Predictive Value of Thymidylate Synthase Expression in Gastric Cancer : A Systematic Review with Meta-analysis

Gastric cancer (GC) is the 4th most common cancer and the 2nd most common cause of cancer mortality in the world (Jemal et al., 2011). It has been proven that fluoropyrimidine can significantly improve survival in GC patients. In the advanced GC, fluoropyrimidine has been widely used as the mainstay of chemortherapeutic agent (Van Cutsem et al., 2006; Koizumi et al., 2008). In localized disease, a large proportion of patients who were at risk of relapse after curative resection have benefited from adjuvant therapy. Adjuvant chemotherapy with fluoropyrimidine is an accepted standard of care in many parts of the world (Sakuramoto et al., 2007). The antitumor effect of fluoropyrimidine mainly stems from its competitive inhibition of thymidylate synthase (TS). TS is a rate-limiting enzyme in the synthesis of 2’deoythymidine-5’monophosphate, which is an essential precursor for DNA biosynthesis (Santi et al., 1974). Intratumoral TS expression in vivo may be pivotal


Introduction
Gastric cancer (GC) is the 4 th most common cancer and the 2nd most common cause of cancer mortality in the world (Jemal et al., 2011).It has been proven that fluoropyrimidine can significantly improve survival in GC patients.In the advanced GC, fluoropyrimidine has been widely used as the mainstay of chemortherapeutic agent (Van Cutsem et al., 2006;Koizumi et al., 2008).In localized disease, a large proportion of patients who were at risk of relapse after curative resection have benefited from adjuvant therapy.Adjuvant chemotherapy with fluoropyrimidine is an accepted standard of care in many parts of the world (Sakuramoto et al., 2007).
The antitumor effect of fluoropyrimidine mainly stems from its competitive inhibition of thymidylate synthase (TS).TS is a rate-limiting enzyme in the synthesis of 2'-deoythymidine-5'-monophosphate, which is an essential precursor for DNA biosynthesis (Santi et al., 1974).Intratumoral TS expression in vivo may be pivotal

Predictive Value of Thymidylate Synthase Expression in Gastric Cancer: A Systematic Review with Meta-analysis
Hua-Bin Hu 1& , Lei Kuang 2& , Xiao-Min Zeng 3 , Bin Li 1 , En-Yi Liu 1 , Mei-Zuo Zhong 1 * in predicting tumor sensitivity to fluoropyrimidine, as TS expression has been revealed to be determinant in such predictions in vitro (Berger et al., 1985;Johnston et al., 1992).
After over 10 years of research, although most studies reported poor survival and reduced response to fluoropyrimidine with high TS expressing in tumors, evidence is insufficient to conclude whether TS acts as a predictive marker in gastric cancer.The purpose of this article was to evaluate the scientific evidence for the effect of TS expression on GC outcome, using a standard metaanalysis of data from published studies.In fact, two major meta-analysis were performed separately, one in advanced GC and the other in localized disease undergoing adjuvant therapy.

Search Strategy and Study Selection
The search for studies was performed using the ).We also reviewed the references reported in the relevant studies to identify additional studies.
Studies that met the following criteria were eligible for inclusion: (1) patients had a diagnosis of gastric cancer; (2) all patients received fluoropyrimidine-containing chemotherapy; (3) overall survival, event-free survival, or treatment response to chemotherapy were analyzed stratified by TS expression; (4) the results are part of an original analysis; (5) when results reported by the same author were acquired from the same patient population in more than one publication, only the study involving the highest number of patients was included.

Data Extraction
Two investigators (HB-H and LK) extracted data from the eligible studies independently and reached consensus to all items.Data retrieved from each report included the first author, year of publication, treatment setting, chemoterapy rigemens, TS evaluation method, cutoff value used to dichotomize TS as "high" and "low", number of patients analyzed, proportion of high TS expression.If data from any of the above categories were not reported in the primary study, items were treated as "not applicable."We did not contact the author of the primary study to request the information.

Statistical Methods
For the quantitative aggregation of the results, statistical analysis of the overall hazard ratio (HR) for overall survival (OS) and event-free survival (EFS) (classified as progression-free survival, disease-free survival, time to progression), the odds ratio (OR) for overall response rate (ORR).By convention, for the high TS expression group, an observed HR >1 implied a worse prognosis, and OR <1 indicated a poor response to fluoropyrimidine-containing regimens.The impact of TS expression was considered to be statistically significant if their 95% CI did not overlap 1.If these statistical variables were not reported explicitly in the individual study, they were estimated by the methods of Parmar et al. (Parmar et al., 1998).
Heterogeneity test based on I 2 statistic was performed in all meta-analysis.I 2 is measured from 0-100% with increasing I 2 values indicating a larger impact of betweenstudy heterogeneity (Higgins et al., 2002).A randomeffects model was applied to pool study results in all meta-analysis reported below (DerSimonian et al., 1986).
Evidence of publication bias was obtained using the Begg's test (p <0.05 was considered to represent

Eligible studies and Characteristics
Thirty-five studies that met the inclusion criteria were identified.Eleven studies were excluded from further analysis (Figure 1).Five were excluded because survival was assessed in heterogeneous patient cohorts (Kuniyasu et al., 1998;Ishikawa et al., 1999;Tsujitani et al., 2000;Terashima et al., 2003;Chung-Kang et al., 2006), comprising both advanced and localized GC patients, and extraction of separate risk estimates of outcome for patients treated in the advanced or adjuvant disease setting was not possible from data available.In four studies (Lenz et al., 1996;Metzger et al., 1998;Napieralski et al., 2005;Fukuda et al., 2006), patients have received neoadjuvant treatment, which may have altered TS expression, and a treatment-related effect cannot be entirely discounted.Two studies provided insufficient outcome data for effect estimation thus were excluded (Liu et al., 2004;Ishizone et al., 2006).Hence, a total of twenty-four studies with 2,079 patients remained eligible for meta-analysis (Table 1).

Evaluation of TS Methodologies
The most widely-adopted technique to determine TS expression for survival analysis was Immunohistochemistry (IHC) (15 of 24 studies).A number of semiquantitative methods were used to dichotomize TS expression.In three studies (Boku et al., 1998;Yeh et al., 1998;Miyamoto et al., 2000), staining intensity grades lower than 1 or 2 represented low levels of TS expression.In five studies (Choi et al., 2001;Tahara et al., 2004;Boku et al., 2007;Lee et al., 2008;Yeh et al., 2010), expression was dichotomized by quantifying the proportion of stained cells using arbitrary thresholds of 20% or 25%.In two studies (Kwon HC et al., 2007;Kim KH et al., 2011), cases were defined as high expression on the condition that the grades of intensity and extent are both 2 or higher.In four studies (Cho et al., 2006;Kim et al., 2009;Choi et al., 2011;Jeong et al., 2011), from multiplying the grades of intensity by either the grades of extent or the percentage of stained cells, a IHC score was derived so as to dichotomize the levels of TS.In the remaining one study (Suda et al., 1999), the high expression were judged when the cytoplasm of cancer cells showed positive signals compared with stromal inflammatory cells.

Results of Meta-Analysis in the Advanced Disease Setting
The pooled HR for OS across twelve advanced studies 1.43 (95%CI: 1.08 -1.90), indicating that patients with high TS expression had a risk of death 1.43 times greater than patients with low TS expression (Figure 2).However, large heterogeneity was found among these studies (I 2 =74.1%).The analysis of chemotherapy regimen subgroup were performed (Figure 2).HR pooled from studies in which all patients received fluoropyrimidine monotherapy was 2.32 (95%CI: 1.53 -3.50,I 2 =40.7%), indicating that the statistical link between high TS expression and poor OS was rather stronger.Unfortunately, limiting analysis to the studies in which fluoropyrimidine based combination chemotherapy was administrated, TS status is not significantly correlated with OS (HR: 0.94, 95%CI: 0.82 -1.08,I 2 =19.9%).Heterogeneity was not detected in both above subgroups.When grouped according to the method of TS assessment used, the pooled HR was 1.34 (95%CI: 0.88 -2.05,I 2 =14.7%) for IHC and 1.49 (95%CI: 1.05 -2.13,I 2 =85.5%) for RTPCR.
No statistically significant effect of TS on EFS was observed (Figure 3), the pooled HR from four studies was 1.36 (95%CI: 0.88 -2.10), with evidence of study heterogeneity (I 2 =68.8%).When the analysis was limited to the studies in which patients received fluoropyrimidine monotherapy, there was a significant association between high TS expression and poor EFS (HR: 1.76, 95%CI: 1.19 -2.60,I 2 =0%).However, these results should be interpreted with caution due to the small number of contributing studies.
Overall response rate stratified by TS expression was reported by evelen studies (Figure 4).There was evidence of a trend towards reduced response to fluoropyrimidine -containing chemotherapy with high TS expressing (OR: 0.57, 95%CI: 0.31 -1.05,I 2 =60.3%), although this was not statistically significant.When we restrict analysis to the studies in which patients received fluoropyrimidine monotherapy, there was statistical evidence that high TS status indicated poorer response (OR: 0.32, 95%CI: 0.11 -0.95,I 2 =73.6%).

Results of Meta-Analysis in the Adjuvant Disease Setting
In the adjuvant disease setting, no significant effect on OS was observed (Figure 5).The pooled HR from nine adjuvant studies was 1.22 (95%CI 0.82 -1.82), with evidence of study heterogeneity (I 2 =73.8%).The result indicated that high TS expression was not significantly associated with OS in adjuvant disease setting.Seven studies used IHC to test the TS expression, in which the pooled HR was 1.03 (95%CI: 0.70 -1.51,I 2 =69.8%).In the remaining two studies by RTPCR, the pooled HR was 2.77 (95%CI: 1.51 -5.08,I 2 =0%).

Discussion
The results of this systematic review and meta-analysis demonstrate the predictive significance of TS expression  doi.org/10.7314/APJCP.2012.13.1.261Predictive Value of Thymidylate Synthase Expression in Gastric Cancer: A Systematic Review with Meta-analysis level in GC patients treated with fluoropyrimidinecontaining chemotherapy.In the advanced setting including 844 patients, the results suggested that high TS expression was an indicator of poor OS in advanced GC patients.Especially in the subgroup of fluoropyrimidine monotherapy administrated, TS expression has even stronger value in predicting OS, EFS and ORR.Thus, for the elder and the patients who can not tolarance for multi-drug chemotherapy, the predictive value of TS expression may help clinicians choose the optimal single agent.However, in the subgroup of fluoropyrimidine based combination chemotherapy used, TS expression did not significantly predict the treatment outcomes.This may account for that the tumours with high TS expression might respond to other drugs, whereas those tumours were refractory to fluoropyrimidine alone.Therefore accordingly, it may contribute to more accurate prediction of treatment outcomes if we evaluate the interaction between TS and other known predictive factors.
In the adjuvant setting including 1,235 patients, high TS expression was not associated with OS.To localized GC who have received curative surgery, OS may be subject to other more important factors, for instance extent of gastric resection and lymphadenectomy.Interestingly, our results showed that high TS expression was significantly correlated with poor EFS in ajuvant studies.
The value of TS expression in predicting poor OS seems stronger in studies using RTPCR than IHC in both advanced and adjuvant settings.This is partially attributable to the thresholds used in TS status assignment, as in many RTPCR studies the dichotomizations were defined by the maximal χ 2 method and dependent on likely response.This may indicate a source of bias (Altman et al., 1994).
In all meta-analysis reported above, no siginificant publication bias was detected according to Begg's test.However, it should be kept in mind that this methodology is not completely bias-free, because there might have been rejection or even non-submission of negative data existed.In addition, another potential source of bias could be introduced and need to be paid attention as inadequate blinding of survival data from assessors of TS expression.Of all the fifteen studies using IHC, three did not point that their evalution of TS expression was done by assessors who were blind to clinical data.
A statistically significant heterogeneity must be addressed in our report.Firstly, some of the heterogeneity observed might account for different thresholds to define TS status and the wide variation in the proportion of high TS expression in each study.Secondly, varied antibodies for IHC and housekeeping genes for RTPCR were used with no consistent criteria.Thirdly, inadequate sample size was also a frequent problem in the studies included in our analysis, with only five of the twenty-four studies reporting outcomes from over 100 patients.Whilst pooling data may in part address deficiencies in individual study sample sizes, smaller studies are more likely to generate heterogeneity.Thus, a random-effects model was used to estimate the effect of TS high expression on outcomes due to these evidences of methodological heterogeneity across studies.This assumes that the studies were random samples from a hypothetical population of studies taking into account variability within and among studies.
We did no attempt to weigh each study by a quality score, since quality assessment tools for examining prognostic and predictive biomarker studies do not currently exist, and are only beginning to be discussed for prognosis studies in general (Hayden et al., 2006).Evidently the design of some studies is not optimal.For example, dissimilar methodologies and no criteria of threshold used in TS status assignment.Moreover, the majority of survival data were based on smallsized sample and retrospective analysis.In the future research, large multi-centre prospective studies should be conducted with the use of standard unbiased methods, with assessors blinded to the clinical data, and include more homogeneous GC patients, to investigate the precise predictive effect of TS expression in GC.

Figure 1 .
Figure 1.The Flow Diagram of Search Strategy

Figure 5 .
Figure 5. Forest Plot of Hazard Ratios for Overall Survival in the Adjuvant Disease Setting