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 Google Scholar Articles by Yu, J. B. Articles by Knisely, J. P.S. PubMed PubMed Citation Articles by Yu, J. B. Articles by Knisely, J. P.S. Related Articles Related Reply Related Article

Impact of Immediate Postoperative Death on the Estimation of a Survival Benefit From Postoperative Radiation Therapy for Cancer of the Gallbladder

Department of Therapeutic Radiology, Yale School of Medicine, Department of Biostatistics, Yale School of Public Health, New Haven, CT

To the Editor:

We read with great interest the article by Dr Wang et al.¹ We commend the authors for investigating the National Cancer Institute Surveillance, Epidemiology, and End Results database (SEER) for this rare cancer, and for the sophisticated analysis brought to bear on the important question of the role of postoperative radiotherapy (PORT). However, we want to comment on the findings of the study as they relate to the limitations of the SEER database.

First, we are troubled by the broad range of surgeries investigated under the all-encompassing category of "surgery." The type of surgeries defined by the authors’ specified SEER codes included a wide range of disparate procedures, such as simple cholecystectomy and "debulking" with or without dissection of lymph nodes. Including such a range of procedures potentially dilutes the utility of the proposed prognostic tool. The significance of PORT is different for a patient after a simple cholecystectomy alone, versus a patient who underwent a radical resection with an extensive nodal dissection.

Second, postoperative pathologic margin status is not recorded in the SEER database. The resection margin status is critical to the success or failure of adjuvant therapy.² Are we to calculate the benefit from PORT in an identical manner for patients who have gross residual disease and those with an R0 resection?

Finally, and perhaps most concerning, is the large selection bias that can occur when attempting to analyze the association of PORT with mortality. Patients who die on the operative table, or die in the immediate postoperative period before they can reasonably receive PORT are automatically assigned to the non-PORT arm. This selection bias artificially lowers the survival of the non-PORT arm and improves by comparison the arm that receives PORT. This is especially clear in Wang et al Figure 1, where the two survival curves diverge at the origin. It is extremely unlikely that PORT contributes to immediate survival even before it is administered.

To further elucidate the significance of the patients who lived at least 4 months, we independently reanalyzed the SEER data. Using the author's published selection criteria, we found 5,419 patients. After excluding patients who had more than one malignancy, patients who received nonexternal beam radiation, and patients who received radiation in a nonpostoperative or unknown manner, we found 4,429 patients—a number closer to the published cohort. Without excluding patients with follow-up of at least 4 months, we calculated overall survival at 1, 3, and 5 years to be 47%, 26%, and 20%, respectively (data were similar to those of the authors). For these patients, we also found, in agreement with the authors, that PORT was associated with improved survival. However, if patients with at least 4 months of follow-up were excluded, patients who underwent PORT had a reduced survival (hazard ratio for death 1.31; 95% CI, 1.18 to 1.45; P < .001). We are not suggesting that PORT causes a reduction in survival in the prospective setting, but rather that patients selected for PORT retrospectively had an association with worse survival than those who underwent surgery alone.

The Cox proportional hazards model used by Wang et al assumes that there is a constant relative hazard (ie, a proportional hazard) of death over time.³ To test this assumption, we independently calculated the Schoenfeld residuals⁴ and found that if all patients are included in the Cox proportional hazards model (including those with at least 4 months of follow-up), we saw that the constant proportional hazards assumption was violated. Excluding patients with at least 4 months of follow-up caused this assumption to become valid.

Any future studies investigating the role of PORT in the SEER database must be explicit about the database's limitations. It is critical that patients who do not survive long enough to receive PORT be excluded from analysis. Certainly, the SEER database continues to be a wonderful resource for the investigation of clinical questions for which prospective data are not forthcoming. Furthermore, PORT should continue to be investigated in prospective and retrospective studies, and we are optimistic about its potential benefits for cancers of the gallbladder. However, we could not let these important points pass without comment. We encourage the authors to re-analyze their data, and to publish the exact criteria by which patients were selected and all resultant patient information. We must not let the fact that there are patients who die in the immediate postoperative period before receiving radiation trick us into thinking that PORT is beneficial for other patients simply because they lived long enough to receive it.

AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

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

REFERENCES

1. Wang SJ, Fuller CD, Kim JS, et al: Prediction model for estimating the survival benefit of adjuvant radiotherapy for gallbladder cancer. J Clin Oncol 26:2112-2117, 2008[Abstract/Free Full Text]

2. Ben-David MA, Griffith KA, Abu-Isa E, et al: External-beam radiotherapy for localized extrahepatic cholangiocarcinoma. Int J Radiat Oncol Biol Phys 66:772-779, 2006[Medline]

3. Cox D: Regression models and life tables. J R Stat Soc B 34:187-220, 1972

4. Schoenfeld D: Partial residuals for the proportional hazards regression model. Biometrika 69:239-241, 1982[Abstract/Free Full Text]

Related Reply

• In Reply
  Samuel J. Wang, C. David Fuller, Jong-Sung Kim, and Charles R. Thomas, Jr
  JCO 2008 26: 4524-4526 [Full Text]

Related Article

• Prediction Model for Estimating the Survival Benefit of Adjuvant Radiotherapy for Gallbladder Cancer
  Samuel J. Wang, C. David Fuller, Jong-Sung Kim, Dean F. Sittig, Charles R. Thomas, Jr, and Peter M. Ravdin
  JCO 2008 26: 2112-2117 [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 Google Scholar Articles by Yu, J. B. Articles by Knisely, J. P.S. PubMed PubMed Citation Articles by Yu, J. B. Articles by Knisely, J. P.S. Related Articles Related Reply Related Article