Scoring Systems For Predicting Outcome Of Chronic Myeloid Leukemia In Adults Are Poorly Informative In Pediatric Patients Treated With Imatinib

Imatinib (IM) front-line treatment impressively improved survival of children with chronic myeloid leukemia (CML). In contrast to adult CML, specific scoring systems predicting the treatment response in individual pediatric patients (pts) are still lacking. Here we analyzed a cohort of pediatric pts with CML applying the established prognostic scores for adults in a comparative fashion. We question the value of four scoring systems (Sokal-, Sokal young-, Hasford-, Eutos-Score) especially with regard to grouping individual children differently or homogeneously into a defined risk category. In addition, we analyzed which scoring system would classify most specifically the prognosis of pediatric CML with regard to early molecular response (MR) on IM.

A total of 90 pts (male/female: 57/33; median age: 11.6 yrs, range: 1-18) with CML-CP enrolled in the prospective trial CML-PAED-II were included in this analysis. Registry data were collected on standardized forms filled in by the treating physicians. On this basis the Eutos-, Sokal- and Hasford-Scores were calculated using internet resources of the ELN (www.leukemia-net.org/content/leukemias/cml/cml_score), whereas the Sokal young Score (Sy) – a score described specifically for adolescents and younger adults (Sokal JE, Blood 1985;66:1352) – was manually calculated. Pts were grouped using the original three risk categories (low=LR, intermediate=IR, high=HR) or two categories, respectively, for the Eutos-Score (LR or HR). Evaluation of therapeutic response was performed by assessing the MR by measurement of the transcript ratio BCR-ABL1/ABL1 in blood specimen sent to the central reference laboratory at month 3 after start of IM treatment. Measurements were expressed according to the International Scale.

By Sokal-Score 59/90 pts were classified as LR, 20/90 pts as IR and 11/90 pts as HR. By Hasford Score 57/90 pts were classified as LR, 25/90 pts as IR, and 8/90 pts as HR. By Eutos Score 73/90 pts were classified as LR and 17/90 pts as HR. As the hematocrit value was not collected systematically at diagnosis, this necessary parameter for calculating the Sy-Score was applicable only in 46/90 pts and thus 44/46 pts were classified as LR, 2/46 as IR, and 0/46 as HR. Comparing results of individual pts only 25/46 pts (54%) were categorized homogeneously as LR by applying all 4 scoring systems, while 54/90 pts (60%) were classified as LR if Sy-Score was excluded. Thus, the remaining 21/46 pts (46%) were grouped heterogeneously by applying each of the 4 prognostic scores, and correspondingly 33/90 pts (37%) were classified heterogeneously within different risk categories by the Eutos, Hasford and Sokal Score. Only 3 pts were categorized homogenously as HR by each of the Sokal, Hasford, and Eutos Score and by applying all 4 scoring systems no patient was concordantly classified as HR. When comparing only the Sokal-Score to the Sy-Score, discordant results were obtained in 19/46 (41%) pts.

BCR-ABL1/ABL1 transcript ratio could be analyzed quantitatively in 72/90 pts at month 3 after treatment initiation. In this cohort we identified 46/72 good responders (ratio BCR-ABL1/ABL1 <10%) and 26/72 poor responders (ratio >10%). Although the Eutos-score performed best in in a logistic regression analysis with an Odds Ratio OR=3.02 to predict an unfavorable course of IM-treated CML in the HR group, the discrimination did not reach statistical significance (p=0.08). However, by reducing the cut-off point for the Eutos Score from 87 to 64 an OR=4.8 with p=0.004 was achieved, thus indicating that a refined risk categorization appears beneficial.

Comparing risk categorization by all four scores in individual pediatric pts, results may vary considerably. Keeping in mind that the number of pts analyzed is still small, especially applying the Sy-Score seems not to provide benefit in this cohort with a median age of only 11 years. Contrasting results in adults, in this pediatric cohort the Sokal- and Hasford-Scores did not predict a poor IM treatment response at month 3 while the Eutos Score achieved borderline significance. Thus, there is an urgent need for the development of a more specific pediatric risk score.

No relevant conflicts of interest to declare.