LONG-TERM SURVIVORS IN CHRONIC GRANULOCYTIC-LEUKEMIA - A STUDY BY THEINTERNATIONAL CGL PROGNOSIS STUDY-GROUP

The purposes of the present work were to identify the initial characte ristics associated with long-term survival in chronic granulocytic leu kaemia (CGL) and to analyse the accuracy of prognostic models in ident ifying long-term survivors. 813 Philadelphia (Ph) chromosome-positive, nonblastic CGL patients from six American and European institutions, the majority treated conventionally, with a minimum follow-up > 10 yea rs, were studied. Stepwise logistic regression was performed to ascert ain the association between the initial clinicohaematological variable s and survival greater than or equal to 8 years, and a prognostic inde x was derived. The usefulness of both Sokal's and the new prognostic i ndex to identify long-term survivors was assessed by calculating their positive and negative predictive accuracies, sensitivity and specific ity. Median survival of the series was 45 months (range 1-255), with 7 84 patients (96.4%) having died and 109 (13.4%) surviving 8 years or l onger. Younger age, smaller spleen, platelets less than or equal to 60 0 x 10(9)/l, and lower blood blast percentage were associated with sur vival greater than or equal to 8 years; platelets less than or equal t o 600 x 10(9)/l and lower blood blast percentage were the predictive f actors in patients 50 years old or younger. Two-thirds of long survivo rs belonged to Sokal's low-risk group, but the positive predictive acc uracy and specificity for prolonged survival of Sokal's index were ver y low. This was also the case for the new predictive index. In 366 pat ients with information on the presence or not of additional karyotypic abnormalities at diagnosis, younger age, platelets less than or equal to 600 x 10(9)/l, lower blood blast percentage, and absence of additi onal cytogenetic abnormalities were associated with survival greater t han or equal to 8 years. The addition of the cytogenetic data improved the positive predictive accuracy and specificity of the model, albeit modestly. We conclude that, although most CGL long survivors show fav ourable prognostic features at presentation, clinicohaematological dat a do not allow a precise definition of such a subpopulation.