Phase II clinical study of erlotinib for treatment of myelodysplastic syndromes


  • Conflict of interest: No conflict of interest for any of the authors.

  • Data presented at American Society of Hematology (ASH) 2011 Meeting, abstract # 1714.


Outcome in patients with myelodysplastic syndrome (MDS) after azanucleoside failure is poor with unmet need for active novel agents. Preclinical data have suggested that erlotinib has in vivo and in vitro off epidermal growth factor receptor (EGFR)-target activity in MDS. We conducted a phase II study with single-agent erlotinib 150 mg/day orally in MDS patients following azanucleoside failure. All intermediate-2 or high-risk MDS patients by International Prognostic Scoring System and only those low/intermediate-1 patients with transfusion-dependent anemia or platelet counts <50 × 109/L or a significant clinical hemorrhage requiring platelet transfusion or ANC <1 × 109/L were eligible, with most of our patients being at high risk. In 35 eligible patients, overall best response was 14% (3 patients having marrow complete response and 2 hematological improvement). Four deaths occurred on study (sepsis, intracranial hemorrhage, sudden death, and acute myeloid leukemia (AML)). The most common observed grade 3/4 toxicities according to CTCAE v3 were diarrhea (17.1%), rash (17.1%), and infection (11.6%), accompanied by fatigue, thrombocytopenia, and anorexia at 5.7% each. Median overall survival was 6.8 months (95% CI 4.9–13.2), and leukemia-free survival was 5 months (95% CI 3.4–7.3). Erlotinib was generally well tolerated, with modest single-agent activity. Given these results and preclinical data suggesting synergistic effect with azanucleosides, the combination should be further explored. Am. J. Hematol. 89:809–812, 2014. © 2014 Wiley Periodicals, Inc.


Myelodysplastic syndromes (MDS) are a heterogeneous group of primary bone marrow neoplastic stem cell disorders, characterized clinically by marrow failure with resultant cytopenia and a variable risk of acute myeloid leukemia (AML) transformation and pathologically by the presence of cytologic dysplasia or increased myeloblasts [1]. The International Prognostic Scoring System (IPSS) is currently the most widely used tool for risk stratification [2]. Based on bone marrow myeloblast percentage, karyotype, and number of cytopenias, patients are stratified into four groups: low, intermediate-1 (int-1), intermediate-2 (int-2), and high risk. Int-2 and high-risk groups are referred to as higher risk MDS, where the median overall survival for untreated patients is approximately 1 year with a high rate of AML transformation. Currently, azanucleoside agents (e.g., azacitidine) are considered standard of care for treating higher risk MDS [3]. Outcome after azanucleosides failure is poor. The median overall survival ranges from 4–8 months [4-6]. The response to intensive chemotherapy after azanucleoside failure is less than 20%. There are no approved agents for this group of patients; therefore, there is a need to explore novel agents post-azanucleoside failure.

Erlotinib, an oral small-molecule tyrosine kinase inhibitor that inhibits intracellular epidermal growth factor receptor (EGFR), has been approved by the FDA for treatment of lung cancer and pancreatic cancer. Preclinical data suggest that erlotinib has in vivo and in vitro efficacy in MDS and AML [7], inducing apoptosis in MDS and AML cell lines and primary myeloblasts and promotion of myeloid differentiation. The antitumor effects of erlotinib have also been demonstrated in an AML xenograft mouse model. In addition, anecdotal case reports have documented the hematological activity of erlotinib in patients with lung cancer and concomitant MDS or AML [7-9]. Here, we report the results of a phase II trial examining the activity of erlotinib in MDS patients after azanucleoside failure.



This study was conducted according to the guidelines of the Declaration of Helsinki, with written informed consent obtained from all patients. This trial was registered at NCT00977548 Eligible patients had confirmed MDS or MDS/myeloproliferative neoplasm diagnosis with int-2 or high-risk MDS by IPSS. Patients with low- or int-1 risk IPSS MDS were eligible only if they had symptomatic anemia/transfusion-dependent anemia or had platelet counts <50 × 109/L or a significant clinical hemorrhage requiring platelet transfusion or ANC <1 × 109/L. Patients with refractory anemia with excess blasts in transformation (RAEB-t) by FAB classification (AML 20–30% myeloblasts by WHO) were also eligible. Adequate kidney function (creatinine levels <2× upper normal limit) and adequate liver function (SGOT or SGPT ≤2× institutional upper limits of normal, and pretreatment bilirubin ≤1.5× institutional upper limit of normal) were required. Prior intensive induction chemotherapy, malignancy in the past 2 years, known history of HIV infection, and ECOG performance status 3/4 were the key exclusion criteria. Prior treatments were required to have been discontinued 28 days before day 1 of treatment, except for erythroid-stimulating agents and colony-stimulating factors, which were required to have been stopped 14 days before, and hydroxyurea, which was required to have been stopped two days before.

Study Design

This was a single-institution two-stage phase II clinical study. Erlotinib was given as an oral 150 mg daily dose for 16 weeks. The dose was adjusted for diarrhea, rash, and pulmonary toxicity. Study assessments included baseline and weekly complete blood counts and repeat bone marrow aspirate and biopsy at weeks 8 and 16; nonresponders were removed from study after 16 weeks. Responding patients (at least hematological improvement (HI)) continued study treatment until evidence of disease progression or relapse. The primary endpoint was the overall response rate (ORR), with complete response (CR), partial response (PR), marrow CR (mCR), or HI as defined by the International Working Group (IWG) 2006 criteria [10]. Secondary endpoints included overall survival, progression-free survival, and leukemia-free survival. The study was approved by our scientific review committee and IRB and monitored by an independent internal monitoring committee.

Statistical Analysis

The primary objective of this study was to test whether treatment of MDS with erlotinib is sufficiently effective to warrant further investigation. The primary endpoint for this evaluation was ORR, that is, the probability of achieving CR, PR, mCR, or HI (any cell line). Previous trials of azacitidine in MDS have shown a response rate (CR + PR) of 15%. Therefore, further investigation of erlotinib would not be warranted if it produced an ORR (CR + PR + mCR + HI) of 10% or less and would be warranted if it produced an ORR of 30% or more. Secondary endpoints included duration of response, time to AML progression or death, and overall survival. The study was not powered for those endpoints. Kaplan–Meier estimates were used for analyses of secondary endpoints. All subjects enrolled on study who received any treatment were evaluated for toxicity. Patients considered evaluable for response must have started active therapy and undergone the necessary response assessments at the predetermined milestones or had evidence of disease progression after at least 1 week of active therapy.

Patients were accrued to this study in two steps. In the first step, 20 eligible patients were to be registered. If fewer than 2 of these 20 achieved CR, PR, mCR, or HI, then the study would be terminated due to lack of efficacy. Otherwise, an additional 15 eligible patients were to be registered. If 8 or more of the 35 patients achieved CR, PR, mCR, or HI, then the predetermined statistical endpoint for overall response warranting further study would have been reached. The study had a critical level (probability of erroneously concluding the regimen warrants further study) of 0.02 if the true total response rate was determined as 10% and power (probability of correctly concluding the regimen warrants further study) of 0.87 if the true total response rate was determined as 30%. If the true total response rate was 10%, then the probability of stopping the study after the first step was 0.39. With 35 patients in the study, the probability of any particular toxicity can be estimated to within at most ±17% (95% CI). Any toxicity having a true occurrence rate of 5% or higher is very likely to be observed in at least one patient (probability ≥ 83%).

Targeted Next-Generation Sequencing

Exploratory sequencing was conducted in one responder and a matched nonresponder. Briefly, a 26-gene amplicon-based targeted next-generation sequencing panel consisting of all known exons of genes previously identified to be recurrently mutated in myeloid malignancies was constructed using the Qiagen GeneRead software suite (Supporting Information Table S1). Sequencing was performed on a Mi-Seq personal sequencer using 80–200 ng of DNA isolated from patient bone marrow mononuclear cells. The average read targeted was 500× to allow for detection of rare clones and to accurately estimate variant allele frequencies. Alignment and variant calling were performed using the GeneRead NGS data analysis suite ( To be a true mutation, variants had to result in a frameshift or nonsense mutation. Missense mutations were only included if they had been previously reported in hematologic malignancies and confirmed to be somatic in the literature.


Between September 2009 and January 2011, 39 patients at Moffitt Cancer Center consented to the study (four patients were found to be ineligible after signing informed consent). One patient had CML blast crisis transformation upon review of cytogenetics, which demonstrated Philadelphia chromosome acquisition, but received <1 week erlotinib (this patient was included in efficacy and toxicity analysis). For the 35 patients treated on protocol, the mean age was 73 years: 80% were male and the majority (92%) were Caucasians. By WHO classification, patients were classified as follows: RCMD (n = 2), RAEB-I (n = 8), RAEB-II (n = 9), CMML (n = 6), or AML (RAEB-t by FAB) (n = 8), and MDS/MPN-U (n = 1) (Table 1). The IPSS risk group was low in 2 patients (4.2%), INT-1 in 6 (17.14%), INT-2 in 13 (37.14%), and high risk in 13 (37.14%). The median number of prior treatments was 2. All patients had received prior azanucleoside therapy (azacitidine or decitabine) (Table 1). The median duration of follow-up was 17 months.

Table 1. Baseline characteristics
 Number of patients (%) (N = 35)
Mean age73 years
Male28 (80%)
Female7 (20%)
White32 (92%)
WHO Subtype 
AML (RAEB-t by FAB)8
MDS/myeloproliferative neoplasm unclassifiable1
Low2 (4.2%),
Int-16 (17.14%)
Int-213 (37.14%)
High13 (37.14%)
Prior Therapy 
Azanucleosides35 (100%)

Best responses for all eligible patients on study by IWG 2006 criteria were 3 mCR, 2 HI, and 11 stable disease, for a combined ORR of 5/35 (14%) (Table 2). Four deaths occurred on study (sepsis, intracranial hemorrhage, sudden death, and AML). Nine patients were not evaluable for response (1 patient with CML transformation, 1 patient withdrew consent, and 7 discontinued study due to adverse event or underlying disease complication before first evaluation). The ORR for the 26 evaluable patients was 5/26 (19%) (3 mCR and 2 HI).

Table 2. Best response by IWG 2006 criteria
Best ResponseNumber of patients (%) (N =3 5)
  1. CR, complete response; HI, hematological improvement; mCR, marrow complete response; SD, stable disease; PD, progressive disease; ORR, overall response rate (CR, mCR, HI).

mCR3 (9%)
HI2 (6%)
SD11 (31%)
PD10 (29%)
Not evaluable9 (26%)
All patients5/35 (14%)
Evaluable5/26 (19%)

The most common observed grade 3/4 toxicities according to CTCAE v3 were diarrhea (17.1%), rash (17.1%), and infection (11.6%), with 5.7% having fatigue, thrombocytopenia, and anorexia (Table 3).

Table 3. Any grade 3/4 adverse events observed on the study
ToxicityNumber of patients (%)
Grade 3Grade 4
Rash/acneiform4 (11.4%)0
Pruritus1 (2.9%)0
Rash//desquamation2 (5.7%)0
Fatigue2 (5.7%)0
Pain2 (5.7%)0
Anorexia2 (5.7%)0
Infection4 (11.6%)0
Edema1 (2.9%)0
Neutropenia1 (2.9%)0
Thrombocytopenia02 (5.7%)
Syncope1 (2.9%)0
Renal failure1 (2.9%)0
Dyspnea1 (2.9%)0

The median overall survival was 6.8 months (95% CI 4.9–13.2), the median progression-free survival was 3.6 months (95% CI 2–4.8), and leukemia-free survival was 5 months (95% CI 3.4–7.3) (Figs. 1 and 2). The median overall survival was 16.5 months (95% CI 2.9–30) for responding patients who had HI+, 7.1 months (95% CI 0–12.5) for patients who had stable disease, and 5 months (95% CI 1–9) for patients who had progressive disease or were not evaluable (P = 0.06).

Figure 1.

Kaplan–Meier estimates of overall survival.

Figure 2.

Kaplan–Meier estimates of progression-free and leukemia-free survival.

To preliminarily explore whether this agent impacted the genomic architecture of the disease during therapy, a representative responder and case-matched nonresponder were later sequenced at the time of study entry and first response or disease progression, respectively. Using a 26-gene targeted next-generation sequencing panel, a responding patient with CMML at the time of study entry was found to have mutations in SETBP1 (D868N), U2AF1 (Q84P), and RUNX1 (P263S) at variant allele frequencies (VAF) of 46%, 23%, and 49%, respectively. At the time of first response, the identical mutations of SETBP1, U2AF1, and RUNX1 were identified at VAF of 33%, 19%, and 49%, respectively. In contrast, the case-matched nonresponder was identified to have a mutation in ASLX1 (Y591*), RUNX1 (FS), and U2AF1 (R83H) at a VAF of 49%, 41%, and 10%, respectively. At time of disease progression, the identical mutations were identified at relatively similar VAF in addition to the acquisition of a MLL (FS) mutation at a VAF of 52%, raising the hypothesis that disease progression was associated with genetic evolution while response was associated with genomic stability.


There is a clear unmet need for MDS patients after azanucleoside failure, particularly for int-2 and high-risk IPSS MDS patients. Several groups have reported poor outcome and lack of effective therapies in this group of patients post-azanucleoside failure. Most of the novel therapies tested in this setting have shown modest activity, with 10–30% response rates [11-14].

Preclinically, erlotinib has been demonstrated to induce dose-dependent apoptosis and differentiation in ex vivo MDS and AML cells and in AML cell lines, but not in nonmalignant myeloid progenitor control cells, with the drug effect being dependent on nucleophosmin expression. In the same paper, the authors reported a patient with metastatic non-small cell lung cancer (NSCLC) and high-risk MDS (RAEB-2) who received erlotinib monotherapy and experienced a transient, objective HI (according to the IWG 2006 criteria) in platelets and neutrophils that was maintained throughout the duration of erlotinib treatment and recurred following its discontinuation [7]. Two other separate reports of erlotinib-treated patients with concurrent NSCLC and AML diagnoses documented achievement of CR of the AML [8, 9]. The precise non-EGFR kinase targets responsible for erlotinib activity in myeloid malignancies are not clear, but may involve aberrant Lyn, Syk, and mTOR-mediated signaling.

In our phase II single-institution study, erlotinib demonstrated modest activity as a single agent in MDS patients after azacitidine failure. The treatment was generally well tolerated, with a toxicity profile similar to that observed in NSCLC patients. Responders (with mCR and HI), as expected, had a better median overall survival. Currently, response rates to all available options of therapy including intensive chemotherapy and novel agents are in the range 10–30%, albeit CR can be inconsistently obtained with intensive chemotherapy [5, 15].

Our correlative studies comparing genetic architecture at study entry and time of response were limited by the number of responders and availability of sequential samples. The findings are only hypothesis generating, but the acquisition of a new mutation in the nonresponder would suggest that disease progression was associated with genetic evolution while response was associated with genomic stability.

Recent data have suggested that the azacitidine and erlotinib combination exerts a synergistic effect against MDS/AML cell lines [16]. The combination blocked cell-cycle progression and induced apoptosis more than either of the two agents alone. The synergy involved the proteasomal degradation of the anti-apoptotic Bcl-2 family members MCL-1 and BCL2L10, whose upregulation is reported to mediate azacitidine resistance [17]. The intracellular accumulation of azacitidine was increased by erlotinib. Similarly, synergism was observed in myeloblasts derived from patients.

In summary, our study demonstrated that erlotinib has modest single-agent activity and good tolerability in MDS patients after azanucleoside failure, similar to several other novel agents being investigated. Future development of this agent in MDS will likely depend on further identification of relevant molecular biomarkers to predict response in selected patient populations, as well as studies that combine erlotinib with azanucleosides, given the observed clinical activity in our study and preclinical data suggesting synergistic effect.


We thank Rasa Hamilton (Moffitt Cancer Center) for editorial assistance.

Funding support

The study was sponsored as an investigator-initiated clinical study by Genentech Inc. Author contributions: RSK, EP, AFL, and JEL conceived and/or designed the work that led to the submission. All authors acquired data and/or played an important role in interpreting the results. All authors must have contributed to the following two categories: (a) drafted or revised the manuscript and (b) approved the final version.