Is there evidence for subclasses of chronic lymphocytic leukemia? A study using numerical classification techniques

Various clinical and biological parameters were analysed among 95 chronic lymphocytic leukemia patients by two numerical classification techniques in an attempt to identify natural subdivisions in the disease. For each patient, a profile made up of 47 dichotomous variables was constructed and the distance between profiles computed with equal weight for each variable. No specific cluster of patients was identified by these techniques. However limited availability of data for some laboratory tests, particularly concerning the immune system, meant that these results had to be excluded from analysis. The profiles of patients who died from leukemia during a mean follow up period of 4 years apparently do not generate a distinct group from the other profiles in spite of some clustering.