Absolute lymphocyte count at day 28 independently predicts event-free and overall survival in adults with newly diagnosed acute lymphoblastic leukemia

Introduction

Acute lymphoblastic leukemia (ALL) develops from acquired somatic mutations that promote survival and depress differentiation of lymphoid hematopoietic progenitors. Although often viewed as a childhood disease, ALL affects ∼1/100,000 people in the United States over age 15 years. Long-term survival rates for adults with ALL remain poor at 40% [1] when compared with pediatric ALL patients, [2] for whom long-term survival approaches 90%. The limited available predictive markers for assessing risk and determining treatment stratification—already in use to treat childhood ALL—may play a role in the poor prognosis of long-term adult ALL survival.

Risk assessment of adult ALL patients considers age, [3] cytogenetics, [4] white blood cell count at diagnosis, [5] and the presence of minimal residual disease (MRD) following treatment [6]. These features inform the treatment plan, including whether allogeneic bone marrow transplant had to be performed in first complete remission (CR1). Identification of additional prognostic markers could help predict relapse, permit better risk stratification, and promote the development of novel therapies.

As an indicator of bone marrow recovery from chemotherapy, the absolute lymphocyte count (ALC) may be reflective of the body's immune surveillance potential against cancer cells. Hematopoietic malignancies are targeted by natural killer (NK) cells and T cells, which recognize myeloid and B cell malignancies [7]. Alloreactive T cells and NK cells also mediate graft-versus-leukemia (GVL) effects, in which they destroy residual tumor cells [7, 8] following hematopoietic stem cell transplantation (HSCT). In pediatric ALL patients receiving allogeneic HSCT, those exhibiting faster ALC recovery had lower relapse rates, decreased graft-versus-host-disease, and higher survival rates [8]. Similarly, pediatric ALL patients receiving unmanipulated haploidentical related SCT with higher ALC at day 30 post-transplant demonstrated prolonged survival and decreased transplant-related mortality [9]. These studies suggest that lymphocytes may play a role in eliminating residual disease and promoting long-term survival. Indeed, childhood ALL studies have demonstrated that higher ALC at day 15 of induction predicts improved 6-year OS [10]. A similar study evaluating the prognostic value of ALC in pediatric ALL found that high ALC at day 29 of induction was predictive of improved 5-year EFS [11]. These studies suggest that ALC levels following induction chemotherapy may be a marker of immune competence, reestablishing its surveillance of malignancy, and eradicating MRD.

Although ALC has been evaluated for prognostic significance in pediatric ALL, it is unclear whether this marker carries the same prognostic import in adult ALL. We hypothesized that higher ALC following induction therapy in adult ALL patients is associated with prolonged OS and EFS.

Materials and Methods

We conducted a retrospective chart review of 230 adult patients (age range 18–64 years) with de novo ALL treated with induction chemotherapy between the years 1993 and 2010 at the Cleveland Clinic and Stanford University. Institutional Review Board approval was obtained at both institutions. Prior studies in the pediatric population identified an ALC of 350 cells/μL as a cutoff, which was predictive of outcome. We assessed this cutoff as well as other ALC metrics including ALC counts, nadirs, and area under the curve at various time points. ALC was measured using an automated differential. Day 28 was counted from the first day of chemotherapy. Gender, age at diagnosis, cytogenetics, and white blood cell count at diagnosis were included in univariate and multivariate analyses. Cytogenetic risk groups were defined by Cancer and Leukemia Group B (CALGB) criteria [12]. OS was measured from start of induction therapy to death from any cause and EFS was measured from the date of CR to the date of relapse or death. The Kaplan–Meier method was used to summarize OS and EFS and the log-rank test was used for univariable analysis. The Cox proportional hazards model, stratified by institution, was used for multivariable and univariable analyses of measured factors. A number of patients had incomplete data and therefore analyses were conducted two ways: using only patients with the relevant data and using multiple imputations (five data sets and assuming multivariate normality). Results were similar in both analyses and therefore only those based on patients with data are presented. For convenience, a recursive partitioning algorithm was used to identify cut-points for the white blood count. All data analyses were conducted using SAS 9.2 (SAS, Cary, NC).

Results

Predictors of Outcome

In univariable analyses, older age at diagnosis, poor risk cytogenetics, and elevated white blood count at diagnosis were all associated with poor outcome (Supporting Information Table S1). Several of the ALC metrics were also found to be associated with outcome; however, the ALC at day 28 with a cut-point of <350 cells/μL was the strongest predictor of OS and EFS (Supporting Information Table S1). Patients with an ALC ≥350 cells/μL at day 28 had a median OS of 47.4 months when compared with 17.6 months for patients with an ALC <350 cells/μL (HR = 1.98, P = 0.007, Fig. 1). Among patients who achieved a CR, the median EFS for those with an ALC ≥350 cells/μL on day 28 was 42.1 months when compared with 13.9 months in patients with an ALC <350 cells/μL (HR = 2.08, P = 0.006, Fig. 1). The improved EFS was due to a lower relapse rate.

In multivariable analyses, the ALC on day 28 (<350 cells/μL vs. ≥350 cells/μL, P ≤ 0.0001 for OS and P = 0.0004 for EFS), white blood count at diagnosis (P = 0.002 and 0.001), and cytogenetics (P = 0.002 and 0.02) were independent prognostic factors for both OS and EFS (Table II). The magnitude of the regression coefficients for both models were similar in magnitude and the models could therefore be described by simply counting the number of poor prognostic features present, where day 28 ALC <350 cells/μL, WBC ≤6.0 or >30.0 K/μL, and abnormal cytogenetics each count as one poor feature. Applying this simplification results in three distinct prognostic groups (Fig. 2). Estimated median OS and EFS for patients with the most favorable profile (0–1 poor risk feature, 30% of patients with data) has not yet been reached; however, 3-year OS and EFS are estimated to be 76 ± 9% and 72 ± 10%, respectively. In contrast, patients with the least favorable profile (all three poor features present, 26%) had an estimated OS and EFS of 9.0 months (95% CI 6.4–15.2) and 6.5 months (95% CI 3.5–8.0), respectively; and 3-year OS rates <6%. Patients with two poor features (45%) had an estimated median OS and EFS of 47.4 months (95% CI 17.5, N/A) and 66.0 months (95% CI 13.1, N/A), respectively and corresponding 3-year OS rates of 58 ± 8% and EFS of 59 ± 8%.

Discussion

This study demonstrates that ALC is predictive of OS and EFS in adult ALL. A number of ALC metrics correlated with OS and EFS, but day 28 ALC dichotomized at the cut-point of 350 cells/μL was the strongest predictor of survival and remained significant in multivariable analysis. The combination of day 28 ALC with additional known prognostic factors (cytogenetics and white blood count) separates patients into three distinct prognostic groups. This study involved two major academic institutions, a large number of patients, and the results were statistically significant. It was retrospective, however, and had to be validated prospectively in a group of uniformly treated patients. A number of patients were missing key data including day 28 ALC. However, there were no significant differences in key predictors such as cytogenetics (P = 0.86) and white blood count (P = 0.98) or outcome (P = 0.24 and 0.52 for OS and EFS, respectively) between patients with and without Day 28 ALC data, suggesting that patients with available day 28 ALC data are not a select subpopulation. In addition, as discussed in the methods, analyses that used multiple imputation methods to handle missing data yielded results similar to those presented. The ALC measurement was also subject to some error because it was based on a manual differential and some of the lymphoblasts may have been counted as lymphocytes. However, the majority of the cases were reviewed by a hematopathologist.

These data, along with the previous reports of ALC as a prognostic factor in pediatric ALL [10], adult lymphoma [13], and adult acute myeloid leukemia [14], demonstrate that ALC can be used as a prognostic marker in adult ALL and may be useful in risk stratification for treatment. Since our study is retrospective, we were unable to determine whether higher ALC causally leads to prolonged survival. Lower ALC may be a secondary measurement for chemotherapy toxicity or complications from treatment, which may result in treatment delays or dose reductions that contribute to decreased survival [15]. ALC may also be linked to characteristics of the leukemia itself; for example, patients with better ALC recovery may present with a lower leukemic stem cell burden at diagnosis, which would lead to improved immune reconstitution following therapy and prolonged survival.

Although we are unable to conclude this from our study, a more direct relationship between higher ALC and prolonged survival may be due to the immune surveillance of reconstituting lymphocytes against malignant cells [7]. Many studies have shown the importance of functional immune cells in maintaining remission. The GVL effect, for example, has been characterized in HSCT with recipients of identical twin donors having significantly higher relapse rates than recipients of allogeneic donors. [7] In addition, inherent leukemia-specific T-cell responses have been shown to be increased in leukemia patients vs. normal individuals, suggesting that, though present, immune control could not control leukemic cell proliferation [16, 17]. Further study is required to determine if HSCT and current treatment regimens help to re-establish the balance in the immune-leukemia equilibrium [7].

Mechanisms of immune surveillance have yet to be fully elucidated. Studies have demonstrated that NK cells are able to target hematopoietic malignancies, directing the GVL effect, and even predicting relapse in patients with low NK leukemia cytolytic activity [18, 19]. The presence of CD8⁺ T cells that can be activated ex vivo against tumor-specific antigens, such as BCR-ABL, PR1, and WT-1, have been identified in chronic myeloid leukemia [20]. Recently, the role of CD4⁺ T cells in oncogene addiction—the dependence of malignant cells on a single oncogene—has been evaluated. The CD4⁺ T cell chemokine expression profile mediates tumor regression, senescence, and the discontinuation of angiogenesis upon inactivation of certain oncogenes (MYC in lymphoma, BCR-ABL in pro-B cell leukemia) [21].

Despite these studies suggesting the importance of an intact immune system in maintaining remission in hematologic malignancies, it is clear that induction chemotherapy compromises the function of these healthy immune cells. The immune profile in ALL patients following induction therapy has not been well-characterized. Eyrich et al. [22] evaluated immune function in children with standard risk ALL during induction chemotherapy and found that normal B cells were most severely depleted with NK and T cells partially recovering by the end of induction. T cell function, including the CD4:CD8 ratio, cytokine production, and TCR repertoire, were minimally affected. Further work is needed to evaluate the types of lymphocytes reconstituting the immune system following induction in adult ALL.

Understanding the effect of chemotherapy and the mechanism of immune surveillance can help to optimize treatment protocols for adult ALL. Current treatment regimens diminish the leukemic burden but may also deplete the immune cells necessary to maintain long-term remission. Modifying the ALC with immune-boosting therapies might improve survival in adult ALL. Growth factors such as IL-2, IL-7, IL-12, and IL-15 can be used to promote lymphocyte recovery post-chemotherapy [7]. Another strategy is to induce antigen specific T-cell expansion against residual leukemic cells after chemotherapy [7]. One approach, evaluated in the lab, has been to target toll-like receptors (TLRs) [23]. TLR9 is one of the most abundant TLRs on pre-B cell ALLs [23]. The classical ligands for TLR9 are CpG oligonucleotides. In a murine-xenograft model, CpG administration decreased leukemia burden [24] and induced durable remission and ongoing immune-mediated protection again leukemia [25].

ALC may be useful not only in risk-stratification of newly diagnosed adults with ALL but also in determining which patients could benefit from the development of novel adjuvant therapies to boost the immune system. This study indicates the importance of a high ALC in predicting survival in adult ALL and suggests the need for additional studies evaluating the relationship between ALC and mechanisms of immune surveillance.