Application of statistical machine learning in biomarker selection.

In the recent JAVELIN Bladder 100 phase 3 trial, avelumab plus best supportive care significantly prolonged overall survival relative to best supportive care alone as first-line maintenance therapy following first-line platinum-based chemotherapy in patients with advanced urothelial cancer (aUC). Discovering biomarkers using genomic profiling to understand potential patient heterogeneity is essential to help improve patient care with precision medicine. For the JAVELIN Bladder 100 trial, it is unclear which variable selection methods can most reliably identify biomarkers to inform patient care because the dataset is characterized by high collinearity and low signal. The aim of this paper was to evaluate available selection methods and their ability to discover prognostic and predictive biomarkers in patients with aUC receiving first-line maintenance therapy. A simulation study evaluated the performance of popular variable selection approaches for high-dimensional data, including penalized regression models, random survival forests, and Bayesian variable selection methods. For Bayesian variable selection methods, a modified Bayesian Information Criterion (BIC) thresholding rule was proposed in addition to the traditional BIC thresholding rule. These methods were applied to the JAVELIN Bladder 100 dataset to investigate potential biomarkers associated with survival benefit. Results from the simulations demonstrated the strengths and limitations of the different methods. The variable selection methods demonstrated low false discovery rates under different conditions. However, their performance declined in the presence of high collinearity. Using the JAVELIN Bladder 100 data, we identified some potentially significant biomarkers across multiple models. Several lasso-related methods were able to identify potentially biologically meaningful variables in the trial. Some variable selection methods (such as stochastic search variable selection and random survival forest) may not be well suited to this type of data due to the presence of extreme collinearity and low signal. Future research should explore novel variable selection methods that may be more suitable for identifying prognostic and predictive biomarkers in this population.Trial registration: ClinicalTrials.gov Identifier: NCT02603432.

Suboptimal adherence to daily growth hormone in a US real-world study: an unmet need in the treatment of pediatric growth hormone deficiency.

OBJECTIVE: To evaluate adherence to, and discontinuation of, somatropin treatment over 4 years in a US population-based study of children with pediatric growth hormone deficiency (pGHD). METHODS: A retrospective cohort analysis of commercially insured patients ≥3 and <16 years, diagnosed with pGHD, newly treated with somatropin was conducted using Optum De-identified Clinformatics Data Mart. Index date was defined as the first prescription for somatropin between 01 July 2002 and 30 September 2019. Five non-exclusive patient cohorts were identified (>3, 12, 24, 36, and 48 months of post-index continuous enrollment). Suboptimal adherence was defined as medication possession ratio <80%. Discontinuation was defined as the date at which a gap of >60 days between somatropin fills first occurred. Cox proportional hazards regression was used to evaluate time to discontinuation. RESULTS: In the 12-month cohort (n = 3091), mean age was 11.3 ± 2.9 years, 75.9% were male, 70.9% white, 9.4% Hispanic, 3.6% Asian, and 3.1% black. The proportion with suboptimal adherence at months 12 and 48 was 19.6% and 35.9%, respectively. Discontinuation occurred in 42.2% of patients. The rate of discontinuation (HR [95% CI]) was higher for age ≥10 (1.74 [1.53-1.98]), females (1.35 [1.21-1.50]), black and Hispanic race/ethnicity (1.50 [1.18-1.90] and 1.27 [1.09-1.49] compared to White) and obesity (1.69 [1.19-2.40]). CONCLUSION: Suboptimal adherence increases with treatment duration, and risk of discontinuation is associated with age, female gender, black or Hispanic race/ethnicity, and obesity. Strategies that facilitate adherence among children at risk of discontinuation may improve clinical outcomes.

Likelihood-Based Methods for Assessing Principal Surrogate Endpoints in Vaccine Trials.

When evaluating principal surrogate biomarkers in vaccine trials, missingness in potential outcomes requires prediction using auxiliary variables and/or augmented study design with a close-out placebo vaccination (CPV) component. The estimated likelihood approach, which separates the estimation of biomarker distribution from the maximization of the estimated likelihood, has often been adopted. Here we develop a likelihood-based approach that jointly estimates the two parts and describe the methods for selecting auxiliary variables as risk predictors and/or biomarker predictors. Through numerical studies, we observe that in a standard trial design without a CPV component, the two methods achieve similar performance in estimation of the risk model and the marker model. However, for trials augmented with a CPV component, using the likelihood-based method achieves better estimation performance compared to the estimated likelihood method. Moreover, in the presence of a large number of covariates from which to select, the ML method achieves comparable or better performance compared to the EL method in both designs. While the CPV component has not yet been implemented in existing vaccine trials, our results have applications in the planning of future vaccine trials. We illustrate the method using data from a dengue vaccine trial.