The prognostic value of morphometric features was studied in a group of 33 children with acute lymphoblastic leukemia (ALL) and compared with clinical and hematologic parameters. Air dried, May-Grünwald-Giemsa-stained specimens were prepared from iliac crest biopsies, and for each patient, 150 blasts and their nuclei were selected according to a stratified selection method and measured on a graphic tablet system. Univariate overall survival analysis showed the French-American-British (FAB) classification to be the strongest clinical parameter (P less than .0001). However, the significance was mainly due to the fact that both L3 cases died; the results for L1 and L2 were less satisfactory, with a survival rate (at 10 years) of 69% for the 26 L1 cases and 80% for the five L2 cases. The nuclear/cytoplasmic (N/C) ratio was the best morphometric feature (P less than .0001) and provided more satisfactory classification results than did FAB: only 2 of the 21 (10%) cases with N/C ratios greater than 0.90 died (7 and 9.5 years after the diagnosis, respectively), and 9 of the 12 (75%) cases with N/C ratios greater than or equal to 0.90 died. For recurrence-free survival analysis, essentially the same results were obtained. The N/C ratio retained its significant prognostic value after recurrence: 11 of the 15 patients with eventual recurrences died; 9 of them had an (original) N/C ratio less than or equal to 0.90. Three of the four recurring cases that survived after recurrence had an N/C ratio greater than 0.90 (P less than .03).(ABSTRACT TRUNCATED AT 250 WORDS)