Omission of Chemoradiation is Associated with Poor Survival in Medicare Patients with Resected Pancreas Cancer
Abstract & Commentary
Synopsis: Outcomes following resection of pancreas cancer have improved recently, more than can be accounted for by a drop in procedure-related mortality. This study from the Harvard School of Public Health performed a retrospective analysis of claims-based Medicare data and correlated it with SEER data to identify prognostic factors that may be contributing to this phenomenon.
Source: Lim JE, et al. Ann Surgery. 2003;237:74-85.
The NCI SEER program collects demographic and tumor registry data, including surgery and radiotherapy information, for areas of the country reflecting approximately 14% of the US population. Chemotherapy data are considered unreliable because outpatient chemotherapy can be missed. Lim and colleagues from Harvard Medical School performed a retrospective cohort study on patients treated from 1991-1996 by merging SEER files with Medicare claims data in order to identify potential prognostic factors related to postresection survival in patients with pancreas cancer. Using ICD-9 and DRG codes, Lim et al correlated inpatient and outpatient files with physician and lab billing data from the Health Care Finance Administration (HCFA) database and identified 6 prognostic factors that were significantly related to survival in a multivariate analysis.
Medicare enrollees residing in 1 of the 11 SEER catchment areas who were treated surgically with curative intent were included in the study. HMO patients were excluded because of incomplete reporting to HCFA during the study period. Curative resections were either radical pancreaticoduodenectomy or a Whipple procedure. A total of 396 patients were identified, with a median age of 72 years. There were 196 women and 200 men. Connecticut contributed the most patients. There were 321 Caucasians, 29 African Americans, 20 Asians, and 26 unspecified. Among the patients studied, 185 (46.7%) received some form of adjuvant therapy, including 125 who received adjuvant chemoradiation, 49 who received RT alone, and 11 who received chemotherapy alone. Further details regarding adjuvant therapy doses and schedules were not provided. Patients who received adjuvant therapy were slightly younger (71.3 vs 73.3 years) and were more likely to have above-median income (59% vs 48%) than patients who were treated with surgery alone. They were also more likely to have had lymph nodal involvement. Factors assessed included: demographic details like age, gender, race, ethnicity, and socioeconomic status; perioperative issues such as type of resection (partial, etc), transfusions, teaching venue, and adjuvant therapy; and histopathologic findings such as tumor size, tumor grade, and TNM stage.
Median follow-up was 38.5 months (range, 0.2-44.8). Median tumor diameter was 3 cm. No regional variation in assignment of adjuvant treatment was identified, and the use of adjuvant therapies did not change over the course of the study period. More patients were resected in teaching facilities as the study period progressed, including 64% of all the patients. Median overall survival for the entire group was 17.6 months. One-year survival was 60.1% and 3-year survival was 34.3%. There was a statistically significant difference in median survival for those patients who received adjuvant therapy compared with those who did not. Post-chemoradiation median survival was 29 months compared with 12.5 months for no adjuvant treatment, 1-year survival was 81% vs 51%, and 3-year survival was 45% vs 30% (P = 0.0003).
Multivariate analysis revealed no significant difference in outcome based on Whipple vs non-Whipple resection, number of units transfused, or T-stage. Six variables were shown to be statistically significantly related to survival, including absence of adjuvant therapy (P = .0002), tumor diameter > 2 cm (P = .004), positive lymph nodes (P = .009), Grade 2 or 3 (P = .01), nonteaching venue (P = .01), and low SES (P = .02). Patients with 1-3 positive lymph nodes did as well as patients with negative lymph nodes.
Lim et al concluded that their findings were consistent with those of the Gastrointestinal Study Group randomized trial results published in 1985,1 but cautioned that certain aspects of their study may limit the accuracy of their conclusions. For example, 27% of patients with pancreas cancer are younger than 65 years, and were excluded from the study population. There was no assessment of surgical margin status, ploidy, or lymphvascular invasion, and there was no central pathology review. The advantage seen with surgery in a teaching center may have been a proxy for high-volume hospitals. The findings relating to SES may have reflected access to care issues, or possibly patient preference. While the most significant predictor of survival following resection was the administration of adjuvant therapy, further studies are needed to confirm these observations.
Comment by Edward J. Kaplan, MD
This study is interesting because it involved a reasonably large number of patients and covered a wide geographic area. The method used by Lim et al was innovative. Their conclusions were not surprising, except that the patients who received adjuvant therapy were more likely to be node positive, but still did better than those who received surgery alone. In their paper, Lim et al made reference to the 2 recently completed European randomized trials that looked at the effect of adjuvant chemoradiation on outcomes in resected pancreas cancer. The EORTC 40891 trial accrued patients from 1987-1995 and included 114 evaluable patients with pancreas cancer. This trial randomized between surgery alone vs surgery followed by 40 Gy split-course RT and 5-FU. Preliminary results were published in 1999 and showed no significant benefit with adjuvant chemoradiation.2 However, given the poor track record for split- course RT, the small sample size, and the fact that 20% of the patients assigned to adjuvant chemoradiation did not receive it, the early results of the EORTC trial are not convincing. The second trial cited was the ESPAC-1 trial from the European Study Group for Pancreatic Cancer. This trial ran from 1994-2000 and enrolled 541 patients. It randomized patients between adjuvant 5-FU vs 20 Gy + 5-FU vs observation and found no survival benefit for adjuvant chemoradiotherapy. According to the first published results from Neoptolemos, each of the 61 cancer centers relied on its own RT quality assurance standards to deliver the radiotherapy according to local practice. They concluded that the ESPAC-1 results "clear the way for focusing on chemotherapy as the principle adjuvant modality in pancreatic cancer . . . " Not surprisingly, an ASCO abstract from 2002 combining the GITSG, EORTC, and ESPAC-1 data along with data from a Norwegian trial into a meta-analysis likewise concluded that chemoradiation offered no survival benefit.4
I have to differ with the statement made in the ESPAC-1 report. I think their results clear the way for the RTOG 9704 randomized trial results, since it is the only modern randomized trial that used RT doses that would be considered efficacious (ie, 50.4 Gy with concomitant radiosensitizing 5-FU vs gemcitabine). This trial closed recently and was omitted in the discussion by Lim et al. I would tend to give much greater credence to its results than I would to those from the 2 European trials, both of which I consider to have been woefully deficient from a radiotherapy perspective.
Dr. Kaplan is Acting Chairman, Department of Radiation Oncology, Cleveland Clinic Florida, Ft. Lauderdale, FL and Medical Director, Boca Raton Radiation Therapy Regional Center, Deerfield Beach, FL.
References
1. GITSG. Arch Surg. 1985;120:899-903.
2. Klinkenbijl JH, et al. Ann Surg. 1999;230:776-784.
3. Neoptolemus JP, et al. Lancet. 2001;358:1576-1585.
4. Dunn JA, et al. Proc ASCO. 2002;abstract 564.
Outcomes following resection of pancreas cancer have improved recently, more than can be accounted for by a drop in procedure-related mortality. This study from the Harvard School of Public Health performed a retrospective analysis of claims-based Medicare data and correlated it with SEER data to identify prognostic factors that may be contributing to this phenomenon.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content