Multidrug Resistance Proteins Do Not Predict Benefit of Adjuvant Chemotherapy in Patients with Completely Resected Non–Small Cell Lung Cancer: International Adjuvant Lung Cancer Trial Biologic Program Free

Several randomized trials showed that adjuvant cisplatin-based chemotherapy improves survival of patients with completely resected non–small cell lung cancer (NSCLC; refs. 1–3). The International Adjuvant Lung Cancer Trial (IALT), the largest of these trials, randomly assigned 1,867 patients either to receive cisplatin-based chemotherapy or to be observed only (1). A pooled analysis of the five largest trials recently confirmed a survival benefit of absolute 5% by adjuvant chemotherapy (4). Thus, adjuvant chemotherapy is now considered as standard treatment for selected patients with completely resected NSCLC, particularly those with stage II and III disease. Sex, histologic type, and other clinical variables do not allow selecting those patients who benefit from adjuvant chemotherapy (1, 4). Therefore, molecular markers should be studied for their potential predictive value.

Of particular interest in this regard are transport proteins, such as those belonging to the ATP-binding cassette multidrug resistance proteins (MRP; refs. 5–7). Overexpression of MRP1 or MRP2 in tumor cells confers resistance to various anticancer drugs, including anthracyclines, Vinca alkaloids, and epipodophyllotoxins. MRP1 and MRP2 have similar substrate specificity but, in contrast to MRP1, MRP2 also mediates resistance to cisplatin (7).

The purpose of our study, which is part of the IALT biologic program (IALT-Bio), was to determine whether MRP1 and MRP2 are of prognostic and/or predictive value in NSCLC patients who were enrolled into the IALT.

Patients and study design. We retrospectively collected paraffin-embedded tumor blocks from 867 patients, of which 783 were of sufficient quality for further analyses (8). Twenty-eight centers in 14 countries contributed samples (see Appendix 1). Approval was obtained by local Institutional Review Boards according to the legal regulations in each participating country. All tumor specimens were obtained at the time of surgery.

Immunohistochemistry. Immunohistochemical analyses were carried out in a single laboratory (Department of Medicine I, Medical University of Vienna, Vienna, Austria) and done as described previously (9).

Tissue sections were deparaffinized and rehydrated. After heat-induced epitope retrieval in 10 mmol/L citrate buffer (pH 6.0; only for MRP2), slides were incubated for 90 min at room temperature with either the monoclonal antibody for MRP1 (clone MRPr1; 1:100 dilution; Alexis) or MRP2 (clone M2III-6; 1:100 dilution; Alexis). Antibody binding was detected by the avidin-biotin-peroxidase method (MRP1) or by means of the UltraVision LP detection system (MRP2; Lab Vision Corp.) according to the manufacturer's recommendations. Color development was done with 3,3′-diaminobenzidine (Dako). The slides were counterstained with Mayer's hematoxylin and mounted.

Positive controls were NSCLC samples known to express the given antigen. Immunostaining was examined by an experienced lung pathologist (K.S.) who was blinded to the clinical outcome of the patients. Immunostaining was classified based on staining intensity and percentage of stained tumor cells. Membranous and cytoplasmatic staining of tumor cells was scored as positive. Staining intensity was determined as 0 (absent), 1 (weak), 2 (moderate), and 3 (strong). The median values of the percentages of MRP1- or MRP2-positive tumor cells were prospectively chosen as cutoff points to classify marker-positive and marker-negative tumors. Choosing the median as cutoff point guarantees objectivity because it is not determined a posteriori as an “optimal” cutoff point. Therefore, one does not need any P value adjustment and questionable bias correction. Furthermore, maximal power for prognostic and predictive studies is achieved when the marker is equally distributed (10).

Statistical analysis. Assuming that the effect of chemotherapy is different between equally sized marker-positive and marker-negative patients, the estimated power for 800 patients is 66% for detecting a 20% absolute difference in 5-year survival benefit with a two-sided type I error of 1% (11). For the prognostic analysis, the power was 99% to detect a 20% absolute difference in 5-year survival between marker-positive and marker-negative patients.

To study selection bias within the IALT-Bio participating centers, the prerandomization characteristics of patients for whom tumor blocks were available or patients without blocks were compared using χ² tests stratified by center, and their overall survival using a Cox model. Baseline data according to the biomarker status were compared using univariate analyses with χ² tests and multivariate logistic models.

Survival rates were estimated using the Kaplan-Meier method. The prognostic values of the biomarker status and chemotherapy on survival were studied using the Cox model. As in the IALT analysis, the Cox model included every factor used in the stratified randomization (center, tumor stage, and type of surgery) plus clinical and histologic prognostic factors (age, sex, WHO performance status, nodal status, lymphoid infiltration, and the revised histopathologic type). All other factors that were statistically related to the biomarker status in the multivariate logistic model (P < 0.05) were added to the survival Cox model. The predictive value of the biomarkers was studied by testing the interaction between the biomarker status and the attributed treatment (chemotherapy or no chemotherapy) in the same Cox model. Sensitivity analyses using Cox models with a lower number of adjustment factors were done and the results were similar. Therefore, only results corresponding to the above model are presented. All reported P values were two sided. In the statistical analysis plan, P values of <0.01 were considered statistically significant to limit the risk of false-positive results. All analyses were done using Statistical Analysis System software, version 8.2 (SAS Institute, Inc.).

We did the immunohistochemical analyses of MRP1 and MRP2 in tumor sections of 783 patients. In one case, staining was not evaluable for both markers. Therefore, 782 patients remained in the study and all further statistical analyses were based on these patients. These patients were also representative of the participating center population because the prerandomization characteristics and overall survival of the patients who were included into the present study did not differ from that of cases of the same centers that were not included (data not shown).

Immunostaining of MRP1 and MRP2 was both membranous and cytoplasmatic and ranged from 0% to 100% (Fig. 1). Histograms of the percentages of stained cells by staining intensity of MRP1 or MRP2 are shown in Fig. 1. The median value of the percentages of stained tumor cells was 20% for MRP1. In case of MRP2, 60% of the samples did not show any immunostaining of tumor cells and were therefore classified as negative. MRP1 expression was considered positive (>20% MRP1-positive tumor cells) in 364 (47%) and negative (≤20% MRP1-positive tumor cells) in 418 (53%) patients. Likewise, MRP2 expression was considered positive (>0% MRP2-positive tumor cells) in 313 (40%) and negative (0% MRP2-positive tumor cells) in 469 (60%) patients. The association of MRP1 or MRP2 with clinical variables is shown in Table 1. MRP1 expression was more frequently positive in patients with lymph node involvement (P = 0.002), tumor stage II and III (P < 0.001), and poor performance status (P < 0.001) but was not significantly associated with any other clinical variable listed in Table 1. MRP2 positivity correlated significantly with adenocarcinomas (P < 0.001) and performance status of 0 (P = 0.007).

Next, we compared MRP1 or MRP2 expression simultaneously with other variables (age, sex, tumor status, nodal status, pathologic tumor stage, histologic type, performance status, and lymphoid infiltration) in multivariate logistic models. In these analyses, MRP1 positivity was associated with stage II and III (P < 0.001), poor performance status (P = 0.002), and middle-aged patients (P = 0.04). Patients with adenocarcinomas (P < 0.001), performance status of 0 (P = 0.02), and vascular invasion (P = 0.04) were more frequently MRP2 positive.

Prognostic analysis. With a median follow-up of 56 months, 404 of 782 (52%) patients had died (205 patients in the chemotherapy group and 199 patients in the control group). The 5-year overall survival rate was 43% [95% confidence interval (95% CI), 39-47%] in the total study population. Older age, male sex, higher lymph node status, nonadenocarcinoma, poor performance status, low lymphoid infiltration, and MRP2 expression were significantly associated with shorter overall survival in the univariate analyses. The Cox models were adjusted for these variables plus tumor stage, type of surgery, and treatment and stratified on centers. In these multivariate analyses, MRP2 was significantly associated with overall survival of the patients (Table 2; Fig. 2B). Patients with MRP2-positive tumors had a significantly shorter overall survival than patients with MRP2-negative tumors [adjusted hazard ratio (HR) for death, 1.37; 95% CI, 1.09-1.72; P = 0.007; Table 2; Fig. 2B]. The 5-year survival rates were 45% (95% CI, 39-50%) in MRP2-negative patients and 40% (95% CI, 34-46%) in MRP2-positive patients, respectively. Median overall survival was 9 months longer in the MRP2-negative group compared with the group of patients with MRP2-positive tumors (54 and 45 months, respectively; Table 2). In contrast, we found no significant correlation between MRP1 expression and overall survival of the patients (Table 2; Fig. 2A).

Predictive analysis. Next, we analyzed the predictive value of MRP1 or MRP2. The benefit of cisplatin-based chemotherapy was not different between MRP1-negative and MRP1-positive patients (test for interaction, P = 0.37; Table 3A), that is, the adjusted HR for death among MRP1-negative patients (0.79; 95% CI, 0.60-1.04) was not significantly different from that of MRP1-positive patients (0.95; 95% CI, 0.71-1.28; Fig. 2C and D). Similarly, the benefit of cisplatin-based chemotherapy was not different between MRP2-negative and MRP2-positive patients (test for interaction, P = 0.56; Table 3B), that is, the adjusted HR for death among MRP2-negative patients (0.93; 95% CI, 0.71-1.21) was not significantly different from that of MRP2-positive patients (0.82; 95% CI, 0.60-1.12; Fig. 2E and F).

In the present study, we assessed expression of MRP1 and MRP2 in a homogenous and well-defined patient population consisting of completely resected NSCLC patients who were enrolled into the IALT. This was the largest group of NSCLC patients ever used for the investigation of MRPs. We showed that MRP2 is of prognostic value in these patients, whereas MRP1 had no effect on survival. Neither MRP1 nor MRP2 predicted benefit of adjuvant cisplatin-based chemotherapy. Therefore, MRP1 or MRP2 may not play a significant role in the drug resistance of completely resected NSCLC.

The MRP family (also known as C group of ATP-binding cassette transporters) currently consists of 13 structurally related ATP-binding cassette transporters that are involved in the transport of various substances (6, 7). Some of these proteins function as efflux pumps, thereby conferring resistance to anticancer drugs. The prognostic and/or predictive value of the different MRPs in malignant diseases is under investigation (6, 7).

In clinical NSCLC specimens, high levels of MRP1 are frequently observed but conflicting results has been published with regard to the prognostic value of MRP1 expression in NSCLC patients (12–19). Whereas MRP1 overexpression of the tumors predicts a worse outcome in two studies (12, 15), MRP1 expression was associated with longer overall survival in another report (19). In patients with advanced NSCLC treated with platinum-based chemotherapy, no significant relationship between MRP1 with response to chemotherapy or survival was observed (18). Interestingly, MRP2 expression was associated with overall survival but not with response to chemotherapy and progression-free survival, which is consistent with our results (18). Therefore, MRP2 expression may be of clinical relevance with regard to prognosis assessment (e.g., for molecular staging of lung cancer). However, before implementation for molecular staging of lung cancer patients, confirmatory data are required.

The IALT-Bio study was conducted according to a detailed protocol, which stressed the importance of collecting most of the tumor specimens within the participating centers, required a large sample size to ensure adequate power both for prognostic and predictive analyses, and specified a statistical plan of analysis. Adjusting on standard prognostic variables and specifying an objective cutoff point for defining positivity strengthen the reported results, notably about the prognostic role of MRP2 (20). For the predictive analyses, the estimated power for detecting a 20% absolute difference in 5-year survival was 66%. Thus, lack of significant predictive values of MRP1 and MRP2 does not exclude the possibility of a small predictive effect, which could only be detected by a much larger study population but would probably be without clinical relevance. Although it was not possible to collect tumor samples for 100% of the patients within centers, the 83% included did not differ from the 17% not included. Therefore, the advantages of conducting this study within a randomized trial preserved in particular patients allocated to chemotherapy are similar to patients in the control group with respect to known and unknown factors.

In the IALT-Bio study, the excision repair cross-complementation group 1 enzyme was recently characterized as predictor of the benefit of adjuvant chemotherapy (8). NSCLC patients with excision repair cross-complementation group 1–negative tumors benefit from adjuvant cisplatin-based chemotherapy following complete tumor resection.

In conclusion, MRPs are frequently expressed in NSCLC patients. MRP2 expression is an independent prognostic factor in patients with completely resected NSCLC but neither MRP1 nor MRP2 was of predictive value in patients enrolled into the IALT.

Austria: R. Pirker, Department of Medicine I, Vienna; G. Dekan, Institute of Clinical Pathology, Vienna.

Belgium: J. Vansteenkiste, University Hospital, Leuven.

Brazil: I. Sathler Pinel, Instituto Nacional de Cancer, Rio de Janeiro; R. Younes, Hospital A.C. Camarco, Sao Paulo.

France: A.A. Kanoui, Centre Physiothérapie du Rouget, Sarcelles; R. Dachez, Laboratoire L.C.L., Paris; S. Deslignères, Hospital Delafontaine, Saint-Denis; O. Languille-Mimoune, Cabinet Pathologie, Paris; P. Sabatier, Centre Hospitalier Victor Dupouy, Argenteuil; T. Le Chevalier, Institut Gustave-Roussy, Villejuif; M. Antoine, Hôpital Tenon, Paris; P. Boz, Cabinet de Pathologie, Papeete; P. Bruneval, Association Promotion Anatomie Pathologique, Paris; M.C. Charpentier, Cabinet Pathologie Tolbiac, Paris; B. Chetaille, Hôpital Sainte Marguerite, Marseille; E. Dulmet, Centre Chirurgical Marie-Lannelongue, Le Plessis Robinson; F. Capron, Groupe Hospitalier Pitié-Salpétrière, Paris; B. Gosselin, C.H.U., Lille; D. Grunenwald, P. Validire, Institut Mutualiste Montsouris, Paris; F. Labrousse, C.H.U., Limoges; N. Pericoli, Roma (Italy); D. Petrot, Cabinet d'Anatomie Pathologique, Niort; N. Rouyer, Cabinet de Pathologie Butet-Rouyer, Nice; B. Milleron, M. Antoine, Hôpital Tenon, Paris; J.F. Morère, M.A. Kambouchner, Hôpital Avicenne, Bobigny; G. Ozenne, Ceditrac-CMC du Cèdre, Bois Guillaume; T. Ducastelle, Laboratoire d'Anatomie et Cytologie, Rouen; E. Quoix, Hôpital Lyautey, Strasbourg; P. Durand de Grossouvre, Laboratoire d'Anatomie Pathologique, Haguenau; B. Gasser, C.H.U., Strasbourg; A. Rivière, Centre François Baclesse, Caen; F. Galateau-Salle, CHU, Caen; C. Tuchais, P. Jallet, G. Bertrand, I. Valo, Centre Paul Papin, Angers.

Germany: W. Eberhardt, University Hospital, Essen; D. Theegarten, Institute of Pathology, Ruhr-University Bochum, Bochum.

Greece: P. Christaki, Papanikolaou General Hospital, Pylea; T. Dosios, V. Kyriakou, Athens University School of Medicine, Athens; E. Papadakis, P. Agelidou, Sotiria Hospital, Athens; K. Zarogoulidis, University Hospital, Thessaloniki.

Italy: A. Masotti, Azienda Ospedaliera Di Verona, Verona.

Lithuania: A. Jackevicius, Institute of Oncology Vilnius University, Vilnius.

Poland: J. Laudanski, L.Chyczewski, M. Kozlowski, J. Niklinski, Medical School, Bialystok. T. Grodski, J. Pankowski, Regional Hospital for Lung Diseases, Szczecin; T. Orlowski, M. Chabowski, R. Langfort, Institute of Tuberculosis and Lung Disease, Warsaw; B. Muszczynska-Bernhard, Dolnoslaskiego Centrum Chorob Pluc, Wroclaw.

Romania: T. Ciuleanu, Oncological Institute “Ion Chiricuta,” Cluj-Napoca.

Slovakia: J. Baumohl, University Teach. Hospital, Kosice.

Spain: F. Cardenal, Hospital Duran I Reynals, Barcelona; R. Bernat, Hospital de Bellvitge, Barcelona; J. Salinas, J.B. Lopez, Hospital Virgen de Arrixaca, El Palmar Murcia.

Sweden: B. Bergman, A. Hussein, Sahlgrenska Hospital, Göteborg.

Yugoslavia: G. Radosavljevic, Institute for Lung Disease, Belgrade.