Implications of key trials in advanced nonsmall cell lung cancer


  • Philip D. Bonomi MD

    Corresponding author
    1. Division of Hematology/Oncology, Rush University Medical Center, Chicago, Illinois
    • Division of Hematology/Oncology, Rush University Medical Center, 1725 West Harrison Street, Suite 821, Chicago, IL 60612
    Search for more papers by this author
    • Fax: (312) 942-3192


Many different targeted therapies with varying mechanisms of action have been added to standard first-line chemotherapy doublets in an effort to improve survival of patients with advanced nonsmall cell lung cancer (NSCLC). Only 2 targeted therapies—bevacizumab and cetuximab—have been associated with superior survival in phase 3 first-line studies. For both agents, the decision to enter phase 3 was based on results from randomized phase 2 trials, unlike other targeted therapies where the decision was made using either phase 1 or single study arm phase 2 results. There is also mounting evidence that patient selection will play a key role in the successful development of any targeted agent. Bevacizumab is indicated for patients with nonsquamous NSCLC who do not have certain comorbidities. Use of cetuximab is not restricted by safety factors, but may be focused on patients whose tumors are epidermal growth factor receptor (EGFR)-dependent; whether EGFR expression or absence of KRAS mutations are appropriate markers is still under study. By including randomized phase 2 trials in the development pathway, and by improving patient selection for individual agents (enriching trials with patients most likely to respond), it may be possible to enhance the success rate of future phase 3 clinical trials and, in turn, define future clinical practice with improved patient outcomes. Cancer 2010. © 2010 American Cancer Society.

Lung cancer is the second most common malignancy in both men and women in the United States, and the leading cause of cancer death. In 2008, an estimated 215,000 people will be diagnosed with lung cancer, representing 15% of all cancer diagnoses, and 161,840 people will die from it, accounting for 28.6% of all cancer deaths.1 Approximately 85% of the cases are classified as nonsmall cell lung cancer (NSCLC), which comprises a heterogeneous group of histologies, including squamous cell carcinoma, adenocarcinoma, and large cell carcinoma.2 The high cancer death rate reflects the large percentage of cases diagnosed at an advanced stage as well as the plateau in effect achieved with cytotoxic chemotherapy in this setting. Historically, platinum doublets—with taxanes, antimetabolites, or vinca agents—have exhibited the best activity in advanced NSCLC, with objective response rates of 20% to 30% and median survival of 8 to 10 months.3, 4 Although doublet chemotherapy significantly improves response rates and 1-year survival compared with single-agent therapy, a meta-analysis showed that adding a third cytotoxic agent produces only a small increase in response rate and no improvement in survival compared with the doublet.5

Recent efforts to improve outcome in NSCLC have focused on numerous targeted therapeutic approaches. Unfortunately, nearly all phase 3 studies in which a targeted therapy was added to first-line chemotherapy have had only limited success6-18 (Table 1). Many of these agents—including the matrix metalloproteinase inhibitor prinomastat, the epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors erlotinib and gefitinib, the retinoid X receptor modulator bexarotene, and the protein kinase C-alpha antisense oligonucleotide aprinocarsen—were evaluated in 2 phase 3 studies, 1 in combination with paclitaxel/carboplatin and the other in combination with gemcitabine/cisplatin. Adding these agents failed to improve overall survival (OS) or measures of disease progression compared with chemotherapy alone. In many cases, toxicity was increased. Moreover, recent press releases from study sponsors indicate that phase 3 trials with the multikinase inhibitors sorafenib and cediranib were stopped early, because interim analyses showed no survival benefit or increased toxicity compared with chemotherapy alone.19, 20 Taken together, this largely negative body of evidence may provide lessons for improving the design of future phase 3 studies. Moreover, it underscores the potential importance of phase 3 studies that show clinical benefit in the advanced NSCLC setting.

Table 1. Unsuccessful Phase 3 Studies of Adding Targeted Therapy to First-Line Chemotherapy of Advanced NSCLC
AgentMechanismPatientsChemotherapyExperimental ArmResults
  1. NSCLC indicates nonsmall cell lung cancer; GC, gemcitabine plus cisplatin; PC, paclitaxel plus carboplatin; OS, overall survival; EGFR, epidermal growth factor receptor; PFS, progression-free survival; PKCα, protein kinase C-alpha; RXR, retinoid X receptor; TLR9, toll-like receptor 9; TTP, time to progression; VC, vinorelbine plus cisplatin.

Prinomastat6, 7MMP inhibitorStudy 1: N=362GC (study 1) or PC (study 2) q3w up to 6 cyclesGC or PC + prinomastat (5-15 mg bid po) until disease progressionNo improvement in OS or PFS; increased musculoskeletal toxicity
Study 2: N=686
BMS-2752918MMP inhibitorN=774PC q3w up to 8 cyclesPC + BMS-275291 (1200 mg/day po) until disease progressionNo improvement in OS or PFS; increased toxicity
PF-35126769, 10TLR9 agonistStudy 1: N=839GC (study 1) or PC (study 2) q3w up to 6 cyclesGC or PC + PF-3512676 (0.2 mg/kg, sc on d 8 and 15 during CT, then weekly)No improvement in OS or PFS; increased toxicity
Study 2: N=828
Gefitinib11, 12EGFR TKIStudy 1: N=1093GC (study 1) or PC (study 2) q3w up to 6 cyclesGC or PC + gefitinib (250 or 500 mg/day po) until disease progressionNo improvement in OS or TTP
Study 2: N=1037
Erlotinib13, 14EGFR TKIStudy 1: N=1172GC (study 1) or PC (study 2) q3w up to 6 cyclesGC or PC + erlotinib (150 mg/day po) until disease progressionNo improvement in OS or TTP; increased rash and diarrhea
Study 2: N=1059
Bexarotene15, 16RXR modulatorStudy 1: N=623VC q4w (study 1) or PC q3w (study 2) for ≥4 cyclesVC or PC + bexarotene (400 mg/day po + lipid-lowering therapy) until disease progressionNo improvement in OS or TTP; increased toxicity including hyperlipidemia; median OS improved by 2-3 months in retrospective cohort with grade 3/4 hypertriglyceridemia
Study 2: N=612
Aprinocarsen17, 18PKCα inhibitorStudy 1: N=616PC or GC q3w up to 6 cyclesPC or GC+aprinocarsen (2 mg/kg on d1-14 by continuous infusion starting 1 or 3 d before chemotherapy)No improvement in OS or PFS; increased toxicity
Study 2: N=670

Challenges in Clinical Development of Drugs for NSCLC

The decision of whether to initiate phase 3 clinical testing is a critical developmental milestone. Several key factors are coming to light at this go-no-go decision point for targeted therapies in NSCLC. The first issue is the heterogeneity of the disorder. NSCLC encompasses a number of histologies: squamous cell carcinoma, adenocarcinoma, large cell carcinoma, and several less prevalent entities.21 Moreover, within each histologic subtype, several variants have been identified. The molecular events leading to the development of each histological form are not necessarily the same; consequently, certain therapies may influence response and outcome differently depending on NSCLC histology. For example, pemetrexed/cisplatin was as effective as gemcitabine/cisplatin in first-line treatment of advanced NSCLC in a phase 3 trial involving 1725 patients.22 Median survival with each regimen was 10.3 months. However, survival differed depending on histologic type: pemetrexed/cisplatin produced significantly longer survival than gemcitabine/cisplatin in patients with adenocarcinoma (12.6 vs 10.9 months, P = .03) and large-cell carcinoma (10.4 vs 6.7 months, P = .03), whereas gemcitabine/cisplatin was favored in patients with squamous cell histology (9.4 vs 10.8 months, P = .05).

Most cases of NSCLC occur in current or former smokers, although approximately 10% of patients are never smokers whose disease may have a unique pathogenic and molecular etiology.23 Somatic activating mutations of the EGFR tyrosine kinase domain—found in a small minority of NSCLC patients—have been detected more frequently in never smokers than in those with a smoking history.24, 25 These activating mutations confer sensitivity to the EGFR tyrosine kinase inhibitors gefitinib and erlotinib. In subset analyses from the phase 3 trials, response rates and survival with erlotinib were greater in the never smokers than in past or current smokers.14, 26, 27 Other demographic factors, such as sex and ethnicity, may influence response rates, but have not been shown to affect survival.

Second, traditional endpoints may not always be suitable to assess how biologic activity translates into clinical efficacy for targeted therapies. The objective response rate with the anti-EGFR monoclonal antibody cetuximab was only 4.5% when it was used as a single agent in a phase 2 study of 66 patients with recurrent NSCLC.28 In the subset with EGFR-positive tumors, the response rate was 5%. It is more important to determine whether a targeted therapy enhances response rate and progression-free survival (PFS) to chemotherapy. In this regard, adding cetuximab to cisplatin/vinorelbine enhanced response rates (35% vs 28%) and prolonged PFS (5.0 vs 4.6 months) compared with chemotherapy alone in a randomized phase 2 study of 86 patients with previously untreated advanced NSCLC.29 Comparable results were obtained in another randomized phase 2 study when cetuximab was added to gemcitabine/platinum.30

The reliability of surrogate clinical endpoints for efficacy is another important issue to consider. Unlike advanced colorectal or breast cancer, where survival times surpass 2 years, overall survival in NSCLC once the disease enters advanced stage hovers around the 10-to-12-month benchmark. Given this relatively short survival period, the limited effect of second- or third-line therapies and the aggressive nature of the disease, OS is still considered the gold standard to assess clinical efficacy in clinical trials in first-line NSCLC. However, to optimize clinical trial designs, it is valuable to consider the relative robustness of surrogates. Buyse and colleagues evaluated whether PFS could be used as a surrogate for OS in clinical trials of advanced NSCLC.31 They conducted a pooled analysis using data from 2838 patients in 7 randomized clinical trials that compared docetaxel-based therapy versus vinca alkaloid-based therapy in the first-line NSCLC setting. Median PFS was 5.5 months and median OS was 10.0 months for the entire cohort. PFS was correlated with OS (r2 = 0.68), and the effect of treatment on PFS was correlated with the effect of treatment on OS (r2 = 0.83). The surrogate threshold effect was a PFS hazard ratio of 0.70, which indicated that a 30% risk reduction in terms of PFS predicts a significant effect of treatment on OS. Therefore, if a drug reduces the PFS hazard by at least 30%, it warrants further investigation in a phase 3 trial with the expectation that it has a relatively good chance of providing a survival benefit.

Third, the transition from phase 2 to phase 3 needs to be optimized. In each of the negative clinical studies listed in Table 1, the decision to conduct the phase 3 trial was based on results from either phase 1 or single study arm phase 2 studies. For example, the decision to evaluate gefitinib in combination with paclitaxel/carboplatin was based on a small phase 1 study of 24 patients with chemotherapy-naïve advanced NSCLC (as well as studies of single-agent gefitinib after failure of first-line chemotherapy).32 Conversely, positive phase 3 trial results have been obtained for 2 targeted therapies in first-line treatment of advanced NSCLC. These studies will be discussed in the next section, but in both cases, the phase 3 trials were preceded by a randomized phase 2 trial.

A unique randomized discontinuation phase 2 design may be helpful in that it may enrich for patients with slowly growing disease who may be most likely to benefit from targeted therapy. In the E2501 study, 342 patients with advanced NSCLC who had failed at least 2 prior chemotherapy regimens were treated with sorafenib for 2 months.33 Patients who responded to sorafenib continued treatment, whereas those who had progressive disease were withdrawn. Only patients with stable disease were randomly assigned to continue sorafenib or switched to placebo and then allowed to return to sorafenib on progression. Preliminary results indicated that patients continuing on sorafenib had higher rates of stable disease 2 months after randomization (29% vs 5%, P = .002) and longer PFS (3.6 vs 1.9 months, P = .01). This randomized discontinuation design may offer another approach for determining whether a targeted therapy is effective before embarking on a large phase 3 study. In the case of sorafenib, however, interim results from a phase 3 study indicated that this agent does not enhance survival when added to first-line paclitaxel/carboplatin chemotherapy.19

Positive Phase 3 Trials: E4599 and FLEX

Despite previous failures, the treatment of advanced NSCLC has progressed in recent years with the incorporation of the antiangiogenic agent bevacizumab and the anti-EGFR agent cetuximab.34-37


E4599 evaluated whether adding bevacizumab—a monoclonal antibody against vascular endothelial growth factor—to paclitaxel/carboplatin would improve OS compared with chemotherapy alone in 878 patients with advanced NSCLC.34 Chemotherapy was administered every 3 weeks for 6 cycles, with bevacizumab 15 mg/kg given intravenously on Day 1 every 3 weeks until disease progression. Based on phase 2 trials that had shown unacceptable risks associated with bevacizumab in certain patient populations, patients with squamous cell tumors, brain metastases, clinically significant hemoptysis, or poor performance status (PS) were excluded. Adding bevacizumab significantly improved median OS (12.3 vs 10.3 months; hazard ratio [HR] = 0.79; 95% CI, 0.67-0.92; P = .003) and PFS (6.2 vs 4.5 months; HR = 0.66; 95% CI, 0.57-0.77; P < .001) compared with chemotherapy alone (Table 2). Objective response rates also increased significantly (35% vs 15%, P < .001 ). The effects of adding bevacizumab were consistent across subgroups defined by disease stage (IIIB vs IV vs recurrent), weight loss (<5% vs ≥5%), and prior radiation therapy. However, the survival benefit of bevacizumab was seen in males but not in females and appeared greater in patients <65 years of age than in older patients.34

Table 2. Benefit of Adding Bevacizumab or Cetuximab to First-Line Chemotherapy in Advanced NSCLC
StudyChemotherapyParameterChemotherapy AloneChemotherapy+ Targeted TherapyHazard Ratio (95% CI)P
  • NSCLC indicates nonsmall cell lung cancer; CI, confidence interval; OS, overall survival; PFS, progression-free survival.

  • a

    Primary endpoint of study.

E459934Paclitaxel-carboplatin±bevacizumab (15 mg/kg)Median OSa10.3 mo12.3 mo0.79 (0.67-0.92)0.003
Median PFS4.5 mo6.2 mo0.66 (0.57-0.77)<0.001
Response rate15%35%<0.001
AVAiL35Gemcitabine-cisplatin±bevacizumab (7.5 or 15 mg/kg)Median PFSa6.1 mo   
 7.5-mg/kg 6.7 mo0.75 (0.62-0.91)0.0026
 15 mg/kg 6.5 mo0.82 (0.68-0.98)0.0031
Response rate20%   
 7.5-mg/kg 34% <0.0001
 15 mg/kg 30% 0.0017
FLEX36Vinorelbine-cisplatin±cetuximabMedian OSa10.1 mo11.3 mo0.871 (0.762-0.996)0.044
Median OS-Caucasians9.1 mo10.5 mo0.803 (0.694-0.928)0.003
Median PFS4.8 mo4.8 mo0.943 (0.825-1.077)NS
Response rate29%36% 0.015
BMS09937Taxane/carboplatin±cetuximabMedian PFS-independent radiology review committeea4.24 mo4.40 mo0.902 (0.761-1.069)0.2358
Median PFS-Investigator3.78 mo4.30 mo0.766 (0.649-0.903)0.0015
Response rate17.20%25.70% 0.0066

The benefit of adding bevacizumab was also reported in a second phase 3 trial, though the results do not totally replicate the success of E4599. In AVAiL, 1043 patients with advanced NSCLC were randomly allocated to receive bevacizumab (either 7.5 or 15 mg/kg) plus gemcitabine/cisplatin or gemcitabine/cisplatin alone.35 The eligibility criteria were similar to E4599; thus, these 2 trials have defined the population of patients that are candidates for bevacizumab therapy. Unlike E4599, where OS was the primary endpoint, AVAiL defined PFS as the primary endpoint. Compared with the chemotherapy group (median PFS, 6.1 months), adding bevacizumab 7.5 mg/kg (6.7 months; HR = 0.75; 95% CI, 0.62-0.90; P = .0026) or 15 mg/kg (6.5 months; HR = 0.82; 95% CI, 0.68-0.98; P = .0301) significantly prolonged PFS, but the magnitude of the effect was smaller than that seen in E4599. Response rates were also increased significantly by both doses of bevacizumab (34% and 30% vs 20%, P < .0001 and P = .0017, respectively). OS was not significantly different between bevacizumab plus gemcitabine/cisplatin versus chemotherapy alone (HR = 0.92 for 7.5 mg/kg and 1.02 for 15 mg/kg).38

It is unclear why the results seen with AVAiL are not as robust as those reported in E4599. One obvious difference between the 2 studies is the chemotherapy regimen. Although antiangiogenic therapy may be expected to potentiate both regimens similarly, that may not necessarily be the case. Other potential factors that could contribute to differing results between studies, such as patient demographics, disease stage, and tumor histology, did not differ appreciably between the 2 trials. However, the duration of treatment in the bevacizumab study arms was longer in E4599 than in AVAiL. Overall, patients in the chemotherapy study arms of both trials received a median of 5 cycles. In E4599, patients in the bevacizumab study arm received a median of 7 cycles, whereas patients in the low-dose and high-dose bevacizumab study arms of AVAiL received a median of 6 and 5 cycles, respectively.34, 35 However, this difference in treatment duration may be a secondary manifestation of the difference in PFS rather than a primary factor contributing to the magnitude of the effect on PFS.


FLEX evaluated the benefit of adding cetuximab to cisplatin/vinorelbine in 1125 patients with advanced NSCLC who were selected on the basis of having EGFR-expressing tumors detected by immunohistochemistry, which was defined as staining in at least 1 cell (85% of the screened patients were deemed eligible).36 Chemotherapy was administered every 3 weeks for up to 6 cycles; cetuximab was administered at an initial dose of 400 mg/m2 and then weekly at 250 mg/m2 until disease progression. The randomization was stratified by Eastern Cooperative Oncology Group (ECOG) performance status (0-1 vs 2) and disease stage (wet IIIB vs IV). Unlike the phase 3 trials involving bevacizumab, patients with squamous cell carcinoma, brain metastases, and PS 2 were trial eligible. The study cohort had a median age of 59 years (range, 18 to 83 years); most were male (70%), and had good PS (83%) and stage IV disease (94%). One third of the patients had squamous cell carcinomas. Patients in both groups received a median of 4 cycles of chemotherapy, with cetuximab administered for a median of 18 weeks.36

Adding cetuximab to cisplatin/vinorelbine significantly improved median OS—the primary study endpoint—compared with chemotherapy alone (11.3 vs 10.1 months; HR = 0.871; 95% CI, 0.762-0.996; P = .044) (Table 2). This survival benefit was seen across all prespecified subgroups based on PS, smoking status, histology, sex, age, and tumor stage. Cetuximab also significantly increased the response rate (36% vs 29%, P = .012), but did not influence PFS as assessed by the investigator (4.8 months in both study arms). Asian patients accounted for 10.8% of the study cohort,and had a significantly better prognosis than Caucasian patients (median OS, 19.5 vs 9.6 months). Adding cetuximab did not improve survival in the Asian subgroup, although this conclusion must be tempered by the small sample size, as well as differences between treatment arms in this subgroup. However, adding cetuximab improved survival in the large cohort of 946 Caucasian patients compared with chemotherapy alone (10.5 vs 9.1 months; HR = 0.803; 95% CI, 0.694-0.928; P = .003).36

Despite the positive results in FLEX, the results from another phase 3 trial (BMS099) with cetuximab were inconclusive.37 Patients were randomly assigned to receive cetuximab in combination with a taxane (paclitaxel or docetaxel, at the discretion of the investigator) and carboplatin, or chemotherapy alone. The randomization was stratified by taxane, ECOG PS (0 vs 1), and study site. The primary endpoint was PFS as determined by an independent radiology review committee. Eligibility criteria were similar to those used in FLEX, except that EGFR testing was not required. The study cohort had a median age of 64 years (range, 37 to 87 years)—somewhat older than the cohort in FLEX—and included a smaller percentage of male patients (59% vs 70% in FLEX), never smokers (8% vs 22%), squamous cell carcinomas (19.5% vs 34%), and PS2 patients (1% vs 17%).

The BMS099 study showed that adding cetuximab increased the objective response rate (25.7% vs 17.2%, P = .0066), but it found no difference between treatments in median PFS as determined by the independent radiology review committee (4.40 vs 4.24 months; HR = 0.902; 95% CI, 0.761-1.069; P = .236). However, PFS as assessed by the investigators showed that adding cetuximab significantly improved PFS compared with chemotherapy alone (4.30 vs 3.78 months; HR = 0.766; 95% CI, 0.649-0.903; P = .0015). Notably, the independent radiology review committee and investigators agreed on progression in only 70% of the cases.37

It is tempting to speculate that the investigator assessment of progression may have been more accurate or more clinically meaningful to survival than that made by the independent radiology review committee. On examining their patients, investigators may conclude that disease has progressed if PS declines by 2 levels or if pain requiring radiation develops—both of which, in my opinion, should be included in the definition of progression. In comparison, the independent radiology review committee discerns progression based solely on radiology findings. In addition, censoring patients who died without evidence of anatomic progression versus classifying them as progressors could influence the relationship between PFS and OS. After longer follow-up, the OS analysis revealed a numeric advantage with cetuximab consistent with the difference seen in the FLEX study, but that did not reach statistical significance (median OS, 9.69 vs 8.38 months; HR = 0.89, 95% CI; 0.75-1.05; P = .17).39

Integrating Trial Results Into Current Clinical Practice and Future Clinical Development

In light of the body of evidence (both negative and positive) across all these trials in advanced NSCLC, the question remains: what have we learned about the integration of these agents in practice? And how can we improve future clinical development strategies?

Current Patient Selection

Patient selection is emerging as the paramount factor in the successful use of targeted therapies in NSCLC and beyond. Bevacizumab is currently indicated for use in combination with paclitaxel/carboplatin for first-line treatment of patients with unresectable advanced, recurrent, or metastatic nonsquamous NSCLC based on results from E4599.40 Because of safety concerns, patients with predominant squamous cell histology, brain metastases, gross hemoptysis, or unstable angina and those receiving therapeutic anticoagulation were excluded from E4599 and are not usually treated with bevacizumab40 (Table 3). However, preliminary evidence from an open-label study (SAiL), an observational cohort study (ARIES), and other ongoing studies (ATLAS, PASSPORT) suggest that bevacizumab may be suitable for use in patients with treated, asymptomatic brain metastases without causing undue risk of central nervous system hemorrhage.41-43 Moreover, these preliminary data suggest that bevacizumab may be safe in patients receiving anticoagulation therapy and does not impair management of hypertension in patients receiving concomitant antihypertensive medication.42, 44

Table 3. Patient Selection Criteria
  • EGFR indicates epidermal growth factor receptor.

  • a

    Preliminary results from ongoing studies and patient registries suggest that bevacizumab may be safe in these patients.40-43

 Squamous cellNoYes
Brain metastasesNo (?)aYes
 Full-dose anticoagulationNoYes
 Low-dose anticoagulationNo (?)aYes
Molecular markers  
 KRAS mutation?No
 High EGFR copy number?Yes

On the basis of FLEX, it is tempting to speculate that cetuximab will be used in patients with EGFR-expressing tumors, without the risk-related restrictions applied to bevacizumab. Thus, the use of cetuximab would not be limited by NSCLC histology or the presence of brain metastases (Table 3). The immediate beneficiaries, therefore, could be those who are not considered eligible for bevacizumab therapy, although other selection criteria (see below) could optimize the identification of patients for cetuximab therapy.

Integrating Both Bevacizumab and Cetuximab

The positive phase 3 results with bevacizumab and cetuximab have prompted several ongoing phase 2 trials to explore the feasibility of administering both targeted therapies in combination with chemotherapy. The Southwest Oncology Group (SWOG) tested paclitaxel 200 mg/m2, carboplatin AUC = 6, and bevacizumab 15 mg/kg q3wk in combination with the standard dosing schedule of cetuximab (400 mg/m2 on Day 1, 250 mg/m2 weekly thereafter) in 104 patients in a single study arm phase 2 trial. Bevacizumab 15 mg/kg and cetuximab 250 mg/m2 were continued until progression.45 The response rate was 48%, and toxicity was considered acceptable. Based on the results of their single study arm phase 2 trial,45 SWOG is planning a phase 3 study comparing paclitaxel/carboplatin/bevacizumab versus these 3 agents plus cetuximab.

Paclitavel-carboplatin plus the antibody doublet are also being tested in an open-label randomized trial, patients with nonsquamous advanced NSCLC and good PS are being allocated to receive paclitaxel/carboplatin every 3 weeks for either 3 or 6 cycles. Both treatment arms receive a standard schedule of cetuximab until disease progression, plus bevacizumab 15 mg/kg on Day 8 of each cycle for 6 cycles. PFS is being evaluated as the primary endpoint, with OS, response rate, and safety as secondary endpoints. Planned accrual is 120 patients, with study completion expected in March 2009.46

Improving Patient Selection for Future Trials

As our understanding of the heterogeneity of this disease grows, the positive results of trials that enrolled “all-comers” no longer serve as terminal points of clinical development, but rather as initial steps toward optimal clinical use of these agents. The key to future development of targeted therapies is likely to be patient selection, with the goal of enriching the study cohort for patients most likely to respond to the experimental agent. This objective may be approached by using clinical trial or patient registry data to compare outcomes for various subgroups stratified by demographics, tumor characteristics, or molecular markers. Consider tumor histology as an example. As noted previously, advanced NSCLC patients with adenocarcinomas or large-cell carcinomas had longer median OS with pemetrexed/cisplatin than gemcitabine/cisplatin, whereas the opposite was true for patients with squamous cell carcinomas.22 In this case, tumor histology can be used to guide selection between these chemotherapy regimens. Moreover, evidence that patients with certain tumor histology may be prone to serious adverse events can lead to excluding that subgroup, as in the case of increased risk of hemorrhagic events with bevacizumab in patients with squamous cell carcinomas.47 Therefore, patients with squamous cell histology were excluded from further clinical evaluation. However, once a survival advantage is established, new clinical studies or observational cohorts can be used to reassess whether certain risk factors should continue to be used to exclude patients from treatment. This approach, as noted above, is currently being used for bevacizumab.41-44

By identifying those most likely to respond to targeted therapy, molecular markers may offer the greatest promise for improving success rate in the clinical development of new targeted agents. Consider the case of anti-EGFR agents. Increased EGFR copy number measured by fluorescent in situ hybridization (FISH) is found in 40% to 60% of advanced NSCLC tumors.26, 48-50 Patients with FISH-positive tumors achieved significantly longer PFS (6 vs 3 months, P = .0011) and OS (15 vs 7 months, P = .046) when treated with cetuximab plus chemotherapy than FISH-negative patients.49 However, EGFR gene copy number predicted longer time to progression but not improved survival when erlotinib was added to paclitaxel/carboplatin in a subset of patients in TRIBUTE.48 After failure of first-line chemotherapy, high EGFR copy number as well as EGFR expression by immunohistochemistry were associated with improved survival with single-agent erlotinib compared with placebo in the BR.21 trial.26, 50 In contrast to the setting of salvage therapy with an EGFR TKI versus placebo, high EGFR copy number did not have predictive value for survival in patients treated with second-line gefitinib versus docetaxel.51 Additional data are needed to determine the predictive value of EGFR gene copy number in patients treated with EGFR inhibitors.

EGFR mutations are found in 13%-22% of NSCLC patients and have been associated with improved response rate but not improved survival to erlotinib and gefitinib.26, 50, 52 Results from the IPASS trial have shown that EGFR mutations have significant predictive value in Asians with previously untreated stage IV NSCLC.53 Approximately 1200 patients were randomized to gefitinib or to paclitaxel-carboplatin. Eligibility requirements included a history of never smoking or remote light smoking and adenocarcinoma histology. Despite the finding that these clinical featues along with Asian ethnicity are associated with a higher response rate to EGFR tyrosine Kinase inhibitors, chemotherapy was associated with significantly longer PFS in patients with wild type EGFR (HR = 2, 85; P < 0:001). In contrast, superior progression free survival was observed with gefitinib in patients who had EGFR mutations. The results of the IPASS trial appear to validate the value of EGFR mutations in predicting PFS in the NSCLC first-line treatment setting53 Analysis of OS is ongoing.

KRAS mutations are found in 15%-20% of patients with advanced NSCLC.50, 52 Patients with KRAS mutations showed lower response rates and shorter survival than those with wild type KRAS in response to erlotinib plus chemotherapy in the TRIBUTE trial and single-agent erlotinib in the BR.21 trial.50, 52 Similarly, data from a patient registry of 5 clinical trials performed in the United States and Europe indicated that KRAS mutations confer resistance to EGFR TKI therapy.54 Although these observations suggest that EGFR TKIs have less effect in tumors with KRAS mutations, there is no significant difference in survival between gefitinib and docetaxel when used for second-line treatment of NSCLC in patients with KRAS mutated tumors.51KRAS mutations are associated with resistance to cetuximab and poorer survival in advanced colorectal patients;55, 56 however, preliminary analyses based on the limited amount of samples collected in BMS099 indicated that that may not be the case in NSCLC.57 Consequently, the predictive value of KRAS mutations with respect to survival and anti-EGFR antibodies cannot be elucidated.

For new targeted therapies, the mechanism of action should first be defined as clearly as possible in preclinical models, and then their single-agent activity should be tested in clinical trials, with blood and tissue specimens stored for future molecular profiling. As new and potentially relevant markers are identified, the stored specimens can be tested to help refine patient selection criteria for future clinical trials in not only the advanced disease setting but also subsequently locally advanced disease and finally in the adjuvant setting. The development of the anti-HER2 monoclonal antibody trastuzumab for breast cancer illustrates this point. Patients with metastatic breast cancer who had high HER2 expression were found to respond best to this agent, and, consequently, high HER2 expression was carried forward as an eligibility criterion in clinical trials conducted in locally advanced disease and in adjuvant therapy.58

Other Issues

The dose and administration schedule need to be optimized to ensure that drug is delivered to the site of action and to produce drug concentrations above those necessary to inhibit the molecular target without causing unacceptable toxicity. These factors should be addressed early in the clinical development program, in phase 1 studies or with preoperative and postoperative molecular pharmacodynamic profiles in early stage patients treated with novel agents before surgery. Ideally, this information should be available before starting randomized phase 2 and phase 3 trials.

The use of maintenance therapy in patients whose disease is controlled on first-line chemotherapy is another consideration. In each of these randomized trials, patients with disease control on a platinum doublet were assigned to immediate treatment with a different single agent on to secondary treatment at the time of disease progression.59, 60 There is considerable controversy regarding routine use of maintenance therapy in stage IV disease. One possible explanation for the lack of consensus includes the wish of some physicians to provide a treatment holiday for incurable patients. Another explanation might be related to inherent weaknesses in the design of these studies. The docetaxel maintenance trial59 did not include a placebo study arm, while treatment with pemetrexed was not required at the time of disease progression for observation patients in the other maintenance study.60

In a phase 3 trial of patients with advanced NSCLC who did not progress on first-line gemcitabine/carboplatin, docetaxel was started either immediately after chemotherapy as a maintenance therapy or delayed until disease progression.59 Immediate treatment with docetaxel significantly improved PFS (6.5 vs 3.8 months, P < .0001), but produced only a trend for better OS (11.9 vs 9.1 months) compared with the delayed approach. Maintenance pemetrexed following a platinum doublet has also been associated with superior PFS (HR = 0.444; P = .00,001) in patients with nonsquamous tumors.60 In phase 3 trials of bevacizumab and cetuximab, these targeted therapies were continued after chemotherapy until evidence of disease progression. It is not known whether patients who respond to initial chemotherapy plus bevacizumab or cetuximab would benefit from switching immediately to a different targeted therapy in a maintenance paradigm.


Targeted therapies have been evaluated in an effort to improve patient outcomes beyond the plateau effect achieved with chemotherapy doublets in first-line treatment of advanced NSCLC. Toward this end, a number of different targeted therapies with varying mechanisms of action have been tested, but with limited success. Only 2 targeted therapies—bevacizumab and cetuximab—when added to first-line chemotherapy have been shown to extend survival of advanced NSCLC patients. Unlike the targeted therapies that have produced negative results, both bevacizumab and cetuximab were tested in combination with chemotherapy in randomized phase 2 trials before advancing to the pivotal phase 3 stage. Accordingly, the go-no-go decision to advance to phase 3 should be based on safety and efficacy data in randomized phase 2 trials rather than phase 1 or single study arm phase 2 studies.

The use of targeted therapies, whether in clinical trials or practice, is currently dictated by eligibility for therapy based on clinical features and demographic factors. In the future, efforts will focus on ways to enrich study cohorts with patients who are most likely to respond to therapy by also considering pathologic features and molecular markers. Improved patient selection is expected to enhance the success rate in NSCLC clinical trials, and in turn, define future clinical practice with improved patient outcomes.


Philip D. Bonomi has received research support from ImClone, Bristol, Myers, Squibb, and Genentech and has received honoraria for speakers' bureau services to Genentech and ImClone.