Outcome of gastric cancer patients after successful gastrectomy

Influence of the type of recurrence and histology on survival

Authors


Abstract

BACKGROUND.

The effect of the location of disease recurrence after curative (R0) gastrectomy on patient survival has not been elucidated. The authors hypothesized that the location of recurrence would have a significant influence on survival.

METHODS.

Medical records of all patients who received treatment for gastric cancer at The University of Texas M. D. Anderson Cancer Center between 1985 and 1998 were reviewed. Patients who underwent R0 resection for gastric cancer and subsequently developed localized (anastomotic) recurrence (LR), lymph node (regional) recurrence (NR), or distant metastases (DM) were analyzed for overall survival (OS). All study factors were entered into a Cox proportional hazards model to provide multivariate hazard ratios. The model was adjusted for the effects of primary site of recurrence, histologic grade, patient age, and location of the primary tumor.

RESULTS.

This retrospective analysis included 227 consecutive patients. The median survival of patients who developed NR (11 months) was similar to that of patients who developed LR (10 months), but both groups had significantly longer median survival compared with patients who developed DM (7 months; log-rank P = .03). Patients who had well differentiated or moderately differentiated tumors had a longer OS (11 months) than patients who had poorly differentiated tumors (8 months; log-rank P = .02). In this cohort, location of the primary cancer and age at recurrence had no significant impact on OS.

CONCLUSIONS.

The data from this study suggested that, among patients who undergo R0 gastrectomy for gastric cancer, LR and NR versus DM should be considered a valid stratification factor for randomized trials based on significant differences in survival. Determining whether this stratification should apply to histologic differentiation will require further investigation in a larger multicenter cohort. Cancer 2006. © 2006 American Cancer Society.

Gastric cancer often is advanced at the time of diagnosis; and, even when a curative (R0) resection, is possible, recurrences are common, occurring in approximately 60% of patients. In the setting of recurrence, gastric cancer rarely is curable. With the recent advances in combination chemotherapy and the incorporation of newer agents, such as taxanes, epirubicin, and irinotecan, a steady rise in response rates have been reported in Phase II trials,1–9 resulting in the launch of many randomized trials. Phase III trials for patients with advanced gastric cancer often include patients who have localized recurrences and/or distant metastases (DM). Many trials do not make a distinction between these 2 populations and, thus, do not tend to stratify patients accordingly. However, it is unclear whether differences in survival exist between patients who have predominantly localized (locoregional) recurrences (LR) and DM. We hypothesized that the clinical biology of cancer would be different if the recurrence is local (anastomotic; LR), a local lymph node recurrence (NR), or DM. It also seemed to be intuitive that patients who had LR and NR would have similar, more favorable outcomes compared with patients who had DM.

We chose to study this question in patients who underwent R0 resection for their gastric cancer and had a documented recurrence. In the same population, we also analyzed the influence of histologic differentiation, age, and location of the primary cancer on overall survival (OS).

MATERIALS AND METHODS

Patient Population

The current analyses include all patients with gastric cancer who received treatment at The University of Texas M. D. Anderson Cancer Center from January 1, 1985, to May 31, 1998. Each patient's medical record was entered into a confidential, retrospective database, and the histologic diagnosis of gastric adenocarcinoma was confirmed by the Department of Pathology. Other gastric tumors, such as neuroendocrine tumors, lymphoma, sarcoma, or adenocarcinoma in situ, were excluded from this analysis. This study was approved by the Institutional Review Board of The University of Texas M. D. Anderson Cancer Center. Patients with resected LR gastric cancer who subsequently developed LR, NR, or DM were included in this analysis.

Data Collection

Medical records were reviewed to obtain patient data, including age at the time of recurrence, self-identified ethnicity, location of the primary cancer (proximal vs mid-distal), and World Health Organization (WHO) gastric pathology classification. Histology was divided into low-grade and high-grade groups for comparison. After surgery, all patients were followed every 3 months for 1 year, then every 6 months for 2 additional years, then every year or until death. Computed tomography scans were obtained with every visit, and endoscopy was alternated with barium swallow every visit. The primary site of recurrence was determined by esophagogastroduodenoscopy, by computed tomography scanning (magnetic resonance imaging on occasion), or at surgery. A database was established and verified from the medical records by 1 oncologist (J.C.Y.). If a patient had local and metastatic recurrences, they were assigned to the metastatic category. Peritoneal or pelvic recurrences were categorized as metastatic.

The WHO gastric histologic classification was used to designate tumor grade. Gastric carcinomas were graded into 4 grades: well differentiated, with well developed tubular glands that mimicked the normal architecture of gastric glands; moderately differentiated, with glandular component and often with a cribriforming or acinar pattern but with architecture that was less well defined than the well differentiated tumors; poorly differentiated, with poor glandular formation often in small clumps or as isolated cells; and undifferentiated, with solid sheets of cohesive cells and without discernible glandular differentiation.

Statistical Analysis

The OS of patients was determined from the time of documented recurrence. The median OS was determined by using the Kaplan-Meier method. The log-rank test was used to compare cumulative OS between patients who developed LR, NR, or DM. A multivariate Cox proportional hazards model was constructed to assess the effects of primary site of recurrence, histologic grade, patient age, and location of primary tumor on survival. Adjusted hazard ratios with associated 95% confidence intervals (95% CI) were calculated. P values were 2-sided, and values <.05 were considered statistically significant. SPSS software (version 10.1; SPSS Inc., Chicago, IL) was used for all analyses.

RESULTS

In this retrospective study, 227 consecutive patients who had undergone R0 resection of primary gastric cancer and subsequently developed recurrence were included. Of these, 156 patients (69%) were men, and 71 patients (31%) were women. The first site of recurrence was an LR in 46 patients, an NR in 32 patients, and DM in 149 patients. Patient characteristics are summarized in Table 1 according to the site of first recurrence.

Table 1. Patient Characteristics by Site of Failure
CharacteristicLocal recurrence (N = 46)Lymph node recurrence (N = 32)Distant metastasis (N = 149)
No.%*No.%No.%
  • NA indicates not applicable; SD, standard deviation.

  • *

    Because percentages are rounded, totals may not add to 100%.

  • Grades 1 and 2 represent well differentiated or moderately differentiated tumors, and grades 3 and 4 represent poorly differentiated tumors.

Sex
 Men327028889664
 Women14304135336
Race
 Caucasian439320638859
 Non-Caucasian3712386040
 Unknown    11
Mean age (SD), y59 (14)NA55 (13)NA57 (14)NA
Location of primary tumor 
 Proximal204311344933
 Mid-distal245217538960
 Unknown24413117
Histologic grade
 Grades 1 and 29209284832
 Grades 3 and 4367820639765
 Unknown123943

Univariate Survival Analyses

The OS of patients with NR was similar to that of patients with LR and was longer than the OS of patients with DM (log-rank P = .03) (Fig. 1, Table 2). The median OS of patients with LR, NR, and DM was 10 months, 11 months, and 7 months, respectively. The 1-year survival rates of patients with LR, NR, and DM were 46%, 47%, and 28%, respectively.

Figure 1.

Survival by site of recurrence. Kaplan-Meier plot compares overall survival, defined as the time after first recurrence, among patients whose first site of recurrence was a local recurrence (LR), an LR lymph node recurrence (NR), or distant metastasis (DM).

Table 2. Univariate Analysis of Factors Affecting Overall Survival From the Time of Recurrence
VariableMedian survival (95% CI), moHR (95% CI)P*
  • 95% CI indicates 95% confidence interval; HR, hazard ratio; LR, local recurrence; NR, lymph node recurrence; DM, distant metastasis.

  • *

    Log-rank P values for site of recurrence, histologic grade, and location of primary analyses. The P value for age was obtained from a proportional hazards model.

  • Grades 1 and 2 represent well differentiated or moderately differentiated tumors, and grades 3 and 4 represent poorly differentiated tumors.

First site of recurrence  .34
 LR10.4 (4.9–15.8)1.0 (Reference) 
 NR11.5 (6.1–17.9)0.93 (0.56–1.55) 
 DM7.4 (6.4–8.3)1.45 (1.01) 
Histologic grade  .020
 Grades 1 and 210.7 (6.3–15.2)1.0 (Reference) 
 Grades 3 and 47.9 (6.2–9.5)1.48 (1.06–2.05) 
Age (as a continuous variable)NA1.01 (0.99–1.02).354
Location of primary tumor  .569
 Proximal7.0 (8.6–14.4)1.0 (Reference) 
 Mid-distal8.1 (6.4–9.7)1.09 (0.81–1.48) 

Patients with well or moderately differentiated cancer had a longer OS than patients with poorly differentiated cancer. The median OS was 11 months for patients with well or moderately differentiated cancer and 8 months for patients with poorly differentiated cancer (log-rank P = .02).

According to the Kaplan-Meier and Cox proportional hazard analyses, the location of the primary cancer had no effect on OS (Table 2). Similarly, the age of the patients at the time of recurrence had no significant impact on OS (hazard ratio, 1.01; 95% CI, 0.99–1.02).

Multivariate Survival Analyses

All study variables were included in a multivariate Cox proportional hazards model to adjust for the effects of covariates (Table 3). In that model, patient who had NR had similar OS compared with patients who had LR. Patients who had DM had a significantly decreased OS compared with patients who had LR or NR (P = .004).

Table 3. Multivariate Analysis of Factors Affecting Overall Survival from the Recurrence of Gastric Cancer
VariableHR (95% CI)P
  • HR indicates hazard ratio; 95% CI, 95% confidence interval; LR, local recurrence; NR, lymph node recurrence; DM, distant metastasis.

  • *

    Grades 1 and 2 represent well differentiated or moderately differentiated tumors, and grades 3 and 4 represent poorly differentiated tumors.

Site of recurrence .003
 LR1.0 (Reference) 
 NR1.0 (0.55–1.79) 
 DM1.81 (1.22–2.68) 
Histologic grade* .004
 Grades 1 and 21.0 (Reference) 
 Grades 3 and 41.71 (1.19–2.46) 
Age (as a continuous variable)1.01 (0.99–1.02).324
Location of primary tumor .238
 Proximal1.0 (Reference) 
 Mid-distal0.82 (0.60–1.14) 

Patients who had poorly differentiated tumors also had a significantly decreased OS compared with patients who had well or moderately differentiated tumors (P = .003). Finally, location of the primary cancer and patient age at the time of recurrence had no significant impact on survival.

DISCUSSION

The prognosis for patients with gastric cancer continues to be dismal. The best outcome can be expected in patients who have a localized cancer and undergo R0 resection. In the West, patients who undergo R0 resection and receive postoperative chemoradiation may be expected to have a 5-year survival rate of 40%. Thus, most patients are expected to die of their disease. Once the recurrence is documented, the patient rarely is curable.

However, specifically for the purposes of patient entry into Phase III trials, no distinction is made based on the geographic location of recurrence. This may not be appropriate. We believe that the clinical biology certainly is different for a patient who has LR cancer than for a patient who has DM to the liver or lung. To clarify any distinctions that may exist, we chose a clean population of patients who had undergone R0 surgical resection of their gastric cancer and who had objectively documented recurrences. This also was an opportunity to collect many other data, such as the influence of histologic differentiation, location of the primary tumor, and age at the time of recurrence. If we could detect substantive differences, then we would be in a position to propose novel stratification factors for future Phase III trials.

Our data suggest that the clinical biology of patients with LR and NR is similar, and these 2 groups fared better than patients with DM in terms of survival. Although we observed differences in survival based on the type of histologic differentiation of the cancer, we are concerned that it would not be prudent to recommend this as a potential stratification factor without further multiinstitutional investigations. One cause for this reservation is the degree of subjectivity involved in the designation of the histologic grade in different institutions. If we embrace histologic grading as an important stratification factor, then it would require a central review by multiple pathologists. This may put additional burden on the infrastructure and remains a subject of further discussion, although the impact of histologic grade on the survival of patients with gastric cancer also has been described previously in other large databases.10–12 Conversely, stratification based on the location of recurrence may be a valid stratification scheme, because it can be documented more objectively.

In conclusion, the type of recurrence after gastric surgery has a significant influence on patient survival and should be considered a valid stratification factor for future Phase III trials. In addition, efforts to reduce recurrence after surgery must continue. These may include further refinements and improvements in preoperative therapy and surgical techniques, and postoperative adjuvant therapy also may alter the patterns of recurrence.13–25

Acknowledgements

We thank to A. Najam, MD and G. Giacco, MS for their contribution to the database and analyses

Ancillary