Preoperative chemoradiation for localized gastric cancer can modify baseline stage, as determined by surgical pathology stage. Therefore, the authors hypothesized that surgical pathology stage would be a better prognosticator of overall survival (OS) than baseline stage.
Patient populations were combined from 2 prospectively conducted, preoperative chemoradiation trials that used the same therapeutic strategy. Patients must have had localized gastric adenocarcinoma and were staged extensively, including endoscopic ultrasonography and laparoscopy. Patients had to be fit for surgery medically with a technically resectable cancer. All patients provided written informed consent. Patients first received induction chemotherapy for up to 2 months followed by chemoradiation (45 grays) and an attempted surgery. OS was correlated with pretreatment and posttreatment parameters, including surgical pathology stage according to American Joint Commission on Cancer criteria.
Of 74 patients who were registered, 69 patients (93%) had undergone surgery. Nineteen patients (26%) had a pathologic complete response (pathCR), and 55 patients (81%) had a curative (R0) resection. None of the pretreatment parameters correlated with OS; however, longer OS correlated with lower pathologic stage (P < .0001), R0 resection (P < .001), clinical response noted prior to surgery (P = .002), pathCR (P = .004), lower pathologic lymph node classification (P = .006), and lower pathologic tumor classification (P = .03). Pathologic stage and R0 resection were independent prognostic factors for OS (multivariate Cox model; both P = .05).
In 2005, the number of newly diagnosed gastric cancers in the United States was estimated to be 22,000, and the estimated deaths from gastric cancer were 11,500.1 Although the incidence of gastric cancers is decreasing, it rarely is detected early, and the prognosis remains poor. In approximately 50% of newly diagnosed patients, the carcinoma is advanced beyond its original local-regional boundaries.2 Surgery remains the primary therapy for localized gastric cancer,3–7 and a resection can result in a cure (R0 resection). However, the rate of achieving R0 resection is < 50% among patients who undergo surgery as their primary therapy.3 Wanebo et al.,8 based on data from the 1980s, reported an R0 resection rate of only 23% among the 13,295 patients they studied, but their data may not be representative of the current state of staging and patient selection.
In a multivariate analysis of prognostic factors for patients with locally advanced gastric carcinoma, Roder et al.9 identified distant metastases and incomplete resection as the major predictors of poor survival. Local-regional recurrence also can be a significant problem, reaching a rate as high as 54% after surgical resection in some series.5–7, 10, 11
Currently, the best predictor of patient outcome when surgery is the primary therapy in terms of overall survival (OS) and disease-free-survival (DFS) is the surgical pathologic stage, as determined by the American Joint Commission on Cancer (AJCC) criteria.12 However, for patients with localized gastric carcinoma, the strategy of preoperative therapy is becoming increasingly popular.13–15 When preoperative therapy is used, there is a distinct possibility that the true pathologic stage may be altered in some patients. This is because of considerable heterogeneity in tumor biology, even for cancers within the same stage. Downstaging may be possible in cancers that inherently are sensitive to chemotherapy and/or chemoradiation. In this circumstance, it is not clear whether OS can be predicted by using the baseline clinical stage or the potentially modified surgical pathology stage, as assessed according to AJCC criteria. We hypothesized that the posttreatment (surgical pathology) AJCC stage, and not the baseline clinical stage, would be the best prognosticator of OS and DFS. To test this hypothesis, we examined OS and DFS in relation to various pretreatment and posttreatment parameters (including surgical pathology stage) by using data from 2 prospectively conducted, preoperative chemoradiation trials.13, 14
MATERIALS AND METHODS
The patient populations in this analysis were from 2 prospective, preoperative chemoradiation therapy trials.13, 14 Both protocols were similar in terms of patient selection and therapeutic strategy, but the chemotherapy combination administered prior to chemoradiation and during chemoradiation varied. This slight variation in chemotherapy in the 2 trials was not considered fundamentally significant for the purposes of the current analysis.
Patient Selection and Evaluation
Patients with localized, histologically confirmed gastric or gastroesophageal adenocarcinoma were eligible. The bulk of cancer had to be in the stomach, although the gastroesophageal junction may have been involved. Part of the staging work-up for each patient included a chest radiograph, computed tomography (CT) scan of the abdomen (pelvis and chest were scanned if needed), upper gastrointestinal barium radiographs, esophagogastroduodenoscopy with endoscopic ultrasonography (EUS), electrocardiogram, SMA-12 serum chemistry panel, complete blood count, and measurement of electrolyte and carcinoembryonic antigen levels. Patients with T2 or T3 carcinoma with (N+) or without lymph node involvement (N0) and patients with T1N1 carcinoma were eligible. T classification was determined by EUS. Prior to starting any therapy and prior to registration on 1 of the 2 trials, laparoscopic staging and J-tube placement were performed in all patients. Nutritional counseling was provided as needed. A multidisciplinary evaluation was performed before patients participated in these studies. All patients signed an Institutional Review Board-approved written informed consent form.
Ineligible for inclusion in this study were patients who had T4, M1, or T1N0 carcinoma; peritoneal carcinomatosis; or uncontrolled medical conditions (e.g., diabetes, New York Heart Association Class III or IV hypertension, or psychiatric illness). Patients who could not comprehend or comply with the study requirements also were ineligible.
Design of Clinical Trials
Patients were to receive up to 2 cycles of induction chemotherapy and then chemoradiotherapy. The 2nd cycle of induction chemotherapy was given if the cancer did not progress and was skipped if the cancer did progress. Patients were taken off protocol if distant metastases developed. Clinical response was judged by using radiographic or endoscopic techniques.
If a patient underwent an R0 resection, then no further therapy was given. In patients who underwent R1 resection (microscopically positive margin), R2 resection (only partial resection), or had M1 carcinoma discovered, palliative care was given.
Step 1: Chemotherapy
In both trials,13, 14 patients received 5-flurouracil (5-FU)/ cisplatin-based chemotherapy. In 1 trial,13 patients received 5-FU, cisplatin, and leucovorin; and, in the 2nd trial,14 patients received 5-FU, cisplatin, and paclitaxel. If the cancer had not progressed, then the 2nd cycle was to be repeated 28 days after the first cycle began. Complete blood counts were obtained weekly, and serum chemistries were monitored before each course.
Step 2: Chemoradiotherapy
In both trials,13, 14 patients received low-dose, continuous-infusion 5-FU with radiation therapy; however, in the 2nd trial,14 patients also received a weekly intravenous bolus of paclitaxel. The interval between Day 1 of the last induction chemotherapy cycle and the first day of chemoradiotherapy was ≥ 28 days. Radiation fields included the entire stomach, perigastric extension (if present), and draining lymph nodes (gastric, celiac, porta hepatis, gastroduodenal, splenic-suprapancreatic, and retropancreaticoduodenal). For lesions that involved the cardia or the gastroesophageal junction, a 5-cm margin of esophagus was included; and, for distal lesions at or near the gastroduodenal junction, a 5-cm margin of duodenum was included. Esophagoscopy, barium swallow radiographs, and CT scans of the chest and abdomen were used to determine the maximum extent of disease relative to the primary tumor and the lymph node groups. Idealized fields were modified as needed to shield at least ≥ 66% of 1 kidney. For proximal lesions, cardiac shielding was recommended along with evaluation of lateral fields for a component of treatment.
Linear accelerators delivered a dose of 45 grays (Gy) (25 fractions of 1.8 Gy) over 5 weeks using either 15-MV or 18-MV photons and a 3-dimensional conformal radiotherapy technique. Sometimes, anterior and posterior fields were angled slightly to avoid the spinal cord and the right kidney.
Step 3: Surgery
The type of surgery performed depended on the location and extent of the primary cancer. The cancer was resected along with a luminal gastric margin of ≥ 5 cm when feasible. Also when feasible, a 2-cm duodenal margin was obtained for distal cancers, and a 3-cm esophageal margin was obtained for proximal cancers. In both cases, frozen-section confirmation of a negative margin was sought. For distal cancers, a subtotal gastrectomy was considered adequate; total gastrectomy was at the discretion of the surgeon. For proximal cancers, total gastrectomy or esophagogastrectomy was performed. En bloc resection of adjacent organs was performed when their involvement was suspected. The spleen was preserved when possible. An attempt was made to perform a spleen-preserving D2-type lymph node dissection. During surgery, the J-tube was revised and replaced for temporary postoperative nutritional support in all patients.
Upper gastrointestinal barium radiographs were taken after each 28-day cycle of induction chemotherapy and just before surgery; and esophagogastroduodenoscopy, CT scan of the abdomen, chest radiography, and all blood tests were repeated before surgery. Previously described criteria for response evaluation were used.13, 14, 16 Briefly, a pathologic complete response (pathCR) was defined as the absence of carcinoma cells in the primary site (including in the examined lymph nodes), and a pathologic partial response (pathPR) was defined as < 10% residual carcinoma cells in the specimen.16
Baseline staging was performed by using EUS and radiologic imaging studies. Endoscopy and radiologic imaging studies were repeated a few days prior to surgery. Finally, posttreatment pathologic stage was determined by examining the resected specimens.
Each patient was assessed at 3 months, 6 months, and 12 months and then every 6 months for 5 years or until the patient died.
The current analyses focused on patients who underwent surgery after preoperative chemoradiation to test the primary hypothesis that OS (and DFS) would be predicted better by surgical pathology stage, as designated by the AJCC criteria, rather than by the baseline clinical stage. Using the data from 2 clinical trials, 2 sets of analyses were performed. First, Kaplan–Meier curves were generated for OS and DFS. Differences in survival distributions with respect to various pretreatment and posttreatment parameters (including surgical pathology stage) were examined for OS and DFS by using log-rank tests and Cox regression. Next, chi-square tests and Fisher exact tests were used to assess relations between different factors and downstaging or pathologic response. For those tests, OS was defined as the time from the start of induction chemotherapy to the date of death or the date of last follow-up. DFS was defined as the time from the start of induction chemotherapy to recurrence of the cancer. The statistical analyses are detailed below.
Data from living patients were censored. The log-rank test was used to test for differences in survival distributions by gender; pathologic response; baseline tumor (T), lymph node (N), and EUS stages; postsurgical T, N, and pathologic stages; and T and N downstaging. Kaplan–Meier survival curves were plotted for patients stratified by resection (R0 vs. non-R0), baseline and pathologic stage (Stages I and II vs. Stages III and IV), and pathologic response (pathCR vs. nonpathCR). All statistical tests were 2-sided, and significance was set at the .05 level.
All surgical reports and necessary specimens were rereviewed for this analysis by 1 pathologist (T.-T.W.). Pathologic response (pathCR vs. pathPR) and the incidence of T and N downstaging were cross-tabulated by age; gender; baseline T, N, and EUS stages; and postsurgical T, N, and pathologic stages. Baseline T classification data were grouped into 2 categories: T1 or T2 vs. T3. Chi-square tests and Fisher exact tests were used to assess the correlation of each of these factors with pathologic response and the incidence of N and T downstaging.
Data were analyzed from 74 patients who were enrolled in the 2 studies. The median age was 59 years (range, 28–80 years), and all patients had a Karnofsky performance scale score of 1. Other baseline patient characteristics are shown in Table 1. Most patients were men (n = 49 patients; 66%). Seventy-two of 74 patients (97%) underwent baseline EUS examinations, and 2 patients did not have EUS examinations because of technical issues. The tumors were poorly differentiated in most patients (n = 60 patients; 82%) and proximal in most patients (n = 59 patients; 80%). Most patients had baseline Stage III EUS (n = 39 patients; 54%), T3 tumor (n = 64 patients; 89%), and/or N1 disease (n = 43 patients; 60%) (Table 1).
Table 1. Patient Characteristics at Baseline (n = 74 Patients)
No. of patients
EUS indicates endoscopic ultrasonography.
For technical reasons, 2 patients did not have EUS staging.
Sixty-nine of 74 patients (93%) underwent surgical exploration. Of the 69 patients who underwent surgery, 55 patients (81%) had a complete R0 resection and 19 patients (28%) achieved a pathCR. In addition, 14 patients (20%) had some response (< pathCR), and 37 patients (54%) had no evidence of a pathologic response. Of the 55 patients who underwent an R0 resection, 42 patients (76%) had T1 or T2 tumors, and 13 patients (24%) had T3 tumors. Thirty-nine of 55 patients (71%) had no lymph node metastases (N0), 14 patients (25%) had N1 disease, and 2 patients (4%) had N2 disease. The degree of response to therapy was similar in both studies.
Postsurgical pathologic stage was determined for all 69 patients who underwent surgery. Nineteen patients (28%) had Stage 0 cancer, 16 patients (24%) had Stage I cancer, 9 patients (13%) had Stage II cancer, 11 patients (16%) had Stage III cancer, and 14 patients (20%) had Stage IV cancer.
Baseline Clinical Stage versus Surgical Pathology Stage as Predictors of OS
In 72 patients who had baseline EUS stage measured, there was no correlation with OS (P = .4) (Fig. 1). However, postsurgical pathologic T, N, and overall stages all were correlated significantly with OS (P < .05). Patients who had lower pathologic T classification (T1 or T2) survived significantly longer than patients who had higher pathologic T classification (T3; median time not reached vs. 60 months; P = .03). The same was true for the pathologic N stage (N0, median time not reached; N1, 60 months; N2, 19 months; P = .006). Most noteworthy, of all factors that were considered, the pathologic AJCC stage was the strongest predictor of OS. Patients who had lower postsurgical pathologic stage cancers (Stages 0–II; median time not reached) survived significantly longer than patients who had higher stage cancers (Stages III and IV; 12 months; P = .001) (Fig. 2). A greater proportion of patients with lower pathologic stage cancers also remained alive at the time of the current analysis (Stage 0, 79%; Stages I and II, 72%; Stages III and IV, 21%; P < .0001) (Table 2).
Table 2. Characteristics that Correlated with Overall Survival*
None of the baseline characteristics examined were associated significantly with pathCR (gender, P = .77; tumor location, P = .84; baseline T classification, P = .99; baseline N classification, P = .99; and baseline EUS stage, P = .94).
The Effects of R0 Resection, PathCR, Cancer Downstaging, and Clinical Response on OS
Patients who underwent an R0 resection survived significantly longer than those who did not undergo an RO resection (median time not reached vs. 7 months; P < .001) (Fig. 3). A greater proportion of patients who underwent complete R0 resection were alive compared with patients who did not undergo a complete resection (69% vs. 0%; significance could not be assessed accurately).
Patients who achieved a pathCR survived significantly longer than those who did not achieve a pathCR (median time not reached vs. 53 months; P = .004). A comparison of survival curves for pathCR versus nonpathCR is shown in Figure 4. A significantly greater proportion of patients who achieved a pathCR were alive at follow-up compared with patients who did not achieve a pathCR (78% vs. 47%; P = .02).
Patients who had their disease downstaged after surgery with respect to T classification (median time not reached vs. 23 months; P = .03) and N classification (median time not reached vs. 54 months; P = .02) also survived significantly longer than patients who did not have their disease downstaged. Patients who had any clinical response survived significantly longer than patients who achieved no response (median time not reached vs. 19 months; P = .002).
Patients who had their disease downstaged after surgery with respect to T classification (median time not reached) also survived significantly longer without disease than patients who did not have their disease downstaged (12 months; P < .001). A significantly greater proportion of patients who had T-stage downstaging survived without disease at follow-up (29% vs. 13%; P = .005). Patients who had any clinical response (median time not reached) survived significantly longer than patients who achieved no response (7 months; P = .004).
Baseline EUS stage had no correlation with DFS (P = .23). Only postsurgical pathologic stage was correlated significantly with DFS. Patients who had lower postsurgical pathologic stage cancers (Stages 0–II; median time not reached) survived without disease for a significantly longer time than patients who had higher stage cancers (Stages III and IV; 4 months; P < .0001). A significantly greater proportion of patients who had lower stage cancers also were alive without disease at follow-up (Stage 0, 79%; Stages I and II, 64%; Stages III and IV, 13%; P < .001) (Table 3).
Table 3. Association of Clinicopathologic Characteristics with Disease-Free Survival*
Disease-free survival did not significantly correlate with gender, age, baseline tumor status, baseline lymph node status, baseline EUS stage, or postsurgical lymph node down-staging. These results were not included in the table.
Values shown are the percentage of patients in the subcategory who remained alive without recurrence at follow-up (i.e., the total percent of patients who had presurgical N0 lymph node status).
Of the patient characteristics examined, gender; age; baseline T, N, and EUS stages; intermediate T or EUS stage; intermediate T or N downstaging; and postsurgical N downstaging did not correlated significantly with DFS (data not shown).
A multivariable Cox model revealed that R0 resection and pathologic stage were independent predictors of OS (P = .05 for both). For DFS, pathologic stage was the only important prognosticator (P = .0031).
The preoperative approach to treating localized gastric cancer provides innumerable strategic advantages over the traditional postoperative approaches. Most patients with localized gastric carcinoma have chronic progressive symptoms that are treated symptomatically for some time. The diagnosis is confirmed when patients are referred to a gastroenterologist who performs an endoscopic biopsy. Once the diagnosis is made, the patients often are referred to a community surgeon who, after obtaining more conventional staging, proceeds to remove the cancer (and performs systematic lymph node dissection; this is variable). Only after the patient recovers, may their case be discussed by a tumor board and subsequently referred to another specialist, such as a radiation oncologist and/or medical oncologist. In this common scenario, the lines of communication are horizontal (from 1 point to another), more like a relay. There may be numerous disadvantages to this approach, including the lack of a multidisciplinary evaluation, the lack of communication prior to giving therapy, and, thus, the lack of a long-term therapy-planning process. The Intergroup 0116 trial17 did not require a multidisciplinary evaluation before surgery; this may have lead to the observed difficulty in postoperative radiation field planning, which affected approximately 34% of patients.
The preoperative approach can overcome some of these disadvantages but creates a new dilemma, in that it can alter the true stage of the cancer based on tumor biology and susceptibility to therapy. In this instance, it is not known whether the surgical pathology AJCC stage can predict outcome. To study adequate numbers of patients who underwent surgery after preoperative chemoradiation, we combined 2 populations of patients in prospectively conducted, preoperative chemoradiation studies that used similar strategies. The only difference between the 2 studies was the addition of paclitaxel to induction chemotherapy and chemoradiation.
Our current data suggest that surgical pathology AJCC stage (whether altered or unaltered by therapy) is prognostic of patient outcome. It also was an independent factor in multivariate analysis for OS and DFS. Patients who underwent an R0 resection survived longer than patients who underwent less than an R0 resection. Patients who achieved a pathCR fared better than patients who did not; however, the clinical factors did not help predict the outcome after preoperative chemoradiation therapy. This inability to optimize therapy in patients with localized gastric cancer is a significant drawback and may result in the administration of frequently toxic therapy to some patients without achieving any benefit (and some of those patients may be spared devastating surgical complications and lifestyle consequences). It is hoped that, by studying cancer biology and patient genetics, we may be able to optimize therapy for these patients.
In conclusion, our data demonstrate that, in circumstances in which preoperative therapy has the potential to change the true stage of cancer, it is the surgical pathology AJCC stage that provides the best prognostic information for OS and DFS. Therefore, surgical pathology stage potentially may serve as an intermediate endpoint for Phase II and III trials for the rapid evaluation of a therapy's impact.
We thank Mr. Simon Lunagomez for his assistance with this article.