LETTERS TO THE EDITOR

To the Editor:

We read with great interest the two recent publications by Neoptolemos and colleagues on the ESPAC-1 trial of adjuvant therapy for resectable adenocarcinoma of the pancreas. ^1,2 We congratulate the authors on completing a randomized trial for a disease that has proven difficult to study. We agree with the authors that patients who undergo an incomplete resection resulting in a positive margin are at increased risk for disease-specific mortality. We are concerned, however, over the conclusion of the manuscript published in this journal–that an R1 resection (microscopically positive margins) occurs exclusively because of the underlying biology of the tumor (namely, “a biologically more aggressive cancer”) and is independent of patient selection and surgical technique. Concerns over how this conclusion was reached include the following:

1. The authors need to clearly state whether they did or did not have a standardized system of pathologic assessment of surgical specimens, and if they did, how they monitored for quality control. This information is not present in either manuscript. The data presented in Table 3 suggest that the pathologic evaluation of surgical specimens may have been incomplete. Only 77 (76%) of 101 margin-positive patients and 333 (76%) of 440 margin-negative patients were included in the multivariate analysis. Although not stated, it is implied that the remaining patients (nearly 25%) were excluded from analysis, presumably because pathologic data were not evaluable or incomplete.

2. Among the 101 patients with positive resection margins, only 48 were randomized in the 2 × 2 factorial design. Interpretation of the analysis of the other 53 patients is limited because nonrandomized treatments were not standardized. Furthermore, information on the treatment received by these patients was not provided.

3. The 25th percentile of follow-up in the 227 living patients is given as 1 month (interquartile range, 1–25 months). Thus, 56 (25%) of 227 patients had no more than 1 month of follow-up. This seems high and suggests a lack of maturity of follow-up.

4. It is not clear why the authors performed separate analyses in margin-positive (R1) and margin-negative (R0) patients. In both Tables 2 and 3, the effects of clinical and tumor characteristics on survival are presented separately for the two groups defined by margin status. Thus, there is interest in assessing whether the effects of the clinical and tumor characteristics vary according to margin status. The statistically appropriate analysis for addressing this question is to assess the significance of an interaction (or product) between margin status and each study factor in a model including both margin-positive and margin-negative patients (this is equivalent to the test of heterogeneity reported for the treatment effects).

5. In Table 3, an initial variable selection was performed to identify significant “independent” predictors of survival duration. However, more appropriate would have been an analysis that adjusts the apparent effect of margin status on survival for any confounding from all study factors. This requires including all variables in the analysis regardless of their significance. Both the univariate (unadjusted) and multivariate (adjusted) hazard ratios for the effect of resection margin status on survival should be given along with 95% confidence intervals.

The authors’ conclusion that a microscopically positive margin following pancreaticoduodenectomy reflects tumor biology (and is independent of preoperative staging or surgical technique) is not as secure as suggested in their manuscripts. ^1,2 It is important to acknowledge the limitations of their study that may impact their analysis and conclusions. These may include the lack of standardized pathology assessment, the lack of protocol-mandated pathology quality control, limitations imposed on the analysis by incomplete data, and technical aspects (outlined above) of the statistical analysis itself.

In patients with adenocarcinoma of the pancreatic head or uncinate process, the most common site of margin positivity following pancreaticoduodenectomy is the retroperitoneal (also termed mesenteric or uncinate) margin adjacent to the proximal 3 to 4 cm of the SMA. A positive retroperitoneal margin may result from: 1) poor patient selection for pancreaticoduodenectomy–for example, surgical resection in patients with tumors that extend to the SMA, usually a result of poor-quality preoperative imaging; 2) failure of the surgeon to separate the specimen from the retroperitoneum in the immediate periadventitial plane of the SMA, an error that can be prevented with proper surgical technique; or 3) the infiltrative nature of pancreatic adenocarcinoma. The latter factor is related to the underlying biology of pancreatic cancer. However, patient selection and surgical technique are under the immediate control of the surgeon and probably have little to do with subtle variations in the biologic behavior of individual tumors.

Douglas B. Evans, MD
Kenneth R. Hess, PhD
Peter W.T. Pisters, MD