Besse and Le Chevalier provide us with an excellent and comprehensive review of available literature summarizing the current state of the art relating to both adjuvant and neoadjuvant (or induction) chemotherapy in resectable non–small-cell lung cancer (NSCLC). They review the efficacy of adjuvant and neoadjuvant treatment in randomized clinical trials (RCTs) that employ cisplatin-based chemotherapy in stages I, II, and IIIA NSCLC, as well as meta-analyses of these RCTs. They present what they believe to be the pros and cons of each approach. Ultimately, the authors are unable to answer the question they are asking—whether adjuvant or neoadjuvant chemotherapy provides the greater benefit in this setting. This is not a criticism of their manuscript, because such an answer is not available on the basis of existing data. Moreover, it may not represent the key question that needs to be addressed at this time.
‘Milestone’ Studies Reconsidered
In a section entitled “Adjuvant Chemotherapy Milestones,” the authors review seven cisplatin-based RCTs addressing the value of adjuvant chemotherapy in various stages of resectable NSCLC.[2-8] These include three highly influential RCTs that reported positive overall results among populations randomized to adjuvant chemotherapy.[4-6] However, in each of these seminal trials only certain subgroups appeared to benefit, while other “milestone” studies were quite negative.[2,3,8] For this reason, the authors depend heavily on recent meta-analyses to justify their conclusions regarding the efficacy of adjuvant chemotherapy.
In the next section, they describe the studies they consider to be “Induction Chemotherapy Milestones.” In my opinion, the results they summarize in this section provide little evidence for efficacy of neoadjuvant chemotherapy.
The first two neoadjuvant RCTs they describe are the Barcelona study (reported by Rosell et al) and the M.D. Anderson study (reported by Roth et al), both of which deal exclusively with stage IIIA NSCLC. They indicate that these studies had “a significant impact on medical practice due to their impressive results.” This conclusion is accurate if it is based exclusively upon early reports of these trials.[9,10] Unfortunately, these “impressive results” did not hold up when mature and long-term data were later reported.[11,12]
In the Barcelona study, patients were randomized to either three cycles of neoadjuvant chemotherapy with cisplatin, mitomycin, and ifosfamide(Drug information on ifosfamide) followed by resection and postoperative radiation therapy, or resection followed by radiation therapy alone. Induction chemotherapy produced an overall response rate of 60% (including 7% complete responses). Resection rates were similar and very high—85% and 90% in the chemotherapy and surgical arms, respectively. In the initial report (at which time the median follow-up was 2 years), the investigators noted an impressive threefold improvement in median survival for those randomized to chemotherapy (26 vs 8 months; P < .001).
In the M.D. Anderson study, patients were randomized to either neoadjuvant chemotherapy with three cycles of cisplatin, etoposide(Drug information on etoposide), and cyclophosphamide(Drug information on cyclophosphamide) followed by resection, or to resection alone. The response rate for chemotherapy was not impressive (35% overall response rate, including 4% complete responses), and successful resections were achieved in 61% and 66% of the chemotherapy and surgery groups, respectively. What appeared to be remarkable at the time of the initial report (when median follow-up was approximately 3 years) was that median survival was almost sixfold greater among patients receiving chemotherapy than in those randomized to surgery alone. In the chemotherapy group, median survival was an astounding 64 months, compared to 11 months in the surgery-alone group (P < .008).
One potential problem with both studies—particularly the M.D. Anderson trial—is that the magnitude of survival differences was far greater than could be reasonably expected based on the modestly effective chemotherapy regimens employed. Moreover, at least in the Barcelona study, reports of molecular markers suggested imbalances in the randomization process, resulting in a higher fraction of more virulent tumors in the surgical group. For example, the surgery arm had a higher proportion of tumors with KRAS mutations (42% vs 15%), as well as a higher proportion of aneuploid tumors (70% vs. 29%). In the M.D. Anderson trial, molecular markers were not reported.
Interpreting the Long-Term Data
Most importantly, with longer observation times, these impressive early results did not hold up. Roth reported long-term results of the M.D. Anderson study, at which time median follow-up was almost 7 years. With prolonged follow-up, the median survival for those randomized to chemotherapy was only 21 months compared to 14 months in the surgical group. While the survival trend continued to favor the chemotherapy group, the survival difference was no longer statistically significant (P = .056).
Similarly, Rosell reported long-term results of the Barcelona trial when median follow-up was also approximately 7 years. Median survival was 22 months among those randomized to neoadjuvant chemotherapy, compared to 10 months for those randomized to surgery. While overall survival differences still favored the chemotherapy arm (P = .005, log-rank test), the surgery group did extremely poorly. Indeed, there was not a single long-term survivor in the surgical group.
Interestingly, the Barcelona group was unable to confirm its own results regarding the efficacy of their treatment program in a subsequent RCT. In this second trial, the control arm received a regimen that was identical to what had been administered in the original Barcelona study, whereas the experimental arm received a higher cisplatin dose than had been used earlier. Median survival for patients on the control arm was 11 months, compared to 22 months for patients receiving this identical regimen in the first Barcelona trial.
In addition to the Barcelona and M.D. Anderson trials, Besse and Le Chevalier list four other RCTs as “induction chemotherapy milestones.” These studies are much larger (including from 270 to 519 patients), and cumulatively involved a total of 1,498 patients.[15-18] Each included a variety of disease stages. Three trials included stages IB, II, and IIIA patients, while the MRC LU22/NVALT2/EORTC 08012 trial also included stage IA patients. All four of these large RCTs demonstrated negative results. Accordingly, the data from individual RCTs provide little evidence that neoadjuvant chemotherapy improves survival in early-stage NSCLC.
Limitations of Meta-analysis
Because the data from both adjuvant and neoadjuvant trials do not provide unequivocal evidence for efficacy of either strategy, Besse and Le Chevalier rely heavily on meta-analyses to justify their conclusions. This is true with regard to both adjuvant[19,20] and neoadjuvant[21,22] approaches to resectable NSCLC.
This reviewer has written previously about the limitations of a meta-analytic approach, and the reader is referred to other publications for more detailed descriptions of the problems of meta-analysis.[23,24] Nevertheless, it is worth noting here that a fundamental problem with the meta-analytic approach is that it is a purely statistical technique that does not adequately address the problem of inconsistent results between individual trials. Moreover, it can provide no insight as to why different studies apparently support inconsistent results when addressing identical therapeutic questions. While meta-analysis was developed as a solution to the problem of insufficient statistical power, such studies have been increasingly used to reconcile divergent and often directly conflicting results from well-powered individual trials. This is clearly how meta-analyses have been used in the area of adjuvant and neoadjuvant chemotherapy in NSCLC, and the answers it provides are not definitive, given the inconsistent results of individual trials.
Lack of Evidence
Besse and Le Chevalier pose the question of whether adjuvant or neoadjuvant strategies should be used in resectable NSCLC, but they are ultimately unable to provide a clear answer. While this reviewer agrees with Besse and Le Chevalier that a powerful rationale exists to justify the neoadjuvant approach, such considerations are not evidence-based, and are also not convincing. Indeed, the best evidence supporting the neoadjuvant approach in resectable patients with NSCLC is based on phase II rather than phase III studies.
In the treatment of breast cancer, several RCTs have compared adjuvant and neoadjuvant approaches. The best known individual RCT is the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-18 protocol reported a decade ago by Fisher et al. In general, these RCTs have shown that survival was not significantly different when adjuvant or neoadjuvant chemotherapy was used. A similar conclusion was reached with a meta-analysis of nine RCTs that compared adjuvant and neoadjuvant chemotherapy in breast cancer. This reviewer suspects that adjuvant and neoadjuvant therapy is likely to provide similar outcome in appropriately selected patients.
In lung cancer, two ongoing RCTs are directly comparing neoadjuvant and adjuvant chemotherapy. One is the NATCH (Neoadjuvant-Adjuvant Taxol Carboplatin(Drug information on carboplatin) Hope) trial, a three-arm study comparing neoadjuvant or adjuvant paclitaxel(Drug information on paclitaxel) and carboplatin to surgery alone in stage IB, II, and IIIA NSCLC. The second study, comparing neoadjuvant to adjuvant cisplatin and docetaxel(Drug information on docetaxel) (Taxotere) in patients with stage I, II, or IIIA lung cancer, is also ongoing. However, no results from either of these trials are currently available.
This reviewer agrees with Besse and Le Chevalier that the issue of adjuvant or neoadjuvant chemotherapy is probably not the key question that needs to be addressed with regard to the optimal management of resectable NSCLC. Instead, the most critical question is to identify those individuals who are most likely to benefit from either adjuvant or neoadjuvant chemotherapy.
In this regard, future adjuvant and neoadjuvant trials should be designed to incorporate advances in molecular biology. The objective is to identify individuals most likely to benefit from either adjuvant or neoadjuvant chemotherapy based on prognostic markers that indicate a high risk of recurrence as defined by molecular techniques. Similarly, an objective will be to test whether predictive markers are able to identify those specific therapies that are most likely to be effective in individual patients. Indeed, broadly inclusive RCTs will likely be used on a more limited basis in the future, as we increasingly seek to customize adjuvant or neoadjuvant therapy on the basis of tumor biology at the level of the individual patient.