Causal deep learning reveals the comparative effectiveness of antihyperglycemic treatments in poorly controlled diabetes.

Type-2 diabetes is associated with severe health outcomes, the effects of which are responsible for approximately 1/4th of the total healthcare spending in the United States (US). Current treatment guidelines endorse a massive number of potential anti-hyperglycemic treatment options in various combinations. Strategies for optimizing treatment selection are lacking. Real-world data from a nationwide population of over one million high-risk diabetic patients (HbA1c ≥ 9%) in the US is analyzed to evaluate the comparative effectiveness for HbA1c reduction in this population of more than 80 different treatment strategies ranging from monotherapy up to combinations of five concomitant classes of drugs across each of 10 clinical cohorts defined by age, insulin dependence, and a number of other chronic conditions. A causal deep learning approach developed on such data allows for more personalized evaluation of treatment selection. An average confounder-adjusted reduction in HbA1c of 0.69% [-0.75, -0.65] is observed between patients receiving high vs low ranked treatments across cohorts for which the difference was significant. This method can be extended to explore treatment optimization for other chronic conditions.

Minimizing bias in massive multi-arm observational studies with BCAUS: balancing covariates automatically using supervision.

BACKGROUND: Observational studies are increasingly being used to provide supplementary evidence in addition to Randomized Control Trials (RCTs) because they provide a scale and diversity of participants and outcomes that would be infeasible in an RCT. Additionally, they more closely reflect the settings in which the studied interventions will be applied in the future. Well-established propensity-score-based methods exist to overcome the challenges of working with observational data to estimate causal effects. These methods also provide quality assurance diagnostics to evaluate the degree to which bias has been removed and the estimates can be trusted. In large medical datasets it is common to find the same underlying health condition being treated with a variety of distinct drugs or drug combinations. Conventional methods require a manual iterative workflow, making them scale poorly to studies with many intervention arms. In such situations, automated causal inference methods that are compatible with traditional propensity-score-based workflows are highly desirable. METHODS: We introduce an automated causal inference method BCAUS, that features a deep-neural-network-based propensity model that is trained with a loss which penalizes both the incorrect prediction of the assigned treatment as well as the degree of imbalance between the inverse probability weighted covariates. The network is trained end-to-end by dynamically adjusting the loss term for each training batch such that the relative contributions from the two loss components are held fixed. Trained BCAUS models can be used in conjunction with traditional propensity-score-based methods to estimate causal treatment effects. RESULTS: We tested BCAUS on the semi-synthetic Infant Health & Development Program dataset with a single intervention arm, and a real-world observational study of diabetes interventions with over 100,000 individuals spread across more than a hundred intervention arms. When compared against other recently proposed automated causal inference methods, BCAUS had competitive accuracy for estimating synthetic treatment effects and provided highly concordant estimates on the real-world dataset but was an order-of-magnitude faster. CONCLUSIONS: BCAUS is directly compatible with trusted protocols to estimate treatment effects and diagnose the quality of those estimates, while making the established approaches automatically scalable to an arbitrary number of simultaneous intervention arms without any need for manual iteration.

Burden of hereditary cancer susceptibility in unselected patients with pancreatic ductal adenocarcinoma referred for germline screening.

BACKGROUND: Recent guidelines recommend consideration of germline testing for all newly diagnosed pancreatic ductal adenocarcinoma (PDAC). The primary aim of this study was to determine the burden of hereditary cancer susceptibility in PDAC. A secondary aim was to compare genetic testing uptake rates across different modes of genetic counselling. PATIENTS AND METHODS: All patients diagnosed with PDAC in the province of British Columbia, Canada referred to a population-based hereditary cancer program were eligible for multi-gene panel testing, irrespective of cancer family history. Any healthcare provider or patients themselves could refer. RESULTS: A total of 305 patients with PDAC were referred between July 2016 and January 2019. Two hundred thirty-five patients attended a consultation and 177 completed index germline genetic testing. 25/177 (14.1%) of unrelated patients had a pathogenic variant (PV); 19/25 PV were in known PDAC susceptibility genes with cancer screening or risk-reduction implications. PDAC was significantly associated with PV in ATM (OR, 7.73; 95% CI, 3.10 to 19.33, P = 6.14E-05) when comparing age and gender and ethnicity-matched controls tested on the same platform. The overall uptake rate for index testing was 59.2% and was significantly higher with 1-on-1 consultations and group consultations compared to telehealth consultations (88.9% vs 82.9% vs 61.8%, P < .001). CONCLUSION: In a prospective clinic-based cohort of patients with PDAC referred for testing irrespective of family history, germline PV were detected in 14.1%. PV in ATM accounted for half of all PVs and were significantly associated with PDAC. These findings support recent guidelines and will guide future service planning in this population.

Multi-Gene Panel Testing of 23,179 Individuals for Hereditary Cancer Risk Identifies Pathogenic Variant Carriers Missed by Current Genetic Testing Guidelines.

Recent advancements in next-generation sequencing have greatly expanded the use of multi-gene panel testing for hereditary cancer risk. Although genetic testing helps guide clinical diagnosis and management, testing recommendations are based on personal and family history of cancer and ethnicity, and many carriers are being missed. Herein, we report the results from 23,179 individuals who were referred for 30-gene next-generation sequencing panel testing for hereditary cancer risk, independent of current testing guidelines-38.7% of individuals would not have met National Comprehensive Cancer Network criteria for genetic testing. We identified a total of 2811 pathogenic variants in 2698 individuals for an overall pathogenic frequency of 11.6% (9.1%, excluding common low-penetrance alleles). Among individuals of Ashkenazi Jewish descent, three-quarters of pathogenic variants were outside of the three common BRCA1 and BRCA2 founder alleles. Across all ethnic groups, pathogenic variants in BRCA1 and BRCA2 occurred most frequently, but the contribution of pathogenic variants in other genes on the panel varied. Finally, we found that 21.7% of individuals with pathogenic variants in genes with well-established genetic testing recommendations did not meet corresponding National Comprehensive Cancer Network criteria. Taken together, the results indicate that more individuals are at genetic risk for hereditary cancer than are identified by current testing guidelines and/or use of single-gene or single-site testing.

Cascade Genetic Testing of Relatives for Hereditary Cancer Risk: Results of an Online Initiative.

In cascade testing, genetic testing for an identified familial pathogenic variant extends to disease-free relatives to allow genetically targeted disease prevention. We evaluated the results of an online initiative in which carriers of 1 of 30 cancer-associated genes, or their first-degree relatives, could offer low-cost testing to at-risk first-degree relatives. In the first year, 1101 applicants invited 2280 first-degree relatives to undergo genetic testing. Of invited relatives, 47.5% (95% confidence interval [CI] = 45.5 to 49.6%) underwent genetic testing, and 12.0% (95% CI = 9.2 to 14.8%) who tested positive continued the cascade by inviting additional relatives to test. Of tested relatives, 4.9% (95% CI = 3.8 to 6.1%) had a pathogenic variant in a different gene from the known familial one, and 16.8% (95% CI = 14.7 to 18.8%) had a variant of uncertain significance. These results suggest that an online, low-cost program is an effective approach to implementing cascade testing, and that up to 5% of the general population may carry a pathogenic variant in 1 of 30 cancer-associated genes.