Children with Critical Illness Carry Risk Variants Despite Non-Diagnostic Whole Exome Sequencing

Rapid genetic sequencing is an established and important clinical tool for management of pediatric critical illness. The burden of risk variants in children with critical illness but a non-diagnostic exome has not been explored. This was a retrospective case-control analysis of research whole exome sequencing data that first underwent a diagnostic pipeline to assess the association of rare loss-of-function variants with critical illness in children with diagnostic and non-diagnostic whole exome sequencing including those with virally mediated respiratory failure. Using a gene-based collapsing approach, the odds of a child with critical illness carrying rare loss-of-function variants were compared to controls. A subset of children with virally mediated respiratory failure was also compared to controls. Cases were drawn from the general pediatric ward and pediatric intensive care unit (PICU) at Morgan Stanley Childrens Hospital of NewYork-Presbyterian (MSCH) - Columbia University Irving Medical Center (CUIMC) and from the Office of the Chief Medical Examiner (OCME) of New York City. Of the 285 enrolled patients, 228 (80.0%) did not receive a diagnosis from WES. After quality control filtering and geographic ancestry matching, an analysis of 232 children with critical illness compared to 5,322 healthy and unrelated controls determined cases to harbor more predicted loss-of-function (pLOFs) in genes with a LOEUF score [≤] 0.680 (p-value = 1.0 x 10-5). After quality control and geographic ancestry matching, a subset of 176 children without a genetic diagnosis compared to 5,180 controls harbored pLOFs in genes without a disease association (OR 1.7, CI [1.2 - 2.3], FDR adjusted p-value = 4.4 x 10-3) but not in genes with a disease association (OR 1.2, CI [0.8 - 1.7], FDR adjusted p-value = 0.40). This enrichment primarily existed among ultra-rare variants not found in public data sets. Among a subset of 25 previously healthy children with virally mediated respiratory failure compared to 2,973 controls after quality control and geographic ancestry matching, cases harbored more variants than controls in genes without a disease association at the same LOEUF threshold [≤] 0.680 (OR 2.8, CI [1.2 - 7.2], FDR adjusted p-value = 0.026) but not in genes with a disease association (OR 0.7, CI [0.2 - 2.2], FDR adjusted p-value = 0.84). Finally, children with critical illness for whom whole exome sequencing data from both biological parents were available, we found an enrichment of de novo pLOF variants in genes without a disease association among 114 children without a genetic diagnosis (unadjusted p-value < 0.05) but not among 46 children with a genetic diagnosis. Children with critical illness and non-diagnostic whole exome sequencing may still carry risk variants for their clinical presentation in genes not previously associated with disease.