A New Model to Predict Remission Status in Acute Myeloid Leukemia (AML) Patients Based on Day 14 Bone Marrow (D14 BM) Biopsy

Introduction: Although day D14 BM after initiation of induction chemotherapy is accepted standard of care in AML patients (pts), it has poor predictive value and low accuracy for refractory disease. Currently there are no established clinical and laboratory factors which accurately predict which AML pts with positive D14 BM require immediate reduction chemotherapy for persistent disease and which pts achieve complete remission (CR)

Methods: We retrospectively analyzed pretreatment factors and post-induction response in AML pts to determine if clinical and laboratory characteristics can improve the predictive value of the D14 BM evaluation.

Results: Among 297 pts with D14 BM biopsies, 183 pts (61%) had positive D14 BM (either ≥ 5% myeloblasts or cellularity ≥ 20%). Of those with a positive D14 BM biopsy, no reinduction chemotherapy was given to 89 pts of which 57 (64%) pts had persistent disease at count recovery and 32 (36%) pts achieved CR. Persistent disease at count recovery after positive D14 BM was more likely associated with higher percentage of D14 myeloblasts, history of relapsed disease, and poor-risk disease category than pts with positive D14 BM who achieved CR. Age, D14 BM cellularity, and WBC at first day of induction chemotherapy had no significant influence on remission status in pts with a positive D14 BM (Table 1). We developed, and tested in a validation cohort, new prediction model using both D14 BM status and clinical/laboratory factors such as the percentage of blasts, history of relapsed disease, and poorer disease risk category. Then we compared results of this prediction model to that of D14 BM alone without the usage of clinical/laboratory prognostic factors. Our prediction model significantly improved the positive predictive value (84% vs.64% P=0.001) and the accuracy of prediction of recovery marrow status (0.88% vs. 80%, P=0.002) in AML pts with positive D14 BM (Table 2).

Conclusion: In this study we developed and validated a new prediction model for interpreting D14 BM biopsies in AML pts after induction chemotherapy. With the addition of readily available clinical and laboratory information, our multivariable model provides a more accurate prediction of recovery bone marrow status and identification of patients with positive D14 BM who may not benefit from early reduction chemotherapy.

Table 1. Clinical and Laboratory Characteristics of AML Patients with Positive D14 BM Who Did Not Receive Immediate Reinduction Chemotherapy.

	Persistent AML at count recovery	CR+CRi	P value
Number	57	32
Age, years                                              Median (min, max)	59 (19, 75)	55.5 (20, 78)	0.064*
WBC Count at Induction, x 109/L        Median (min, max)	5.7 (0, 285)	3 (0, 95)	0.25*
D14 Cellularity, %                                 Median (min, max)	10 (5, 40)	10 (3, 60)	0.079*
D14 Blasts, %                                                           0-10                                                                                10%-30                                                                                      >30	13 20 24	15 13 4	0.0023*
Disease Status Before Induction Therapy           De novo                                                                                Relapsed	19 38	26 6	<0.0001**
Antecedent Hematologic Disorder                              Yes                                                                                         No	25 32	6 26	0.021**
Risk                                                          Favorable/Normal                                                                                      High	13 44	19 13	0.0013***

AML, acute myeloid leukemia; CR, complete remission; CRi, complete remission with incomplete count recovery; D, day; WBC, white blood cell.

* Wilcoxon rank sum test. ** Fisher’s exact test.*** Chi-square test

Table 2. Performance of Models in Prediction of Recovery Bone Marrow using the D14 BM with and without Clinical/laboratory Diagnostic Factors (N = 203)

	Sensitivity	Specificity	PPV	NPV	Accuracy	P value*
Without prognostic factors	0.88	0.77	0.64	0.93	0.80	0.002
With prognostic factors**	0.78	0.93	0.84	0.90	0.88

* Comparison of accuracies by two-sample binomial test. ** The percentage of myeloblasts in the D14 BM, disease status, and risk category were used as prognostic factors.NPV, negative predictive value; PPV, positive predictive variable.