Staging systems and laboratory features help predict survival in chronic lymphocytic leukaemia but they do not distinguish patients who will progress from those whose disease will remain indolent. CD38 expression has emerged as an independent prognostic factor, yet there is debate as to what level of CD38 affects prognosis. We plotted the hazard ratios for the treatment-free interval (TFI) between the higher and lower groups defined by CD38 cut-offs from 0 to 100%. The maximum hazard ratio was achieved for a cut-off of 7%. We examined by triple colour analysis the values for CD38 in 289 untreated patients using both >or=30 and >or=7% as thresholds for prognosis. Using a >or=7% threshold (but not >or=30%), we showed a significant correlation with advanced stage and male sex. The interval from diagnosis to first therapy or TFI was longer (median 36 months) in patients with <7% CD38 positive cells than those with >or= 30% (8.7 months) or with intermediate values between 7 and 29% (P<0.00005). The <7% threshold also identified patients in stage A with a long TFI (P=0.0001). Multivariate analysis showed that CD38 has independent prognostic value for TFI in all patients. In 135 patients tested for deletions of p53, 13q14 and 11q23 and for trisomy 12, we showed a correlation between 13q14 deletion and <30%/<7% CD38 positive cells and a tendency for trisomy 12 to be associated with >or=30%/>or=7% CD38 positive cells. We conclude that 7% may be a more useful threshold for disease progression than higher values of CD38.