Numerical syndromology: a mathematical approach to the nosology of complex phenotypes

Numerical taxonomy is defined by Sneath and Sokal as the grouping of taxonomic units on the basis of their character states by numerical methods of multivariate data analysis, and syndromology as the study of multiple congenital anomaly (MCA) syndromes and of their nosology. We present here an application of those methods to the analysis of overlapping syndromes. The main advantage of numerical taxonomy is that it allows simultaneous objective and unweighted analysis of multiple traits, giving the possibility to test mathematically the clinical hypotheses about the heterogeneity between closely resembling syndromes and uncovering objective patterns of anomalies, to be compared with the subjective pattern recognition process which characterizes most of the diagnostic approach in syndromology. In this paper, we explored 5 syndromes whose most severe expression belongs to the cerebroacrovisceral early lethality (CAVE) phenotype: hydrolethalus, severe Smith-Lemli-Opitz, orofaciodigital type VI (Varadi-Papp), holoprosencephaly-polydactyly, and Pallister-Hall syndromes. Fifty-five published cases, including many overlapping cases, were submitted to principal factor analysis followed by hierarchical clustering and graphical scaling. Results show that the 5 syndromes clearly constitute independent phenotypic entities, that some of the original diagnoses have to be reconsidered, and that many of the overlapping cases may be unambiguously set in one category. Hypothalamic hamartoblastoma appears to be a nonspecific dysplasia occurring in any of the 5 disorders.