During the last 4 yr the classification of osteogenesis imperfecta (01) proposed by Sillence et a1 [1979] has become widely used and forms the basis for communication among most geneticists about this heterogeneous disorder. It was recognized then that the analysis of different types was incomplete and as clinical studies have progressed [Levin et al, 1978;Rowe et al, 1981; Sillence, 19811 and as biochemical investigations have proceeded it has become clear that the classification must be extended. As a result, both type I and type IV have been expanded to include varieties in which dentinogenesis imperfecta is either present or absent; type I11 has been expanded to include autosomal dominant and autosomal recessive varieties, and now Sillence et a1 (this issue of the journal) have suggested that type 11, the perinatally lethal form of the disorder, is heterogeneous.
In that paper, Sillence et a1 propose that there are at least three separate variants of 01 type I1 that can be distinguished radiographically. The first they characterize by broad crumpled femora and continuously beaded ribs, the second by broad crumpled femora but minimal or no rib fractures, and the third by narrow fractured femora and thin beaded ribs. Such a classification could prove to be extremely useful if it predicted recurrence risks and longevity. Unfortunately, all infants in the series appear to have been ascertained through death (this is certainly true of the original group from Victoria [Sillence et al, 19791) so that a true measure of longevity is difficult to determine. Furthermore, since there are no biochemical data from any of the infants included in this study, there are no clear indications of which among them, if any, may have disease as the result of dominant new mutations. Ideally, at the time of birth it should be possible to predict the natural history for an infant with type I1 01 and to