at follow-ups using a clinic iPad. The most common symptoms were bowel problems, swelling/fluid, pain during examination, vaginal or vulvar dryness, and vaginal bleeding. Implementation challenges included improving patient compliance and ensuring that reports were reviewed by all clinical teams. We developed screening e-mails detailing patients whose assessments were due, planned training sessions for multidisciplinary teams, and incorporated feedback on methods for reviewing symptoms reports. CONCLUSION The GST demonstrated feasibility, a high completion rate, and minimal time commitment. It was an effective electronic reporting mechanism for patients, enabling the medical team to develop specific strategies for alleviating bothersome symptoms during follow-up.

Deep Learning-Based Dynamic Risk Prediction of Venous Thromboembolism for Patients With Ovarian Cancer in Real-World Settings From Electronic Health Records

PURPOSE Patients with epithelial ovarian cancer (EOC) have an elevated risk for venous thromboembolism (VTE). To assess the risk of VTE, models were developed by statistical or machine learning algorithms. However, few models have accommodated deep learning (DL) algorithms in realistic clinical settings. We aimed to develop a predictive DL model, exploiting rich information from electronic health records (EHRs), including dynamic clinical features and the presence of competing risks. METHODS We extracted EHRs of 1,268 patients diagnosed with EOC from January 2007 through December 2017 at the National Cancer Center, Korea. DL survival networks using fully connected layers, temporal attention, and recurrent neural networks were adopted and compared with multi-perceptron–based classification models. Prediction accuracy was independently validated in the data set of 423 patients newly diagnosed with EOC from January 2018 to December 2019. Personalized risk plots displaying the individual interval risk were developed. RESULTS DL-based survival networks achieved a superior area under the receiver operating characteristic curve (AUROC) between 0.95 and 0.98 while the AUROC of classification models was between 0.85 and 0.90. As clinical information benefits the prediction accuracy, the proposed dynamic survival network outperformed other survival networks for the test and validation data set with the highest time-dependent concordance index (0.974, 0.975) and lowest Brier score (0.051, 0.049) at 6 months after a cancer diagnosis. Our visualization showed that the interval risk fluctuating along with the changes in longitudinal clinical features. CONCLUSION Adaption of dynamic patient clinical features and accounting for competing risks from EHRs into the DL algorithms demonstrated VTE risk prediction with high accuracy. Our results show that this novel dynamic survival network can provide personalized risk prediction with the potential to assist risk-based clinical intervention to prevent VTE among patients with EOC.

Development of an Automatic Rule-Based Algorithm for the Detection of Ovarian Cancer Recurrence From Electronic Health Records

PURPOSE As the onset of cancer recurrence is not explicitly recorded in the electronic health record (EHR), a high volume of manual chart review is required to detect the cancer recurrence. This study aims to develop an automatic rule-based algorithm for detecting ovarian cancer (OC) recurrence on the basis of minimally preprocessed EHR data. METHODS The automatic rule-based recurrence detection algorithm (Auto-Recur), using notes on image reading (positron emission tomography-computed tomography [PET-CT], CT, magnetic resonance imaging [MRI]), biomarker (CA125), and treatment information (surgery, chemotherapy, radiotherapy), was developed to detect the first OC recurrence. Auto-Recur contains three single algorithms (images, biomarkers, treatments) and hybrid algorithms (combinations of the single algorithms). The performance of Auto-Recur was assessed using sensitivity, specificity, and accuracy of the recurrence time detected. The recurrence-free survival probabilities were estimated and compared with the retrospective chart review results. RESULTS The proposed Auto-Recur considerably reduced human resources and time; it saved approximately 1,340 days when scaled to 100,000 patients compared with the conventional retrospective chart review. The hybrid algorithm on the basis of a combination of image, biomarker, and treatment information was the most efficient (sensitivity: 93.4%, specificity: 97.4%) and precisely captured recurrence time (average time error: 8.5 days). The estimated 3-year recurrence-free survival probability (44%) was close to the estimates by the retrospective chart review (45%, log-rank P value = .894). CONCLUSION Our rule-based algorithm effectively captured the first OC recurrence from large-scale EHR while closely approximating the recurrence-free survival estimates obtained by conventional retrospective chart reviews. The study findings facilitate large-scale EHR analysis, enhancing clinical research opportunities.

Breast, Cervical, and Colorectal Cancer Screening Among New Jersey Medicaid Enrollees: 2017-2022

PURPOSE The COVID-19 pandemic disrupted cancer screenings in the United States, with disproportionate impact on health disparity populations. The objective of this study was to examine the impact of the pandemic on routine screening for breast, cervical, and colorectal cancer among Medicaid enrollees. MATERIALS AND METHODS This study is a retrospective, descriptive analysis to estimate the rate of breast, colorectal, and cervical cancer screenings among Medicaid enrollees age 50-75 years in New Jersey. Secondary enrollment and claims from the 2017-2022 Medicaid Management Information System were used. The results were stratified by screening type and socioeconomic factors. Bivariate analysis assessed between-group differences. RESULTS Although April 2020 had the lowest screening rates in the 6-year period, rates for all three cancer types rebounded to prepandemic levels by late summer 2020. In 2022, breast cancer screening rates exceeded previous peaks. However, cervical and colorectal screening rates did not resume their prepandemic trajectories. Key findings comparing 2022 with 2019 were (1) across all three cancer screening groups, the younger group (50-64 years) had a higher screening rate than the older group (65-75 years); (2) Hispanic enrollees consistently had the highest screening rates; (3) the screening rate among dually eligible enrollees increased throughout the pandemic; and (4) there was wide screening variation by geographic region. CONCLUSION Multilevel, multisectoral approaches, including policy and health system strategies, are critical to addressing gaps in care for Medicaid enrollees. Future efforts should focus on bolstering cervical and colorectal cancer screening rates and ensuring equitable access to cancer screening and treatment.

Association of Multiple High-Risk Factors on Observed Outcomes in Real-World Patients With Advanced Ovarian Cancer Treated With First-Line Therapy

PURPOSE To identify risk factors for disease progression or death and assess outcomes by risk categories in real-world patients with advanced ovarian cancer. METHODS This retrospective study included adult patients from a nationwide electronic health record–derived deidentified database with stage III/IV ovarian cancer who received first-line therapy and had ≥12 weeks of follow-up after index date (end of first-line therapy). Factors predictive of time to next treatment and overall survival (OS) were assessed. Patients were grouped according to the cumulative number of high-risk factors present (stage IV disease, no debulking surgery or neoadjuvant therapy and interval debulking surgery, visible residual disease after surgery, and breast cancer gene [ BRCA] wild-type disease/unknown BRCA status), and time to next treatment and OS were assessed. RESULTS Region of residence, disease stage, histology, BRCA status, surgery modality, and visible residual disease were significant predictors of time to next treatment; age, Eastern Cooperative Oncology Group performance status, disease stage, BRCA status, surgery modality, visible residual disease, and platelet levels were significant predictors of OS (N = 1,920). Overall, 96.4%, 74.1%, and 40.3% of patients had at least 1, 2, or 3 high-risk factors, respectively; 15.7% of patients had all four high-risk factors. Observed median time to next treatment was 26.4 months (95% CI, 17.1 to 49.2) in patients with no high-risk factors and 4.6 months (95% CI, 4.1 to 5.7) in patients with four high-risk factors. Observed median OS was shorter among patients with more high-risk factors. CONCLUSION These results underscore the complexity of risk assessment and demonstrate the importance of assessing a patient's cumulative risk profile rather than the impact of individual high-risk factors. They also highlight the potential for bias in cross-trial comparisons of median progression-free survival because of differences in risk-factor distribution among patient populations.

Benefit From Fractionated Dose-Dense Chemotherapy in Patients With Poor Prognostic Ovarian Cancer: ICON-8 Trial

PURPOSE An international meta-analysis identified a group of patients with advanced epithelial ovarian cancer (EOC) with a very poor survival because of two unfavorable features: (1) a poor chemosensitivity defined by an unfavorable modeled CA-125 ELIMination rate constant K (KELIM) score <1.0 with the online calculator CA-125—Biomarker Kinetics, and (2) an incomplete debulking surgery. We assumed that patients belonging to this poor prognostic group would benefit from a fractionated densified chemotherapy regimen. METHODS The data set of ICON-8 phase III trial (ClinicalTrials.gov identifier: NCT01654146 ), where patients with EOC were treated with the standard three-weekly, or the weekly dose-dense, carboplatin-paclitaxel regimens and debulking primary surgery (immediate primary surgery [IPS] or delayed primary [or interval] surgery [DPS]), was investigated. The association between treatment arm efficacy, standardized KELIM (scored as favorable ≥1.0, or unfavorable <1.0), and surgery completeness was assessed by univariate/multivariate analyses in IPS and DPS cohorts. RESULTS Of 1,566 enrolled patients, KELIM was calculated with the online model in 1,334 with ≥3 CA-125 available values (85%). As previously reported, both KELIM and surgery completeness were complementary prognostic covariates, and could be combined into three prognostic groups with large OS differences: (1) good if favorable KELIM and complete surgery; (2) intermediate if either unfavorable KELIM or incomplete surgery; and (3) poor if unfavorable KELIM and incomplete surgery. Weekly dose-dense chemotherapy was associated with PFS/OS improvement in the poor prognostic group in both the IPS cohort (PFS: hazard ratio [HR], 0.50; 95% CI, 0.31 to 0.79; OS: HR, 0.58; 95% CI, 0.35 to 0.95) and the DPS cohort (PFS: HR, 0.53; 95% CI, 0.37 to 0.76; OS: HR, 0.57; 95% CI, 0.39 to 0.82). CONCLUSION Fractionated dose-dense chemotherapy might be beneficial for patients belonging to the poor prognostic group characterized by lower tumor chemosensitivity assessed with the online calculator CA-125—Biomarker Kinetics and incomplete debulking surgery. Further investigation in the future SALVOVAR trial is warranted.

Prompt Engineering for Eastern Cooperative Oncology Group Status Extraction: Comparing Large Language Model Techniques

PURPOSE Eastern Cooperative Oncology Group (ECOG) performance status is critical for cancer patient management, yet it is often documented only in unstructured clinical notes. This study compares several approaches to extract ECOG status from oncology notes, focusing on advanced prompting techniques for large language models (LLMs). METHODS We evaluated four ECOG extraction approaches on unstructured clinical notes from patients with non–small cell lung cancer, multiple myeloma, or ovarian cancer (2017-2021). The approaches were a rule-based natural language processing algorithm, simple LLM prompting, and two advanced prompts (chain-of-thought and Double Filtering) using a domain-tuned LLM (LLAMAv3.2). Performance was measured on a binary outcome (any ECOG documented v none) and a three-class outcome (ECOG 0-1 v ≥2 v none) and via an adapted QUEST questionnaire for human evaluation. RESULTS Both CoT and double filtering technique (DFT) achieved 94% accuracy, outperforming the rule-based method (91%) and simple prompting (86%). DFT had the highest specificity (0.91) and positive predictive value (PPV; 0.93), whereas CoT attained the highest sensitivity (0.98). In the QUEST evaluation, DFT and CoT scored higher on output quality, reasoning, bias reduction, and user satisfaction than the simple prompt. DFT received the top satisfaction rating. In the three-class analysis, DFT and CoT again performed best (accuracy 0.91 v 0.87) and DFT was most sensitive for ECOG ≥2 cases. Estimates for ECOG ≥2 remained imprecise because of the small sample (n = 20). All methods sometimes hallucinated ECOG status. CONCLUSION Advanced LLM prompting improved ECOG extraction over basic methods. DFT and CoT each showed specific strengths (DFT had higher PPV and user satisfaction; CoT achieved higher sensitivity). These approaches appear to be generalizable across cancer types. Key implementation considerations include computational cost and human oversight. Overall, advanced prompting can standardize ECOG documentation, accelerate patient cohort identification, and inform personalized treatment planning.

Cost Study of the PlasmaJet Surgical Device Versus Conventional Cytoreductive Surgery in Patients With Advanced-Stage Ovarian Cancer

PURPOSEAdjuvant use of Neutral Argon Plasma (PlasmaJet Surgical Device) during cytoreductive surgery (CRS) for advanced-stage epithelial ovarian cancer improves surgical outcomes. The aim of this study is to examine the costs of adjuvant use of the PlasmaJet during surgery compared with conventional CRS in advanced-stage epithelial ovarian cancer.MATERIALS AND METHODSThe patients were randomly assigned to surgery with or without the PlasmaJet. Analysis of the intra- and extramural health care costs was performed. Costs were divided into three categories: costs of the diagnostic phase (T1), inpatient care up to discharge including costs of surgery (T2), and outpatient care including chemotherapy until 6 weeks after the last cycle of chemotherapy (T3).RESULTSOverall, 327 patients underwent CRS (surgery with PlasmaJet: n = 157; conventional surgery: n = 170). The mean total health costs were significantly higher for CRS with adjuvant use of PlasmaJet compared with conventional CRS (€19,414 v €18,165, P = .017). Costs are divided into costs of the diagnostic phase (€2,034 v €1,974, P = .890), costs of inpatient care (€10,956 v €9,556, P = .003), and costs of outpatient care (€6,417 v €6,628, P = .147).CONCLUSIONMean total health care costs of the use of PlasmaJet in CRS were significantly higher than those for conventional CRS. This difference is fully explained by the additional surgery costs of the use of PlasmaJet. However, surgery with the use of the PlasmaJet leads to a significantly higher percentage of complete CRS and a halving of stomas. A cost-effectiveness analysis will be performed once survival data are available (funded by ZonMw, Trial Register NL62035.078.17).

Utilization of an Electronic Patient-Reported Outcome Platform to Evaluate the Psychosocial and Quality-of-Life Experience Among a Community Sample of Ovarian Cancer Survivors

PURPOSE Novel distress screening approaches using electronic patient-reported outcome (ePRO) measurements are critical for the provision of comprehensive quality community cancer care. Using an ePRO platform, the prevalence of psychosocial factors (distress, post-traumatic growth, resilience, and financial stress) affecting quality of life in ovarian cancer survivors (OCSs) was examined. METHODS A cross-sectional OCS sample from the National Ovarian Cancer Coalition-Illinois Chapter completed web-based clinical, sociodemographic, and psychosocial assessment using well-validated measures: Hospital Anxiety/Depression Scale-anxiety/depression, Post-traumatic Growth Inventory, Brief Resilience Scale, comprehensive score for financial toxicity, and Functional Assessment of Cancer Therapy-Ovarian (FACT-O/health-related quality of life [HRQOL]). Correlational analyses between variables were conducted. RESULTS Fifty-eight percent (174 of 300) of OCS completed virtual assessment: median age 59 (range 32-83) years, 94.2% White, 60.3% married/in domestic partnership, 59.6% stage III-IV, 48.8% employed full-time/part-time, 55.2% had college/postgraduate education, 71.9% completed primary treatment, and median disease duration 6 (range < 1-34) years. On average, OCS endorsed normal levels of anxiety (mean ± standard deviation = 6.9 ± 3.8), depression (4.1 ± 3.6), mild total distress (10.9 ± 8.9), high post-traumatic growth (72.6 ± 21.5), normal resilience (3.7 ± 0.72), good FACT-O-HRQOL (112.6 ± 22.8), and mild financial stress (26 ± 10). Poor FACT-O emotional well-being was associated with greater participant distress ( P < .001). Partial correlational analyses revealed negative correlations between FACT-O-HRQOL and anxiety ( r = –0.65, P < .001), depression ( r = –0.76, P < .001), and total distress ( r = –0.92, P < .001). Yet, high FACT-O-HRQOL was positively correlated with post-traumatic coping ( r = 0.27; P = .006) and resilience ( r = 0.63; P < .001). CONCLUSION ePRO assessment is feasible for identification of unique psychosocial factors, for example, financial toxicity and resilience, affecting HRQOL for OCS. Future investigation should explore large-scale, longitudinal ePRO assessment of the OCS psychosocial experience using innovative measures and community-based advocacy populations.