Chinese Patients with Gastric Cancer Need Targeted Adjuvant Chemotherapy Schemes

Gastric cancer (GC) is one of the most common cancers in China. Although nationwide retrospective studies have indicated that mortality from GC is declining, it still ranks in third place, behind bronchial lung cancer and liver cancer, in cancer deaths. According to the GLOBOCAN 2008 statistics, there were almost 989,000 new cases worldwide, while approximately 463,000 new cases arose in China, accounting for 48.6%. Simultaneously, approximately 737,000 deaths caused by GC occurred around the world in 2008, nearly 352,000 deaths in China, accounting for 47.8% (Chen, 2008; Zhou et al., 2012). In China, adjuvant chemotherapy (AC) is a routine auxiliary treatment for GC. After curative gastrectomy, patients obtained greater survival benefits from AC than from surgery alone through reduced tumor relapse rates and prolonged patient life spans, with a small but


Introduction
Gastric cancer (GC) is one of the most common cancers in China. Although nationwide retrospective studies have indicated that mortality from GC is declining, it still ranks in third place, behind bronchial lung cancer and liver cancer, in cancer deaths. According to the GLOBOCAN 2008 statistics, there were almost 989,000 new cases worldwide, while approximately 463,000 new cases arose in China, accounting for 48.6%. Simultaneously, approximately 737,000 deaths caused by GC occurred around the world in 2008, nearly 352,000 deaths in China, accounting for 47.8% (Chen, 2008;Zhou et al., 2012).
In China, adjuvant chemotherapy (AC) is a routine auxiliary treatment for GC. After curative gastrectomy, patients obtained greater survival benefits from AC than from surgery alone through reduced tumor relapse rates and prolonged patient life spans, with a small but because of China's late start and ethical concerns. Therefore, we attempted to estimate the status of AC as a treatment for GC in China compared with other regions and to explore ways of creating targeted AC schemes for Chinese patients with GC.

Study retrieval and eligibility criteria
Two reviewers (Q Ding and K Su) independently searched a series of databases for studies investigating AC for GC, including the Cochrane Library (1992( to Mar. 2012, MEDLINE (1960to Mar. 2012, EMBASE (1976to Mar. 2012, as well as Chinese databases such as National Knowledge Infrastructure (1979to Mar. 2012) and the VIP database (1989( to Mar. 2012. Medical Subject Headings (MeSH) and keywords were used, including "stomach neoplasm," "adjuvant chemotherapy," "gastric cancer", and "adjuvant treatment." In addition, the reference lists of the retrieved full-text papers were also searched to ensure that there were no omissions.
The following inclusion criteria for the literature were determined by consulting clinicians: 1) patients with adequate organ function and a histologically proven diagnosis of GC; 2) studies comparing surgery plus AC with surgery alone; 3) an endpoint of a hazard ratio (HR) of mortality, with the HR reported or data sufficient for calculating the HR being necessary; and 4) in English or Chinese with a published English abstract. We excluded studies about radiotherapy and/or immuno-chemotherapy combined with chemotherapy, trials of repetition and pseudo-randomized trials.

Study selection and data extraction
The titles and abstracts of the retrieved articles were read by both reviewers (JQ Li and MJ Tang) to identify studies according to the eligible criteria above. Then, we attempted to obtain full-text articles using the databases or the Internet or through correspondence with the authors. Based on the qualified results, important information from the included studies was separately extracted by two reviewers (JQ Li and MJ Tang) using a predefined data extraction form; this information included the authors, years of publication, case sources, regimens, dosages, schedules, numbers of patients, recruitment periods, stages, and median follow-up durations.

Analysis of bias risk
The quality of methodological bias for the included studies was assessed by referring to the Cochrane Handbook for Systematic Reviews of Interventions (version 5.0.2) (Higgins et al., 2011), including evaluation of randomization, allocation concealment, blinding and intention-to-treat (ITT) analysis. Divergence between the reviewers was reconciled by discussion with a third reviewer (JW Zhang), whenever it arose. When necessary, corresponding authors were contacted to clarify details necessary to optimize the relevant data. In addition, some studies were performed using minimization methods to improve the balance of the baselines (Scott et al., 2002). Though it was a type of non-random method, we considered these trials eligible due to their reliable designs and we rated their randomization as high-level.
Assessment of Grades of Recommendation, Assessment, Development, and Evaluation (GRADE), recommended by the Cochrane Collaboration, provides a quantitative quality evaluation system for systematic reviews and guidelines (Guyatt et al., 2011a). Evidence derived from RCTs was considered to be highly qualified. The assessment was implemented according to explicit criteria concerning study design, risk of bias, imprecision, inconsistency, indirectness, and magnitude of effect. In addition, when death details were not provided in the original research, we estimated them according to the survival rate.

Statistical analysis
Review Manager 5.1 was used for the statistical analysis and for the quality assessment of individual studies. Stata 11 was used to detect publication. Gradeprofile 3.6 was employed to rate the quality of the evidence. First, we calculated the log-hazard ratio (log HR) of mortality and its standard error (SE) for each study based on the method described (Parmar et al., 1998), unless the study provided results from a univariate Cox regression analysis with log HR and its SE. Second, heterogeneity was estimated using the Chi-square-based Z statistic for statistical significance. If P>0.05 indicated little heterogeneity, we used a fixed-effect model in generic inverse variance to analyze the data; if not, a random effect model was adopted. The amount of heterogeneity was estimated using the I2 statistic. If I2>50%, it indicated that substantial heterogeneity existed. When I2<75%, the heterogeneity between studies could be accepted. Publication bias and selection bias were tested with Stata 11, using funnel plots with Begg's test. If P<0.05, it revealed the existence of publication bias and selection bias. Finally, the grading strength of the evidence was assessed, followed by the creation of SoF (summary of findings) tables in detail (Guyatt et al., 2011b). The number needed to treat (NNT) was calculated to reveal the curative effects of AC for GC in patients from different geographic areas.

Synthesis of results
Anti-metabolites, anti-tumor antibiotics, alkylating agents, anti-tumor plant medicines, anti-tumor hormonal medicines, anti-tumor auxiliary drugs and miscellaneous anti-tumor drugs were all commonly used as chemotherapeutic agents for resected GC. Among them, the frequency of applications combining anti-metabolites with anti-tumor antibiotics was highest. To clarify the effects of various combinations, we stratified the 39 trials into 3 subgroups based on the agents: a subgroup containing anti-metabolites plus others with anti-tumor antibiotics; a subgroup containing anti-metabolites plus others without anti-tumor antibiotics; and a subgroup without anti-metabolites. There was one article with 3 study groups (Nakajima et al., 1980) divided into 2 individual studies, as well as 1 Chinese article (Zhang et al., 2005). The relevant forest plot data are shown in Figure 2. Using the Z statistical test for analysis, good heterogeneity was calculated within the 3 subgroups (P=1.00, I2=0% . Although the 2 anti-metabolites-based subgroups exhibited no significant differences, the pooled data supported that AC could reduce the risk of death as a protective treatment for the disease. The starting time of AC was a confusing problem for the clinicians. Postoperative AC was commonly applied, while some patients were begun a month after surgery or even later. Due to confusion over the starting time, the trials were stratified into 4 groups, including administration within a month, administration a month or more later, peri-operative administration and administration not mentioned. The results are displayed in Table 2. No obvious significant differences in the overall estimates were tested in the subgroups (P>0.05). Our findings suggested that the curative effect of AC was not influenced by the time at which the drugs were   Because of the different populations' varying races and living habits, we divided the patients into subgroups according to geographic area. The results are shown in Figure 3. The results of the Z statistical test showed that intra-group heterogeneity was good (P>0.1). In the 4 subgroups, the HR of mortality was 0. Only the HR of mortality in Asian countries other than China was statistically difference between the treatment and control groups (Z=2.29, P=0.02), indicating that patients in several Asian countries, such as Japan, Korea, and China, could benefit more from AC than patients in non-Asian countries, including the United States, the United Kingdom, Italy, etc. Diversity appeared not only in morbidity and mortality but also in the effects of chemotherapy drugs across different areas. Thus, based on the stratification above, we divided the trials sequentially according to the drugs administered. Certain potential protective effects of the AC drugs for GC patients in each region were tested (HR<1) without significant differences (P>0.05), in addition to the subgroups of Asian countries other than China without anti-metabolites exhibited a marginal benefit in the treated group (Z=1.97, P=0.05). Nonetheless, the risk of death was reduced more by AC combinations among Asians than among non-Asian patients.
The effects of D2 lymphadenectomy-based AC on GC constituted another controversial focus between the East and West. The Japanese guidelines and clinical trials reported that patients receiving AC could achieve better survival rates than with surgery alone after D2 lymphadenectomy, while many studies revealed that patients in Europe and the United States failed to benefit more from D2 lymph node dissection. To illustrate the effects of D2 lymphadenectomy-based AC, 13 trials of D2 lymphadenectomy-based AC were extracted and are showed in Table 3. Though there was no statistical significance (P>0.05) in any region, the pooled data indicated that D2 lymphadenectomy-based AC was effective (HR=0.89, 95%CI: [0.80, 0.99], Z=2.10, P=0.04), suggesting that AC drugs should be adjusted to adapt to D2 lymphadenectomy, such as intravenous fluorouracil (5-FU) in a range from 350 to 1500 mg/m2. The data originated from Japan, demonstrating that oral S-1 40 mg/ m2 after D2 lymphadenectomy was another good choice for advanced GC and for a reduction of adverse events (Sakuramoto et al., 2007). Based on the results above, sensitivity analysis was conducted. First, the Chinese studies were eliminated because their imperfect design might have led to obvious bias. We found that the pooled data on D2 lymphadenectomy-based chemotherapy were not stable. After the 4 Chinese trials were removed from analysis, the HR of mortality changed to 0.91 (95% CI: [0.81, 1.02]) without significant difference (Z=1.58, P=0.11) between the treatment and control groups. Second, we eliminated 2 Japanese studies due to the high survival rates that they reported. Interestingly, we found the same index also changed. The HR of mortality was 0.91 (95% CI: [0.81, 1.03]) and was not significantly different (Z=1.54, P=0.12).

Analysis of bias risk for eligible RCTs and GRADE assessment
To clarify the credibility of the conclusions of the included individual studies, quality assessment was implemented, as shown in Figures 4 and 5. The publication bias based on region is shown in Figure 6 and in Table  4. The GRADE evaluation is displayed in Table 5. The GRADE assessment confirmed that the strength of the evidence from the Chinese studies was moderate because similar domestic research seldom focused on comparisons of the curative effects between surgery and AC and surgery only due to certain late starts and ethical concerns.

Discussion
China is a high-risk region for GC. The number of deaths in China from GC account for approximately 23% of all deaths from cancer, with nearly 227,000 deaths every year since AC was applied nationwide in China as a routine auxiliary approach for GC. Raw data from RCTs of AC originated from foreign trials conducted between 1970 and 2004, which might not have been optimized for Chinese GC patients because of differences in race and living habits. Therefore, we conducted this systematic review to identify the effects of AC in Chinese patients with GC compared to other Asian countries, including Japan and South Korea, as well as European countries, such as the United Kingdom and Italy, and the United States, with the aim of exploring ways of creating targeted AC schemes for Chinese patients with GC.
Some RCTs indicated that patients receiving AC obtained no greater survival benefits. The Eastern Cooperative Oncology Group (ECOG) found no treatment benefit from AC with 5-FU plus 1-(2-chloroethyl)-3-(4methylcyclohexil)-1-itrosourea (Me-CCNU), and they concluded, based on a benefit-risk analysis, that this combination was not recommended for patients after resection because of its toxicity (Engstrom et al., 1985). The British Stomach Cancer Group (BSCG) performed another prospective RCT with 138 patients in the treated group and 145 in the control group. After a mitomycin, doxorubicin, and 5-FU (MAF) regimen was given, the 5-year survival rate was 19% in the treatment group and 20% in the control group without statistical significance (P=0.69) (Hallissey et al., 1994). However, AC has been regarded as efficacious in other trials. A phase Ⅲ RCT (ML17032) assessing capecitabine and cisplatin (XP) compared to 5-FU and cisplatin (FP) for advanced GC revealed that the former treatment led to a higher remission rate of 42% vs. 29%, as well as longer survival of 10.5months vs. 9.3 months (Kang et al., 2009). Neri et al. concluded that treatment was the only significant prognostic factor after administering epidoxorubicin, leucovorin and 5-FU (ELF) to the treatment group with a 5-year follow-up (Neri et al., 2001). Similarly, our data were consistent with the viewpoint that AC is an effective intervention for GC patients as a protective factor. Suspecting that an earlier starting time for AC would lead to a better theoretical response, our data indicated that the curative effects of AC were not influenced by the time at which the drugs were administered. Starting time might be not an independent risk factor for mortality. However, some recent research has indicated that the curative effects with respect to GC are not exactly the same in different geographic areas. The divergence originated from a study of lymph node dissection and application of S-1. Sakuramoto et al. showed that S-1 was an effective adjuvant regimen for East Asian patients after D2 lymph-node dissection for locally advanced GC, with 3-year overall survival rates of 80.1% in the S-1 group (95%CI: [76.1, 84.0]) and 70.1% in the surgery-only group (95%CI: [65.5, 74.6]) (Sakuramoto et al., 2007). Some trials conducted in Europe have proved the classical Japanese D2 resection offered no survival advantage over D1 surgery among European patients (Bonenkamp et al., 1999;Cuschieri et al., 1999), while D2 dissection is the standard surgical technique used in Japan. S-1-based chemotherapy and the combination of S-1 and cisplatin are the most reasonable first-line schemes for unresectable advanced GC used in Japan (Kobayakawa et al., 2011), but their application had been delayed in western countries, not only because they do not provide increased survival but also because of postoperative complications and mortality. Another trial, conducted by the First-Line Advanced GC Study group (FLAGS trial), indicated that S-1 plus cisplatin improved safety significantly but did not prolong survival in advanced GC and gastroesophageal adenocarcinoma when compared with cisplatin plus 5-FU (Ajani et al., 2010). Based on these differences, we found that Asians could obtain a greater reduction in mortality risk from D2 lymphadenectomy-based AC, compared to non-Asians, suggesting that relevant studies could be conducted among Chinese patients with GC for further data.
Some individual studies abroad have determined that patients in particular statuses would benefit more from AC compared to other patients with GC. Kulig et al. indicated that a postoperative etoposide, adriamycin and cisplatin (EAP) regimen offered no survival advantage in GC patients, but their subgroup analysis revealed a survival benefit from chemotherapy in patients with tumors infiltrating the serosa and in patients with 7-15 metastatic lymph nodes (Kulig et al., 2010). A phase Ⅲ trial performed by Al-Batran et al. found that patients aged 65 years old or older would benefit more from 5-FU, leucovorin and oxaliplatin (FLO) than from 5-FU, leucovorin and cisplatin (FLP) (Al-Batran et al., 2008). These multiple findings supported the idea that qualified trials could be performed in China for further investigation.
Our quality assessment determined that the strength of the evidence from foreign studies, conducted in Europe, the United States and Asian countries other than China, was high, while the strength of the evidence from Chinese studies was moderate. Because of late starts and certain ethical issues, standardized RCTs concerning AC treatment for GC have been insufficient among Chinese studies. Given the positive effects of AC, we should focus on optimizing targeted AC schemes for Chinese GC patients based on therapeutic actuality rather than violating our ethics to perform similar trials.
In Conclusion, AC, as an effective intervention for GC seems beneficial for the Chinese patients, even more than for Asians in general. Its effects were not influenced by the starting time of the administration of AC doses, such as 5-FU administered intravenously in the range from 350 to 1500 mg/m2 after D2 lymphadenectomy. Lymphadenectomy-based D2 and S-1 might be a safer and better choice for Asian patients than for non-Asians. Based on these results, it is necessary to optimize schemes for Chinese patients with GC.