Scandinavian Journal of Gastroenterology. 2014; 49: 690–704
ORIGINAL ARTICLE
Adjuvant chemotherapy after curative resection for gastric cancer: a meta-analysis
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
JISEN CAO, FENG QI & TONG LIU Department of General Surgery, Tianjin Medical University General Hospital, Tianjin, China
Abstract Objective. The aim of this article is to review up-to-date clinical data published in the literature in regard to adjuvant chemotherapy in patients with gastric cancer after radical surgical resection. Materials and methods. Medline, Embase, PubMed, the Cochrane Library and CBMDisc were searched to identify data published regarding this issue from 1966 to 2013. All the calculations and statistical tests were done using RevMan5.2 software. Results. A total of 29 trials with 8580 patients met all inclusion criteria. Among them, 27 studies reported survival rates at the end of follow-ups, 64.2% alive among 3981 patients in the adjuvant chemotherapy arm and 57.3% alive among 4027 patients in the observation arm. Statistical results showed that the observation arm had a shorter disease-free survival (RR: 1.11, 95% CI: 1.07–1.15), and the treatment arm had a lower recurrence rate (RR: 0.79, 95% CI: 0.74–0.84). Leucopenia, anemia, nausea and vomiting, diarrhea, alopecia and infection occurred more frequently in the treatment arm. Adjuvant chemotherapy decreased the occurrence of peritoneum relapse [RR = 0.77, 95% CI (0.66–0.90)], lymphoid nodes relapse [RR = 0.58, 95% CI (0.45–0.75)] and local relapse [RR = 0.57, 95% CI (0.41–0.80)]. Conclusions. Adjuvant chemotherapy can improve the survival rate and disease-free survival rate and reduce the relapse rate after curative resection. Adjuvant chemotherapy cannot induce thrombocytopenia and mucositis or affect liver function. The tumor in situ recurrence and peritoneum, lymph nodes relapse decrease after chemotherapy, and patients benefit from adjuvant chemotherapy regardless of the numbers of positive lymph node, depth of local invasion, Asian or non-Asian, the length of follow-up, and numbers of cycles.
Key Words: adjuvant chemotherapy, gastric cancer, meta-analysis
Introduction Gastric cancer is one of the common malignant tumors in the world with a high incidence and mortality. Almost a million new cases of gastric cancer are reported worldwide each year [1]. Surgical resection remains the mainstay of curative treatment for patients with gastric cancer. However, the 5-year– survival rate has been disappointing. Even for those patients who undergo radical resection, the rate of recurrence is still high [2]. Therefore, an increase in the postoperation cure rate is a major concern. The evaluation of adjuvant chemotherapy after surgery for gastric cancer is underway in many clinical trials, particularly in the light of evidence of a benefit
from such therapies in common tumors of the gastrointestinal tract, such as colorectal cancer. Adjuvant chemotherapy for gastric cancer has been administered for over 40 years; however, varied results were documented, and whether patients can benefit from chemotherapy after curative resection has yet to be clarified. Several meta-analyses of adjuvant chemotherapy for gastric cancer have been conducted. One metaanalysis of 11 randomized trials of adjuvant chemotherapy following curative resection for localized gastric cancer did not demonstrated a survival benefit (OR = 0.88, 95% CI 0.72–1.08) [3]. The study by Earle et al., which included 13 randomized controlled trials, suggested that adjuvant chemotherapy may
Correspondence: Feng Qi, Department of General Surgery, Tianjin Medical University General Hospital, Tianjin, China. Tel: +86 022 60362255. E-mail: [email protected]
(Received 3 March 2014; accepted 12 March 2014) ISSN 0036-5521 print/ISSN 1502-7708 online 2014 Informa Healthcare DOI: 10.3109/00365521.2014.907337
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Meta-analysis of adjuvant chemotherapy produce a small survival benefit (OR = 0.80, 95% CI 0.66–0.97) [4]. In addition, the meta-analysis conducted by Mari et al. [5] showed that chemotherapy reduced the risk of death by 18% (OR = 0.82, 95% CI 0.75–0.89). But two reports concerning biological immunotherapy and radiotherapy were incorrectly included in the analysis. Panzini et al. [6] also reported that adjuvant chemotherapy raised survival rate (OR = 0.72, 95% CI 0.62–0.84), but patients in an American study [7] did not receive radical surgery on the basis of definite pathology. The analyses above did not exhaustively analyze the influence of adjuvant chemotherapy on postoperative statistical indicators, such as disease-free survival rate, relapse rate, and median survival time. And flaws in the conduct of several analyses made it difficult to draw firm conclusions, including the exclusion of a small but clinically meaningful survival benefit. Besides, with the use of some new chemotherapeutics nowadays, it is worthwhile to reevaluate the role of adjuvant chemotherapy after curative resection for gastric cancer. In this study, we reevaluated the role of adjuvant chemotherapy after curative resection for gastric cancer, included new studies, and illustrated patient types that benefit from adjuvant chemotherapy. Particularly, the safety of chemotherapeutics and tolerance of patients were evaluated exhaustively. Methods Search strategy Computer searches were performed by two independent reviewers. The following databases were searched from inception to May 2013: Medline, Embase, PubMed, the Cochrane Library and CBMDisc. Combinations of the following search terms were used: adjuvant chemotherapy AND (gastric cancer OR stomach cancer OR stomach neoplasm OR gastric carcinoma) AND randomized clinical trial. The languages were set as English and Chinese. References of the included literatures were also tracked. The authors of the papers were contacted to follow-up on the data and the complete text when necessary. Inclusion and exclusion criteria All included papers had to satisfy the following three requirements: the study had to be a randomized controlled trial (RCT), comparing the treatment effects of surgery plus adjuvant chemotherapy with surgery alone; all patients included in the studies had to have primary gastric cancer confirmed by pathology; patients in the experimental group received
691
adjuvant chemotherapy including two intravenous or oral ways not limited the dosage, composition, number of treatments, starting time, and so on; and patients in the control group received surgery alone. We used the following exclusion criteria: the studies were not clinically relevant, such as animal studies; distant metastasis had occurred in the patients; patients got gastroesophageal junction tumors; and the patients had received radiotherapy, immunotherapy, intraperitoneal chemotherapy or molecular-targeted therapy. For studies conducted by the same research institute at different times, the newer, more complete paper was used. Data extraction and quality assessment Data extraction and quality assessment were carried out independently by two investigators. Discrepancies between the two investigators were resolved by discussion or consensus with a senior investigator. The following data from eligible articles were extracted: first author, year of publication, country of study, patient numbers, regimens of adjuvant chemotherapy, the serious adverse events, characteristics of two arms of RCT and follow-up. The quality assessments of include studies were evaluated with criteria of Cochrane Reviewer Handbook 5.2. There were five aspects related to the quality assessment for RCTs used to determine: 1) whether the RCTs was truly random; 2) whether correct concealment of allocation was used; 3) whether proper blinding was conducted; 4) whether loss of follow-up in each group was stated and whether intention-to-treat analysis was performed; 5) whether the baseline of groups was similar. A trial with five “yes” to the questions was regarded as A grade. Studies with answers in partly accordance with the five criteria were taken for B grade. If a study did not fulfill the criteria, it was regarded as C grade. To avoid compounding bias, low-quality studies would not take part in further meta-analysis. Statistical methods The meta-analysis was performed using RevMan5.2 software provided by the Cochrane Collaboration. Subgroup analysis was conducted according to the number of positive lymph nodes, depth of local involvement, population from west or east, the length of follow-up and number of cycles. Heterogeneity was assessed using a Chi-square test. When a p-Value of the Chi-square test was 50% was considered substantial
692
J. Cao et al.
heterogeneity. Reasons for significant heterogeneity were explored by using sensitivity analyses that removal of certain studies from the analysis, as was suggested by the forest plot. Publication bias was assessed with the funnel plot. All p-Values in this study were two-sided. Results
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Quantity and quality of studies Reading of titles or abstracts from 435 articles resulted in 41 potentially relevant literatures [8–48]. After carefully reading the full texts of the 41 researches, 3 studies [37,43,48] were excluded because they were non-randomized concurrent control trials and 2 papers [17,20] were excluded due to receiving immunotherapy of patients. Three trials [10,13,18] were also excluded for patients in the study received palliative resection and another study included gastroesophageal junction tumor [36]. In addition, three studies that were renewed later were excluded [12,21,27]. The remaining 29 RCTs were included for data extraction and quality assessment. In these studies, 8580 patients were randomly allocated, of whom 4260 patients received adjuvant chemotherapy and 4320 patients were in the control group. The quality assessments of included studies were evaluated with criteria of Cochrane Reviewer Handbook 5.2: 14 studies for criteria A [15,16,19,23,24,29,32,33,39,41,42,44,47], 15 studies for criteria B [8,9,11,14,22,25,26,28,31,34,35,38,40,45,46] and 1 study for criteria C [30]. Overall survival rate Twenty-six studies were compared for the difference of survival rate between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.20, I2 = 18%) (Figure 1A). Fixed effects model was used, and adjuvant chemotherapy was also associated with statistically significant reduction in hazard of death as compared with control [RR = 1.09, 95% CI (1.06– 1.23)]. Funnel plot analysis of potential publication bias was showed in Figure 1B. Disease-free survival rate Fourteen studies were compared for the difference of disease-free survival rate between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.09, I2 = 35%) (Figure 2). Fixed effects model was used, and adjuvant chemotherapy showed
significant survival improvement [RR = 1.11, 95% CI (1.07–1.15)]. Relapse rate Sixteen studies were compared for the relapse rate between adjuvant chemotherapy group and surgeryalone group. There was no substantial statistical heterogeneity among the trials (p = 0.07, I2 = 37%) (Figure 3). Fixed effects model was used and adjuvant chemotherapy showed significant survival improvement [RR = 0.79, 95% CI (0.74–0.84)]. Toxicity of chemotherapy Eight studies reported the comparison of toxicity of chemotherapy between two groups, and the results suggested that myelosuppression and gastrointestinal complication were of great proportion. In our study, we analyzed the difference of leucopenia, thrombocytopenia, anemia, nausea and vomiting, diarrhea, abnormality of liver function, mucositis, alopecia and infection between adjuvant chemotherapy group and surgery-alone group (Figure 4). Leucopenia in seven studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.20, I2 = 30%). Fixed effects models were used, and there is a significant difference between the two groups [RR = 13.29, 95% CI (5.67–31.18)]. Thrombocytopenia in six studies was compared between adjuvant chemotherapy group and surgery-alone group. There was substantial statistical heterogeneity among the trials (p = 0.06, I2 = 52%). Fixed effects models were used, and adjuvant chemotherapy group showed no significant difference for thrombocytopenia [RR = 3.04, 95% CI (0.99–9.32)]. The occurrence of anemia in six studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.48, I2 = 0%). Fixed effects models were used, and there is a significant difference between two groups [RR = 1.98, 95% CI (1.13–3.48)]. The occurrence of nausea and vomiting in six studies was compared between adjuvant chemotherapy group and surgeryalone group. There was substantial statistical heterogeneity among the trials (p = 0.0003, I2 = 79%). Random effects models were used, and the incidence of nausea and vomiting in the adjuvant chemotherapy group was more significant [RR = 3.48, 95% CI (2.42–5.01)]. The occurrence of diarrhea in five studies was compared between adjuvant chemotherapy group and surgery-alone group. There was substantial statistical heterogeneity among the trials (p = 0.07,
Meta-analysis of adjuvant chemotherapy
693
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
A
p p
B 0
SE (log[RR])
0.2
0.4
0.6
0.8
1
RR 0.05
0.2
1
5
20
Figure 1. A. Forest plot of the risk ratio (RR) of the overall survival with adjuvant chemotherapy versus surgery alone for gastric cancer. B. Funnel plot of potential publication bias with adjuvant chemotherapy versus surgery alone for gastric cancer.
I2 = 54%). Random effects models were used, and the incidence of diarrhea in the adjuvant chemotherapy group was more significant [RR = 3.90, 95% CI (2.22–6.86)]. The occurrence of abnormality of liver
function in four studies was compared between adjuvant chemotherapy group and surgery-alone group. There was substantial statistical heterogeneity among the trials (p = 0.03, I2 = 67%). Fixed effects models
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
694
J. Cao et al.
p p Figure 2. Forest plot of the risk ratio (RR) of the disease-free survival with adjuvant chemotherapy versus surgery alone for gastric cancer.
were used, and adjuvant chemotherapy group showed no significant difference for abnormality of liver function [RR = 0.86, 95% CI (0.61–1.22)]. The occurrence of mucositis in three studies was compared between adjuvant chemotherapy group and surgery-alone group. There was substantial statistical heterogeneity among the trials (p = 0.09, I2 = 59%). Fixed effects models were used, and adjuvant chemotherapy group showed no significant difference for mucositis [RR = 3.56, 95% CI (0.75–16.84)]. The occurrence of alopecia in 2 studies was compared
between adjuvant chemotherapy group and surgeryalone group. There was no substantial statistical heterogeneity among the trials (p = 0.36, I2 = 0%). Fixed effects models were used, and the incidence of alopecia in the adjuvant chemotherapy group was higher than control group [RR = 48.82, 95% CI (9.78–243.66)]. The occurrence of infection in two studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.93, I2 = 0%). Fixed effects models were used,
p p Figure 3. Forest plot of the risk ratio (RR) of the relapse with adjuvant chemotherapy versus surgery alone for gastric cancer.
Meta-analysis of adjuvant chemotherapy
p
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
p
p p
p p
p p
p p
p p
p p
p p
p p
p p p
Figure 4. Forest plot of the toxicity of chemotherapy with adjuvant chemotherapy versus surgery alone for gastric cancer.
695
696
J. Cao et al.
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
p p
p p
p p
p p
p p p
Figure 5. Forest plot of the relapse sites with adjuvant chemotherapy versus surgery alone for gastric cancer.
Meta-analysis of adjuvant chemotherapy and the incidence of infection in the adjuvant chemotherapy group was higher than control group [RR = 3.49, 95% CI (1.66–7.33)].
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Relapse site In our study, we analyzed different relapse sites between adjuvant chemotherapy group and surgeryalone group (Figure 5). The proportion of liver relapse in seven studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.14, I2 = 38%). Fixed effects models were used, and adjuvant chemotherapy could not reduce the occurrence of liver relapse significantly compared to control [RR = 0.82, 95%
697
CI (0.60–1.11)]. The proportion of peritoneum relapse in 10 studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.86, I2 = 0%). Fixed effects models were used, and adjuvant chemotherapy could reduce the occurrence of peritoneum relapse significantly compared to control [RR = 0.77, 95% CI (0.66–0.90)]. The proportion of lymphoid nodes relapse in nine studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.85, I2 = 0%). Fixed effects models were used, and adjuvant chemotherapy could reduce the occurrence of lymphoid nodes relapse significantly compared to control [RR = 0.58, 95% CI
p p
p p
p p p
Figure 6. Subgroup analysis by the number of positive lymph nodes.
698
J. Cao et al.
(0.45–0.75)]. The proportion of local relapse in nine studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.80, I2 = 0%). Fixed effects models were used, and adjuvant chemotherapy could reduce the occurrence of local relapse significantly compared to control [RR = 0.57, 95% CI (0.41–0.80)].
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Subgroup analysis Studies included were analyzed in subgroups separately according to the number of positive lymph nodes, depth of local infiltrate, patients from Asian or non-Asian, the length of follow-up and number of cycles. Of the studies included in the subgroup analysis, there were 14 studies in which the number of positive lymph nodes ‡2/3 and 11 studies in which the number of positive lymph nodes 6 years [p = 0.07, I2 = 39%, RR = 0.85, 95% CI (0.80–0.91)] or £5 years [p = 0.58, I2 = 0%, RR = 0.92, 95% CI (0.86–0.98)], and fixed effects models were used (Figure 10).
Sensitivity analysis In our study, we selected studies of grade A to make sensitivity analysis (Figure 11). The results showed that there was no substantial statistical heterogeneity among the trials (p = 0.30, I2 = 12.7%), and patients got significant benefit from adjuvant chemotherapy (RR = 0.86, 95% CI (0.80–0.92). Discussion Gastric cancer is still a major health problem and a leading cause of cancer-related death, although its incidence decreased worldwide [49]. As the majority of gastric cancer patients are at the advanced stage at
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
700
J. Cao et al.
p p
p p
p p p Figure 9. Subgroup analysis by the length of follow-up.
the time of diagnosis, their overall prognosis is suboptimal despite aggressive treatment, with an overall survival rate of 20–30% after radical surgical resection in Europe and about 60% in Japan [50,51]. However, it is still in argument whether patients get definite benefit from adjuvant chemotherapy, meanwhile, it is not clear which kind of patients can get the best curative effect. Besides, with the occurrence of some new chemotherapeutics nowadays, it is of great importance to reevaluate the role of chemotherapy for gastric cancer after radical surgery. On the basis of studies in the past, this meta-analysis recorded the data of new published clinical trials to investigate
whether gastric cancer patients after curative resection can benefit from adjuvant chemotherapy. The quality of included studies in this meta-analysis was evaluated strictly. Except for one study in Brazil [30], all other trials used random allocation methods. Almost half of the studies involved clarified concrete randomization, some of which illustrated the methods of allocation concealment. In addition, all studies elaborated the exclude criteria and statistical methods, and the baseline consistency of some factors was analyzed to guarantee the comparability between adjuvant chemotherapy group and control group. All the patients in the trials included were followed
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Meta-analysis of adjuvant chemotherapy
701
p p
p p
p p p Figure 10. Subgroup analysis by the number of cycles.
up, and more than 50% studies extensively described the quantity and reasons of patients who were lost to follow-up (proportion
ORIGINAL ARTICLE
Adjuvant chemotherapy after curative resection for gastric cancer: a meta-analysis
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
JISEN CAO, FENG QI & TONG LIU Department of General Surgery, Tianjin Medical University General Hospital, Tianjin, China
Abstract Objective. The aim of this article is to review up-to-date clinical data published in the literature in regard to adjuvant chemotherapy in patients with gastric cancer after radical surgical resection. Materials and methods. Medline, Embase, PubMed, the Cochrane Library and CBMDisc were searched to identify data published regarding this issue from 1966 to 2013. All the calculations and statistical tests were done using RevMan5.2 software. Results. A total of 29 trials with 8580 patients met all inclusion criteria. Among them, 27 studies reported survival rates at the end of follow-ups, 64.2% alive among 3981 patients in the adjuvant chemotherapy arm and 57.3% alive among 4027 patients in the observation arm. Statistical results showed that the observation arm had a shorter disease-free survival (RR: 1.11, 95% CI: 1.07–1.15), and the treatment arm had a lower recurrence rate (RR: 0.79, 95% CI: 0.74–0.84). Leucopenia, anemia, nausea and vomiting, diarrhea, alopecia and infection occurred more frequently in the treatment arm. Adjuvant chemotherapy decreased the occurrence of peritoneum relapse [RR = 0.77, 95% CI (0.66–0.90)], lymphoid nodes relapse [RR = 0.58, 95% CI (0.45–0.75)] and local relapse [RR = 0.57, 95% CI (0.41–0.80)]. Conclusions. Adjuvant chemotherapy can improve the survival rate and disease-free survival rate and reduce the relapse rate after curative resection. Adjuvant chemotherapy cannot induce thrombocytopenia and mucositis or affect liver function. The tumor in situ recurrence and peritoneum, lymph nodes relapse decrease after chemotherapy, and patients benefit from adjuvant chemotherapy regardless of the numbers of positive lymph node, depth of local invasion, Asian or non-Asian, the length of follow-up, and numbers of cycles.
Key Words: adjuvant chemotherapy, gastric cancer, meta-analysis
Introduction Gastric cancer is one of the common malignant tumors in the world with a high incidence and mortality. Almost a million new cases of gastric cancer are reported worldwide each year [1]. Surgical resection remains the mainstay of curative treatment for patients with gastric cancer. However, the 5-year– survival rate has been disappointing. Even for those patients who undergo radical resection, the rate of recurrence is still high [2]. Therefore, an increase in the postoperation cure rate is a major concern. The evaluation of adjuvant chemotherapy after surgery for gastric cancer is underway in many clinical trials, particularly in the light of evidence of a benefit
from such therapies in common tumors of the gastrointestinal tract, such as colorectal cancer. Adjuvant chemotherapy for gastric cancer has been administered for over 40 years; however, varied results were documented, and whether patients can benefit from chemotherapy after curative resection has yet to be clarified. Several meta-analyses of adjuvant chemotherapy for gastric cancer have been conducted. One metaanalysis of 11 randomized trials of adjuvant chemotherapy following curative resection for localized gastric cancer did not demonstrated a survival benefit (OR = 0.88, 95% CI 0.72–1.08) [3]. The study by Earle et al., which included 13 randomized controlled trials, suggested that adjuvant chemotherapy may
Correspondence: Feng Qi, Department of General Surgery, Tianjin Medical University General Hospital, Tianjin, China. Tel: +86 022 60362255. E-mail: [email protected]
(Received 3 March 2014; accepted 12 March 2014) ISSN 0036-5521 print/ISSN 1502-7708 online 2014 Informa Healthcare DOI: 10.3109/00365521.2014.907337
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Meta-analysis of adjuvant chemotherapy produce a small survival benefit (OR = 0.80, 95% CI 0.66–0.97) [4]. In addition, the meta-analysis conducted by Mari et al. [5] showed that chemotherapy reduced the risk of death by 18% (OR = 0.82, 95% CI 0.75–0.89). But two reports concerning biological immunotherapy and radiotherapy were incorrectly included in the analysis. Panzini et al. [6] also reported that adjuvant chemotherapy raised survival rate (OR = 0.72, 95% CI 0.62–0.84), but patients in an American study [7] did not receive radical surgery on the basis of definite pathology. The analyses above did not exhaustively analyze the influence of adjuvant chemotherapy on postoperative statistical indicators, such as disease-free survival rate, relapse rate, and median survival time. And flaws in the conduct of several analyses made it difficult to draw firm conclusions, including the exclusion of a small but clinically meaningful survival benefit. Besides, with the use of some new chemotherapeutics nowadays, it is worthwhile to reevaluate the role of adjuvant chemotherapy after curative resection for gastric cancer. In this study, we reevaluated the role of adjuvant chemotherapy after curative resection for gastric cancer, included new studies, and illustrated patient types that benefit from adjuvant chemotherapy. Particularly, the safety of chemotherapeutics and tolerance of patients were evaluated exhaustively. Methods Search strategy Computer searches were performed by two independent reviewers. The following databases were searched from inception to May 2013: Medline, Embase, PubMed, the Cochrane Library and CBMDisc. Combinations of the following search terms were used: adjuvant chemotherapy AND (gastric cancer OR stomach cancer OR stomach neoplasm OR gastric carcinoma) AND randomized clinical trial. The languages were set as English and Chinese. References of the included literatures were also tracked. The authors of the papers were contacted to follow-up on the data and the complete text when necessary. Inclusion and exclusion criteria All included papers had to satisfy the following three requirements: the study had to be a randomized controlled trial (RCT), comparing the treatment effects of surgery plus adjuvant chemotherapy with surgery alone; all patients included in the studies had to have primary gastric cancer confirmed by pathology; patients in the experimental group received
691
adjuvant chemotherapy including two intravenous or oral ways not limited the dosage, composition, number of treatments, starting time, and so on; and patients in the control group received surgery alone. We used the following exclusion criteria: the studies were not clinically relevant, such as animal studies; distant metastasis had occurred in the patients; patients got gastroesophageal junction tumors; and the patients had received radiotherapy, immunotherapy, intraperitoneal chemotherapy or molecular-targeted therapy. For studies conducted by the same research institute at different times, the newer, more complete paper was used. Data extraction and quality assessment Data extraction and quality assessment were carried out independently by two investigators. Discrepancies between the two investigators were resolved by discussion or consensus with a senior investigator. The following data from eligible articles were extracted: first author, year of publication, country of study, patient numbers, regimens of adjuvant chemotherapy, the serious adverse events, characteristics of two arms of RCT and follow-up. The quality assessments of include studies were evaluated with criteria of Cochrane Reviewer Handbook 5.2. There were five aspects related to the quality assessment for RCTs used to determine: 1) whether the RCTs was truly random; 2) whether correct concealment of allocation was used; 3) whether proper blinding was conducted; 4) whether loss of follow-up in each group was stated and whether intention-to-treat analysis was performed; 5) whether the baseline of groups was similar. A trial with five “yes” to the questions was regarded as A grade. Studies with answers in partly accordance with the five criteria were taken for B grade. If a study did not fulfill the criteria, it was regarded as C grade. To avoid compounding bias, low-quality studies would not take part in further meta-analysis. Statistical methods The meta-analysis was performed using RevMan5.2 software provided by the Cochrane Collaboration. Subgroup analysis was conducted according to the number of positive lymph nodes, depth of local involvement, population from west or east, the length of follow-up and number of cycles. Heterogeneity was assessed using a Chi-square test. When a p-Value of the Chi-square test was 50% was considered substantial
692
J. Cao et al.
heterogeneity. Reasons for significant heterogeneity were explored by using sensitivity analyses that removal of certain studies from the analysis, as was suggested by the forest plot. Publication bias was assessed with the funnel plot. All p-Values in this study were two-sided. Results
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Quantity and quality of studies Reading of titles or abstracts from 435 articles resulted in 41 potentially relevant literatures [8–48]. After carefully reading the full texts of the 41 researches, 3 studies [37,43,48] were excluded because they were non-randomized concurrent control trials and 2 papers [17,20] were excluded due to receiving immunotherapy of patients. Three trials [10,13,18] were also excluded for patients in the study received palliative resection and another study included gastroesophageal junction tumor [36]. In addition, three studies that were renewed later were excluded [12,21,27]. The remaining 29 RCTs were included for data extraction and quality assessment. In these studies, 8580 patients were randomly allocated, of whom 4260 patients received adjuvant chemotherapy and 4320 patients were in the control group. The quality assessments of included studies were evaluated with criteria of Cochrane Reviewer Handbook 5.2: 14 studies for criteria A [15,16,19,23,24,29,32,33,39,41,42,44,47], 15 studies for criteria B [8,9,11,14,22,25,26,28,31,34,35,38,40,45,46] and 1 study for criteria C [30]. Overall survival rate Twenty-six studies were compared for the difference of survival rate between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.20, I2 = 18%) (Figure 1A). Fixed effects model was used, and adjuvant chemotherapy was also associated with statistically significant reduction in hazard of death as compared with control [RR = 1.09, 95% CI (1.06– 1.23)]. Funnel plot analysis of potential publication bias was showed in Figure 1B. Disease-free survival rate Fourteen studies were compared for the difference of disease-free survival rate between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.09, I2 = 35%) (Figure 2). Fixed effects model was used, and adjuvant chemotherapy showed
significant survival improvement [RR = 1.11, 95% CI (1.07–1.15)]. Relapse rate Sixteen studies were compared for the relapse rate between adjuvant chemotherapy group and surgeryalone group. There was no substantial statistical heterogeneity among the trials (p = 0.07, I2 = 37%) (Figure 3). Fixed effects model was used and adjuvant chemotherapy showed significant survival improvement [RR = 0.79, 95% CI (0.74–0.84)]. Toxicity of chemotherapy Eight studies reported the comparison of toxicity of chemotherapy between two groups, and the results suggested that myelosuppression and gastrointestinal complication were of great proportion. In our study, we analyzed the difference of leucopenia, thrombocytopenia, anemia, nausea and vomiting, diarrhea, abnormality of liver function, mucositis, alopecia and infection between adjuvant chemotherapy group and surgery-alone group (Figure 4). Leucopenia in seven studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.20, I2 = 30%). Fixed effects models were used, and there is a significant difference between the two groups [RR = 13.29, 95% CI (5.67–31.18)]. Thrombocytopenia in six studies was compared between adjuvant chemotherapy group and surgery-alone group. There was substantial statistical heterogeneity among the trials (p = 0.06, I2 = 52%). Fixed effects models were used, and adjuvant chemotherapy group showed no significant difference for thrombocytopenia [RR = 3.04, 95% CI (0.99–9.32)]. The occurrence of anemia in six studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.48, I2 = 0%). Fixed effects models were used, and there is a significant difference between two groups [RR = 1.98, 95% CI (1.13–3.48)]. The occurrence of nausea and vomiting in six studies was compared between adjuvant chemotherapy group and surgeryalone group. There was substantial statistical heterogeneity among the trials (p = 0.0003, I2 = 79%). Random effects models were used, and the incidence of nausea and vomiting in the adjuvant chemotherapy group was more significant [RR = 3.48, 95% CI (2.42–5.01)]. The occurrence of diarrhea in five studies was compared between adjuvant chemotherapy group and surgery-alone group. There was substantial statistical heterogeneity among the trials (p = 0.07,
Meta-analysis of adjuvant chemotherapy
693
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
A
p p
B 0
SE (log[RR])
0.2
0.4
0.6
0.8
1
RR 0.05
0.2
1
5
20
Figure 1. A. Forest plot of the risk ratio (RR) of the overall survival with adjuvant chemotherapy versus surgery alone for gastric cancer. B. Funnel plot of potential publication bias with adjuvant chemotherapy versus surgery alone for gastric cancer.
I2 = 54%). Random effects models were used, and the incidence of diarrhea in the adjuvant chemotherapy group was more significant [RR = 3.90, 95% CI (2.22–6.86)]. The occurrence of abnormality of liver
function in four studies was compared between adjuvant chemotherapy group and surgery-alone group. There was substantial statistical heterogeneity among the trials (p = 0.03, I2 = 67%). Fixed effects models
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
694
J. Cao et al.
p p Figure 2. Forest plot of the risk ratio (RR) of the disease-free survival with adjuvant chemotherapy versus surgery alone for gastric cancer.
were used, and adjuvant chemotherapy group showed no significant difference for abnormality of liver function [RR = 0.86, 95% CI (0.61–1.22)]. The occurrence of mucositis in three studies was compared between adjuvant chemotherapy group and surgery-alone group. There was substantial statistical heterogeneity among the trials (p = 0.09, I2 = 59%). Fixed effects models were used, and adjuvant chemotherapy group showed no significant difference for mucositis [RR = 3.56, 95% CI (0.75–16.84)]. The occurrence of alopecia in 2 studies was compared
between adjuvant chemotherapy group and surgeryalone group. There was no substantial statistical heterogeneity among the trials (p = 0.36, I2 = 0%). Fixed effects models were used, and the incidence of alopecia in the adjuvant chemotherapy group was higher than control group [RR = 48.82, 95% CI (9.78–243.66)]. The occurrence of infection in two studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.93, I2 = 0%). Fixed effects models were used,
p p Figure 3. Forest plot of the risk ratio (RR) of the relapse with adjuvant chemotherapy versus surgery alone for gastric cancer.
Meta-analysis of adjuvant chemotherapy
p
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
p
p p
p p
p p
p p
p p
p p
p p
p p
p p p
Figure 4. Forest plot of the toxicity of chemotherapy with adjuvant chemotherapy versus surgery alone for gastric cancer.
695
696
J. Cao et al.
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
p p
p p
p p
p p
p p p
Figure 5. Forest plot of the relapse sites with adjuvant chemotherapy versus surgery alone for gastric cancer.
Meta-analysis of adjuvant chemotherapy and the incidence of infection in the adjuvant chemotherapy group was higher than control group [RR = 3.49, 95% CI (1.66–7.33)].
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Relapse site In our study, we analyzed different relapse sites between adjuvant chemotherapy group and surgeryalone group (Figure 5). The proportion of liver relapse in seven studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.14, I2 = 38%). Fixed effects models were used, and adjuvant chemotherapy could not reduce the occurrence of liver relapse significantly compared to control [RR = 0.82, 95%
697
CI (0.60–1.11)]. The proportion of peritoneum relapse in 10 studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.86, I2 = 0%). Fixed effects models were used, and adjuvant chemotherapy could reduce the occurrence of peritoneum relapse significantly compared to control [RR = 0.77, 95% CI (0.66–0.90)]. The proportion of lymphoid nodes relapse in nine studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.85, I2 = 0%). Fixed effects models were used, and adjuvant chemotherapy could reduce the occurrence of lymphoid nodes relapse significantly compared to control [RR = 0.58, 95% CI
p p
p p
p p p
Figure 6. Subgroup analysis by the number of positive lymph nodes.
698
J. Cao et al.
(0.45–0.75)]. The proportion of local relapse in nine studies was compared between adjuvant chemotherapy group and surgery-alone group. There was no substantial statistical heterogeneity among the trials (p = 0.80, I2 = 0%). Fixed effects models were used, and adjuvant chemotherapy could reduce the occurrence of local relapse significantly compared to control [RR = 0.57, 95% CI (0.41–0.80)].
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Subgroup analysis Studies included were analyzed in subgroups separately according to the number of positive lymph nodes, depth of local infiltrate, patients from Asian or non-Asian, the length of follow-up and number of cycles. Of the studies included in the subgroup analysis, there were 14 studies in which the number of positive lymph nodes ‡2/3 and 11 studies in which the number of positive lymph nodes 6 years [p = 0.07, I2 = 39%, RR = 0.85, 95% CI (0.80–0.91)] or £5 years [p = 0.58, I2 = 0%, RR = 0.92, 95% CI (0.86–0.98)], and fixed effects models were used (Figure 10).
Sensitivity analysis In our study, we selected studies of grade A to make sensitivity analysis (Figure 11). The results showed that there was no substantial statistical heterogeneity among the trials (p = 0.30, I2 = 12.7%), and patients got significant benefit from adjuvant chemotherapy (RR = 0.86, 95% CI (0.80–0.92). Discussion Gastric cancer is still a major health problem and a leading cause of cancer-related death, although its incidence decreased worldwide [49]. As the majority of gastric cancer patients are at the advanced stage at
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
700
J. Cao et al.
p p
p p
p p p Figure 9. Subgroup analysis by the length of follow-up.
the time of diagnosis, their overall prognosis is suboptimal despite aggressive treatment, with an overall survival rate of 20–30% after radical surgical resection in Europe and about 60% in Japan [50,51]. However, it is still in argument whether patients get definite benefit from adjuvant chemotherapy, meanwhile, it is not clear which kind of patients can get the best curative effect. Besides, with the occurrence of some new chemotherapeutics nowadays, it is of great importance to reevaluate the role of chemotherapy for gastric cancer after radical surgery. On the basis of studies in the past, this meta-analysis recorded the data of new published clinical trials to investigate
whether gastric cancer patients after curative resection can benefit from adjuvant chemotherapy. The quality of included studies in this meta-analysis was evaluated strictly. Except for one study in Brazil [30], all other trials used random allocation methods. Almost half of the studies involved clarified concrete randomization, some of which illustrated the methods of allocation concealment. In addition, all studies elaborated the exclude criteria and statistical methods, and the baseline consistency of some factors was analyzed to guarantee the comparability between adjuvant chemotherapy group and control group. All the patients in the trials included were followed
Scand J Gastroenterol Downloaded from informahealthcare.com by University of Maastricht on 07/03/14 For personal use only.
Meta-analysis of adjuvant chemotherapy
701
p p
p p
p p p Figure 10. Subgroup analysis by the number of cycles.
up, and more than 50% studies extensively described the quantity and reasons of patients who were lost to follow-up (proportion
