George Simon and Gerold Bepler were at H. Lee Moffitt Cancer Center and Research Institute when this work was performed.
George R. Simon generated the study concept and contributed to data tabulation, statistical analysis, and patient accrual; he is a guarantor of the article and takes responsibility for the integrity of the work as a whole, from inception to publication.
Fax: (843) 792-0644
Michael J. Schell PhD,
Biostatistics Programs, H. Lee Moffitt Cancer Center and Research Institute, Tampa, Florida
Excision repair cross complementing 1 (ERCC1) and ribonucleotide reductase M1 (RRM1) are molecular determinants that predict sensitivity or resistance to platinum agents and gemcitabine, respectively. Tailored therapy using these molecular determinants suggested patient benefit in a previously reported phase 2 trial. Here, we report an individual patient analysis of prospectively accrued patients who were treated with the “personalized therapy” approach versus other “standard,” noncustomized approaches.
Patients who had nonsmall cell lung cancer (NSCLC) with extranodal metastatic disease and an Eastern Cooperative Oncology Group performance status of 0/1 were accrued to 4 phase 2 clinical trials conducted at the H. Lee Moffitt Cancer Center: Trial A (first-line carboplatin/gemcitabine followed by docetaxel), Trial B (docetaxel and gefitinib in patients aged ≥70 years), Trial C (combination therapy with carboplatin/paclitaxel/atrasentan), and Trial D (personalized therapy based on ERCC1 and RRM1 expression). Patients with low RRM1/low ERCC1 expression received gemcitabine/carboplatin, patients with low RRM1/high ERCC1 expression received gemcitabine/docetaxel, patients with high RRM1/low ERCC1 expression received docetaxel/carboplatin, and patients with high RRM1/high ERCC1 expression received vinorelbine/docetaxel. Patients who were treated on Trials A, B, and C were pooled together and analyzed as the “standard therapy” group. Patients accrued to Trial D were called the “personalized therapy” group. Individual patient data were updated as of February 8, 2011. Overall survival (OS) and progression-free survival (PFS) were estimated using the Kaplan-Meier method.
There were statistically significant improvements between the personalized therapy group versus the standard therapy group in response (44% vs 22%; P = .002), OS (median: 13.3 months vs 8.9 months; P = .016), and PFS (median: 7.0 months vs 4.3 months; P = .03).
In the United States, there were 212,500 cases of lung cancer diagnosed in the year 2010, and lung cancer accounted for >25% of all cancer-related deaths.1 Among lung cancers, nonsmall cell lung cancer (NSCLC) comprises approximately 85% of all newly diagnosed cases. Approximately 40% of all patients with NSCLC present with extranodal metastatic disease, for which combination chemotherapy is considered palliative and produces low measurable response rates (range, 20%-35%) and rare complete responses. In these patients, disease progression is inevitable, and proves to be fatal in the majority of patients.2-4
In most patients, a doublet chemotherapy regimen is considered the standard of care for the first-line treatment of advanced NSCLC.2, 4 It produces a median overall survival (OS) of 8 to 10 months and a 1-year survival rate of 31% to 36%. Recent clinical trials have demonstrated that adding bevacizumab5 or cetuximab6 to specific chemotherapy doublets improves survival. No single doublet regimen has emerged as the best choice in terms of efficacy, although recent trials have indicated that pemetrexed-cisplatin combinations may be better in nonsquamous NSCLC than gemcitabine-cisplatin. The reverse is true for squamous cell carcinoma.3 Nevertheless, the selection of chemotherapy for the majority of patients remains arbitrary and typically is dictated by the oncologist's personal preference, convenience of delivery, and regimen-specific toxicity.
Excision repair cross complementing gene 1 (ERCC1) is a DNA damage repair gene that encodes the 5′ endonuclease of the nuclear excision repair complex and plays an important role in DNA damage repair. Platinum compounds are heavy metal complexes that form adducts with, and cross-links between, DNA molecules and, thus, effectively block DNA replication and transcription. Repair of these adducts and cross-links are dependent on ERCC1. The nuclear excision repair complex recognizes and removes these adducts and, thus, triggers resistance to platinum agents.7 The ribonucleotide reductase M1 (RRM1) gene is the regulatory subunit of ribonucleotide reductase (RR). RRM1 is also the predominant cellular determinant of chemotherapeutic efficacy for gemcitabine.8
We previously reported a single-institution, phase 2 trial in patients with advanced NSCLC in which the selection of a chemotherapy doublet was based on tumor RRM1 expression (determining whether or not gemcitabine was used) and ERCC1 expression (determining whether or not platinum was used). The outcome of patients on the trial was encouraging, with improved responses (44%; 95% confidence interval [CI], 31%-59%) and improved median OS (13.3 months; 95% CI, from 11.5 to <24 months) and progression-free survival (PFS) (6.6 months; 95% CI, 4.7-8.8 months).9 In the current report, we demonstrate that the response and survival of patients who received treatment with this personalized approach compared favorably with the response and survival of patients who received noncustomized (for the purposes of this report called “standard”) treatments in a prospectively accrued cohort of patients.
MATERIALS AND METHODS
Patients were accrued prospectively to 4 phase 2 clinical trials conducted at the H. Lee Moffitt Cancer Center and Research Institute (Tampa Fla; National Clinical Trials [NCT] ClinicalTrials.Gov identifiers NCT00226590, NCT00231465, and NCT00215930 and 1 study identified as MCC-13303). Patients were eligible for inclusion in these trials if they had stage IV or stage IIIB NSCLC with malignant pleural effusions. Patients with brain metastases were included if they had received treatment for such metastases and were deemed stable. Only patients with an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1 were eligible.
Staging studies had to include a physical examination, computed tomography studies of the chest and upper abdomen, and computed tomography or magnetic resonance imaging studies of the brain, if clinically indicated. Patients underwent 18F-fluorodeoxyglucose positron-emission tomography for staging at the investigator's discretion and were previously untreated for advanced disease. The University of South Florida Institutional Review Board (IRB) approved all trials included in this report, and all patients signed an IRB-approved informed consent.
In the first trial (Trial A), which enrolled patients between March 2001 and February 2003 (NCT00226590), patients received second-line therapy (docetaxel) soon after first-line therapy (carboplatin and gemcitabine) was concluded.10 In the second-trial (Trial B), which enrolled only patients aged ≥70 years between March 2003 and May 2005 (NCT00231465), a novel combination of a chemotherapeutic drug (docetaxel) and an oral epidermal growth factor tyrosine kinase inhibitor (gefitinib) was tested as first-line therapy.11 In the third trial, between July 2003 and August 2005 (Trial C), patients were enrolled in a phase 1/2 trial combining chemotherapy (carboplatin and paclitaxel) with a novel antiangiogenesis agent (atrasentan; MCC-13303).12 In these 3 trials, patients were considered eligible for analyses if they received at least 1 cycle of treatment. For purposes of this analysis, patients enrolled in these studies were considered to have received “nonpersonalized” therapy and are grouped together as the “standard therapy” group.
Finally, a single-institution phase 2 trial (NCT00215930; Trial D) accrued patients between February 2004 and December 2005 and evaluated the feasibility and efficacy of selecting double-agent chemotherapy based on tumor expression of RRM1 and ERCC1 in previously untreated patients with advanced NSCLC.7 This study required a pretreatment biopsy. Tumoral ERCC1 and RRM1 expression levels were measured by real-time quantitative reverse transcriptase polymerase chain reaction, as reported previously; and chemotherapy was assigned based on the expression of these molecular determinants.8, 13ERCC1 and RRM1 are known predictors of platinum resistance and gemcitabine resistance, respectively. In the current study, patients were included for analyses if they had a biopsy for gene measurements. Predetermined values for RRM1 and ERCC1 were used to dichotomize between high versus low expression levels. ERCC1 and RRM1 values were expressed as unit-less ratios of the housekeeping gene 18s RNA. If the RRM1 expression level was ≤16.5, then gemcitabine was used in the treatment doublet; if the ERCC1 expression level was ≤8.7, then carboplatin was used in the treatment doublet. Docetaxel was used if the ERCC1 or RRM1 levels were high in lieu of carboplatin or gemcitabine, respectively. These levels were selected based on our previously published experience.8, 13 This strategy resulted in 4 possible gene expression strata with the following doublet therapies: the low RRM1 and low ERCC1 group received gemcitabine, 1250 mg/m2 on days 1 and 8, and carboplatin at an area under the plasma concentration time curve (AUC) of 5 on day 1 every 21 days (the gemcitabine/carboplatin group); the low RRM1 and high ERCC1 group received gemcitabine 1250 mg/m2 on days 1 and 8 and docetaxel 40 mg/m2 on days 1 and 8 every 21 days (the gemcitabine/docetaxel group); the high RRM1 and low ERCC1 group received docetaxel 75 mg/m2 on day 1 and carboplatin at an AUC of 5 on day 1 every 21 days (the docetaxel/carboplatin group); and the high RRM1 and high ERCC1 group received vinorelbine 45 mg/m2 on days 1 and 15 and docetaxel 60 mg/m2 on days 1 and 15 every 28 days (the docetaxel/vinorelbine group). Patients enrolled in this study are considered to have received “personalized therapy” and are referred to as the “personalized therapy” group.
In the 4 trials described above, disease response was assessed sequentially after every 2 cycles by computed tomography studies of the chest, upper abdomen, and other areas as indicated. Other imaging modalities that were used for disease response assessment included magnetic resonance imaging studies of the brain and soft tissues. Patients without disease progression were continued on therapy for at least 4 cycles. Patients enrolled on the docetaxel and gefitinib trial received the combination of docetaxel and gefitinib for 4 cycles if there was no progression and then were maintained with gefitinib alone until they developed disease progression.11 Subsequent clinical management in all studies was at the discretion of the treating physician.
Follow-up data for OS, DFS, and sites of tumor recurrence were obtained at regular intervals and updated. We recommended that patients have follow-up visits every 3 months until progression. All patients who were not being followed routinely at the H. Lee Moffitt Cancer Center were contacted, and their current status was ascertained as of February 8, 2011.
OS was calculated from the date of first chemotherapy to the date of death, and PFS was calculated from date of first chemotherapy to the date of either recurrence or death, whichever came first. For patients who remained alive or progression free, data were censored at the time of the last follow-up visit.
Baseline characteristics were reported as counts and percentages dichotomized by the 2 study groups (the personalized therapy group and the standard therapy group). The Mantel-Haenszel chi-square test for 2 × 2 tables and the Fisher exact test for 2 × 3 tables were used to explore the association of study groups with baseline characteristics. All outcome analyses were performed according to the intent-to-treat principle. OS and PFS were estimated using the Kaplan-Meier method, and differences in OS and PFS between study groups were assessed with 2-sided log-rank tests. A Cox proportional hazards model was used to assess the effect of baseline characteristics, including study group, on survival. Variables with P values ≤.25 in univariable analysis were selected for construction of the multivariable model. A backward elimination procedure was used, and variables with P values ≤.15 were retained in the model. All statistical analyses were performed using the SAS statistical software package (version 9.2; SAS Institute Inc., Cary, NC).
Between March 2001 and December 2005, 181 patients were accrued to the 4 clinical trials. A summary of the 4 clinical trials is outlined in Table 1. The salient patient and disease characteristics of the standard therapy and personalized therapy groups are summarized in Table 2. There were no statistically significant differences between the 2 groups except for age. More patients aged >70 years were enrolled in the standard therapy group, because 1 of the studies (NCT00231465) enrolled only elderly patients (ie, age ≥70 years). Patients were treated for 2 cycles beyond maximal response; then treatment was stopped, and patients were observed until they developed disease progression. The median number of cycles delivered in both arms was 4 cycles.
Table 1. Summary of the Evaluated Clinical Trials
Personalized Therapy Group
Standard Therapy Group
Abbreviations OS, overall response; NR, not reported; PFS, progression-free survival.
There was a statistically significant improvement in response (complete and partial responses) in the personalized therapy group compared with the standard therapy group (44% vs 22%; P = .002). Achieving a response had an unambiguous effect on survival; the hazard ratio (HR) for nonresponse was 1.63 (95% CI, 1.15-2.29) for OS and 1.82 (95% CI, 1.30-2.54) for PFS.
There was a statistically significant improvement in OS (P = .01) in the personalized therapy group over the standard therapy group (median survival, 13.3 months [95% CI, 10.3-19.3 months] vs 8.9 months [95% CI, 6.4-11.2 months]) (Fig. 1A). The effect of age on survival was evaluated carefully, because the standard therapy group had more patients aged ≥70 years compared with the personalized therapy group, and this difference was statistically significant (P < .0001). The median OS for patients aged ≥70 years in the entire study population was 12.1 months (95% CI, 5.6-16.1 months) compared with 10.3 months for patient aged <70 years (95% CI, 8.5-12.0 months; P = .897). The median PFS was 5.1 months for the older age group (95% CI, 2.8-7.0 months) compared with 4.8 months for the younger age group (95% CI, 3.8-6.0 months; P = .398). Thus, the survival of older patients in this analysis was not inferior to the survival of younger patients.
The median PFS was 7.0 months (95% CI, 4.7-9.0 months) in the personalized therapy group versus 4.3 months (95% CI, 3.0-5.3 months) in the standard therapy group (P = .030) (Fig. 1B), although nearly all patients eventually progressed or died.
Univariable and Multivariable Cox Proportional Hazards Models for Overall and Progression-Free Survival
Female sex, a performance status of 0 (vs 1), and personalized therapy group were predictors of improved survival in both univariable and multivariable Cox proportional hazards model analyses for OS (Table 3) and PFS (Table 4).
Table 3. Univariable and Multivariable Cox Proportional Hazard Models for Overall Survivala
The results from this individual patient analysis suggest that ERCC1/RRM1-tailored chemotherapy improves survival in patients with advanced NSCLC. Several phase 3 randomized trials currently are underway testing the efficacy of customized chemotherapy versus standard treatment-selection approaches in both adjuvant and advanced settings.14 Cobo et al15 conducted a randomized trial to test the hypothesis that ERCC1-tailored therapy would improve objective responses. Patients were assigned randomly at a 1:2 ratio to either the control arm or the genotypic arm. Patients in the control arm received docetaxel plus cisplatin. In the genotypic arm, patients with low ERCC1 levels received docetaxel plus cisplatin, and those with high levels received docetaxel plus gemcitabine. Of the 346 patients in that study who were assessable for response, objective responses were noted in 53 patients (39.3%) in the control arm and in 107 patients (50.7%) in the genotypic arm (P = .02). However, there were no survival differences in the 2 study arms, which the study was not designed to demonstrate. A phase 3 randomized trial is currently underway that will test the hypothesis that ERCC1/RRM1-tailored therapy will improve survival over the standard chemotherapy regimen of carboplatin and gemcitabine (NCT00499109).
Another interesting observation is that the OS curves for personalized and standard therapy separated out early and remained separate for the entire duration of the follow-up period (9 years) (Fig. 1A), despite the finding that the PFS curves essentially merged together at the end of 2 years (Fig. 1B). This leads us to speculate that the benefits of the 4 cycles of personalized chemotherapy, usually completed in the first 4 months after diagnosis, may continue even after the initial therapy has been concluded. We and others have demonstrated previously that higher expression of ERCC1 and RRM1 predicts for a more indolent phenotype.13, 16, 17 Treatment with cisplatin in patients with low ERCC1 expression will induce the increased expression of ERCC1 that, in turn, will bestow the surviving tumor with an indolent phenotype. Treatment with gemcitabine will have a similar effect on patients with low RRM1 expression. Indeed, in our population of patients with stage IV disease, 11.3% of patients remained alive after 7 years in the personalized therapy arm compared with 4.5% of patients in the control arm. To confirm this hypothesis, serial biopsies will need to be performed in future studies to better understand how the expression of these and other genes change in response to treatment.
In conclusion, our data suggest that ERCC1/RRM1-tailored selection of first-line therapy improves survival over standard treatment-selection approaches. The limitations of our analyses are that this was a pooled analyses of patients accrued to 4 different phase 2 trials. There was significant overlap between the accrual periods of these 4 trials, with patients accrued between the years 2001 and 2005. Although there was parity between the performance status and other eligibility criteria across all 4 studies (as outlined in Table 2), patients in the personalized therapy group required a second biopsy to enable molecular analyses. This could have led to a cohort of “more motivated” patients enrolling in the personalized therapy group, potentially introducing a bias in favor of the personalized therapy group. Hence, the results of this analysis can only be considered corroboratory but not confirmatory. Confirmation of these data must come from the currently ongoing randomized phase 3 trials.
This work was partly supported by grant NIH 1 R21 CA106166-01 from the National Institutes of Health.
CONFLICT OF INTEREST DISCLOSURES
Gerold Belper has a licensing agreement and pending patent for Therapeutic Decisions and has received research funding from Sanofi-Aventis and Eli Lilly.