Gene Expression as a Predictive Marker of Outcome in Stage IIB-IIIA-IIIB Non-Small Cell Lung Cancer After Induction Gemcitabine-Based Chemotherapy Followed By Resectional Surgery

The overall 5-year survival of patients with non-small cell lung cancer (NSCLC) has remained at <15% for the past 20 years. For patients with clinical stage IIB (T1–2N1M0 and T3N0M0), 5-year survival is ∼25%; for those with stage IIIA (T3N1M0 and T1–2-3N2M0), it is 13%; and for those with stage IIIB (T4N0–1-2M0), it bottoms out at 7% (1). Small randomized studies of neoadjuvant chemotherapy in stage IIIA (2) or stage IIB-IIIB (3) showed remarkable improvement in survival over patients treated either with surgery alone or with surgery followed by adjuvant radiotherapy. Event-free survival was similar in the two studies: 12.7 (2) and 20 (3) months in the neoadjuvant chemotherapy arm and 5.8 (2) and 5 (3) months in the surgery arm.

The efficiency of removal of cisplatin DNA adducts by the nucleotide excision repair (NER) system is assumed to be one of the determinants of cisplatin resistance (4, 5). Excision repair cross-complementation group 1 (ERCC1) is a single-stranded DNA endonuclease, which forms a tight heterodimer with xeroderma pigmentosum complementation group F. Its role in NER is to incise DNA on the 5′ side of lesions such as cisplatin DNA adducts. Overexpression of ERCC1 and other NER genes has been associated with repair of cisplatin-induced DNA damage and clinical resistance to cisplatin (4, 5, 6, 7, 8), and repair of cisplatin DNA adducts does not occur in the absence of functional ERCC1 (4, 5, 6, 7, 8). Ribonucleotide reductase is responsible for the reduction of ribonucleotides to their corresponding deoxyribonucleotides, providing a balanced supply of precursors for DNA synthesis and repair. Alterations in ribonucleotide reductase levels can have significant effects on such biological properties of cells as tumor promotion and tumor progression and can potentiate metastasis. It has been reported that transcription coupled-NER-deficient cells, with underexpression of xeroderma pigmentosum group D (XPD), are hypersensitive to cisplatin regardless of their global genome-NER status.

On the basis of these data, we have examined for the first time the potential correlation and predictive value of ERCC1, ribonucleotide reductase subunit M1 (RRM1), and XPD mRNA expression in resected specimens from stage IIB, IIIA, and IIIB NSCLC patients treated with neoadjuvant gemcitabine/cisplatin followed by surgery.

The present study composed of 67 consecutive stage IIB-IIIA-IIIB NSCLC patients from one single institution (Hospital Vall d’Hebron, Barcelona, Spain). Initially, surgical resection of both primary disease and hilar/mediastinal nodal disease was recommended for patients with stage IIB-IIIA NSCLC, and patients with potentially resectable T4 N0 lesions (stage IIIB) were also eligible for this study. All of the patients were medically fit to undergo surgical resection, and they underwent surgery between September 1998 and December 2002, after receiving neoadjuvant gemcitabine/platinum chemotherapy. Baseline patient characteristics are shown in Table 1. Patients received three cycles of neoadjuvant chemotherapy; 63 received cisplatin 100 mg/m² day 1 and gemcitabine 1250 mg/m² day 1 and 8 every 21 days; and 4 patients with creatinine clearance <60 ml/min received carboplatin area under the curve = 5 day 1 and gemcitabine 1000 mg/m² day 1 and 8 every 21 days. Clinical tumor response was evaluated after three chemotherapy cycles according to the Eastern Cooperative Oncology Group criteria for solid-tumor response. A thoracotomy was performed within 4–5 weeks after the last chemotherapy cycle; the surgical procedure was based on the extent of tumor at the time of the initial presentation.

Total RNA was recovered from formalin-fixed, paraffin-embedded surgical specimens using proteinase K digestion and phenol-chloroform extraction as described previously (9). After cDNA synthesis XPD, ERCC1, and RRM1 expression was analyzed with real-time quantitative PCR. Quantification of gene expression was performed using the ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA). The primers and 5′-labeled probe were as follows: β-actin, forward 5′ TGA GCG CGG CTA CAG CTT 3′, reverse 5′ TCC TTA ATG TCA CGC ACG ATT T 3′, and probe 5′ ACC ACC ACG GCC GAG CGG 3′; ERCC1, forward 5′ GGG AAT TTG GCG ACG TAA TTC 3′, reverse 5′ GCG GAG GCT GAG GAA CAG 3′, probe 5′ CAC AGG TGC TCT GGC CCA GCA CAT A 3′; XPD, forward 5′ GCT CCC GCA AAA ACT TGT GT 3′, reverse 5′ CAT CGA CGT CCT TCC CAA A 3′, probe 5′ ACC CTG AGG TGA CAC CCC TGC G 3′; and RRM1, forward 5′ ACT AAG CAC CCT GAC TAT GCT ATC C 3′, reverse 5′ CTT CCA TCA CAT CAC TGA ACA CTT T 3′, probe 5′ CAG CCA GGA TCG CTG TCT CTA ACT TGC A 3′.

Relative gene expression quantification was calculated according to the comparative threshold cycle method (2^−ΔΔCt) using β-actin as an endogenous control and commercial RNA controls (Stratagene, La Jolla, CA) as calibrators.

Survival curves were obtained by the Kaplan-Meier method, and the difference in survival in subgroups was analyzed using either the log-rank test or Tarone-Ware test (SPSS 11.5 software). We performed survival analysis using Cox proportional hazards models. We assessed the fit of the hazard models by plotting the cumulative hazard of death against the Cox-Snell residuals. In addition, we assessed the effect of gene mRNA expression on the relative risk (RR) of death using gene mRNA expression quartiles. All of the statistical tests were two-sided, and the level of significance was set at 5%.

Radiographic and surgical results are summarized in Table 1. Pathological complete response occurred in 5 patients (1 patient with radiographic stable disease and 4 patients with radiographic partial response to chemotherapy). Median overall survival for all of the patients was 37.80 months [95% confidence interval (CI), 27.04–48.56 months], and median event-free survival for all of the patients was 11.84 months (95% CI, 5.48–18.21 months). Thirty-day postoperative mortality was 7.5% (5 patients). This included a bronchopleural fistula resulting in death in 1 patient who underwent a bilobectomy; pneumonia and respiratory failure occurred in 2 patients who underwent a right pneumonectomy; a fourth patient died from respiratory failure after a left pneumonectomy; and the fifth cause of death was pneumonia after the patient had undergone a right upper lobectomy.

The 52 completely resected patients attained a median survival of 45.1 months (95% CI, 24.5–65.6 months). The 10 patients with incomplete resection and the 5 unresectable patients had a significantly shorter survival [6.8 months (95% CI, 5.5–8.1 months) and 5.6 months (95% CI, 3.6–7.5 months), respectively; log-rank P = 0.0001]. Twelve of the 67 patients received postoperative radiotherapy: 5 for residual N2 disease, 3 for rib involvement, and the remaining 4 for unresectable disease. Relapse occurred in 38 patients (57%). Sites of initial failure included only local-regional in 37% (14 of 38), only systemic in 18% (7 of 38), and both local and systemic in 45% (17 of 38). Of the 24 patients who relapsed at distant sites, 10 had brain metastases.

Significant correlations were found between ERCC1 mRNA expression and XPD mRNA expression (r = 0.48; P < 0.0001) and between RRM1 mRNA expression and XPD mRNA expression (r = 0.48; P < 0.0001). A nearly significant correlation was found between ERCC1 mRNA expression and RRM1 mRNA expression (r = 0.22; P = 0.07). For RRM1 mRNA levels, patients in the bottom quartile had a decreased risk of death compared with those in the top quartile (RR = 0.30; 95% CI, 0.10–0.91; P = 0.033; Table 2). For XPD mRNA levels, patients in the bottom quartile (0.08–0.71) had a decreased risk of death compared with those in the top quartile (1.71–3.78; RR = 0.40; 95% CI, 0.12–1.37; P = 0.145). For ERCC1 mRNA levels, patients in the bottom quartile (2.73–4.96) had a higher risk of death compared with those in the top quartile (7.45–12.31; RR = 1.51; 95% CI, 0.55–4.10; P = 0.422)

Significant differences in median survival were observed between the 17 patients in the bottom quartile of RRM1 expression and the 15 in the top quartile (52 versus 26 months; P = 0.018). Of these 17 patients in the bottom quartile, 5 underwent a pneumonectomy with a median survival of 31.12 months, and 12 underwent a lobectomy with a median survival of 51.97 months (P = 0.191). When the characteristics of these 17 patients were examined, no differences in the initial staging were observed in comparison with the remaining 50 patients in the other 3 quartiles. However, for patients in the bottom quartile, radiographic response tended to be higher (65% versus 54%; P = 0.24), complete resection was attained more often (94% versus 72%; P = 0.03), and a lobectomy was performed more often (71% versus 34%; P = 0.004). Furthermore, pathological complete response was observed in 29% of patients in the bottom quartile versus 0% of the remaining patients (P = 0.00001). No significant differences were observed according to ERCC1 or XPD mRNA levels (data not shown).

In the present study, a good correlation has been found among ERCC1, RRM1, and XPD gene expression levels, indicating that a single assessment of one of these genes could provide information about all three. This confirms our previous findings of a strong correlation between ERCC1 and RRM1. We carried out three different studies, examining individually the role of ERCC1, RRM1, and then both, mRNA expressions in paraffin-embedded pretreatment bronchial biopsies from gemcitabine/cisplatin-treated advanced NSCLC patients. In the first study, median overall survival was significantly longer in patients with low ERCC1-expressing tumors than in those with high ERCC1-expressing tumors (15 versus 5 months; P = 0.009). ERCC1 mRNA expression, performance status, and weight loss were significant independent prognostic factors (10). Then we analyzed RRM1 expression in pretreatment bronchial biopsies from a second group of gemcitabine/cisplatin-treated advanced NSCLC patients. Median time to progression was 7.6 months for patients with low RRM1 levels and 4.3 months for those with high levels (P = 0.005). Median survival was 15.5 months for patients with low RRM1 levels and 6.8 months for those with high levels (P = 0.002; Ref. 11). In a third group of patients, we studied both ERCC1 and RRM1 mRNA expression in pretreatment bronchial biopsies again from gemcitabine/cisplatin-treated advanced NSCLC patients who were part of a large randomized trial (12). There was a strong correlation between ERCC1 and RRM1 mRNA expression levels (r = 0.4; P < 0.001). Patients with low RRM1 mRNA levels had significantly longer median survival than those with high levels (13.7 versus 3.6 months; P = 0.009). There were no significant differences according to ERCC1 mRNA levels, although there was a tendency to longer median survival among patients with low ERCC1 levels. Median survival was significantly longer among patients with low levels of both RRM1 and ERCC1 (not reached) than among those with high levels of both genes (6.8 months; P = 0.016; Ref. 13). These observations confirm the preclinical data in which the overexpression of ribonucleotide reductase confers gemcitabine resistance in the human oropharyngeal carcinoma KB cells (14).

In the present study, patients with the lowest RRM1 levels attained better survival, which was also correlated with a high complete resection rate, with a median survival of 52 months. Martini et al. (15) demonstrated for the first time that complete resection was essential for effective disease control and long-term survival in stage IIIA N2 disease after neoadjuvant chemotherapy. Eighty-nine patients with complete resection had a median survival of 27 months, a 3-year survival of 41%, and a 5-year survival of 26%, whereas 47 patients with incomplete or no resection had a median survival of 12 months and a 5% survival at 3 and 5 years (15). Several other studies (16, 17, 18) have reported a 40% or more 3-year survival in patients with complete resection. Pathological complete remission was also higher for the group of patients with the lowest RRM1 levels (29%), both compared with other patients in this trial (0%) and with the average reported in the literature (12%; Ref. 19). We did not observe differences in median survival according to surgical procedure, although a tendency toward longer survival was observed in those patients who underwent a lobectomy. Pneumonectomy has been found to be an adverse prognostic factor in multivariate analyses (20).

Further prospective research is being carried out to test whether patients with low RRM1 mRNA levels (bottom quartile) will obtain the maximum survival benefit from cisplatin doublets, whereas those with high levels (top three quartiles) could obtain better outcomes when treated with noncisplatin doublets.

Dr. Thomas Lynch: Dr. Gandara, you have done a lot of work in terms of thinking about how one selects chemotherapy. What do you think is the most promising course?

Dr. David Gandara: Although all of these data are very interesting, Dr. Rosell and I have had practical problems in terms of how do you apply these to a patient population. The problem of course is that our patients are very heterogeneous in regard to all of these pathways and that it really takes a large prospectively designed trial to validate some of these of markers. SWOG is now doing a repeat of the Noda study in small cell lung cancer [N. Engl. J. Med., 346:85–91, 2002]. As part of that study we are selecting genomic DNA from all the patients to look at polymorphisms for UGT1A1 and for all of the DNA repair genes. Among 620 patients, we will probably be able to sort out at least that component of the polymorphisms. The NCI Japan has agreed to provide genomic DNA from similar patients treated there as part as a collaborative project, so we can also see about interracial differences.

Dr. Eric Rowinsky: Given the extremely low rates of polymorphism in the population in regard to any target protein, do you really think any of these studies would be able to discern if there are truly functional differences in activity in those patients? I think the numbers required, if you really tried to do some calculations, are astounding. I don’t think that can be answered in clinical trials and possibly could be answered in the laboratory if we understood the polymorphisms.

Dr. Gandara: In actuality, we have put incidence into the statistical design, and we have very good data for the incidence of the UGT1A1 polymorphism. If European heritage is 13% for the 77 genotype, Hispanic 19%, African Americans 15%, and Japanese 2%, that would predict toxicity from irinotecan, so that could explain a lot of differences between Japan and the US. On the basis of the incidence of the polymorphisms, we can discern it with a 620 patient trial.

Dr. Bruce Johnson: I believe the data to support the study design examining the role of ERCC1 was done mostly with response to gemcitabine, is that correct? Why use docetaxel as a control arm? Most of your results are based on about 50 patients and with your planned multivariate analysis, I would assume the logistics of doing this trial are incredibly difficult. You have to be able to see a difference between 180 patients. Did you think about doing another prospective study with a group that is homogeneously treated, rather than going straight to a randomized study?

Dr. Rosell: Again, we have much preclinical evidence that is used in the design of this trial. It is still relatively difficult to organize clinical trials with mandatory biopsy and a central review because in stage IV, as you know, at least one subset of the patients never have biopsy, they have only positive cytology. This kind of a study should be requested even with targeted therapies, to have the tumor or the tissues analyzed at baseline as a possible parameter. All these studies were analyzed in a few patients because there were no means to find more tissue. Even from the 600-patient Italian study, only 100 tumor samples were available. This is one of the major pitfalls that we have today. If I had the opportunity to reshape the study, I would put RRM1 as a marker, to test if gemcitabine could be valid in patients with overexpression with RRM1 with just changing the cisplatin to an antimicrotubule drug. We are also in this study validating the importance of BRCA1.

Dr. Paul Bunn: What really needs to happen is to validate these assays. These quantitative PCR tests may not give you the same result in two different laboratories.

Dr. Rosell: In the study I have presented they have the same results. But, a major caveat: this is almost impossible unless there is a consensus meeting of people involved in the laboratory.

Dr. Bunn: We need to do a large randomized trial: it is very costly and the question is whether there are really enough data to put resources into that type of trial. There should first be a prospective trial with a reasonable number of patients, where you are going to give them all of the same treatment to see if this really does predict in a large set of patients. I would submit that trial should have two laboratories doing the laboratory tests, and they should be blinded. If you can show that the results are reproducible and predictive, then you could go to a prospective trial. It is very difficult to get these samples.

Dr. Rosell: To avoid potential confusion, I will clarify: RNA isolation can be performed in any laboratory. Quantitative PCR analysis can be performed in any laboratory. It is not necessary to validate. The calibration can be different, because the values you are providing are completely different from another laboratory. There could be consensus on the values, but that is not absolutely necessary.