Originally published as JCO Early Release 10.1200/JCO.2008.19.0322 on October 6 2008

This Article Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Rights & Permissions Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Pereira, B. Articles by Ychou, M. Search for Related Content PubMed PubMed Citation Articles by Pereira, B. Articles by Ychou, M. Related Articles Related Reply Related Article Social Bookmarking                            What's this?

Neoadjuvant Chemoradiotherapy in Esophageal Cancer: Is It Still the Question?

Bruno Pereira, Sophie Gourgou-Bourgade, David Azria, Marc Ychou

Centre Régional de Lutte contre le Cancer Val d'Aurelle-Paul Lamarque, Montpellier, France

To the Editor:

In the March 1st issue of this Journal, Tepper et al published a study showing that the results from this phase III trial indicate a benefit in overall survival (OS) and progression-free survival (PFS) for patients with esophageal cancer treated with cisplatin, fluorouracil, radiotherapy, and surgery as compared with surgery alone. Recruitment was stopped because of a low rate of accrual, and a statistical analysis was performed on 56 patients after a median follow-up of 6 years. The investigators based their conclusions from the results of the exact stratified log rank test for small sample sizes after observing only 39 events, which is 10% of the initially planned number of events.

Conclusions based on the log-rank test do not sufficiently address the significance of the result, especially when the sample size is much lower than that of the initial design. Clinicians need to be able to answer the following question: "Can we really consider trimodality therapy superior to surgery alone in this setting with an analysis performed on only 56 patients rather than the 475 originally planned ?" Although we agree that this trial does provide important information on the probable superiority of the trimodality regimen, the authors should have studied more thoroughly the risk of error associated with this early analysis.

On the basis of their study design, with 475 planned patients and 371 events, the investigators expected to detect an overall survival hazard ratio (HR) of 1.4 with 90% power and a two-sided = .05. Formal interim analyses were planned for OS at 50% and 75% of the expected number of events.

The estimated 95% CI for the HR was 1.46 to 5.69, which does not contain the value of 1.4. Also, the estimated lower 95% CI for median OS was 2.40 for the trimodality arm, whereas the estimated upper 95% CI for median OS in the surgery-alone arm was 2.59, showing that these CIs overlap, suggesting a nonsignificant difference. The 5-year 95% CI for OS also overlaps. So there are sufficient arguments to question the statistically significant difference observed with a P value of .002.

Because the statistical methods used a flexible spending function approach for planning interim analyses, it is possible to adjust the value of for the formal statistical comparison. The statistical result of P = .002 thus cannot be compared with the classical significance level of .05.

At the planning stage for the first interim analysis at 187 events (50%), the value of needed to reject the null hypothesis is .0031. In the published analysis, the authors indicate the superiority of trimodality with 39 events and P = .002. Had we performed a formal statistical analysis at the 10% information fraction, the value of would have been .000001. Because the P value of .002 is greater than this cutoff value, the conclusion of a statistically significant difference is not warranted. At a P value of .002, a 46% information fraction is needed to conclude a statistically significant difference. Under the alternative hypothesis HR = 1.4 and 39 events, the post hoc power of the trial is less than 20%. Simulations indicate that similar results were observed in terms of the and β errors.

We agree that the results of this study are interesting and should have been published, but we do not agree that "the results indicate a benefit in OS and PFS with trimodality." Even with an expected HR of 2.5, which should not be too far from the HR estimated from the data in this study, an analysis performed after 53 events would have had the desirable statistical properties, so an analysis after 39 events is still underpowered. Changing statistical hypotheses during the course of the trial, however, is always open to criticism, but this was not undertaken here because the trial was closed to accrual in 2000 and there is a long follow-up period. Interim analyses are increasingly being planned for clinical phase III trials, but as can be seen from the calculations in this trial, it is nearly impossible to reject the null hypothesis after such a small information fraction. In the particular situation of the Tepper et al study, had the results been "negative," the probable conclusion would have been to say that the study was underpowered. In the context of a "positive" underpowered result, convincing arguments for changing clinical practice will need to be strong to overcome the impact of statistical reasoning based on the originally planned experimental designs. Given these concerns, the place of neoadjuvant chemoradiotherapy in operable esophageal carcinoma is still unresolved.

AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.

NOTES

published online ahead of print at www.jco.org on October 6, 2008

REFERENCE

1. Tepper J, Krasna MJ, Niedzwiecki D, et al: Phase III trial of trimodalidy therapy with cisplatin, fluorouracil, radiotherapy, and surgery compared with surgery alone for esophageal cancer: CALGB 9781. J Clin Oncol 26:1086-1092, 2008[Abstract/Free Full Text]

CiteULike    Complore    Connotea    Del.icio.us    Digg    Facebook    Reddit    Technorati    Twitter    What's this?

Related Reply

• In Reply
  Joel E. Tepper and Donna Niedzwiecki
  JCO 2008 26: 5134 [Full Text]

Related Article

• Phase III Trial of Trimodality Therapy With Cisplatin, Fluorouracil, Radiotherapy, and Surgery Compared With Surgery Alone for Esophageal Cancer: CALGB 9781
  Joel Tepper, Mark J. Krasna, Donna Niedzwiecki, Donna Hollis, Carolyn E. Reed, Richard Goldberg, Krystyna Kiel, Christopher Willett, David Sugarbaker, and Robert Mayer
  JCO 2008 26: 1086-1092 [Abstract] [Full Text]

This Article Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Rights & Permissions Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Pereira, B. Articles by Ychou, M. Search for Related Content PubMed PubMed Citation Articles by Pereira, B. Articles by Ychou, M. Related Articles Related Reply Related Article Social Bookmarking                            What's this?