2018/03/28 by Vahé Asvatourian, Clélia Coutzac, Asvatourian, Vahé +9
Computer Science · Mathematics · #Advanced Causal Inference Techniques #Computational Drug Discovery Methods #FOS: Computer and information sciences #Machine Learning (stat.ML) #Methodology (stat.ME) #Statistical Methods in Clinical Trials
paper · pdf · doi:10.48550/arxiv.1803.10535
openalex publication_date 2018/03/28 · openalex created_date 2022/10/02 · openalex updated_date 2026/07/28
Recently, the intervention calculus when the DAG is absent (IDA) method was\ndeveloped to estimate lower bounds of causal effects from observational\nhigh-dimensional data. Originally it was introduced to assess the effect of\nbaseline biomarkers which do not vary over time. However, in many clinical\nsettings, measurements of biomarkers are repeated at fixed time points during\ntreatment exposure and, therefore, this method need to be extended. The purpose\nof this paper is then to extend the first step of the IDA, the Peter Clarks\n(PC)-algorithm, to a time-dependent exposure in the context of a binary\noutcome. We generalised the PC-algorithm for taking into account the\nchronological order of repeated measurements of the exposure and propose to\napply the IDA with our new version, the chronologically ordered PC-algorithm\n(COPC-algorithm). A simulation study has been performed before applying the\nmethod for estimating causal effects of time-dependent immunological biomarkers\non toxicity, death and progression in patients with metastatic melanoma. The\nsimulation study showed that the completed partially directed acyclic graphs\n(CPDAGs) obtained using COPC-algorithm were structurally closer to the true\nCPDAG than CPDAGs obtained using PC-algorithm. Also, causal effects were more\naccurate when they were estimated based on CPDAGs obtained using\nCOPC-algorithm. Moreover, CPDAGs obtained by COPC-algorithm allowed removing\nnon-chronologic arrows with a variable measured at a time t pointing to a\nvariable measured at a time t' where t'< t. Bidirected edges were less present\nin CPDAGs obtained with the COPC-algorithm, supporting the fact that there was\nless variability in causal effects estimated from these CPDAGs. The\nCOPC-algorithm provided CPDAGs that keep the chronological structure present in\nthe data, thus allowed to estimate lower bounds of the causal effect of\ntime-dependent biomarkers.\n