Background: Machine learning is a broad term encompassing a number of methods that allow the investigator to learn from the data. These methods may permit large real-world databases to be more rapidly translated to applications to inform patient-provider decision making. Methods: This systematic literature review was conducted to identify published observational research of employed machine learning to inform decision making at the patient-provider level. The search strategy was implemented and studies meeting eligibility criteria were evaluated by two independent reviewers. Relevant data related to study design, statistical methods and strengths and limitations were identified; study quality was assessed using a modified version of the Luo checklist. Results: A total of 34 publications from January 2014 to September 2020 were identified and evaluated for this review. There were diverse methods, statistical packages and approaches used across identified studies. The most common methods included decision tree and random forest approaches. Most studies applied internal validation but only two conducted external validation. Most studies utilized one algorithm, and only eight studies applied multiple machine learning algorithms to the data. Seven items on the Luo checklist failed to be met by more than 50% of published studies. Conclusions: A wide variety of approaches, algorithms, statistical software, and validation strategies were employed in the application of machine learning methods to inform patient-provider decision making. There is a need to ensure that multiple machine learning approaches are used, the model selection strategy is clearly defined, and both internal and external validation are necessary to be sure that decisions for patient care are being made with the highest quality evidence. Future work should routinely employ ensemble methods incorporating multiple machine learning algorithms.

Introduction: With a gap in a full understanding of the mechanisms by which survival is extended for patients with cancer who are treated with novel biologic and targeted agents, there is the risk that discordant progression-free and overall survival outcomes are observed due to poor clinical trial design or biases in the interpretation of data. This study was designed to examine the role of study quality and design on the outcomes observed with biologic and targeted agents.

PURPOSE: Comprehensive tumor biomarker testing is a fundamental step in the selection of highly effective molecularly driven therapies for a variety of solid tumors. The primary objective of this study was to examine racial differences in biomarker testing and clinical trial participation in the United States using a real-world database. METHODS: Patients in a real-world deidentified database diagnosed with advanced/metastatic non–small-cell lung cancer (NSCLC), metastatic colorectal cancer (CRC), or metastatic breast cancer were eligible. Biomarker testing and clinical trial participation was compared between Black and White racial groups using chi-squared test and stepwise logistic regression controlling for baseline covariates. RESULTS: A total of 23,488 patients met eligibility criteria. Next-generation sequencing (NGS) testing rates differed significantly between White versus Black race before first-line therapy (36.6% v 29.7%, P < .0001) and at any given time (54.7% v 43.8%, P < .0001) in the nonsquamous NSCLC cohort. Similar disparities in NGS testing rates at any time during the study were observed among patients with CRC (White 51.6%; Black 41.8%, P < .0001). No differences were observed in the breast cancer cohort. Patients of Black race were less likely to be treated in a clinical trial in the overall NSCLC cohort when compared with White counterparts (3.9% v 2.1%, P = .0002). A statistically significant relationship between biomarker/NGS testing and clinical trial enrollment was observed in all cohorts ( P < .003) after adjusting for covariates. CONCLUSION: In a real-world database, significant disparities in NGS-based testing rates were observed between Black and White races in NSCLC and CRC. NGS and any biomarker testing were both associated with trial enrollment in all cohorts. There is a need for interventions to promote access to comprehensive testing for patients with advanced/metastatic tumors.

The study evaluated the diagnostic value and cost-effectiveness of next-generation sequencing (NGS)-based testing versus various combinations of single-gene tests (SGTs) for selection of first-line treatment for patients with advanced/metastatic non-squamous non-small-cell lung cancer in the United States. A dynamic decision analysis model was developed comparing NGS versus SGT from a payer perspective. Inputs were obtained from published sources and included diagnostic performance, biomarker-positive disease rates, biomarker-directed recommendations for treatment, and survival outcomes. Costs were reported in 2020 US dollars. In the base case, NGS improved the detection of actionable biomarkers by 74.4%, increased the proportion of patients receiving biomarker-driven therapy by 11.9%, and decreased the proportion of patients with biomarker-positive disease receiving non-biomarker-driven first-line treatment by 40.5%. The incremental cost-effectiveness ratio per life-year gained of NGS testing versus SGT was $7224 (excluding post-diagnostic costs); the incremental cost-effectiveness ratio for NGS-directed therapy was $148,786 versus SGT-directed therapy. Sensitivity analyses confirmed the robustness of these findings; survival outcomes and targeted therapy costs had the greatest impact on results. Testing strategies with NGS are more comprehensive in the detection of actionable biomarkers and can improve the proportion of patients receiving biomarker-driven therapies. NGS testing may provide a cost-effective strategy for advanced/metastatic non-squamous non-small-cell lung cancer; however, the value of NGS-directed therapy varies by the willingness-to-pay threshold of the decision-maker.

The aim of this study was to conduct a retrospective database analysis to describe the chemotherapy treatment patterns and outcomes of patients with gastric cancer.

Background: Contradictory and limited data are available about the presentation and outcomes of patients with RET-fusion positive metastatic NSCLC as compared to patients without RET fusions. This observational study utilizing a linked electronic health records (EHR) database to genomics testing results was designed to compare characteristics, tumor response, progression-free (PFS) and overall survival (OS) outcomes by RET fusion status among patients with metastatic NSCLC treated with standard therapies. Methods: Adult patients with metastatic NSCLC with linked EHR and genomics data were eligible who received systemic anti-cancer therapy on or after January 1, 2011. Adjusted, using all available baseline covariates, and unadjusted analyses were conducted to compare tumor response, PFS and OS between patients with RET-fusion positive and RET-fusion negative disease as detected by next-generation sequencing. Tumor response outcomes were analysed using Fisher’s exact test, and time-to-event analyses were conducted using Cox proportional hazards model. Results: There were 5807 eligible patients identified (RET+ cohort, N = 46; RET- cohort, N = 5761). Patients with RET fusions were younger, more likely to have non-squamous disease and be non-smokers and had better performance status (all p < 0.01). In unadjusted analyses, there were no significant differences in tumor response (p = 0.17) or PFS (p = 0.06) but OS was significantly different by RET status (hazard ratio, HR = 1.91, 95% CI:1.22–3.0, p = 0.005). There were no statistically significant differences by RET fusion status in adjusted analyses of either PFS or OS (PFS HR = 1.24, 95% CI:0.86–1.78, p = 0.25; OS HR = 1.52, 95% CI: 0.95–2.43, p = 0.08). Conclusions: Patients with RET fusions have different baseline characteristics that contribute to favorable OS in unadjusted analysis. However, after adjusting for baseline covariates, there were no significant differences in either OS or PFS by RET status among patients treated with standard therapy prior to the availability of selective RET inhibitors.