ObjectiveThe classification of complex or rare patterns in clinical and genomic data requires the availability of a large, labeled patient set. While methods that operate on large, centralized data sources have been extensively used, little attention has been paid to understanding whether models such as binary logistic regression (LR) can be developed in a distributed manner, allowing researchers to share models without necessarily sharing patient data.Material and methodsInstead of bringing data to a central repository for computation, we bring computation to the data. The Grid Binary LOgistic REgression (GLORE) model integrates decomposable partial elements or non-privacy sensitive prediction values to obtain model coefficients, the variance-covariance matrix, the goodness-of-fit test statistic, and the area under the receiver operating characteristic (ROC) curve.ResultsWe conducted experiments on both simulated and clinically relevant data, and compared the computational costs of GLORE with those of a traditional LR model estimated using the combined data. We showed that our results are the same as those of LR to a 10−15 precision. In addition, GLORE is computationally efficient.LimitationIn GLORE, the calculation of coefficient gradients must be synchronized at different sites, which involves some effort to ensure the integrity of communication. Ensuring that the predictors have the same format and meaning across the data sets is necessary.ConclusionThe results suggest that GLORE performs as well as LR and allows data to remain protected at their original sites.

ObjectiveBirth defects are an increasing health priority worldwide, and the subject of a major 2010 World Health Assembly Resolution. Excess cancer risk may be an added burden in this vulnerable group of children, but studies to date have provided inconsistent findings. This study assessed the risk for cancer in children and young adolescents with major birth defects.Methods and FindingsThis retrospective, statewide, population-based, cohort study was conducted in three US states (Utah, Arizona, Iowa). A cohort of 44,151 children and young adolescents (0 through 14 years of age) with selected major, non-chromosomal birth defects or chromosomal anomalies was compared to a reference cohort of 147,940 children without birth defects randomly sampled from each state’s births and frequency matched by year of birth. The primary outcome was rate of cancer prior to age 15 years, by type of cancer and type of birth defect. The incidence of cancer was increased 2.9-fold (95% CI, 2.3 to 3.7) in children with birth defects (123 cases of cancer) compared to the reference cohort; the incidence rates were 33.8 and 11.7 per 100,000 person-years, respectively. However, the excess risk varied markedly by type of birth defect. Increased risks were seen in children with microcephaly, cleft palate, and selected eye, cardiac, and renal defects. Cancer risk was not increased with many common birth defects, including hypospadias, cleft lip with or without cleft palate, or hydrocephalus.ConclusionChildren with some structural, non-chromosomal birth defects, but not others, have a moderately increased risk for childhood cancer. Information on such selective risk can promote more effective clinical evaluation, counseling, and research.

The proposed method serves as a proof of concept, in which a publicly available dataset was used to evaluate the performance. The authors do not intend to suggest that this method can resolve policy and engineering issues related to the federated use of institutional data, but they should serve as evidence of the technical feasibility of the proposed approach.Conclusions WebDISCO (Web-based Distributed Cox Regression Model; https://webdisco.ucsd-dbmi.org:8443/cox/) provides a proof-of-concept web service that implements a distributed algorithm to conduct distributed survival analysis without sharing patient level data.

BackgroundLow-density lipoprotein receptor-related protein 1b (encoded by LRP1B) is a putative tumor suppressor, and preliminary evidence suggests LRP1B-mutated cancers may have improved outcomes with immune checkpoint inhibitors (ICI).MethodsWe conducted a multicenter, retrospective pan-cancer analysis of patients with LRP1B alterations treated with ICI at Duke University, Johns Hopkins University (JHU) and University of Michigan (UM). The primary objective was to assess the association between overall response rate (ORR) to ICI and pathogenic or likely pathogenic (P/LP) LRP1B alterations compared with LRP1B variants of unknown significance (VUS). Secondary outcomes were the associations with progression-free survival (PFS) and overall survival (OS) by LRP1B status.ResultsWe identified 101 patients (44 Duke, 35 JHU, 22 UM) with LRP1B alterations who were treated with ICI. The most common tumor types by alteration (P/LP vs VUS%) were lung (36% vs 49%), prostate (9% vs 7%), sarcoma (5% vs 7%), melanoma (9% vs 0%) and breast cancer (3% vs 7%). The ORR for patients with LRP1B P/LP versus VUS alterations was 54% and 13%, respectively (OR 7.5, 95% CI 2.9 to 22.3, p=0.0009). P/LP LRP1B alterations were associated with longer PFS (HR 0.42, 95% CI 0.26 to 0.68, p=0.0003) and OS (HR 0.62, 95% CI 0.39 to 1.01, p=0.053). These results remained consistent when excluding patients harboring microsatellite instability (MSI) and controlling for tumor mutational burden (TMB).ConclusionsThis multicenter study shows significantly better outcomes with ICI therapy in patients harboring P/LP versus VUS LRP1B alterations, independently of TMB/MSI status. Further mechanistic and prospective validation studies are warranted.

Both weight loss and low-carbohydrate diets (LCD) without weight loss prolong survival in prostate cancer models. Few human trials have tested weight loss or LCD on prostate cancer.Experimental Design: We conducted a multi-site randomized 6-month trial of LCD versus control on PSA doubling time (PSADT) in patients with prostate cancer with biochemical recurrence (BCR) after local treatment. Eligibility included body mass index (BMI) ≥ 24 kg/m 2 and PSADT 3 to 36 months. The LCD arm was instructed to eat 20 g/carbs/day; the control arm instructed to avoid dietary changes. Primary outcome was PSADT. Secondary outcomes included weight, lipids, glucose metabolism, and diet.Results: Of 60 planned patients, the study stopped early after an interim analysis showed futility. Twenty-seven LCD and 18 control patients completed the study. At 6 months, although both arms consumed similar protein and fats, the LCD arm reduced carbohydrates intake (À117 vs. 8 g, P < 0.001) and lost weight (À12.1 vs. À0.50 kg, P < 0.001). The LCD arm reduced HDL, triglycerides, and HbA1c with no difference in total cholesterol or glucose. Mean PSADT was similar between LCD (21 months) and control (15 months, P ¼ 0.316) arms. In a post hoc exploratory analysis accounting for prestudy PSADT, baseline PSA, primary treatment, and hemoconcentration, PSADT was significantly longer in LCD versus control (28 vs. 13 months, P ¼ 0.021) arms. Adverse events were few, usually mild, and returned to baseline by 6 months.Conclusions: Among BCR patients, LCD induced weight loss and metabolic benefits with acceptable safety without affecting PSADT, suggesting LCD does not adversely affect prostate cancer growth and is safe. Given exploratory findings of longer PSADT, larger studies testing LCD on disease progression are warranted.