Outcomes Based on CNS Disease Status in Pediatric Acute Myeloid Leukemia and the Role of Peripheral Blood Contamination in Diagnostic Lumbar Punctures; A Report from the Children's Oncology Group Studies AAML0531 and AAML1031

Matthew A. Kutny, Todd A Alonzo, Robert B Gerbing, Yi-Cheng Wang, Richard Aplenc, Alan S Gamis, James Feusner, William G. Woods, Soheil Meshinchi and Donna Johnston


The diagnosis and treatment of CNS leukemia in AML differs between pediatric and adult oncology. Adult AML patients generally receive no prophylactic intrathecal (IT) treatment. Here we evaluate an approach utilized on 2 recent Children's Oncology Group (COG) clinical trials for pediatric AML which included CNS prophylaxis and a low threshold to initiate CNS leukemia directed therapy.

A total of 2119 patients with de novo AML were evaluable from COG trials AAML0531 and AAML1031. Patients with FLT3/ITD on AAML1031 are excluded as that cohort remains eligible for ongoing accrual. The median (range) follow-up time was 7.1 (0 - 10.5) years for AAML0531 and 2.5 (0 - 5.7) years for AAML1031. All patients received IT cytarabine once with each induction and intensification cycle (except for one of the intensification cycles with high dose cytarabine). An LP to determine CNS status was performed at start of induction therapy. CNS positive (CNSpos) leukemia was treated with twice weekly IT cytarabine until 2 negative CSF samples were obtained (minimum 4 and maximum 6 ITs). CNSpos was defined similarly on both trials (Table 1). In order to understand the role of traumatic LP in CNS status and outcomes, CNS status in this analysis was further classified according to criteria used for ALL (Table 1). Comparing the ALL criteria to the AML CNS definition, CNS1 patients are CNS negative (CNSneg) and CNS3 patients are CNSpos, however among patients with CNS2 by ALL criteria, the AML CNS definition does not take into account the number of CSF WBCs in an atraumatic LP and also uses a higher CSF RBC number (100 versus 10) to require use of the Steinherz/Bleyer algorithm.

The rate of CNS relapse (isolated CNS and combined CNS and bone marrow) was evaluated separately on each trial since the protocols differed in number of chemotherapy courses and SCT allocation. On AAML0531, the 5 year CNS relapse rate was lower for CNS1 (4.5 ± 1.7%) compared to CNS2 (9.1 ± 4.5%) or CNS3 (17.7 ±9.4%), P<0.001. This finding was replicated in the AAML1031 outcomes; CNS relapse rates were 4.9 ± 1.8% (CNS1), 11.7 ± 5.0% (CNS2) and 19.3 ±8.8% (CNS3), P<0.001. Patients who were CNSpos had a higher 5 year CNS relapse rate on both studies (CNSpos 13.2 ± 5.1% vs. CNSneg 4.9 ± 1.7% on AAML0531, P<0.001; CNSpos 12.9 ± 4.3 % vs. CNSneg 5.6 ± 1.8% on AAML1031, P<0.001). Event free survival (combined AAML0531 and AAML1031) differed by CNS category (CNS1 51.2%, CNS2 43.5%, CNS3 37.7%, P<0.001) and AML CNS group (CNSneg 41.2%, CNSpos 49.9%, P<0.001), but overall survival did not differ significantly.

We evaluated the role of peripheral blood contamination. For patients with CSF RBC ≥100, the majority of patients were classified by ALL criteria as CNS2 or CNS3, and the rate of CNS3 disease increased approximately two fold compared to those with CSF RBC <100 (Fig 1a). When considering the COG AML definition, CNSpos rate was substantially higher among those with CSF RBC 10-99 (41%) compared to those with CSF RBC <10 (24%) or RBC ≥100 (27%; Fig 1b). This difference is likely due to the presence of peripheral blood blasts without "correction" for this blood contamination because the Steinherz/Bleyer algorithm was not used to determine CNSpos when CSF RBC <100.

The median blood WBC count was significantly lower in patients with CNS1 (14,650; 95% CI 13,200-15,900) compared to CNS2 (46,200; 95% CI 40,500-53,700) or CNS3 (32,400; 95% CI 21,900-43,500). Further, when this analysis was restricted to patients with an atraumatic tap (CSF RBC <10), those with CNS1 still had significantly lower WBC compared to those with CNS2 or CNS3 (median WBC [95% CI]: CNS1 15,100 [13,400-16,800]; CNS2 53,700 [45,800-62,000]; CNS3 35,500 [21,000-50,100]). This suggests that the biologic drivers resulting in leukocytosis may also increase the risk of CNS leukemia.

This analysis demonstrates that CNS disease in AML is associated with a higher CNS relapse risk but caution must be used in interpreting disease status at diagnosis. Due to the worse prognosis associated with CNS disease, we advocate continued use of diagnostic lumbar punctures and intensified treatment for pediatric patients with CNS disease. However, we also speculate that deferring the initial diagnostic lumbar puncture until clearance of peripheral blasts (which usually occurs by day 5-8 of induction) may minimize the risk of false positive diagnosis of CNS disease resulting from high peripheral WBC and blast contamination of CSF samples.

Disclosures No relevant conflicts of interest to declare.

  • * Asterisk with author names denotes non-ASH members.