Statistical analysis of a biomarker is often complicated
because the detection range of the assay of the biomarker is
restricted. The limits of detection (LOD) are the floor and/or
ceiling values of the biomarker that can be accurately measured
by a particular assay type. Any value of the biomarker beyond
the range of LOD, either smaller than the lower LOD or larger
than the upper LOD, cannot be determined accurately. Those
observations, so-called non-detects (NDs), cannot simply be
excluded from the analysis, because NDs are not 'missing at
random'. They can be considered as censored data, and can
therefore be best analyzed by using statistical methods for
survival analysis. lodGWAS is a flexible package for running genome-wide association
analysis of a biomarker that accounts for the problem of
limit of detection of the assay. It treats non-detected values
as censored, and performs a parametric survival analysis on
the phenotype of interest.
The analysis itself is carried out by the function lod_GWAS.
A second function, lod_QC, can be used to check
if the coding of LOD values in the phenotype file is correct.