Absolute lymphocyte count is a novel prognostic indicator in ALL and AML

Implications for risk stratification and future studies

Abstract

BACKGROUND.

Leukemia is the leading cause of disease-related death in children, despite significant improvement in survival and modern risk stratification. The prognostic significance of absolute lymphocyte counts (ALC) was evaluated in young patients with acute myeloblastic leukemia (AML) and acute lymphoblastic leukemia (ALL).

METHODS.

In all, 171 consecutive de novo cases of AML and ALL, age ≤21 years, were analyzed. Age, initial white blood cell count, cytogenetics, and bone marrow response were compared with lymphocyte, neutrophil, and platelet counts at weekly intervals during induction chemotherapy.

RESULTS.

ALC is a significant independent predictor of relapse and survival. For example, in patients with AML an ALC on Day 28 of induction (ALC-28) <350 cells/μL predicts very poor survival, with a 5-year relapse-free survival (RFS) of only 10% (hazard ratio [HR] 3.7, P = .003). In contrast, an ALC-15 >350 cells/μL carries an excellent prognosis, with a 5-year overall survival (OS) of 85% (HR 0.2, P = .012). Similarly in ALL, an ALC-15 <350 cells/μL predicts poor survival, with a 6-year RFS of 43% (HR 4.5, P = .002), whereas an ALC-15 >350 cells/μL predicts excellent outcome, with a 6-year OS of 87% (HR 0.2, P = .018). Importantly, ALC remains a strong predictor in multivariate analysis with known prognostic factors.

CONCLUSIONS.

ALC is a simple, statistically powerful measurement for patients with de novo AML and ALL. The results, when combined with previous studies, demonstrate that ALC is a powerful new prognostic factor for a range of malignancies. These findings suggest a need for further exploration of postchemotherapy immune status and immune-modulating cancer therapies. Cancer 2008. © 2007 American Cancer Society.

Leukemia is the most common pediatric cancer and is the most common cause of disease-related death in childhood. Acute lymphoblastic leukemia (ALL) accounts for 80% of pediatric leukemias.1 Although most patients enter remission initially, over 25% of patients will ultimately experience a relapse, and those patients with an early bone marrow relapse have <10% survival.2 Similarly, 50% of acute myeloblastic leukemia (AML) patients will relapse.3 Although salvage regimens including bone marrow transplant are modestly successful in a subset of young patients, durable remissions remain elusive for most with recurring AML.4 These high recurrence rates, despite multifactorial risk stratification, establish the need for additional prognostic indicators.

Risk Stratification

Risk stratification allows therapy to be selected based on specific prognostic indicators. Current ALL therapeutic regimens risk-stratify patients at diagnosis based on the National Cancer Institute/Rome criteria (age, initial white blood cell count [WBC])5 and immunophenotype (T vs B cell) with additional stratification for cytogenetics (trisomies 4 of 10 of 17, t(12;21), t(9;22), t(4;11), hypodiploid <44 Chr). After initiation of therapy, modifications in stratification are made based on response to treatment (ie, prednisone window, Day 7 of 14 bone marrow morphology).2, 6 Similarly, the major prognostic factors in pediatric AML are morphology/phenotype (M3), cytogenetics (trisomy 21, monosomy 7, 5q−), and response to therapy. In an effort to identify more patients who will ultimately experience recurrence, measurement of minimal residual disease (MRD) after induction appears to be an important prognostic factor in both ALL and AML and is being evaluated in current clinical trials.7, 8 Importantly, MRD measurements primarily reflect the chemosensitivity of the leukemia cells themselves and do not assess host factors, such as chemosensitivity of nonmalignant cells and immune status.9

Lymphocytes in Cancer

Recent studies have reinforced the belief that defective function or numbers of lymphocytes reduce the ability of the patient's immune system to mount an effective response to cancer cells. In particular, autologous T and natural killer (NK) cells have been shown to play a role in prolonged remission in AML.10 Several studies have shown that higher lymphocyte counts predict improved survival in patients who have undergone autologous or allogeneic stem cell transplant for various malignancies.11–14 These results suggest that rapid lymphocyte engraftment in these patients could increase a graft-versus-cancer effect and/or decrease morbidity related to poor engraftment. Interestingly, Behl et al.15 found that absolute lymphocyte counts (ALC) after induction chemotherapy predicts improved survival in adult patients with AML, suggesting that lymphocyte levels are critical in the nontransplant setting also. Two meeting abstracts suggested that lymphocyte counts after chemotherapy may also be predictive of survival in the pediatric age group.16, 17

On the basis of these observations, we analyzed 171 consecutive acute leukemia patients, ≤21 years old, to evaluate the role of lymphocyte counts in young patients with both ALL and AML. Here we report that absolute lymphocyte count (ALC) during induction is a novel independent prognostic indicator of relapse-free survival (RFS) and overall survival (OS) for both ALL and AML. Our findings suggest that a routine complete blood count (CBC) with differential could be used to improve leukemia risk stratification and justify future studies exploring the role on host immune status in acute leukemias.

MATERIALS AND METHODS

Patients

The overall patient population at our institution is atypical because there are a large number of relapse and refractory patients referred to our center. Therefore, for this study we only analyzed 171 consecutive patients with de novo ALL or AML, 0–21 years old, who began treatment at the University of Texas M. D. Anderson Cancer Center from 1995–2005. Twenty-eight patients were excluded because of a lack of adequate laboratory data for this analysis.

Treatment

Pediatric ALL therapy generally consisted of 3- or 4-drug induction (prednisone or dexamethasone, vincristine, asparaginase, with daunorubicin for high-risk patients) according to protocols CCG-1952 or 1961 or 1991. A subset of our young adult ALL patients (n = 25) received hyper-CVAD therapy (dexamethasone, vincristine, doxorubicin, cyclophosphamide, then methotrexate, cytarabine, ×8 cycles).18, 19 The use of hyper-CVAD is not a standard therapy in any of the major pediatric oncology groups; however, it is standard therapy given by the adult leukemia service at our institution. Induction therapy for AML consisted of anthracycline (daunorubicin or idarubicin) and cytarabine with or without etoposide.3, 20, 21

Analysis and Statistics

With Institutional Review Board (IRB) approval we performed a chart review and recorded: age, sex, initial WBC, cytogenetics, bone marrow blast percentage on Days 0 and 7 of treatment (ALL only), treatment protocol (ALL only), date of relapse, date of last contact, survival status, and ALC, absolute neutrophil count (ANC) and platelet count (PLT) on Days 0, 15, 21, and 28 (ALC-0 through 28).

OS, RFS, and time to relapse were estimated using the methods of Kaplan and Meier.22 Differences between survival curves were assessed for statistical significance with the log-rank test. Potential risk factors were considered in a Cox proportional hazards regression model in univariate and multivariate fashions, and hazard ratios (HRs) were estimated with 95% confidence intervals.23 These analyses were performed on the entire ALL and AML cohorts, and also separately for reduced cohorts, which excluded very high- and low-risk patients from unique subgroups (eg, infant ALL, HR cytogenetics, APML, Down syndrome, etc.) with similar results.

RESULTS

Demographics

The patient characteristics are shown in Table 1. Our cohorts have more than the usual number of young adult patients in the 18–21-year-old group, as we have included those treated by the adult leukemia service. Our AML cohort has a 6-year OS of 49% and our ALL cohort has a 6-year OS of 72%, similar to published outcomes for this period. The distribution of prognostic factors, including AML FAB classification, is also similar to published studies in predominantly adolescent and young adult populations (Table 1).

Table 1. Characteristics of ALL and AML Cohorts

Description of cohort	AML patients, N = 54	ALL patients, N = 89
	Median (Range)	Median (Range)
NA indicates not available.
Age at diagnosis, y	18 (0.17–21)	11 (0.25–21)
Initial WBC, ×10^3 cells/μL	11.6 (0.3–373)	11 (1–700)
Bone marrow blasts Day 7 (%)	NA	5.5 (0–95)
Follow-up, mo	30 (2–112)	41 (3–104)
Overall survival: 6 y (95% CI)	49% (31–65%)	72% (58–83%)
Relapse-free survival: 6 y (95% CI)	41% (26–55%)	64% (50–76%)
Prognostic Factors	No. (%)	No. (%)
Age at diagnosis
<1 y, high risk in ALL	2 (3.7%)	2 (2%)
1–<10 y, standard risk	9 (17%)	38 (43%)
≥10 y, high risk in ALL (AML?)	43 (80%)	49 (55%)
Initial WBC, high ≥5 × 10^4cells/μL	17 (31%)	25 (28%)
Phenotype
Favorable	2 (3.7%): APML-M3	83 (93%): precursor B
Unfavorable		6 (7%): T
Cytogenetics
Favorable	4 (7%): Trisomy 21, t(15;17)	11 (12%): t(12:21), >50 Ch
Unfavorable	4 (7%): monosomy 7, 5p-	15 (17%): t(9:22), t(4;11), <44 Ch
Bone marrow Day 7 ≥5% blasts	NA	34 (38%)
FAB morphology		NA
M0	3 (5%)
M1	12 (22%)
M2	14 (26%)
M3	4 (7%)
M4	10 (19%)
M5	6 (11%)
M6	0 (0%)
M7	5 (9%)

Known Prognostics Factors: WBC, Day 7 Marrow Response, Cytogenetics, and Age

Our patient cohort confirmed several known prognostic factors in acute leukemias. For ALL, high initial WBC, age >10 years,24 and high percentage of bone marrow blasts on Day 7 were predictive of poor survival (Table 3). In AML, unfavorable cytogenetics were predictive of poor survival (Table 2).

Table 2. AML Statistical Analysis

* P < .05. † P < .005.
AML Overall Survival	6-Year OS Rate	Hazard Ratio (95% CI)	P
ALC-15 <350 cells/μL	28% vs 85% (5yr)	4.9 (1.4–17)	.012*
ALC-21 <350 cells/μL	32% vs 67%	2.6 (1–6.8)	.048*
ALC-28 <350 cells/μL	27% vs 59%	3.4 (1.2–9)	.017*
ALC <350 at all time points	31% vs 88%	4.6 (1.1–20)	.011*
Multivariate analysis
ALC-15 <350 cells/μL		5.8 (1.4–24)	.014*
Favorable cytogenetics		0.9 (0.3–3)	.874
Unfavorable cytogenetics		14.7 (2.8–77)	.012†
AML Relapse-free Survival	6-Year RFS Rate	Hazard Ratio (95% CI)	P
ALC-15 <350 cells/μL	23% vs 61% (5yr)	2.7 (1.1–6.6)	.025*
ALC-21 <350 cells/μL	20% vs 59%	2.7 (1.2–6)	.018*
ALC-28 <350 cells/μL	10% vs 51% (5yr)	3.7 (1.6–8.8)	.003†
ALC <350 at all time points	23% vs 68% (5yr)	3.3 (1.1–9.6)	.028*
Multivariate analysis
ALC-28 <350 cells/μL		3.8 (1.5–10)	.014*
Favorable cytogenetics		0.6 (0.1–2.9)	.486
Unfavorable cytogenetics		3.9 (1.1–14.1)	.360

Hematologic Recovery

To evaluate whether hematologic recovery affects survival we analyzed the association between ANC, PLT, and survival. Interestingly, a platelet count ≥100,000 on Day 28 was able to predict good OS for patients with AML (HR 0.4, P = .045, Table 2). Platelet recovery has been reported to be prognostic in older adults with ALL25; however, PLTs were not predictive of relapse or survival in our young patients with ALL. Prolonged thrombocytopenia likely reflects host chemosensitivity, and therefore further analysis of relapse versus toxicity is warranted in AML. As has been observed previously in pediatric leukemias, there was no association between ANC recovery and survival (Tables 1, 2).25–27

ALC Predicts Survival in AML

ALC emerged as a significant predictor for both AML RFS and OS. Prior studies have arbitrarily used an ALC of 500 cells/μL or 1000 cells/μL. We allowed our data to direct this cutoff. In our AML cohort, who had the largest number of events, 350 cells/μL was the median ALC-15 value for patients who recurred or died. Using 350/μL, we found that AML patients with an ALC-15 <350 cells/μL had a poor 5-year RFS of 23% (HR 2.7, P = .025) and an OS of 28% (HR 4.9, P = .012) (Table 2, Fig. 1A,B). In contrast, an ALC-15 ≥350 cells/μL predicts an excellent 5-year OS of 85% (HR 0.2, P = .012). This defines a very large group (50%, 25 of 50) for which the outcome for AML is excellent based on a single routine lab test. Strikingly, an ALC <350 cells/μL later in induction, on Day 28 (ALC-28), was able to identify 22% (10 of 47) of our AML patients who had a dismal prognosis of only 10% 5-year RFS (HR 3.7, P = .003; Table 2, Fig. 1C). Furthermore, ALC measurements on Days 15, 21, and 28 were all able to predict AML survival in a statistically significant manner (HR 2.6–4.9, P = .003–0.048, Table 1). Multivariate analysis demonstrated that these ALCs continued to be a significant prognostic factor in AML patients after adjusting for favorable and unfavorable cytogenetics, age, and initial WBC (Tables 1, 2). Three of the patients died of complications from treatment and had an ALC-15, -21, -28 ≥350 cells/μL.

Figure 1. Kaplan-Meler estimates of relapse-free survival (RFS) and overall survival (OS) in AML patients. Patients with an ALC <350 cells/μL versus patients with an ALC ≥350 cells/μL. (A) For ALC-15, the 5-year RFS was 23% vs. 61% respectively, p = 0.025, with a median RFS of 13 months vs. not reached. (B) The 5-year OS was 25% vs. 85% respectively, p = 0.012, with a median OS of 24 months vs. not reached (C) For ALC-28, the 5-year RFS was 10% vs. 51% respectively, p = 0.003. The median RFS was 10 months vs. not reached.

ALC Predicts Survival in ALL

Remarkably similar to AML, ALC-15 is a significant predictor of both ALL RFS and OS. Specifically, ALL patients with an ALC-15 <350 cells/μL had a 6-year RFS of only 43% (HR 4.5, P = .002) and an OS of 55% (HR 4.1, P = .018) (Table 3, Fig. 2A,B). In contrast, 60% of ALL patients had an ALC-15 ≥350 cells/μL and an RFS of 80% (HR 0.22, P = .002) and a very good 6-year OS of 87% (HR 0.24, P = .018). Even more striking is that nearly half (43%) of our ALL patients had an ALC ≥350 cells/μL on at least 1 measurement (Day 15, 21, or 28) that was predictive of an excellent 6-year RFS of 89% (HR 0.17, P = .017) (Fig. 2C) and a 4-year OS of 96% (HR 0.16, P = .084) (Table 3) in a group with 70% high-risk patients. Indeed, low ALCs measured at any timepoint during induction (15, 21, or 28) were independently predictive of recurrence (HR 2.9–6, P = .002–0.030; Table 3). Finally, ALC-15 was a strong predictor of time to recurrence (TTR). Patients with ALL who had an ALC-15 <350 cells/μL had a 6-year relapse-free rate (RFR) of 47% versus 82% (HR 4.8, P = .003). These survival differences increased with longer follow-up, suggesting that ALC during induction influences both early and late recurrence and survival. Using multivariate models, we found that a low ALC-15 remained strongly associated with poor RFS after adjusting for age, initial WBC, treatment regimen, bone marrow blast percent on Day 7, and unfavorable cytogenetics (RFS HR 5.2, P = .008; Table 3). This high HR implies that a patient with an ALC-15 <350 cells/μL has an 83% higher probability of recurrence compared with a patient with an ALC-15 ≥350 cells/μL.28 Although treatment intensity might be expected correlate inversely with ALC and outcome, we found that treatment on a high-risk pediatric protocol (primarily CCG-1961) or Hyper-CVAD did not alter the significance of the ALC-15 (HR 7.8, P = .005; Table 3). Although patients treated on the high-risk pediatric protocols had a reasonably high HR 3.7, as expected, the P-value was not significant in our cohort. Two of the patients died of complications from treatment. One of the patients had ALC-15, -21, -28 ≥350 cells/μL.

Figure 2. Kaplan-Meler estimates of relapse-free survival (RFS) and overall survival (OS) in ALL patients. Patients with an ALC <350 cells/μL versus patients with an ALC ≥350 cells/μL. (A) For ALC-15, the 6-year RFS was 43% vs. 80% respectively, p = 0.002, with a median RFS of 30 months vs. not reached. (B) The 6-year OS was 55% vs. 87% respectively, p = 0.018. (C) For ALC at all time points (15, 21 + 28), the 5-year RFS was 51% vs. 89% respectively, P = 0.018.

Table 3. ALL Statistical Analysis

* P < .05. † P < .01.
ALL Overall Survival	6-Year OS Rate	Hazard Ratio (95% CI)	P
ALC-15 <350cells/μL	55% vs 87%	4.1 (1.3–13)	.018*
ALC-21 <350cells/μL	62% vs 83%	2.3 (0.7–8.0)	.191
ALC-28 <350cells/μL	71% vs 83%	2.8 (0.8–11)	.122
ALC <350 at all time points	62% vs 96% (4 y)	6.1 (0.8–48)	.084
Multivariate Analysis
ALC-15 <350cells/μL		5.2 (1.2–23)	.027*
Age at diagnosis ≥10 y		1.2 (0.3–5.1)	.795
Initial WBC ≥50,000cells/μL		3.2 (1.0–11)	.059
Unfavorable cytogenetics		1.1 (0.3–4.2)	.912
ALL Relapse-free Survival	6-Year RFS Rate	Hazard Ratio (95% CI)	P
ALC-15 <350 cells/μL	43% vs 80%	4.5 (1.7–12)	.002†
ALC-21 <350 cells/μL	47% vs 80%	2.9 (1.1–7.5)	.030*
ALC-28 <350 cells/μL	57% vs 76%	3.3 (1.1–9.5)	.030*
ALC <350 at all time points	51% vs 89%	6 (1.4–26)	.018*
Multivariate analysis
ALC-15 <350cells/μL		4.5 (1.5–14)	.009†
Age at diagnosis ≥10 y		2.9 (1.0–7.8)	.041*
Initial WBC ≥50,000cells/μL		2.7 (1.0–6.9)	.040*
Unfavorable cytogenetics		1.2 (0.4–3.6)	.738

DISCUSSION

ALC Is an Independent Prognostic Indicator in Acute Leukemias

In this exploratory study we found that an ALC <350 cells/μL during induction predicts poor prognosis in children and young adults with acute leukemias. Our data demonstrate that ALC has prognostic significance for RFS and OS in both AML and ALL, using multivariate models. For example, based on ALC-15 greater than or less than 350 cells/μL, AML patients had an improved 5-year OS by a difference of 57% (28% vs 85%, HR 4.9, P = .012, Fig. 1B), making ALC a very clinically relevant prognostic factor. Similarly in ALL, ALC-15 differentiates survival groups by a difference of 34% (Fig. 2A,B). In contrast, the initial WBC separates the good and poor outcome patients by only 7%. These data suggest that ALC-15 is a very strong independent prognostic indicator and may be used to significantly alter postinduction risk stratification for both AML and ALL. To compare with previous studies, we analyzed our data using an ALC <500/μL and also found consistently high HRs and statistical significance in multivariate analysis.

Importantly, both high- and low-risk groups can be defined using ALC measurements. In ALL, ALC-15 ≥350 cells/μL defines a subgroup with 6-year OS of 87%, similar to patients currently stratified as standard risk on current protocols. However, this very good prognosis is independent of current risk stratification because 55% (27 of 49) of these good ALC-15 patients were stratified as high-risk by age or WBC. In contrast, 39% of ALL patients had low ALC-15 and had a relatively poor 6-year RFS of 46%. Even more striking, ALC-15 in AML patients defines a subgroup of 50% of our patients with an excellent 5-year OS of 85% (Fig. 1B). Finally, ALC-15 has the sensitivity to predict ≈70% of all recurrences in both ALL and AML patients.

Behl et al.15 demonstrated that ALC is predictive of outcome in older adults with AML (median age 59 years). Through comparison with our dataset, we observe that in our younger population (median age 18 years) ALC has a more dramatic predictive ability (HR 4.9 vs 2.1) and a lower ALC cutoff (350 vs 500 cells/μL). This higher predictive significance may point to a more profound role of lymphocytes in the younger patient with AML or differing host sensitivity to chemotherapy. Behl et al. did not review adult ALL patients, so it remains to be seen if ALC is prognostic in older adults with ALL.

Immune Status in Acute Leukemia

Initial WBC, cytogenetics, and, indirectly, age (eg, t(9;22) increases with age) predict outcome based on the biologic characteristics of the leukemia blasts. Similarly, MRD evaluation is a direct measurement of the chemosensitivity of these blasts. In contrast, ALC is a novel host factor with a surprising predictive ability. Although it is not yet known whether these lymphocytes are quiescent or functional lymphocytes, we believe that it is unlikely that quiescent lymphocytes would impact so profoundly recurrence and survival.

We speculate that these lymphocytes may represent an effective front line in the host's immune response to leukemic blasts, at a time when immune tolerance may be decreasing and tumor antigenicity may be increasing. Therefore, ALC during induction may be a surrogate marker for host immune robustness. Indeed, Lowdell et al.29 demonstrated that autologous activity by NK cells is what determines the prolonged remission in patients with AML. Studies conducted by Ohnishi et al.30 analyzed the subset population of lymphocytes in 30 patients with AML in complete remission. Their data also suggest that cytotoxic T-cells and NK cells play a role in immune surveillance after the administration of chemotherapy. A recent abstract showed that NK cell numbers are preserved during induction for pediatric ALL, whereas nonmalignant B cells, and to a lesser degree T cells, are significantly decreased.31 Furthermore, in pediatric patients with ALL in remission, Komada et al.32 found that overall lymphocyte counts were depressed and that percentages of CD8+ T cells were increased. Finally, NK cell numbers in autologous grafts have been shown to correlate with ALC posttransplant, whereas CD19+ and CD3+ cells do not.33 Therefore, a thorough analysis of lymphocyte subsets and function during induction may identify patients with an unfavorable immune profile. Such patients may benefit from novel immunomodulatory therapies. Previous studies of several chemotherapeutic agents (eg, cyclophosphamide, temozolomide, or taxotere) have been shown to promote a desirable cytokine profile by inhibiting regulatory T-cells when administered at subtumoricidal levels.34–37

Economic/Global Implications

Our findings suggest that a routine CBC with differential can be used to further refine postinduction risk stratification. Although developing countries often follow treatment protocols from developed countries, limited resources can prevent delivery of optimal care, resulting in poorer survival in ALL, 10% to 60% versus >80%.38–40 This disparity may become more significant as developed nations use more sophisticated studies for early identification of high-risk patients, eg, quantitative polymerase chain reaction (PCR) and flow cytometry-based MRD analysis. Although we advocate for the global use of MRD technology, in some circumstances these assays may not be technologically and/or economically feasible. In these areas the use of ALC to guide postinduction risk stratification may be of particular importance. Indeed, current fees for MRD assays at our institution approach $4000 per sample, compared with $75 for a CBC with differential. Risk-directed therapy after early stratification of patients with a universal test such as ALC may lead to significant improvement in survival in developing countries.

ALC as a Prognostic Factor in a Range of Malignancies

In this study we report the predictive ability of ALC in children and young adults with ALL and AML. In addition, we also demonstrated the predictive ability of ALC in Ewing sarcoma, a bone tumor.41 Similarly, we have discovered that this predictive ability holds true for young patients with non-Hodgkin lymphoma (unpublished data, G. De Angulo). These studies, when combined with several studies in adults by Porrata et al. and Behl et al., suggest that there is a generalized posttherapy ALC phenomenon, relevant to a wide range of malignancies and ages. The importance of these lymphocytes urges a reevaluation of the role of the immune response in cancer treatment and outcome.

Summary

This study demonstrates that ALC during induction chemotherapy is a significant independent predictor of outcome in children and young adults in our ALL and AML cohorts. Because ALC data are easily obtained for nearly all patients, confirmation in larger cohorts of patients should be obtained. If confirmed, this simple measurement from the CBC may provide critical information for more accurate early risk stratification. The improvement in risk-adapted stratification for patients with poor lymphocyte counts may optimize treatment and improve remission durability and cure. Finally, the mounting evidence for a powerful role for host immune status in the outcome of ALL, AML, NHL, and Ewing sarcoma, in children and adults, implies an underlying phenomena that encourages a reexamination of immune-modulating therapies in a wide range of malignancies.