Prospective Phase II Study on 5-Days Azacitidine for Treatment of Symptomatic and/or Erythropoietin Unresponsive Patients with Low/INT-1–Risk Myelodysplastic Syndromes

During the last years, the most relevant progress in biology and treatment of myelodysplastic syndromes (MDS) has been marked by results of epigenetic studies (1, 2).

Aberrant DNA methylation in the CpG sites of several tumor suppressor genes (TSG), even in the absence of LOH, is considered one of the dominant pathogenetic mechanisms in MDS and one of the main causes of progression to acute myelogenous leukemia (AML). Furthermore, clinical responses of MDS to drugs that reverse aberrant hypermethylation, such as deoxycytidine or azacitidine, strongly suggest that aberrant hypermethylation plays a causative role in disease and not just a side effect resulting from a deregulated proliferation or DNA damage (3, 4).

Hypomethylating agents significantly modified the therapeutic approach to MDS, primarily in older patients with high-risk disease for whom intensive chemotherapy and allogeneic stem cell transplantation are not an option (5, 6). At a molecular level, the mechanisms underlying the effect of epigenetic therapy are not completely understood, although it is well known that DNA methyltransferase inhibitors can induce the expression of the methylation-silenced gene products (7). For instance, epigenetic regulation of phosphoinositide-phospholipase C (PI-PLC) β 1 promoter, which affects key molecules involved in cell cycle and differentiation (8, 9) such as cyclin D3, seems to play an important role in the demethylating activity of azacytidine (10–12). Indeed, PI-PLCβ1 is a central molecule in normal cell proliferation and differentiation (13, 14), as well as in hematologic malignancies, such as MDS (15, 16). In fact, MDS showing a PI-PLCβ1 monoallelic deletion are associated with a higher risk of AML evolution (17), and PI-PLCβ1 deregulation affects both myeloid and erythroid differentiation of low-risk MDS cells (18, 19).

Clinically, in high-risk MDS, azacytidine has been the first and only agent able to induce a survival benefit as compared with conventional care regimens. In this setting, the median overall survival (OS) was 24.5 months in azacytidine-treated cases versus 15 months in those treated with conventional care regimens (P = 0.0001; refs. 20, 21).

In low-risk MDS, the role of hypomethylating agents is less understood. The main goals of hypomethylating therapy in low-risk disease should be to reduce transfusion dependency (22), improve quality of life (23), and hopefully the survival, but it is still unclear if this therapeutic approach would be cost-effective. Most data on the use of azacytidine in low-risk MDS are retrospective and come from studies mainly conducted in larger cohorts of patients with high-risk disease. In the prospective phase III study [Cancer and Leukemia Group B (CALGB) 9221], 23 patients with low-risk MDS treated with azacytidine, at a dose of 75 mg/d for 7 days monthly, achieved an overall response rate (ORR) of 59% and a longer OS, as compared with the patients receiving only supportive care (24, 25). In another independent study, including 23 to 65 patients with low-risk MDS across 3 azacytidine alternate dosing cohorts, the ORR ranged between 44% and 56% and the rate of erythrocyte transfusion independence was 46% to 60% (26). More recently, a multicenter retrospective analysis conducted on 74 patients with MDS with International Prognostic Scoring System (IPSS) risk low or Int-1 treated with azacytidine at a dose of 75 mg/sqm for 7 to 10 days for a median of 7 cycles, reported 50% to 77% of responses, with a median duration of 6 months (27).

Altogether, these experiences suggest that azacytidine seem useful also for patients with low-risk MDS. However, the CALGB study 9221 included both high- and low-risk patients and the other cited studies were retrospective. Therefore, they suffer from relevant biases derived from the heterogeneity of enrolled patients, the different schedule and duration of treatments and, moreover, from the absence of biologic studies focused on identification of factors possibly correlated with sensitivity or drug resistance.

In this study, we prospectively evaluated the efficacy and safety of azacytidine, administered at a lower cumulative monthly dose [5-days azacytidine (5d-AZA); 75 mg/sqm/d for 5 days monthly], in patients with IPSS low- or Int-1–risk MDS who were symptomatic and/or unresponsive to previous treatments. Furthermore, we studied the genetic profile by single-nucleotide polymorphism (SNP) arrays and the molecular effects of azacytidine on PI-PLCβ1 promoter methylation and gene expression, to identify biologic factors possibly correlated with response to azacytidine.

The study is a phase II prospective, multicenter, single-arm trial. It was approved by the Ethical Committee of Spedali Civili University Hospital of Brescia (Brescia, Italy; EudraCT Number 2007-003943-55) and registered at ClinicalTrial.gov as NCT00897130.

Azacytidine was administered at a dose of 75 mg/sqm/daily s.c. for 5 consecutive days every 28 days for a total of 8 cycles. Response was firstly assessed after the first 4 cycles and in case of any hematologic response or stable disease, patients received additional 4 cycles. Final response was checked at the end of the eighth cycle.

Azacytidine dose was adjusted according to absolute neutrophil count (ANC) and platelets values. In particular, in case of ANC between 0.5 and 1.5 × 10⁹/L and/or platelets between 25 and 50 × 10⁹/L the dose was reduced to 75%. When ANC and platelet counts were less than 0.5 × 10⁹/L or 25 × 10⁹/L, respectively, dose reduction was of 50%. The first cycle of 5d-AZA was administered in all the patients, at the planned dose, independently of the peripheral blood counts.

Erythropoiesis-stimulating agents (ESA) were avoided during the study; however, granulocyte or granulocyte macrophage colony-stimulating factors (G-CSF or GM-CSF) were allowed in case of severe neutropenia (<0.2 × 10⁹/L) and/or systemic infection. Red blood cell (RBC) were transfused in case of hemoglobin (Hb) level < 8 g/dL, or whenever clinically indicated. Platelets support was allowed if platelet count was less than 20 × 10⁹/L, or whenever indicated. Antibiotics and antifungal prophylaxis were given if neutrophils were less than 0.5 × 10⁹/L. Systemic antibiotics were given in case of fever (>38°C for >24 hours), whereas systemic antifungals were given in case of fever persisting for more than 5 days on antibiotics or whenever indicated. Other experimental drugs or agents were not allowed and if another experimental drug or agent was administered, the patient discontinued azacytidine, and went off-study.

Physical examination, peripheral blood count, and biochemical assessment were carried out every 2 weeks or more often, depending on patients' conditions. Bone marrow aspirate samples for SNP microarray analysis were collected at study entry and after the fourth and eighth cycles. Peripheral blood samples for PI-PLCβ1 gene expression analysis were collected monthly during the treatment. Adverse events were graded according to the National Cancer Institute-Common Toxicity Criteria for Adverse Events (NCI-CTCAE) and were assessed at each patient visit. A Treatment Advisory Committee (TAC), made by investigators of the Writing Committee, was established to support decision-making and to check any relevant problem about the study protocol.

Student t test was used to compare continuous values and Fisher exact test was used to compare differences in percentage.

Patients could be enrolled if 18 years or older and if they had IPSS low- or Int-1–risk MDS and one or more of the following: (i) symptomatic anemia requiring RBC transfusion-supportive therapy, previously unresponsive to erythropoietin (EPO) or not expected to respond to EPO (i.e., with baseline serum EPO > 500), (ii) thrombocytopenia requiring platelet transfusion with or without muco-cutaneous hemorrhagic syndrome, (iii) persistent (>3 months) ANC less then 1.5 × 10⁹/L, with or without infections, requiring or not myeloid growth factor therapy. In addition, a performance status [Eastern Cooperative Oncology Group (ECOG)] of 0–2, an estimated life expectancy of at least 3 months, and a serum bilirubin level ≤ 1.5 upper limit of the normal (ULN), serum aspartate aminotransferase (AST) or alanine aminotransferase (ALT) ≤ 2 × ULN, and serum creatinine level ≤ 1.5 × ULN were required. A negative serum β-human chorionic gonadotropin pregnancy test 24 hours before beginning of therapy with azacytidine was requested for fertile women. All patients provided written informed consent.

The main objective of the study was to evaluate the efficacy (hematologic response) and toxicity of azacytidine in patients with low-risk MDS (IPSS 0–1).

Second, we investigated the molecular genetic profile by SNP microarray and the effects of 5d-AZA on PI-PLCβ1 for evaluating possible biologic markers able to predict the hematologic response.

The response to treatment was assessed according to International Working Group (IWG) criteria as reported by Cheson and colleagues (28), 1 month after the end of the eighth cycle of treatment. Briefly, we considered the following: (i) response criteria for altering the natural history of MDS, such as complete or partial remission (CR/PR), marrow remission, cytogenetic response, stable disease, failure, disease progression, or relapse after CR or PR; (ii) response criteria for hematologic improvement (HI), including erythroid response (HI-E), platelet response (HI-P), neutrophil response (HI-N), progression, or relapse after any hematologic improvement. Patients were considered to be responders if they had response duration of 8 weeks or more.

PI-PLCβ1 promoter methylation and gene expression were quantified in patients with MDS at baseline and monthly during azacytidine treatment. Genomic DNA and total RNA were extracted from peripheral blood mononuclear cells collected within 48 hours from the last azacytidine administration, by the DNeasy and RNeasy Mini Kits (Qiagen), according to the manufacturer's protocol. DNA was subjected to bisulfite modification and PI-PLCβ1 methylation was assessed as previously described (11). RNA was retro-transcribed and PI-PLCβ1b (Hs01001939_m1; Applied Biosystems) mRNA was quantified by a TaqMan Real-Time PCR method, using the GAPDH gene (Hs99999905_m1; Applied Biosystems) as the internal reference and a pool of 15 healthy subjects as the calibrator (29). Data were statistically analyzed by the GraphPad Prism software (v.3.0). Baseline levels were calculated as a percentage ratio of PI-PLCβ1 compared with healthy subjects, whereas the amount of PI-PLCβ1 during treatment was calculated as a percentage ratio of PI-PLCβ1 during azacytidine treatment compared with baseline. Statistically, a difference of at least 20% compared with baseline was considered significant, according to the Dunnett test post-ANOVA. Therefore, patients showing an increase in PI-PLCβ1 expression of at least 20% as compared with baseline at a specific cycle of azacytidine treatment were considered as having a positive “molecular response” at that cycle.

Genomic DNA was extracted using the DNA Blood Mini Kit (Qiagen) from mononuclear cells isolated from bone marrow aspirate samples before treatment with azacytidine. DNA was quantified using the NanoDrop Spectrophotometer and quality was assessed using the NanoDrop and by agarose gel electrophoresis. DNA samples were genotyped with Genome-Wide Human SNP 6.0 array microarrays (Affymetrix Inc.) following the manufacturer's instructions and as previously described (30). Briefly, copy number aberrations (CNA) and LOH were analyzed using Genotyping Console v4.1 (Affymetrix Inc.) by a comparison with a set of 794 HapMap normal individuals and a set of 25 normal samples run in our laboratory to reduce the noise of raw copy number data. Affymetrix CNCHP files were also imported and analyzed by Chromosome Analysis Suite (ChAS) Software v1.2 (Affymetrix Inc.) using the following segment filters: minimum number of markers 25 and minimum size of 50 kbp for both gain and loss. LOH was defined as a region greater than 5 Mbps.

Between September 2008 and February 2010, 7 Italian Haematological Centres recruited 32 patients with IPSS low- or Int-1–risk MDS (31) who met the eligibility criteria and gave their informed consent to be enrolled into the study (32–34).

The main characteristics of these 32 patients are reported in Table 1. Briefly, the median age was 71 years (range, 56–84) and according to World Health Organization (WHO) classification, 15 (47%) patients had refractory anemia (RA), 6 (19%) RA with ringed sideroblasts (RARS), 6 (19%) refractory cytopenia with multilinear dysplasia (RCMD), and 5 (16%) RA with excess of blasts (RAEB-1). Most patients had a good-risk karyotype (72%) by conventional cytogenetics, were RBC transfusion-dependent (81%), receiving a median of 4 RBC U/mo (range, 0–8), and were previously unresponsive to treatment with ESAs (77%). Only 4 patients (12%) were platelet transfusion-dependent.

All patients started azacytidine at a dose of 75 mg/sqm/daily s.c. for 5 consecutive days and their response is reported in Table 2. Out of the 32 enrolled patients, 2 died after the first cycle, 1 withdrew the informed consent after the second cycle, and 29 (90%) completed the first 4 cycles. At this time, 10 (35%) patients achieved hematologic improvement (HI-E in 7 and HI-N or HI-P in 3 cases), 19 (65%) had stable disease. Among the 29 patients eligible for therapy continuation after the fourth cycle, 1 withdrew the informed consent after the fourth cycle and 2 died after the fourth and sixth cycle, respectively. Therefore, 26 patients (81%) completed the treatment plan (8 cycles) and out of them: 5 (19%) achieved CR, 10 (38%) achieved hematologic improvement (HI-E in 8 cases and HI-N/P in the other 2 cases), and 11 (42%) maintained stable disease. Among the 5 patients who achieved CR after the eighth cycle, 3 had previously achieved hematologic improvement at the fourth cycle, whereas the other 2 had stable disease (Table 2). From a clinical point of view, out of the 26 patients completing 8 cycles, 21 and 2 were transfusion-dependent for RBC and platelets at baseline, respectively. Of them, 7 (33%) and 2 (100%) became transfusion-independent at the end of the treatment program, respectively. No patient progressed to high-risk MDS or AML during the treatment period. Thus, on an intention-to-treat, the ORR was 47% (15 of 32 cases), whereas assessing the data on the patients who completed the 8 planned cycles, the ORR was 58% (15 of 26 cases). As reported in Table 3, in the 15 responsive patients, the median duration of hematologic response assessed 1 month after the eighth cycle was 10 months (range, 2–37). At the time of the last follow-up, 17 (65%) of the 26 patients who completed 8 cycles are alive. Three patients have maintained their response achieved within the 8 cycles, after 37, 34, and 33 months, without other treatments or supportive therapy; 6 patients are alive, but lost their hematologic response; 8 patients maintained stable disease; 3 patients progressed to RAEB-2 or developed myelofibrosis. At the last follow-up, the median survival from the start of 5d-AZA and from diagnosis was 28.5 months (range, 10–41) and 55 months (range, 22–180), respectively. Nine of the 26 patients completing 8 cycles died. We did not find a correlation between response and the following baseline variables: age, sex, type and duration of MDS, karyotype, type and duration of previous treatments, or supportive therapies (data not shown).

Furthermore, the median survival from the start of azacytidine for responders was not different from nonresponders, including patients with stable disease (data not shown).

Azacytidine was generally well tolerated. As shown in Table 4, the most common hematologic toxicity was neutropenia, observed in 15 of 32 patients (47%) and 9 (28%) of them had a grade ≥ 3 WHO neutropenia. Thrombocytopenia occurred in 6 of 32 cases (19%) and was severe (≥ 3 WHO) in 4 of 6 (12%) of them. The hematologic toxicity was transient and in these cases azacytidine dose was adjusted according to the previously reported recommendations.

Nonhematologic toxicity was predominantly mild (grade ≤ 1) and mainly consisted of cutaneous erythema at the site of injection (62%), asthenia (31%), diarrhea/constipation (28%), and fever (9%). Only 5 patients had grade ≥ 3 WHO nonhematologic toxicities, who were reported as serious adverse events (SAE), and had favorable outcome in 3 of 5 cases (Table 4).

Out of the 32 enrolled patients, 2 patients died after the first cycle due to septic shock or gastrointestinal hemorrhage and these events were reported as SAE unlikely related to the study drug (Table 2). In the first case, the ANC was >2 × 10⁹/L at the time of death, and in the second case the patient had a severe thrombocytopenia (platelets < 10 × 10⁹/L) at baseline and had previously undergone a gastro-resection. Two other patients died because of pneumonia at the fourth cycle or respiratory distress at the sixth cycle (Table 2). The pneumonia was reported possibly related to azacytidine, because of progressive drug-induced neutropenia (0.7 × 10⁹/L); the respiratory distress was reported as SAE unlikely related to azacytidine, because the patient was suffering from chronic obstructive pulmonary disease, before starting azacytidine.

At diagnosis, neither the level of PI-PLCβ1 promoter methylation nor mRNA expression were significantly correlated to a clinical response at the fourth or the eighth cycle, in terms of CR, hematologic improvement, or stable disease, although PI-PLCβ1 methylation at diagnosis was higher in patients who later showed CR or hematologic improvement, as compared with those who showed stable disease (Fig. 1A).

However, PI-PLCβ1 was predictive of response during treatment, in that the demethylation of PI-PLCβ1 during the first cycles of 5d-AZA was associated with a favorable outcome (Fig. 1B). In fact, as compared with baseline value, 5d-AZA induced a statistically significant decrease in PI-PLCβ1 promoter methylation in responders [Student t test; P < 0.05 vs. baseline; 95% confidence interval (CI), −0.37 to +0.14], whereas in nonresponders we did not observe any significant difference (Student t test; P > 0.05 vs. baseline; 95% CI, −0.21 to +0.18). In addition, 5d-AZA induced a statistically significant increase of PI-PLCβ1 mRNA in responders (Student t test; P < 0.05 vs. baseline; 95% CI, +0.49 to +1.12), but not in nonresponders (Student t test; P > 0.05 vs. baseline; 95% CI, −0.18 to +0.14).

As for the 6 patients who went off-study during treatment, PI-PLCβ1 mRNA increased in 1 case and decreased in 3 cases, whereas the 2 remaining subjects died after the first cycle.

As reported in Fig. 2A, PI-PLCβ1 molecular response preceded the hematologic response in all but one of the responders (14 of 15; 90%) and in only one of the nonresponders (1 of 11; 9%; P < 0.001). In 9 of 14 (64%) responsive patients, the first molecular increase in PI-PLCβ1 level was observed between the third and fourth cycle, therefore anticipating the clinical evaluation. In addition, the 8 cases who showed the loss of the response after the end of therapy (eighth cycle) displayed a significant reduction of PI-PLCβ1 levels, below the pretreatment values, already before the clinical loss of the response (Fig. 2B).

To overcome the technical limitations of metaphase cytogenetics, such as its relatively low resolution and the need for dividing cells, and to reveal all unbalanced gene defects as well as LOH, a whole-genome scanning by SNP arrays (Affymetrix Genome-Wide Human SNP 6.0) was successfully conducted in 26 of 32 patients with (81%) MDS. In 6 of 32 (19%) patients, the analysis failed for lack of sufficient DNA or for bad quality of the extracted DNA. SNP array karyotyping identified genomic abnormalities in all analyzed patients (100%), compared with only 9 of 26 (35%) with metaphase cytogenetics, showing that SNP array analysis is more sensitive in identifying genetic alterations (P = 0.02; Fisher exact test; Supplementary Table S1). CNAs ranged from loss or gain of complete chromosome arms to focal deletions and gains targeting one or few genes (median size for loss was 123 kb, range, 51–76,763 kb; median size for gain was 100 kb, range, 50–146,091 kb). Overall, the median number of abnormalities per sample was 12 (range, 5–57) and gains outnumbered deletions almost 2:1. Macroscopic alterations affecting a complete chromosome or its arms were detected in 6 of 26 (23%) and included trisomy of chromosome 8 in 3 cases (12%), losses of the long arms of the chromosomes 5 and 7 in a single case (4%) and a gain of chromosome 14q, that was not previously identified by metaphase cytogenetics analysis, in another case (4%). Moreover, a patient showed a complex karyotype (defined as more than 3 macroscopic CNAs) including the gains of chromosomes 3q, 9p24-23 and losses of chromosomes 6p and 7q. Importantly, SNP array analysis of 17 patients with normal or unsuccessful metaphase cytogenetics enabled the detection of genetic lesions in all of these cases. Many new abnormalities detected by SNP array analysis were microscopic genomic gains and deletions and copy-neutral LOH (>5 Mbps) affecting the chromosomes 2, 3, 4, 6, 9, 10, and 15 (Supplementary Fig. S2).

Microscopic CNAs were not clusterized in specific genomic regions but were spread through the genome, suggesting a general genomic instability rather than a specific pathogenetic mechanism generating alterations. Noteworthy, in more than 25% of MDS cases we found amplifications of the small nucleolar RNA, C/D box 115-12 gene (SNORD115-12) on chromosome 15q12 involved in the regulation of alternative splicing and of the defender against cell death 1 (DAD1) on chromosome 14q11, a negative regulator of programmed cell death.

To investigate whether the findings from SNPs array may be correlated with an early response to 5d-AZA (hematologic improvement/CR) or stable disease at fourth cycle, we first stratified the patients according to the presence of macroscopic copy number changes or LOH alterations (Supplementary Table S1). Among the 6 patients with MDS (23%) with macroscopic CNAs, 2 (33%) achieved HI/CR, whereas 8 of 20 (40%) without alterations (P = 1.0; Fisher exact test), suggesting that macroscopic CNAs are not associated with response to 5d-AZA. More interestingly, patients with LOH (n = 5; 19%) have a trend for worse response in terms of stable disease (80% vs. 57% of patients without LOH), even if this does not reach a statistically significant result (P = 0.61; Fisher exact test), mainly due to the small number of analyzed patients. Taking advantage of the SNP array analysis, we subsequently stratified patients with MDS based on the number of microscopic CNAs choosing as cutoff of their median number (n = 12). However, we did not find differences in the response (HI/CR) to azacytidine between patients with CNAs ≥ 12 and those with CNAs < 12 (39% vs. 38%, respectively; P = 1.0; Fisher exact test).

In patients with low-risk MDS, azacytidine has been reported to induce transfusion independence and/or hematologic responses in 40% to 70% of cases (25–27). This high variability probably reflects the fact that the low-risk MDS patients' population is clinically and biologically heterogeneous (35–37) and explains why the effects of azacytidine in this setting of patients are less known and why dose and duration of treatment and profile of patients with low-risk MDS who can benefit of azacytidine are still undetermined.

To contribute to answer some of these questions, we conducted a prospective phase II study, to assess the clinical and biologic effects of a 5d-AZA in patients with IPSS low/Int-1–risk MDS who were unresponsive to prior therapies with ESAs or who were not appropriate candidates for ESAs (i.e., serum erythropoietin > 500 mUI/mL) in presence of anemia, with or without other cytopenias. Therefore, the patients selected for our study had an unfavorable prognosis and no alternative approach to supportive care, except hypomethylating agents.

The hematologic response was assessed according to the IWG criteria (28): the ORR was 58% (15 of 26 patients who completed eighth azacytidine cycles) and the rate of RBC transfusion-independence after the eighth cycle was 33%. As observed in previous experiences, mainly conducted in patients with high-risk MDS, the response rate increased with increasing cycles, from 38% after the first 4 cycles to 58% after the eighth cycle but, more interestingly, we observed that 3 (20%) out of 15 responders had maintained hematologic improvement, without any other treatment or supportive therapy after 33, 34, and 37 months from discontinuation of treatment. Probably, these responses are among the longest post-azacytidine therapy reported in the literature, suggesting that at least in a very small proportion of patients with low-risk MDS maintenance therapy with azacytidine may not be necessary to maintain hematologic response.

Our data confirm the results of previous retrospective studies (26, 27) but, about the transfusion independence, they are more favorable than those recently reported by Tobiasson and colleagues, who, in their prospective study, reported a transfusion independence rate of 13%, although they enrolled the same category of patients (low-risk MDS) and used the same schedule (38) As for the other 2 studies using azacytidine in a similar setting of patients, Lyons and colleagues reported hematologic improvement in 44% to 56% of patients with any French–American–British (FAB) subtype MDS, unselected for IPSS, previous resistance to ESAs or severe cytopenia, and treated with 3 different azacytidine schedules (26). Similarly, Musto and colleagues retrospectively reported an ORR of 46% in a population of low-risk MDS treated with different azacytidine schedules over a period of 2 years (27).

According to the treatment plan, we discontinued 5d-AZA after eight cycles. Although a predetermined number of azacytidine cycles is not typical, we have chosen this treatment schedule to draw information on the duration of response after therapy discontinuation. Indeed, while in high-risk MDS the duration of response is generally short, in low-risk MDS it is not really known. Therefore, this information can be relevant for planning new clinical studies and for the optimization of treatment with azacytidine in patients with low-risk MDS. In our study, we documented a median response duration of 10 months (range, 2–37) and in 3 cases the duration of response was particularly long, without any other treatment or supportive therapy. We do not know if prolonging the treatment would have induced much more benefit in terms of achievement and depth of hematologic response. We speculate that this would be less probable in light of the results of PI-PLCβ1 promoter methylation and gene expression during treatment. In fact, although PI-PLCβ1 methylation was unable to clearly distinguish between responders (CR/HI) and nonresponders (stable disease) at diagnosis, all patients who responded to 5d-AZA showed an early decrease of PI-PLCβ1 methylation and an early increase of PI-PLCβ1 mRNA during the therapy, as compared with the pretreatment values, whereas only one of the nonresponders showed the same feature. Furthermore, in 9 of 14 (64%) responders the increase of PI-PLCβ1 mRNA occurred early during treatment (within the fourth cycle of therapy). A demethylation of PI-PLCβ1 promoter, which induces gene expression, has been correlated with response to azacytidine in high-risk MDS (10–12). Here, we showed that PI-PLCβ1 promoter is demethylated also in low-risk cases, where it can be used for monitoring the effect of 5d-AZA during treatment. This contributes to elucidate the mechanisms underlying the effect of azacytidine and confirms the role of PI-PLCβ1 as a key molecule in hematologic malignancies and myeloid differentiation. Indeed, PI-PLCβ1 increase has been specifically associated not only with expression of cyclin D3 (12), but also with the recruitment of MZF-1 transcription factor (39), therefore confirming the hypothesis of a contribution of PI-PLCβ1 in azacytidine-induced myeloid differentiation (18).

Taken together, our results strongly suggest that the early evaluation of PI-PLCβ1 expression could be a reliable dynamic marker of response to azacytidine. Indeed, among the responsive patients, a significant decrease of PI-PLCβ1 expression was observed in those patients that lost the clinical response after the end of treatment, but not in the patients maintaining a durable hematologic response. As for this last point, the long-lasting responses that we observed may shed some light on the potential activity of azacytidine in terms of controlling disease progression and suggest that, at least in a restricted number of patients with low-risk MDS, survival could also be a possible therapeutic goal, besides improvement in transfusion dependence and quality of life.

On the other hand, SNP array analysis, did not allow us to clearly identify those patients who achieved a benefit from 5d-AZA. To identify genetic surrogate markers for response and survival in patients with MDS following azacytidine therapy, we applied SNP array technology. This approach was more sensitive than conventional cytogenetics in identifying genomic abnormalities, including both copy number changes and LOH, in all analyzed patients (100%). However, the type and the number of CNAs did not correlate with the response (HI/CR) to azacytidine. Interestingly, a trend for worse response in terms of stable disease has been found in patients with LOH (80% vs. 57% of patients without LOH). Whether the lesions detected by SNP-array have a prognostic value on clinical outcome is still controversial in MDS (40). In Gondek and colleagues' report with 141 patients with MDS and MDS/MPN, survival duration was not affected by abnormal CNAs (41), by contrast, in Yi and colleagues' study the presence of abnormal SNP lesions correlated with lower CR and PR rate following decitabine therapy and with worse OS even if it was not confirmed in the multivariate analysis (42).

It must be considered that azacytidine, although effective and generally well tolerated, is not without toxicity. Therefore, the early identification of patients who could benefit most from azacytidine would be useful to select those patients for whom the treatment may prove cost-effectiveness. Earlier discontinuation of azacytidine may avoid or reduce the risk of administering an ineffective drug for a prolonged period and to expose patients to unnecessary toxicity in absence of clinical and hematologic benefit. In our study, the most relevant toxicity was hematologic. Azacytidine may worsen preexisting neutropenia and thrombocytopenia and consequently may increase the risk of developing additional complications. It should be underlined that we observed 4 early deaths, which were related to infections in 3 cases and hemorrhage in 1 case. From a clinical point of view, it is very probable that these deaths were related to the underlying hematologic disease or comorbidities and not due to azacytidine.

In conclusion, in this prospective phase II study, we observed that a lower intensity schedule of azacytidine is effective in about 60% of patients with low-risk MDS, refractory to ESAs or with severe cytopenia. Extreme caution must be adopted, particularly in elderly patients with comorbidity. Because of hemato-toxicity associated with the drug, risks of death due to bleeding and infection are not negligible, especially during the initial cycles of therapy. As for the biologic studies, SNP array analysis showed that low-risk MDS is a very heterogeneous group of diseases, but it was unable to identify genetic alterations possibly correlated with hematologic response to azacytidine. On the other hand, our results showed that the evaluation of PI-PLCβ1 gene expression during treatment with azacytidine may be a reliable and early marker of response to azacytidine and can contribute to early identification of responsive patients, even though it does not predict the depth and duration of response. That is why further studies are needed to validate these data on larger series of patients. If confirmed, our results could also be useful to optimize the treatment with azacytidine and to design further investigational studies both in low- and high-risk MDS.

G. Martinelli has honoraria from Speakers Bureau of Novartis and Bristol-Myers Squibb and is a consultant/advisory board member of Roche and Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Malagola, D. Russo

Development of methodology: C. Filì, M.Y. Follo, C. Finelli, F. Cattina, C. Skert, L. Manzoli, L. Cocco, D. Russo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Filì, M. Malagola, M.Y. Follo, C. Finelli, F. Cattina, C. Clissa, A. Candoni, R. Fanin, M. Gobbi, M. Bocchia, M. Defina, P. Spedini, L. Manzoli, L. Cocco

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Filì, M.Y. Follo, C. Finelli, I. Iacobucci, G. Martinelli, C. Skert, L. Manzoli, L. Cocco, D. Russo

Writing, review, and/or revision of the manuscript: C. Filì, M. Malagola, M.Y. Follo, C. Finelli, R. Fanin, M. Bocchia, L. Cocco, D. Russo

Study supervision: M. Gobbi, D. Russo

This work was supported in part by PRIN-Cofin 2007, Italian MIUR-FIRB, Celgene, and Lions Bassa Bresciana Association.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.