Introduction

Gastric cancer remains one of the most common causes of cancer-related deaths worldwide [1]. The primary cause of gastric cancer-related death is recurrence. However, little research exists concerning the time-varying pattern of the recurrence risk for gastric cancer after surgery. Furthermore, the risks of recurrence were almost universally described using survival curves, which lacked information regarding the changes in the recurrence probability over time [2, 3]. Nevertheless, use of the hazard function statistical method can explain changes in the recurrence rate over time. Other investigators have used this method to find a pattern in the time distribution of the recurrence hazard for breast cancer patients [4, 5], which may indirectly elucidate the biological behavior of breast cancer. In addition to their theoretical value, these findings may spur novel therapeutic approaches and appropriate follow-up strategies [6]. We speculated that there may a similar pattern for gastric cancer.

Patients and methods

This study consisted of 1,222 gastric cancer patients who underwent D2 radical surgery with R0 resection and did not receive any neoadjuvant chemotherapy or radiotherapy from January 1994 to December 2006 at the Sun Yat-Sen University Cancer Center in Guangzhou, China. This study was conducted in accordance with the Declaration of Helsinki, and all patients signed a consent form approved by the Research Ethics Committee of the Sun Yat-sen University Cancer Center. The eligibility criteria included a histologically confirmed R0 resection and a postoperative survival time of 3 months, with at least 12 months of follow-up data regarding tumor recurrence and death.

In this study, recurrence-free survival (RFS) was defined as the time from surgery to the earliest occurrence of relapse (local or distant) or to death from the tumor [5]. The survival distributions were estimated using the Kaplan–Meier product-limit method and were compared using the log-rank test. Cox proportional hazards regression was used to model the relationship of RFS with the clinico-pathological parameters. For a graphical display of the RFS, the annual hazard rates were estimated using a Kernel-smoothing method [5, 7]. All of the statistical analyses were performed using the Stata statistical software package 10.0 (Stata Corporation Ltd, College Station, TX, USA).

Results

Nine factors were found to have a statistically significant association with the RFS upon univariate analysis: age, tumor location, tumor size, histological grade, lymphatic/venous invasion, pathological T (pT), pathological N (pN), number of retrieved lymph nodes and chemotherapy (Table 1). All of these variables were included in a multivariate Cox proportional hazards regression analysis, which revealed that all of these variables were independent prognostic factors for RFS (Table 2).

Table 1 Univariate survival analysis of clinic–pathologic variables in 1,222 cases of gastric cancer patients
Full size table
Table 2 Multivariate survival analysis results
Full size table

The annual recurrence hazard curve for all of the patients showed a regular curve: one early peak and a late rise within the 10-year period after the surgery. The first major recurrence surge peaks at 1.5 years after the surgery, covering the first 3 years after surgery, which we call the early peak. Then, the hazard curve begins to fall until 7.5 years after the surgery, at which point the curves begin to rise again (Fig. 1). This recurrence pattern was observed in the subgroup analysis according to several clinico-pathological parameters. In patients with a high risk of relapse (e.g., the tumor size >5 cm (Fig. 2a), the poorly differentiated/undifferentiated adenocarcinoma (Fig. 2b), lymphatic/venous invasion positive (Fig. 2c), T3 and T4 (Fig. 2d), node positive (Fig. 2e), retrieved lymph nodes <15 (Fig. 2f) and without chemotherapy (Fig. 2g)), the hazard pattern was more prominent, and the hazard curve was steeper.

Fig. 1
figure 1

Annual recurrence hazard rate for 1222 gastric cancer patients. HR hazard rate, CI confidence interval

Full size image
Fig. 2
figure 2

Annual recurrence hazard rate for 1222 gastric cancer patients by tumor size, histological grade, lymphatic/venous invasion, depth of invasion, nodal status, number of retrieved lymph nodes and chemotherapy a Tumor diameter ≤5 cm versus tumor diameter>5 cm b Well-/moderately differentiated adenocarcinoma versus poorly differentiated/undifferentiated adenocarcinoma/signet ring cell carcinoma/mucinous adenocarcinoma c Lymphatic/venous invasion negative versus lymphatic/venous invasion positive d T1 and T2 versus T3, T4a, T4b e Node negative versus node positive f The number of retrieved lymph nodes ≥15 versus <15 g Chemotherapy versus no chemotherapy. HR, hazard rate; CI, confidence interval

Full size image

Discussion

To the best of our knowledge, this study is the first retrospective analysis of the postoperative recurrence pattern of gastric cancer. This pattern is similar to the double-peaked recurrence pattern of breast cancer, but with some differences. The occurrence time for the early peak in gastric cancer is the same as that in breast cancer. However, the late recurrence occurs later in gastric cancer and appears as a rising trend rather than a peak within 10 years after surgery [6]. In our attempt to explain this phenomenon, we believed that the hypothesis of tumor dormancy might satisfactorily fit our findings [5, 6, 8, 9]. This hypothesis assumes the presence of micrometastatic foci in various biological steady states in the preclinical stage, most of which do not progress to tumor growth. However, surgery may perturb this orderly and stable process and stimulate the dormant micrometastatic foci to grow, which results in the sudden acceleration of the metastatic process and eventually leads to recurrence [10]. This phenomenon may account for the early recurrence peak.

According to the tumor dormancy hypothesis, surgery may accelerate metastatic development by triggering tumor growth. However, we cannot deny the role of surgery in the treatment of the primary tumor. Indeed, surgery is the main treatment for operable gastric cancer [11]. Therefore, we should learn more regarding the biological behavior of gastric cancer from the results of our findings to help enlighten us about treatment. In a further subgroup analysis, we found that the recurrence rate was closely related to both the surgery and the chemotherapy after surgery. For surgery, if there are <15 retrieved lymph nodes, the curve of recurrence was steeper, indicating that the patients were more likely to relapse. This finding supports the NCCN guidelines concerning gastric cancer surgery, which state that the number of retrieved lymph nodes should be ≥15, which is consistent with other reports [12]. Therefore, we should be attentive to the quality of the surgery. For chemotherapy, the curves of the patients who received chemotherapy were flatter, indicating that chemotherapy after surgery can reduce the hazard rate. This finding further emphasizes the need for chemotherapy [13].

Furthermore, the subgroup analysis revealed some high-risk factors, including a larger tumor size, poorly differentiated/undifferentiated adenocarcinoma, lymphatic/venous invasion, the T3 and T4 stage and a node-positive status. The curves of the patients with these factors were steeper, indicating that these patients were more likely to relapse. The results of a multivariate Cox proportional hazards regression analysis also support this finding. These findings are similar to those of several other studies [3, 14]. Therefore, chemotherapy or other treatments after surgery are more important for patients who present these high-risk factors.

Furthermore, we also found that the first recurrence peak nearly covered the 3 years after surgery. However, chemotherapy cannot be used continuously for 3 years to cover the first recurrence peak because of its side-effects; both the NCCN guidelines and the Japanese gastric cancer association (JGCA) suggest that the length of chemotherapy should not exceed 1 year. Therefore, we must find additional treatments to further reduce the risk of recurrence. Immunotherapy is a potentially valuable treatment that can reduce the risk of recurrence [15].

Finally, using the time-varying pattern of the recurrence risk for gastric cancer, we may be able to formulate individual treatments and follow-up.