- Top of page
- MATERIALS AND METHODS
- FUNDING SUPPORT
- CONFLICT OF INTEREST DISCLOSURES
Gastric carcinoma ranks as the 13th most commonly diagnosed cancer, with approximately 11,000 patients expected to die of the disease in the United States in 2014. Microscopically, gastric cancer can be divided into 2 main histological categories, intestinal type and diffuse type, according to the classification scheme originally described by Lauren in 1965. Intestinal-type tumor cells are said to form irregular glandular structures and are frequently associated with intestinal metaplasia and atrophic gastritis in the adjacent mucosa, whereas diffuse histology is characterized by a poorly cohesive cellular pattern with little or no gland formation. In many instances, the majority of tumor cells of diffuse histology secrete copious mucus, which may accumulate in the intracellular compartment and push the nucleus to the periphery of the cell, giving rise to the term of signet ring cell carcinoma. Another morphologic variant of diffuse-type gastric cancer is linitis plastica (“leather bottle”), which is characterized by diffuse tumor cell infiltration into the stomach wall accompanied by extensive fibrosis.
Clinically, diffuse-type gastric cancer is distinguished from intestinal-type disease by several characteristics, including relatively younger patients, a higher prevalence of lymph node invasion, and lower R0 surgical resection rates. The optimal treatment for patients with diffuse-type gastric cancer has remained largely undefined because to our knowledge no trials to date have been dedicated to the study of this histology despite findings that the incidence of diffuse-type gastric carcinoma has been increasing and currently comprises 25% to 45% of gastric cancer cases in Western countries.[3-8] Recently, a French multicenter retrospective study of approximately 1000 patients with signet ring cell carcinoma demonstrated worse survival associated with perioperative chemotherapy compared with adjuvant chemotherapy. This finding is of particular importance given that the perioperative chemotherapy approach has been increasingly adopted for either intestinal-type or diffuse-type advanced gastric cancer in Europe. In North America, in which D2 lymph node dissection is not commonly performed, adjuvant chemotherapy combined with radiotherapy (RT) has been recommended as a standard option for patients with gastric cancer of stage IB or higher since the publication of the Intergroup (INT)−0116 trial in 2001. However, the efficacy of adjuvant chemoradiation for diffuse-type gastric cancer has recently been put into question by the updated INT-0116 results, which demonstrated a lack of long-term survival benefit for women and for those with diffuse histology in an unplanned and likely underpowered subgroup analysis.
In the current study, we used the Surveillance, Epidemiology, and End Results (SEER) database, which covers up to 25% of the patient population in the United States, to examine whether the findings of the updated INT-0116 subgroup analysis are consistent with outcomes from a large population sample.
- Top of page
- MATERIALS AND METHODS
- FUNDING SUPPORT
- CONFLICT OF INTEREST DISCLOSURES
Using a large population sample, the results of the current study demonstrated a significant survival benefit associated with adjuvant RT in patients with diffuse-type gastric cancer. Although the reason for this is a subject for speculation, the associated survival benefit could be attributed to the unfavorable locoregional control specific to diffuse-type gastric cancer. One of the poor prognostic factors among these patients is the higher prevalence of lymph node invasion found in diffuse-type gastric cancer,[7, 13] supported by the finding of lymph node involvement in >55% of the patients in the current study. In an analysis of lymph node-positive patients who underwent D2 lymph node dissection followed by adjuvant chemotherapy, the Korean ARTIST trial recently demonstrated a prolonged disease-free survival with the addition of adjuvant RT. In light of the finding of diffuse-type histology in approximately 60% of all patients in the ARTIST trial, a potential benefit for adjuvant RT could be proposed for patients with diffuse-type gastric cancer with positive lymph nodes. Another poor prognostic feature of diffuse-type gastric cancer is a lower R0 surgical resection rate despite more extensive surgery, most likely due to the infiltrating characteristic that leads to more positive vertical resection margins.[7, 13, 15] Compared with surgeries performed in large tertiary centers, the percentage of suboptimal surgeries may be even greater in community hospitals, which have contributed approximately 50% of the gastrectomy cases to the SEER database. It is not inconceivable that adjuvant RT may provide a locoregional control benefit and, ultimately, survival benefit in those patients who undergo suboptimal surgery or surgical resections with positive margins.
Several reasons may be proposed for the discrepancy between the results of the current study and those from the subgroup analysis of the updated INT-0116 trial. First, the patient population or treatment received could be different. The patients in the INT-0116 trial likely came from select academic centers and the treatments may not be reflective of those in the community setting. Although the prevalence of lymph node invasion in patients with diffuse-type histology was not specified in INT-0116 trial and therefore could not be compared with that in the SEER study, the percentage of R0 resections in the SEER database is likely greater than that noted in the INT-0016 study because approximately one-half of the gastrectomy cases included in the SEER database came from community hospitals, in which experience in oncologic gastrectomy may be inadequate. However, direct evidence is largely lacking given that SEER did not collect information regarding gastrectomy surgical margin status, one of the important indicators of optimal surgery. Second, although most clinical trials require central pathology review by highly specialized pathologists, it could be speculated that the lack of a central pathology review in SEER may have resulted in a greater variability in interpretations of diffuse-type histology or signet ring carcinoma, defined as the presence of at least 50% of signet ring cells in the pathologic specimen. One study from Stanford University showed a concordance rate of 77% between different pathologists regarding Lauren histology classification. Therefore, the diffuse-type histology in SEER registry data may have included a small portion of intestinal-type histology for which adjuvant RT has been shown to confer a significant long-term survival benefit. If this is the case, the current study may have overestimated the benefit of adjuvant RT among patients with diffuse-type gastric cancer. Conversely, it is likely that such commingling of diagnoses also occurs frequently in the community hospitals that contribute the majority of gastrectomy cases to the SEER database and our result is likely to be more generalizable.
In addition to the limitations in pathology review as discussed earlier, readers are advised to note the following limitations related to cancer registry databases. First, the current analysis of the nonrandomized patient population could not exclude the possibility of selection bias, in which patients with older age, more advanced stage of disease, and poor performance status were more likely forgo adjuvant treatment. The current study used propensity score analysis to balance possible measurable confounders. However, there were residual imbalances with regard to age and AJCC stage III/IV (M0) distribution between the group treated with adjuvant RT and the propensity score-matched group treated without RT. When the small yet statistically significant imbalances were further adjusted for in the Cox regression analysis, the survival benefits associated with adjuvant RT remained significant. In addition, we sought to mitigate the impact of performance status selection bias by censoring the current study data to include only those patients who survived >6 months postoperatively. In performing this sensitivity analysis, our aim was to eliminate some of those patients who did not receive adjuvant RT because of either poor overall condition after surgery or death occurring before they were considered for adjuvant treatment. Despite the possible selection bias due to age, cancer stage, and performance status, the OS in the current study was in keeping with that in the INT-0116 trial, suggesting a limited impact of the selection bias on clinical outcome. For example, in the current study, patients in the gastrectomy-alone group and those in the adjuvant therapy group had 3-year OS rates of 31% and 45%, respectively, whereas the INT-0016 subgroup analysis of patients with diffuse-type histology demonstrated 3-year OS rates of 33% and 41%, respectively.
Second, there is missing information in the SEER database regarding the intent of RT. Although speculative, it is plausible that RT was delivered with a curative intent given the exclusion of patients with metastatic disease in the current study. In addition, the current study cohort demonstrated that approximately 40% of the eligible patients with diffuse-type gastric cancer (782 of 1889 patients) actually received adjuvant RT. Although the omission may point to a possible selection bias in the current study sample, it may also reflect differences in patterns of care in the community. This speculation was supported by our previous finding that only 45% of patients with nonmetastatic gastric cancer, either of the intestinal or diffuse type, received adjuvant RT when we used the SEER data from 2002 to 2005 to examine the use of the INT-0116 treatment scheme since its publication. The relatively low percentage of adjuvant RT cases in the current study sample may also be partly attributable to the previously reported shortcomings in SEER coding, specifically the failure to record adjuvant RT in some of the patients. Several studies have demonstrated underreporting or underascertainment of RT administration in approximately 15% to 30% of patients with breast cancer in the SEER database. In a similar vein, the group treated without RT in the current SEER study may have included some patients who actually received adjuvant RT. If this is the case, the results of the current study may have actually underestimated the survival benefit of adjuvant RT for patients with diffuse-type gastric cancer, because the OS rates in the group treated without RT would be improved by the inadvertent inclusion of patients who may have received adjuvant RT.
In addition, there is missing information in the SEER database regarding the use of chemotherapy, resulting in a potentially significant confounder in the current study. It is possible that patients may have received adjuvant chemotherapy or perioperative chemotherapy, although our cohort was limited to those treated between 2002 and 2005 (ie, before the adoption of the perioperative chemotherapy approach as used in the MAGIC trial published in 2006). Although the MAGIC trial did not perform a subgroup analysis of patients with diffuse-type histology, the survival results from all patients included in the MAGIC trial were similar to those found in the current SEER study. The 5-year OS rate was 23% in the MAGIC trial versus 25% in the current study among patients who did not receive RT, and 36% versus 32%, respectively, in patients who received adjuvant RT. Although perioperative chemotherapy has been the European standard of care for patients with advanced gastric cancer since the publication of the MAGIC trial, several recent studies suggested unfavorable results when this approach was used in patients with diffuse-type gastric cancer. To our knowledge, the largest study to date was a French multicenter retrospective analysis of approximately 1000 patients with signet ring cell carcinoma. It demonstrated worse survival and more extended surgery associated with perioperative chemotherapy than with adjuvant chemotherapy. The findings of low rates of downstaging and high rates of disease progression during the preoperative chemotherapy period suggested inherent chemoresistance in signet ring cell carcinoma and a deleterious effect of delaying definitive surgery for the ineffective preoperative chemotherapy.
Although the results from the current observational study have drawbacks and should be interpreted with caution, the results from the INT-0116 trial subset analysis are also limited. In our opinion, further phase 3 randomized trials designed exclusively for diffuse-type gastric cancer are necessary to provide high-level evidence regarding the optimal treatment. Multiple centers in France have participated in a phase 3 randomized trial evaluating a strategy of primary surgery followed by adjuvant chemotherapy versus perioperative chemotherapy for patients with resectable gastric signet ring cell carcinomas. In Korea, a phase 3 trial (ARTIST II) subsequent to the published ARTIST trial currently is underway to enroll 1000 patients with gastric cancer with positive lymph nodes to evaluate the role of adjuvant RT in the setting of D2 gastrectomy followed by adjuvant chemotherapy. Although the ARTIST II trial is not designed to analyze either intestinal-type or diffuse-type histology, it may provide sufficient power for a subgroup analysis regarding the role of adjuvant RT for diffuse-type gastric cancer if the ARTIST II trial were to include patients with diffuse-type gastric cancer in >50% of the accrual as the original ARTIST trial did.
The current large, population-based study suggests a potential benefit of adjuvant RT for patients with diffuse-type gastric cancer. The optimal treatment will likely remain controversial until high-level evidence becomes available from phase 3 randomized trials designed exclusively for diffuse-type gastric cancer, and a standard practice becomes widely adopted within the community.