Dataset Information

Machine Learning for Exposure-Response Analysis: Methodological Considerations and Confirmation of Their Importance via Computational Experimentations.

ABSTRACT: Exposure-response (E-R) is a key aspect of pharmacometrics analysis that supports drug dose selection. Currently, there is a lack of understanding of the technical considerations necessary for drawing unbiased estimates from data. Due to recent advances in machine learning (ML) explainability methods, ML has garnered significant interest for causal inference. To this end, we used simulated datasets with known E-R "ground truth" to generate a set of good practices for the development of ML models required to avoid introducing biases when performing causal inference. These practices include the use of causal diagrams to enable the careful consideration of model variables by which to obtain desired E-R relationship insights, keeping a strict separation of data for model-training and for inference generation to avoid biases, hyperparameter tuning to improve the reliability of models, and estimating proper confidence intervals around inferences using a bootstrap sampling with replacement strategy. We computationally confirm the benefits of the proposed ML workflow by using a simulated dataset with nonlinear and non-monotonic exposure-response relationships.

SUBMITTER: Harun R

PROVIDER: S-EPMC10221670 | biostudies-literature | 2023 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine Learning for Exposure-Response Analysis: Methodological Considerations and Confirmation of Their Importance via Computational Experimentations.

Harun Rashed R Yang Eric E Kassir Nastya N Zhang Wenhui W Lu James J

Pharmaceutics 20230430 5

Exposure-response (E-R) is a key aspect of pharmacometrics analysis that supports drug dose selection. Currently, there is a lack of understanding of the technical considerations necessary for drawing unbiased estimates from data. Due to recent advances in machine learning (ML) explainability methods, ML has garnered significant interest for causal inference. To this end, we used simulated datasets with known E-R "ground truth" to generate a set of good practices for the development of ML models ...[more]

PMID: 37242624

Similar Datasets

Project description:IntroductionThe advent of the coronavirus disease 2019 (COVID-19) pandemic has led to the development of vaccines against severe acute respiratory syndrome coronavirus 2. Prospective evidence regarding safety for pregnant people and their developing fetuses is lacking. The aim of the COVID-19 Vaccines International Pregnancy Exposure Registry (C-VIPER) is to estimate the relative risk of obstetric, neonatal, and infant outcomes by comparing participants vaccinated against COVID-19 during pregnancy to a reference group of people enrolled in the Pregistry International Pregnancy Exposure Registry (PIPER) who remained unvaccinated during pregnancy.MethodsThe C-VIPER and the PIPER are international, non-interventional, real-world cohort studies. Participants receiving a COVID-19 vaccine during pregnancy will be matched in the analyses by country and gestational age at enrollment to unvaccinated individuals. Self-enrolled and self-consented participants complete online questionnaires at enrollment, during pregnancy, and for 12 months after the delivery of a live infant. Where possible, outcomes are verified by medical records. The study aims to recruit at least 500 pregnancies for each approved or authorized vaccine and will last for 5 years for each product.ConclusionsBy collecting data for each vaccine brand, the C-VIPER will be able to determine individual safety profiles. The study design allows for analysis of the effects of exposure to COVID-19 vaccines during specific etiologically relevant periods of gestation. Although the sample size may be too small to detect associations with rare outcomes, the study will be used to generate hypotheses for future research. Ultimately, the C-VIPER should provide data that will allow pregnant people and their healthcare providers to make informed decisions about COVID-19 vaccination.Clinical trial registrationClinicalTrials.gov NCT04705116. Registered on 12 January, 2021. EU PAS EUPAS39096. Registered on 20 January, 2021.

Project description:PurposeAlthough abstinence is recommended in pregnancy, many pregnancies are exposed to alcohol. Observational studies of the effects of low to moderate prenatal alcohol exposure (PAE) and neurodevelopmental outcomes have yielded inconsistent results, with some studies finding an increased risk of adverse neurobehavioral and cognitive outcomes, and other studies finding no changes or reduced risk of the same outcomes. The purpose of this narrative review is to summarize these inconsistencies and apply a methodological framework to discuss how different parameters contribute to the findings. The authors also provide recommendations on how to advance future research in this area.Search methodsThe PubMed, Web of Science, and Embase databases were searched, along with reference lists of selected systematic reviews and meta-analyses. Search terms used were (infant or child or children or adolescent or offspring) AND (low or light or mild or moderate or low-to-moderate) AND (drinking or alcohol or drinks) AND (pregnancy or prenatal or fetal) AND (neurodevelopment or behavioral or psychological or cognitive or developmental) NOT (mice or rat or fish or animal) NOT (meta-analysis or review). Peer-reviewed original research studies were included if they analyzed associations between an exposure defined and characterized as low/light or moderate PAE with offspring neurodevelopmental outcomes. Animal studies, studies that did not provide clear cutoff points to classify PAE categories, studies lacking an abstinence control group, and studies that did not present a multivariable-adjusted measure of association were excluded.Search resultsThe searches identified 2,422 papers, with 36 papers meeting eligibility criteria. These studies were carried out across nine countries and included samples ranging from approximately 500 to 40,000 participants. Cognitive, academic, socioemotional, and behavioral outcomes were assessed from infancy through age 19.Discussion and conclusionWhen the findings from the selected articles were summarized by geographic region, exposure definition, or neurodevelopmental outcome, no consistent observations or patterns emerged between low to moderate PAE and offspring outcomes. Although some studies found positive (i.e., beneficial) associations between low to moderate PAE and outcomes (primarily outcomes related to cognition) and others found negative (i.e., detrimental) associations (primarily for behavioral outcomes), most findings were null (i.e., showed no effect of PAE). The heterogeneity in study results is likely due to methodological issues, including residual confounding, effect measure modification, and exposure misclassification that make synthesis of studies difficult. Alternative study designs, including longitudinal trajectory analysis, sibling design, negative controls, and instrumental variable analyses, may reduce biases and are discussed. To date, the consequences of light to moderate levels of PAE on neurodevelopment remain unresolved; studies that advance methodological rigor will be important contributions to the field.

Dataset Information

Machine Learning for Exposure-Response Analysis: Methodological Considerations and Confirmation of Their Importance via Computational Experimentations.

Publications

Machine Learning for Exposure-Response Analysis: Methodological Considerations and Confirmation of Their Importance via Computational Experimentations.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets