|Home | About | Journals | Submit | Contact Us | Français|
It is important to examine variation in the treatment effects of patients with esophageal cancer in order to generalize treatment outcomes. We aimed to investigate the range of prognostic differences among hospitals in the treatment of locally advanced esophageal cancer. The JCOG0303 study compared the efficacy of radiotherapy plus low-dose cisplatin and 5-fluorouracil with that of high-dose cisplatin and 5-fluorouracil for unresectable esophageal cancer. Of 32 institutions participating in the JCOG0303 study, the 18 institutions that enrolled three or more patients were included in this study. We predicted the 1-year survival in each institution by using a mixed-effect model. We found that the predicted 1-year survival in the 18 institutions with three or more patients was a median of 60.9%, with a range of 60.9–60.9%. This study is the first to investigated heterogeneity of survival in patients who received definitive chemoradiotherapy for locally advanced esophageal squamous cell carcinoma.
The prognosis of carcinoma of the esophagus is generally poor, especially when the cancer is unresectable (1). Curative resection is not feasible in patients with locally advanced esophageal cancer, particularly clinical T4 or unresectable regional lymph node metastasis. Therefore, the treatment outcome of such cases remains unfavorable (2,3). Definitive chemoradiotherapy (CRT) is one of the most effective treatment options that are currently available for locally advanced esophageal cancer (4–6).
We previously conducted a randomized Phase II study to compare radiation therapy (RT) plus low-dose cisplatin and 5-fluorouracil (LDPF) with RT plus standard-dose cisplatin and 5-fluorouracil (SDPF) for patients with clinical T4 or M1Lym esophageal squamous cell carcinoma (i.e. the JCOG0303 study) (7). A total of 142 patients from 32 institutions were enrolled. We found that RT plus LDPF chemotherapy showed no advantage over RT plus SDPF. The median survival time was 13.1 months for patients treated with SDPF and 14.4 months for patients treated with LDPF. No difference was detected in toxicity between the study arms.
This study had several limitations. First, registration took longer than 5 years. One of the reasons for slow accrual was that confirmed clinical T4 or unresectable regional lymph node metastasis cases that met the eligibility criteria were not common, even in a high-volume center. Second, it was possible that some physicians enrolled patients with borderline-resectable cases while others did not. There was a 6% conversion to esophagectomy after definitive CRT. Third, a clinical diagnosis of unresectable was defined according to classical radiological criteria (8). With the progress in radiological imaging, a diagnosis of unresectable differs between institutions. In this study, we did not conduct independent re-evaluation for an unresectable diagnosis. Therefore, the heterogeneity of unresectable diagnoses between hospitals may have affected the outcomes of the JCOG0303 study. Although these limiting factors may have affected treatment outcomes owing to the potential for heterogeneity at each institution, the data are not indicative of institutional differences in the efficacy of CRT for esophageal cancer. It is important to examine the variation in survival of patients with esophageal cancer in order to evaluate the generalizability of treatment outcomes. This study aimed to investigate prognostic differences of the JCOG0303 study between hospitals.
Only those institutions with three or more patients in the JCOG0303 study were included in this analysis, because including those institutions with too few patients would reduce the precision of the analysis. Eligibility criteria and main outcomes have previously been described (7). This study used a mixed-effects model (9). The variance in overall survival (OS) between institutions was determined using the Cox model with random effects. This model included institutions as a random effect and other clinical background factors (main location of the tumor, metastases [M] and the stricture of the esophagus) as fixed effects. A normal distribution was assumed for variance between institutions. The predicted hazard at each institution determined by the Cox model was translated into the predicted 1-year OS to determine whether there was heterogeneity between institutions. Factors that showed P < 0.10 by Cox multivariate analysis were included in the model as a fixed effect (Table 1). SAS 9.2® (SAS Institute, Cary, NC, USA) NLMIXED procedures were used.
Among 142 patients from 32 institutions in the JCOG0303 study, 123 patients from 18 institutions were analyzed. The baseline characteristics of these 123 patients are shown in Table 2. Histopathological findings of these patients indicated squamous cell carcinoma. The characteristics of patients in the original dataset and the selected dataset were similar. We found that the proportion of predicted 1-year survival was 60.9% (range, 60.9–60.9%) at all institutions. No difference in 1-year survival was observed between institutions (Fig. 1).
Unexpectedly, no differences in 1-year survival were observed between institutions in the JCOG0303 study. This study is the first to evaluate the inter-institutional heterogeneity of definitive CRT for patients with locally advanced esophageal squamous cell carcinoma by using prospective data.
Inter-institutional heterogeneity of survival has previously been reported (10). Several factors can cause prognostic differences between hospitals. First, imaging diagnosis of TNM staging is made based on the findings of a classical examination, such as conventional computed tomography. More advanced high-resolution computed tomography and positron emission tomography have become increasingly common in recent years (11,12). The difference in modalities used in each institution can be a cause of prognostic differences. Second, high-risk surgery, such as esophagectomy, in low-volume centers increases operative mortality (13). Therefore, differences in operative mortality can also be a cause of prognostic differences between institutions. Third, although recent advances in RT with systematic quality assurance prevent insufficient irradiation (14), the quality of RT can affect patient prognosis at each institution.
In the present study, we assumed that a diagnosis of unresectable cancer was difficult to determine, and that this difficulty could cause a prognostic difference. Unresectable cases may include definitive unresectable cases and borderline-resectable cases, the latter of which can potentially become resectable with neoadjuvant therapy. Therefore, we assumed that the ratio of these two types of cases among the enrolled patients could be different at each institution. However, we found no difference in 1-year survival among the participating institutions. A potential explanation for this result is that the patient population enrolled at each institution was almost homogeneous, and that there was no inter-evaluator difference in a diagnosis of unresectable cancer; however, selection bias should be considered. Another explanation for our finding is that heterogeneity is low in non-surgical treatment.
Although we failed to show inter-institutional heterogeneity in the present study, we previously reported variation in survival and perioperative complications between hospitals in another JCOG study evaluating esophagectomy for Stage II/III disease (15). There are several differences between this study and our previous study. The inter-institutional heterogeneity observed in our previous study may have been related to the invasiveness of esophagectomy and the occurrence of perioperative complications. Additionally, subsequent therapies may have been performed after protocol treatment in the previous study, which could be a reason for inter-institutional heterogeneity; subsequent therapies are rarely performed for T4 or unresectable regional lymph node metastatic disease. Another possibility is that supportive management is more standardized for CRT than for perioperative care. Such standardization may have reduced inter-institutional heterogeneity in the present study. RT quality assurance should also be considered. A review of the results of the JCOG0303 study was conducted, and most of the cases were found to have acceptable quality (16). Considering these factors, it seems reasonable that the present study did not show inter-institutional heterogeneity.
This study has several limitations. First, patients in both treatment arms were analyzed together. Although there was no difference in OS between the two arms, a difference in treatment methods may have affected heterogeneity. Second, random effect predictors could have varied depending on the procedure used in the mixed-effects model, because explicit integral computation could not be performed in this study. Low variance might not have been detected, which could explain the lack of inter-institutional heterogeneity in this study. Third, each investigator of the JCOG0303 study was selected, which could explain why a homogenous patient population was selected. Other limitations of our study, such as the small sample size, made further enrollment in this specialized target difficult. However, our findings might be valuable for future studies, which are required to establish further evidence in this area. Additional T4 or borderline-resectable cases could be more heterogeneous, and generalized criteria need to be established.
In conclusion, the predicted prognosis of T4 or unresectable regional lymph node metastasis esophageal cancer by using CRT is the same in this study as in the JCOG0303 study. To conduct future studies including more ‘borderline-resectable cases' with conversion surgery, we need to address the problem of heterogeneity.
This work was supported by the National Cancer Center Research and Development Fund [26-A-4].
Yuko Kitagawa has received grants from Nippon Kayaku, Kyowa kirin, Bristol-Meyers wquibb and Pfeizer-Japan
We thank the members of the JCOG Data Center/Operations Office for their support.