Boostrapping at Several Frequencies Supports Plasma Insulin As a Potential Marker of Breast Cancer Risk

Among 12 endocrine variables in blood from clinically healthy adult women sampled systematically around the clock and the year, pattern discrimination methods single out certain hormones, in certain seasons as classifiers for a high vs. low risk of developing breast cancer (RDBC). The best classifiers differ among seasons. The relation between plasma insulin, a primary classifier, and RDBC is an indirect one in all 4 seasons. These data were re-analyzed by circadian bootstrapping (BS), assuming that deviations of each sample from 2 means, that of the RDBO-group and that of the individual, are correlated along the 24-hour scale in keeping with an auto-regressive model including only the 1st order partial correlation coefficient (after failure of rejecting the null hypothesis in a test for zero 2nd order partial correlation). Results from BS validate differences in circadian MESOR (midline estimating statistic of rhythm, M) for insulin in spring and summer, complementing earlier results from circannual BS validating a difference in circannual M for insulin between the low and high RDBC groups. A two-way analysis of variance further validates an RDBG effect for insulin in spring and summer. BS Of circadian and/or circannual characteristics is a mathematical resampling procedure here suggested to introduce inferential considerations into discriminant analysis and recommended for broad routine use in nonparametric hypothesis testing and biomedical signal simulation, complementing rhythm parameter compari- sons and analysis of variance in the identification of cost-effective time-specified classifiers of RDBC and other diseases.