Clearance of leukaemic blasts from peripheral blood during standard induction treatment predicts the bone marrow response in acute myeloid leukaemia: a pilot study


Dr Giacomo Gianfaldoni, Department of Haematology, University of Florence, Azienda Ospedaliera-Universitaria Careggi, 50134 Florence, Italy.


Although several parameters are useful for risk stratification of patients with acute myeloid leukaemia (AML), there are no firm criteria for predicting response to induction treatment of individual patients. Daily flow cytometry (FC) analysis, carried out during induction treatment in 30 AML patients, showed that the clearance of blasts from peripheral blood (PBC) correlated closely with response, as assessed by bone marrow FC on day 14, and by morphologic analysis at haematopoietic recovery. Therefore, a major treatment outcome can be predicted very early in AML patients, thus providing an opportunity for tailoring treatment modalities from the outset.


Achieving complete remission (CR) is a prerequisite for long-term survival in acute myeloid leukaemia (AML). Among the predictors of CR, age, cytogenetics and secondary versusde novo disease have proven to be the most informative. These factors provide a pretreatment stratification in risk groups with different probabilities to obtain CR: however, they are not sufficient to predict the individual response to the first course of treatment. Moreover, given that, in most cases, induction therapy must be initiated as soon as possible, it is not generally feasible to collect the necessary information to allow stratification at diagnosis. Attempts to predict intrinsic drug resistance by quantitating expression of the multidrug resistance glycoprotein MDR1 using flow cytometry (FC) techniques showed conflicting results (Leith et al, 1999; Broxterman et al, 2000). A recent study showed some predictive value of ex vivo chemosensitivity tests, but these tests requires culture facilities and take several days (Staib et al, 2005). In practice, therefore, individual chemosensitivity can be evaluated, through the assessment of residual disease, only after therapy has been completed. Typically, the response is evaluated, upon full recovery of peripheral blood (PB) counts, by bone marrow (BM) assessment. An earlier BM evaluation has been shown to predict those patients that will or will not achieve CR (Kern et al, 2003; Haferlach et al, 2004). The present study of adult AML patients attempted to obtain a very early prediction of response to chemotherapy during the first course of treatment.


Between May 2004 and January 2006, 30 consecutive newly diagnosed non-M3 AML [according to the French–American–British (FAB) classification] patients (Table I), aged <66 years were treated with a ‘3 + 7’ induction course (cytarabine 100 mg/m2, infused over 3 h, every 12 h, days 1–7; idarubicin 12 mg /m2, infused over 30 min, days 1–3) and were evaluable for BM response. After obtaining informed consent, extensive FC studies were carried out. A population of cells with leukaemia-associated aberrant immuno-phenotype (LAIP) was identified in each patient from the initial BM aspirate. LAIP-positive absolute blast counts were determined on PB immediately before starting therapy and every day until day 8. The clearance of peripheral blast cells (PBC) was expressed as the ratio, converted to logarithmic scale, between baseline value (day 1) and daily absolute blast count. At day 14, FC analysis was performed on BM in order to identify LAIP-positive residual blasts. The degree of BM clearance was expressed as the ratio, converted to logarithmic scale, between the percentage of LAIP-positive blasts determined at diagnosis and day 14 (LD14).

Table I.   Characteristics of patients.
Patients, n30
  1. FAB, French–American–British classification; WBC, white blood cell; LAIP, leukaemia-associated aberrant immuno-phenotype; LD14, ratio between the percentage of LAIP + blasts at diagnosis and at day 14, converted to a logarithmic scale; CR, complete remission; NCR, non-complete remission.

  2. WBC count, Peripheral blasts and LAIP + peripheral blasts refer to day 1, immediately before starting induction.

  3. Cytogenetic risk was assessed according to South West Oncology Group criteria (Slovak et al, 2000).

  4. The definition of CR used established criteria (Cheson et al, 1990). Patients with unfavourable karyotype and/or secondary AML were considered at high risk for primary refractory disease; patients with favourable cytogenetics were classified as low risk; the remainders at intermediate risk.

  5. Exclusion criteria were AML FAB M3 with t(15;17) and/or pml/rarα rearrangements, previously treated AML, absence of morphologically identifiable blast cells in peripheral blood, Eastern Cooperative Oncology Group Performance Status >3 and/or severe cardiac, pulmonary, hepatic or renal impairment not dependant on disease. Patients with preceding myelodysplastic syndrome were not excluded.

Median age, years (range)50 (23–65)
Gender, no.
FAB subtypes, n
Cytogenetic risk group, n
Prior myelodysplastic syndrome, n5
WBC count, ×109/l median (range)11·25 (1·32–125·24)
Peripheral blasts, % median (range)52 (5–94)
LAIP + peripheral blasts, % median (range)20·17 (0·42–88·53)
LAIP + peripheral blasts, ×109/l median (range)2·757 (0·015–74·472)
Bone marrow blasts at diagnosis, % median (range)75 (30–100)
LAIP + bone marrow blasts at diagnosis, % median (range)24·7 (3·31–80·11)
LAIP + bone marrow blasts at day 14, % median (range)0·52 (0·01–49·11)
LD14 median (range)1·85 (0·06–3·30)

Results and discussion

After a single ‘3 + 7’ course, 17 patients achieved CR and 13 patients did not (non-complete remission, NCR). Of these 13 NCR patients, five obtained a partial remission and eight were refractory. According to conventional criteria (cytogenetics and preceding myelodysplastic syndrome), there were 11 high-risk patients, of whom four achieved CR; 14 intermediate risk patients, of whom eight achieved CR; five low risk patients, all of whom achieved CR.

A rapid reduction of peripheral blasts was seen in all patients who eventually achieved CR; by contrast, a slower reduction was seen in NCR patients. On each day, PBC values in CR and NCR groups showed minimal overlap (Fig 1A). The medians of log reduction in the two groups were significantly different on each day, starting as early as day 2 (Fig 1B). In order to compare the rate of PBC between the two groups, we first established that the rate was constant in each patient (data not shown). Next, a multiple regression approach was used to compare the rates estimated in NCR and CR patients, adjusted for baseline and assigned risk. The rate of PBC appeared higher in CR than NCR patients with an estimated difference between groups equal to 0·26 [95% confidence interval (CI) 0·15–0·37; P-value < 0·001]. The difference in rate of PBC between the CR and the NCR groups was not attributable to differences in baseline PB leukaemic burden or in assigned risk category.

Figure 1.

  (A) Peripheral blasts log reduction in complete remission (CR) and non-complete remission (NCR) groups on each day of treatment. The ranges of distribution of the ratios between baseline and daily absolute blast count, converted to a logarithmic scale (Log Reduction), have minimal overlap between CR and NCR groups. In patients who achieved CR, blasts were often already undetectable by day 7 or 8 and so we excluded these time-points from analysis. Box-and-whisker plot: the line at the top of each box denotes the 75th percentile; the line at the bottom of each box, the 25th percentile; the line in the middle of each box, the 50th percentile or median. The whiskers extend to the most extreme data points, which are no more than 1·5 times the interquartile range distant from the box. (B) Log Reduction of peripheral blasts over days of therapy in CR and NCR groups. CR, complete remission; NCR, non-complete remission. For each day of treatment, we used the Wilcoxon (W) Rank Sum test to compare the magnitude of log decrease in CR and NCR groups. The 95% CI was calculated by Bootstrap method. (C) Log Reduction of leukaemic blasts from peripheral blood during induction treatment correlates with bone marrow response as assessed by FC on day 14. Higher values of peripheral blast decrease (Log Reduction) were associated with larger values of bone marrow blast reduction on day 14 (LD14). Slope is the regression coefficient between Log Reduction and LD14 adjusted for baseline and assigned risk category. Day 2: slope = 0·23 (95% CI: 0·15, 0·30); day 3: slope = 0·55 (95% CI: 0·36, 0·74); day 4: slope = 0·81 (95% CI: 0·61, 1); day 5: slope = 0·93 (95% CI: 0·71, 1·15); day 6: slope = 0·99 (95% CI: 0·71, 1·27).

The cut-off point in PBC that would optimise sensitivity and specificity was determined for each day (data not shown). Using a log reduction of 2 as the cut-off point on day 5, the sensitivity was 1 and the specificity was 0·80, with 92% of patients correctly classified. Specifically, CR was not achieved in any of 11 patients who had a PBC below 2 logs on day 5, whereas CR occurred in 17 of 19 patients who had a PBC >2 logs on day 5.

A multiple regression approach was used to describe the relationship between LD14 and log reduction at each time point. Higher values of PBC were associated with larger LD14 on each day (Fig 1C). This correlation was significant on each day and it increased monotonically over days. Therefore, it appears that PB may be in equilibrium with BM in each AML patient, and that PB clearance gives evidence of BM clearance.

Since our approach uses a FC analysis on PB, it has the advantage of being minimally invasive, relatively simple, reproducible and rapid. The fact that a high degree of correlation of PBC with CR was obtained, even in a small cohort of patients, demonstrated that the PBC provides a prediction of response in real time for each patient.

Such an early detection of response to the first standard induction course could provide the opportunity to apply ‘immediate intensification’ to probable non-responsive patients. This approach may make it possible to obtain, with a single course of treatment, a rate of CR comparable with that achieved using intensified induction for all patients (Bishop et al, 1996; Russo et al, 2005) but with a reduction in the overall toxicity. There are two potential advantages (a) no intensification for patients who do not need it; (b) patients subjected to ‘immediate intensification’ might achieve CR earlier. We cannot be sure whether our results may be protocol-dependant. For this reason, and in order to increase the predictive value of PBC by adjusting the cut-off value in log reduction, we plan to extend our analysis from this pilot study to a larger cohort of patients.


We thank Prof. Lucio Luzzatto for his support and valuable collaboration in reviewing the manuscript.

Authors’ contributions

Giacomo Gianfaldoni designed research, performed cytofluorimetric analysis, analysed the data and wrote the paper. Francesco Mannelli designed research, performed cytofluorimetric analysis, analysed the data and wrote the paper. Michela Baccini performed statistical analysis. Elisabetta Antonioli performed cytofluorimetric analysis. Franco Leoni evaluated patients and critically reviewed the paper. Alberto Bosi evaluated patients and reviewed the paper.