Pemetrexed versus erlotinib in pretreated patients with advanced non–small cell lung cancer: A Hellenic Oncology Research Group (HORG) randomized phase 3 study


  • Presented in part as an abstract at the 46th Annual Meeting of the American Society of Clinical Oncology; June 4-8, 2010; Chicago, Illinois.



In this superiority study, pemetrexed was compared with erlotinib in pre-treated patients with metastatic non–small cell lung cancer (NSCLC).


Patients with stage IIIB/IV NSCLC who progressed after first-line or second-line treatment were randomized to receive either pemetrexed or erlotinib. In total, 21.7% of patients in the pemetrexed arm and 23.5% of patients in the erlotinib arm had squamous cell histology, and treatment was third line in 39.2% and 46.4% of patients, respectively. The primary study endpoint was time to tumor progression (TTP). Epidermal growth factor receptor/v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (EGFR/KRAS) mutation status also was investigated.


There was no difference in terms of the TTP (P = .195), the objective response rate (P = .469), or overall survival (P = .986) between the 2 treatment arms. In patients who had squamous cell histology, erlotinib resulted in a superior TTP compared with pemetrexed (4.1 months vs 2.5 months, respectively; P = .006). The incidence of grade 3 and 4 neutropenia, thrombocytopenia, and asthenia was significantly higher in the pemetrexed arm, whereas the incidence of grade 3 and 4 skin rash was higher in the erlotinib arm.


Both pemetrexed and erlotinib had comparable efficacy in pre-treated patients with metastatic NSCLC, and the current results indicated that genotyping of tumor cells may have an important effect on treatment efficacy. Cancer 2013;119:2754–2764. © 2013 American Cancer Society.


Platinum-based chemotherapy represents the standard of care for patients with stage IIIB (with pleural effusion) or stage IV non–small cell lung cancer (NSCLC).[1, 2] For nonresponding or progressing patients, second-line chemotherapy with docetaxel, pemetrexed, or erlotinib offers control of disease-related symptoms, improves quality of life, and prolongs overall survival (OS).[3-5]

Second-line treatment of advanced NSCLC with docetaxel is associated with improved 1-year survival and quality of life compared with ifosfamide, vinorelbine, or best supportive care.[3, 4] Pemetrexed, an antimetabolite that inhibits key folate-dependent enzymes required for de novo nucleotide biosynthesis,[6] has demonstrated clinical efficacy comparable to that of docetaxel, but with a more favorable toxicity profile.[5] There is evidence that the activity of pemetrexed is correlated with adenocarcinoma histology.[7] Erlotinib, an inhibitor of the human epidermal growth factor receptor (EGFR) tyrosine kinase (TKI),[8, 9] reportedly confers a survival benefit over placebo in pretreated patients with advanced NSCLC.[10] The improved outcome with EGFR TKIs is mainly confined to never smokers and patients who have tumors that harbor EGFR mutations,[11-17] whereas the role of v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation status remains unclear.[18-20] Because pemetrexed and erlotinib have different toxicity profiles and routes of administration but both are active in the second-line setting for NSCLC, we decided to directly compare these 2 agents.


Eligibility Criteria

Pemetrexed-naive and TKI-naive patients with documented stage IIIB (with pleural effusion) or stage IV NSCLC who experience disease progression after 1 or 2 chemotherapy lines were enrolled. Patients aged <65 years must have received a platinum-based regimen, which was not mandatory for older patients. After July 2008, an amendment was implemented excluding patients who had squamous cell histology.[7] Other inclusion criteria were: a performance status (PS) from 0 to 2; adequate bone marrow tests (absolute neutrophil count ≥1500/μL, hemoglobin ≥9 g/dL, and platelet count ≥100,000/μL), renal function tests (creatinine ≤2 mg/dL), and liver function tests (total bilirubin ≤1.5 times the institutional upper limit of normal [ULN], aspartate aminotransferase and alanine aminotransferase levels ≤2 times the institutional ULN); a life expectancy ≥3 months; and clinically stable with irradiated brain metastases. Exclusion criteria included: a second primary tumor, active infection, severe heart disease, and uncontrolled diabetes mellitus. All patients provided written informed consent, and the protocol was approved by the Ethics and Scientific Committees of the participating centers. This study is registered as National Clinical Trial NCT00440414 (

Baseline and Follow-Up Assessments

Baseline assessment included a medical history; evaluation of PS; physical examination; 12-lead electrocardiogram; blood tests (complete blood cell count with differential, platelet count, and blood chemistry); computed tomography scans of the chest, abdomen, and brain; and a whole-body radionuclide bone scan. Patients were assessed with physical examination, complete blood cell count, blood chemistry, electrocardiogram, and a chest x-ray before each treatment cycle.

Response was evaluated every 2 months during erlotinib administration, which corresponded to 3 pemetrexed cycles. Objective tumor response was evaluated according to the Response Evaluation Criteria in Solid Tumors[21] and had to be confirmed 4 weeks later. All computed tomography scans were reviewed by an independent radiologist. Toxicity was evaluated before each chemotherapy cycle according to the National Cancer Institute's Common Toxicity Criteria, version 2.[22] All patients were followed until disease progression or death.


Patients were randomly assigned to receive either pemetrexed (Alimta; Eli Lilly, Indianapolis, Ind; 500 mg/m2 over a 1-hour as an intravenous infusion on day 1, every 3 weeks) or erlotinib (Tarceva; F. Hoffmann-La Roche, Basel, Switzerland; 150 mg/day orally). Patients who were allocated to the pemetrexed arm received supplementation with folate and vitamin B12. Patients in the pemetrexed arm were allowed a maximum of 6 chemotherapy cycles, whereas treatment with erlotinib was continued until disease progression, unacceptable toxicity, or withdrawal of the patient's consent. Primary granulocyte-colony–stimulating factor (G-CSF) prophylaxis was not permitted. Crossover between treatment arms was allowed on disease progression at the physician's discretion.

Treatment Modification

Treatment was delayed for 1 week if neutrophil and platelet counts were <1000/μL and <50,000/μL, respectively, and was withheld until the leukocyte counts recovered to ≥4000/μL and platelets recovered to ≥100,000/μL. The occurrence of grade 4 neutropenia and/or thrombocytopenia, febrile neutropenia, or grade 3/4 nonhematologic toxicity, diarrhea, or skin rash was addressed by a dose reduction of 25% for pemetrexed or 100 mg daily for erlotinib. If patients developed grade 2 through 4 neutropenia, subsequent cycles were administered with prophylactic G-CSF support. The study treatment was withheld until nonhematologic toxicity was grade <2. Patients were withdrawn from the study if they required 2 successive dose reductions.

Biomarker Analysis

EGFR and KRAS mutation analysis was performed as previously described.[23] DNA samples were extracted from microdissected cells from formalin-fixed, paraffin-embedded tumor specimens using an Eppendorf piezoelectric microdissector (Eppendorf AG, Hamburg, Germany). The used primers and probe sets (Table 1) were designed according to the Reference Sequences provided on the LocusLink web site (

Table 1. Primers Used for Amplification of Polymerase Chain Reaction Products in Mutation Analysis
 Forward (External)Forward (Internal)Reverse (External)Reverse (Internal)
  1. Abbreviations: EGFR, epidermal growth factor receptor; KRAS, v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog.


Exons 18 through 21, coding for the tyrosine kinase domain of EGFR, and exon 2 of KRAS were amplified by polymerase chain reaction (PCR) and then subjected to bidirectional automatic sequencing. All tests were performed in duplicate if sufficient material was available. PCR primers were designed using human genome reference sequences acquired from the University of California Santa Cruz Genome Browser (National Center for Biotechnology Information accession no. NM_005228 and NM_033360), as previously described (see Table 1).[23] Sequence variants were determined using Seqscape software (Applied Biosystems, Foster City, Calif).

Statistical Analysis

In this randomized, open-label, multicenter, phase 3 study, patients were centrally randomized by computer at a 1:1 ratio and stratified according to PS, disease stage, age (<65 years vs ≥65years), and response to first-line treatment. The primary study endpoint was the time to tumor progression (TTP). Secondary endpoints included progression-free survival (PFS), OS, the objective response rate (ORR), and safety. We assumed that, to detect a 35% increase in the median TTP of 2.2 months for the erlotinib arm[10] to 3 months for the pemetrexed arm with 80% power (α = .05; 2-sided log-rank test) at the statistically significant level of 5%, 163 enrolled patients would be required in each arm.

Analysis was performed on an intent-to-treat basis. TTP was measured from study entry to the day of the first evidence of disease progression, and OS was measured from the date of study entry to the date of either death or last contact. The probability of survival was estimated using Kaplan-Meier methodology, and differences were tested using the log-rank test. Cox regression was used for multivariate and univariate analysis for a unified view of the effects of covariates on TTP and OS. All tests were 2-sided.


Patient Characteristics

Between January 2006 and April 2010, 357 patients from 9 centers were enrolled in the pemetrexed (n = 178) and erlotinib (n = 179) arms. The Consolidated Standards for Reporting Trials (CONSORT) diagram (Fig. 1) for the trial indicates that 12 patients in the pemetrexed arm and 13 patients in the erlotinib arm did not receive the allocated treatment because of early death or withdrawal of consent. In total, 332 patients (n = 166 in each arm) were evaluable for response and toxicity. Five patients were lost to follow-up after the first cycle of treatment and were censored. Patients' clinicopathologic characteristics (Table 2) indicate that 85% had a PS of 0/1, 22.5% had squamous cell carcinoma histology, and 43% had received 2 prior regimens (mainly a taxane/platinum combination).

Figure 1.

This is a Consolidated Standards for Reporting Trials (CONSORT) diagram of the current trial.

Table 2. Baseline Patient Characteristics
 No. of Patients (%) 
 Pemetrexed, n = 166Erlotinib, n = 166P
  1. Abbreviations: Tx, treatment.

Age: Median [range], y66 [42-86]65 [37-83] 
Age, y   
<6569 (41.6)81 (48.8).226
≥6597 (58.4)85 (52.2) 
Men138 (83.1)135 (81.3).67
Women28 (16.9)31 (18.7) 
Performance status   
037 (22.3)44 (26.5).065
198 (59)104 (62.7) 
231 (18.7)18 (10.8) 
Squamous36 (21.7)39 (23.5).694
Non-squamous130 (79.3)127 (76.5) 
IIIB19 (11.4)12 (7.2).187
IV147 (88.6)154 (92.8) 
Treatment line   
Second101 (60.8)89 (53.6).183
Third65 (39.2)77 (46.4) 
Response to previous Tx   
Response36 (21.7)26 (15.7).159
No response130 (78.3)140 (84.3) 
Smoking status   
Active/ex-smokers128 (77.1)124 (74.7).466
Never smokers24 (14.5)29 (17.5) 
Unknown14 (8.4)13 (7.8) 

Treatment Compliance

Forty-eight patients (29%) in the pemetrexed arm completed 6 cycles of treatment, and 71 patients (42.8%) in the erlotinib arm completed 4 months of treatment. Treatment discontinuation in the pemetrexed and the erlotinib arms was attributed to: death (7.2% and 19%, respectively; P = .002), disease progression (53% and 69.6%, respectively; P = .002), withdrawal of consent (3.6% and 0.6%, respectively; P = .065), adverse events (0.6% and 5.7%, respectively; P = .008), or other reasons unrelated to treatment or disease (6.6% and 5.1%, respectively; P = .245). G-CSF support was required for 27% of patients in the pemetrexed arm.

In total, 669 pemetrexed chemotherapy cycles were administered (median, 3 cycles; range, 1-6 cycles), whereas the median duration of treatment with erlotinib was 3.2 months (range, 0.2-48 months). Treatment delay was required in 8.4% of pemetrexed cycles for the following reasons: hematologic or nonhematologic (1.2% and 1.2%, respectively) toxicity, and late admission (6%) for reasons unrelated to disease or treatment. Dose reductions were required in 2.1% of pemetrexed cycles and in 5 patients (3%) in the erlotinib arm.

Response to Treatment

Nineteen patients in the pemetrexed arm (11.4%; 95% confidence interval [CI], 6.6%-16.3%) and 15 patients in the erlotinib arm (9%; 95% CI, 4.7%-13.4%) achieved a partial response (P = .469). In the pemetrexed arm, stable disease was observed in 39 patients (23.5%), and progressive disease was observed in 108 patients (65.1%); whereas, in the erlotinib arm, 29 patients(17.5%) had stable disease, and 122 patients (73.5%) had progressive disease.

After a median follow-up of 27.3 months and 29.0 months for the pemetrexed and erlotinib arms, respectively, disease progression was documented in 303 patients (91%; pemetrexed arm, n = 150; erlotinib arm, n = 153). The median TTP was 3.0 months (range, 0.3-27.3 months) and 3.9 months (range, 0.2-47.8 months) for the pemetrexed and erlotinib arms, respectively (P = .195) (Figs. 2a, 3). Similarly, the median PFS was 2.9 months (range, 0.4-27.3 months) and 3.6 months (range, 0.2-47.8 months) for the pemetrexed and erlotinib arms, respectively (P = .136). At the time of analysis, 245 patients (74%) had died (pemetrexed arm, n = 115; erlotinib arm, n = 130), with a median OS of 10.1 months (range, 0.4-44.6 months) and 8.2 months (range, 0.2-51.4 months), respectively (P = .986) (Fig. 2b). The estimated 1-year survival rate was 43.6% for patients in the pemetrexed arm and 39.5% for patients in the erlotinib arm.

Figure 2.

These Kaplan-Meier curves illustrate (a) the time to tumor progression (TTP) and (b) overall survival (OS) according to the allocated treatment. ERL, indicates erlotinib; PEM, pemetrexed.

Figure 3.

This is a forest plot of the time to tumor progression (TTP). CI indicates, confidence interval; PS, performance status; HR, hazard ratio.

Thirty-two patients (19.2%) were crossed over from pemetrexed to erlotinib, and 7 patients (4.2%) were crossed over from erlotinib to pemetrexed. Overall, 45 patients (27.1%) who received pemetrexed and 52 patients (31.3%) who received erlotinib received subsequent therapy for progressive disease.

Patients with squamous cell carcinoma who received erlotinib had a significantly higher median TTP, but not ORR or OS, compared with patients who received pemetrexed (4.1 months [range, 0.2-33.5 months] vs 2.5 months [range, 0.4-20.4 months], respectively; P = .006). No difference was observed in terms of ORR, median TTP, or median OS in patients who had nonsquamous cell tumors, according to treatment line or smoking status between the treatment arms (data not shown).

Treatment Efficacy According to Biomarker Status


EGFR mutation status was assessed in 123 patients (pemetrexed arm, n = 62; erlotinib arm, n = 61) who had adequate tumor tissue available (see Fig. 4, REMARK diagram). EGFR mutations were detected in 5 and 6 patients enrolled in the pemetrexed and erlotinib arms, respectively. No differences in terms of ORR, TTP, or OS were observed between the 2 treatment arms irrespective of EGFR status. Patients with wild-type EGFR (wt-EGFR) who received pemetrexed experienced a higher disease control rate (DCR) compared with patients who received erlotinib (42.1% vs 21.8%, respectively; P = .022). In the erlotinib arm, patients who had EGFR mutations (mut-EGFR) had a higher ORR compared with the wt-EGFR subgroup (33.3% vs 7.3%; P = .042) and had a trend toward longer OS (23.0 months vs 9.7 months; P = .067) (Table 3), with an estimated 1-year survival rate of 83.3% and 36.5%, respectively.

Figure 4.

This is a Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) diagram of the analyzed tumor specimens. ND indicates not determined; EGFR, epidermal growth factor receptor; KRAS, v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog.

Table 3. Efficacy of Erlotinib According to EGFR/KRAS Status
 Erlotinib-Treated Patients
 EGFR Status, n = 61 KRAS Status, n = 51 
VariableWT-EGFR, n = 55Mut-EGFR, n = 6PWT-KRAS, n = 38Mut-KRAS, n = 13P
  1. Abbreviations: DCR: disease control rate; EGFR, epidermal growth factor receptor; KRAS, v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog; Mut, mutant; ORR: objective response rate; OS, overall survival; TTP, time to tumor progression; WT, wild type.

ORR (95% CI), %7.3 (0.4-14.1)33.3 (0-71.0).04210.5 (0.8-20.3)0.223
DCR (95% CI), %21.8 (10.9-32.7)50 (10.0-90.0).12828.9 (14.5-43.4)7.7 (0-22.2).119
TTP (95% CI), mo2.9 (1.7-4.2)9.6 (0-23.7).1044.2 (3.6-4.6)2.3 (0.7-4.0).181
Range, mo0.4-47.85.3-23.1 0.8-47.80.5-14.7 
OS (95% CI), mo9.7 (6.4-12.9)23.0 (17.1-28.9).06711.9 (9.1-14.7)3.9 (1.3-6.5).001
Range, mo0.6-47.85.9-33.6 1.5-47.80.8-14.7 
1-Year survival, %36.583.3 48.115.4 


KRAS mutation status was assessed in 109 tumor specimens (pemetrexed arm, n = 58; erlotinib arm, n = 51). KRAS mutations (mut-KRAS) were detected in 17 and 13 patients enrolled in the pemetrexed and erlotinib arms, respectively. Patients with mut-KRAS who received pemetrexed experienced a significantly better median OS (P = .009) and a higher DCR (P = .009) than patients who received erlotinib. Within the erlotinib arm, patients with wt-KRAS tumors had a significantly better median OS than the mut-KRAS subgroup (P = .001) (Table 3).

In univariate analysis, a PS of 0/1 (hazard ratio [HR], 1.747; 95% CI, 1.265-2.412; P = .001), stage IIIB disease (HR, 1.737; 95% CI, 1.149-2.625; P = .009), and second-line treatment (HR, 1.620; 95% CI, 1.281-2.049; P = .0001) were correlated with better TTP; whereas PS (HR, 2.399; 95% CI, 1.699-3.387; P = .0001) and disease stage (HR, 1.996; 95% CI, 1.217-3.273; P = .006) also were significant prognostic factors for OS (Table 4). Multivariate analysis revealed that PS, disease stage, and line of treatment were independent factors associated with decreased TTP and OS (Table 5).

Table 4. Univariate Analysis of Time to Tumor Progression and Overall Survival
 Time to Tumor ProgressionOverall Survival
FactorHR (95% CI)PHR (95% CI)P
  1. Abbreviations: CI, confidence interval; HR, hazard ratio; PS, performance status.

Treatment Group: Pemetrexed vs erlotinib1.165 (0.924-1.460).1970.998 (0.775-1.285).986
Age: <65 y vs ≥65 y1.221 (0.970-1.536).0891.105 (0.859-1.421).438
Sex: Men vs women1.225 (0.902-1.663).1941.263 (0.899-1.776).179
PS: 2 vs 0/11.747 (1.265-2.412).0012.399 (1.699-3.387).0001
Histology: Squamous vs non-squamous1.102 (0.836-1.451).4911.167 (0.874-1.558).295
Stage: IV vs IIIB1.737 (1.149-2.625).0091.996 (1.217-3.273).006
Line of therapy: Third line vs second line1.620 (1.281-2.049).00011.267 (0.984-1.631).066
Smoking status: Smokers vs never smokers1.302 (0.949-1.787).1021.403 (0.980-2.009).065
Table 5. Multivariate Analyses of Time to Tumor Progression and Overall Survival
 Time to Tumor ProgressionOverall Survival
  1. Abbreviations: CI, confidence interval; HR, hazard ratio; PS, performance status.

VariableHR (95% CI)PHR (95% CI)P
PS: 2 vs 0-11.579 (1.127-2.213).0082.258 (1.576-3.235).0001
Disease stage: IV vs IIIB1.809 (1.180-2.773).0071.915 (1.163-3.151).011
Line of therapy: >Second line vs second line1.669 (1.302-2.139).00011.290 (0.991-1.680).059


Grade 3 and 4 neutropenia and thrombocytopenia, and grade 2/3 anemia occurred significantly more frequently in the pemetrexed arm (P = .001, P = .013, and P = .0001, respectively) (Table 6). No cases of febrile neutropenia were reported. The incidence of nonhematologic toxicity was low, with grade 3/4 fatigue occurring more frequently in patients who received pemetrexed (P = .002). Grade 2 through 4 skin rash was observed in 16.8% of patients who received erlotinib. There were 2 treatment-related deaths in the pemetrexed arm (1 because of sepsis, without neutropenia, that occurred 10 days after administration of the first chemotherapy cycle and 1 because of central nervous system stroke that occurred 18 days after administration of the sixth chemotherapy cycle). G-CSF was administered, either because of treatment-related neutropenia or for secondary prophylaxis, in 27.1% of patients who received pemetrexed.

Table 6. Treatment Toxicity
 No. of Patients (%)
ToxicityPemetrexed, n = 166Erlotinib, n = 166
Grade 1Grade 2Grade 3Grade 4Grade 1Grade 2Grade 3Grade 4
  1. a

    For grade 3 and 4 neutropenia, P = .001.

  2. b

    For grade 2 and 3 anemia, P = .0001.

  3. c

    For grade 3 and 4 thrombocytopenia, P = .013.

  4. d

    For grade 3 and 4 fatigue, P = .002.

Neutropenia16 (9.6)9 (5.4)9 (5.4)a2 (1.2)a1 (0.6)2 (1.2)
Anemia98 (59)36 (21.7)b2 (1.2)b47 (28.3)11 (6.6)b1 (0.6)b
Thrombocytopenia12 (7.2)5 (3)3 (1.8)c3 (1.8)c6 (3.6)2 (1.2)
Febrile neutropenia1 (0.6)
Nausea4 (2.4)2 (1.2)5 (3)4 (2.4)2 (1.2)
Vomiting3 (1.8)4 (2.4)5 (3)1 (0.6)1 (0.6)
Diarrhea3 (1.8)1(0.6)14 (8.4)2 (1.2)1 (0.6)
Mucositis2 (1.2)2 (1.2)—5 (3)1 (0.6)1 (0.6)
Constipation3 (1.8)1 (0.6)1 (0.6)1 (0.6)
Fatigue31 (18.7)23 (13.9)10 (6)d2 (1.2)d5 (3)11 (6.6)1 (0.6)d
Neurotoxicity3 (1.8)2 (1.2)1 (0.6)1 (0.6)
Skin rash19 (11.4)19 (11.4)8 (4.8)1 (0.6)


The current study failed to meet its primary endpoint, because there was no difference in TTP between pemetrexed and erlotinib when administered to pretreated patients with advanced/metastatic NSCLC. In addition, there was no difference in terms of PFS, ORR, or OS between the 2 treatment arms. Despite the exclusion of patients who had squamous cell histology from enrollment after the Scagliotti et al study,[7] the proportion of patients with this histology (22.5%) remained high, which introduced a bias concerning the efficacy of pemetrexed in histologically unselected NSCLC patients. The required small (3.4 weeks) difference in TTP according to the statistical design, although numerically small, would be meaningful, which was the case for docetaxel[3, 4] and erlotinib[10] in a similar patient population.

The results of the current study are compatible to those from the TITAN study, which compared second-line erlotinib with docetaxel or pemetrexed in patients who progressed during or rapidly after a platinum-based doublet regimen.[24] However, the reported ORR, PFS, and OS were inferior, which probably can be attributed to the poor prognosis of those patients because of platinum resistance. Additional differences concerning patients enrolled on the TITAN study and the current study are: 1) 0% versus 42.7%, respectively, received the allocated treatment as third line; 2) 13% versus 0%, respectively, were of Asian origin; and 3) 36% versus 22.5%, respectively, had squamous cell tumors. In addition, the INTEREST noninferiority trial, which compared second-line gefitinib with docetaxel in both Asian and European pretreated NSCLC patients, met its primary endpoint[25]; whereas the V-15-32[26] and the ISTANA trials,[27] which also compared second-line gefitinib with docetaxel, demonstrated that there was no difference in terms of survival between the treatment arms despite the higher ORR in favor of gefitinib.

In the current study, exploratory subgroup analyses revealed some interesting observations that merit further investigation: 1) Patients with squamous cell histology who received erlotinib derived a significant TTP benefit compared with those who received pemetrexed. Although this observation confirms the limited efficacy of pemetrexed in patients with nonsquamous histology,[7] it is surprising, because EGFR mutations are mainly observed in adenocarcinomas; thus, we cannot exclude a random event. However, it should be mentioned that erlotinib has demonstrated efficacy in unselected patients with NSCLC.[10, 16, 17] 2) There was no statistically significant difference in terms of TTP or OS between patients who received the allocated treatment in the second-line or the third-line setting, suggesting that third-line treatment with these agents may be offered to patients who have a good PS.

The exploratory analysis according to EGFR status revealed that the incidence of tumors with EGFR-activating mutations was comparable to that previously reported in Caucasians,[18, 25, 28] although higher frequencies have been reported.[16, 17, 29] Overall, patients who had mut-EGFR tumors experienced higher ORR and TTP, but not OS, compared with patients who had wt-EGFR tumors, as previously demonstrated.[18, 28] The 2 agents demonstrated comparable efficacy in patients irrespective of their EGFR mutation status (Fig. 5a,b), similar to previous reports comparing TKIs and chemotherapy in pretreated patients.[17, 24, 28] However, a preliminary report from the TAILOR study favored chemotherapy in patients with wt-EGFR tumors.[30] The observed higher ORR and the trend toward improved OS in patients with mut-EGFR tumors in the erlotinib arm support the predictive value of EGFR mutations for response to TKIs[12-15, 17, 19, 28, 29] and their use in the first-line setting.[12-15, 31] However, the strength of these observations is limited by the small number of tumors analyzed for EGFR mutations in our study.

Figure 5.

This forest plot illustrates (a) the time to tumor progression (TTP) and (b) overall survival (OS) according to epidermal growth factor receptor (EGFR) and v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation status. CI indicates, confidence interval; mut, mutated; wt, wild type; HR, hazard ratio.

It has been reported that the presence of KRAS mutation is a possible negative predictive factor for response to EGFR TKIs[16, 18-20]; however, conversely, a positive correlation with pemetrexed efficacy seems to exist.[32] In the current study, no patients with mut-KRAS tumors responded to erlotinib, although there were 2 responders in the pemetrexed arm; in addition, these patients derived a significant benefit in terms of DCR and OS when they received pemetrexed compared with erlotinib. Because the patients with wt-KRAS tumors who received erlotinib had significantly better ORR and OS than the patients with mut-KRAS tumors, it is reasonable to hypothesize that KRAS mutations may be a negative predictive factor for TKI treatment and positive predictive factor for pemetrexed treatment.[32] Nevertheless, these observations should be considered with caution because of the small number of patients in our study who had mut-KRAS tumors.

The toxicity observed in the current study was predictable and manageable, with hematologic toxicity and fatigue mainly observed in the pemetrexed arm and skin toxicity occurring in the erlotinib arm. There were no cases of febrile neutropenia, but prophylactic G-CSF was received by patients who developed grade 2 through 4 neutropenia, because a high proportion of patients were treated in third-line setting.

The recently reported superiority of TKIs over chemotherapy in the first-line setting for patients with mut-EGFR NSCLC eventually will lead to patients requiring second-line treatment upon disease progression. Until new anti-EGFR agents or strategies to overcome resistance to TKIs prove to be effective, the administration of single-agent chemotherapy is a valid option for this setting. In this context, the current study indicates that pemetrexed and erlotinib are equally effective for patients who experience disease progression after first-line chemotherapy. This appears to occur irrespective of EGFR status and may be relevant for daily clinical practice, because this information is often missing. In addition, KRAS status of the tumor may also correlate with the efficacy of TKIs and pemetrexed. Therefore, the data presented here may be relevant for daily clinical practice and decision making.


This work was supported in part by a research grant from the Cretan Association for Biomedical Research (CABR).


The authors made no disclosures.