ABSTRACT
Despite its potentials benefits, using prediction targets generated based on latent variable (LV) modeling is not a common practice in supervised learning, a dominating framework for developing prediction models. In supervised learning, it is typically assumed that the outcome to be predicted is clear and readily available, and therefore validating outcomes before predicting them is a foreign concept and an unnecessary step. The usual goal of LV modeling is inference, and therefore using it in supervised learning and in the prediction context requires a major conceptual shift. This study lays out methodological adjustments and conceptual shifts necessary for integrating LV modeling into supervised learning. It is shown that such integration is possible by combining the traditions of LV modeling, psychometrics, and supervised learning. In this interdisciplinary learning framework, generating practical outcomes using LV modeling and systematically validating them based on clinical validators are the two main strategies. In the example using the data from the Longitudinal Assessment of Manic Symptoms (LAMS) Study, a large pool of candidate outcomes is generated by flexible LV modeling. It is demonstrated that this exploratory situation can be used as an opportunity to tailor desirable prediction targets taking advantage of contemporary science and clinical insights.
Subject(s)
Supervised Machine Learning , Latent Class AnalysisABSTRACT
PTEN hamartoma tumor syndrome (PHTS) is a complex neurodevelopmental disorder characterized by mechanistic target of rapamycin (mTOR) overactivity. Limited data suggest that mTOR inhibitors may be therapeutic. No placebo-controlled studies have examined mTOR inhibition on cognition and behavior in humans with PHTS with/without autism. We conducted a 6-month phase II, randomized, double-blinded, placebo-controlled trial to examine the safety profile and efficacy of everolimus (4.5 mg/m2) in individuals (5-45 years) with PHTS. We measured several cognitive and behavioral outcomes, and electroencephalography (EEG) biomarkers. The primary endpoint was a neurocognitive composite derived from Stanford Binet-5 (SB-5) nonverbal working memory score, SB-5 verbal working memory, Conners' Continuous Performance Test hit reaction time and Purdue Pegboard Test score. Forty-six participants underwent 1:1 randomization: n = 24 (everolimus) and n = 22 (placebo). Gastrointestinal adverse events were more common in the everolimus group (P < 0.001). Changes in the primary endpoint between groups from baseline to Month 6 were not apparent (Cohen's d = -0.10, P = 0.518). However, several measures were associated with modest effect sizes (≥0.2) in the direction of improvement, including measures of nonverbal IQ, verbal learning, autism symptoms, motor skills, adaptive behavior and global improvement. There was a significant difference in EEG central alpha power (P = 0.049) and central beta power (P = 0.039) 6 months after everolimus treatment. Everolimus is well tolerated in PHTS; adverse events were similar to previous reports. The primary efficacy endpoint did not reveal improvement. Several secondary efficacy endpoints moved in the direction of improvement. EEG measurements indicate target engagement following 6 months of daily oral everolimus. Trial Registration Information: ClinicalTrials.gov NCT02991807 Classification of Evidence: I.
