**4. Conclusion**

(86%) gastric, 43 (8.8%) GEJ, and 26 (5.3%) esophageal cancers, obtained from commercial providers, was subsequently included for HER2 testing and analyses. This group was used to evaluate the concordance between different HER2 assay methods, which were performed in two central laboratories and local laboratory. The researchers observed high agreement rates between two different FISH methods, FDA‐approved Dako HER2 IQFISH pharmDx FISH assay and PathVysion HER2 FISH assay (Abbott Molecular, Inc.), for detecting HER2 ampli‐ fication. The concordance rate was also high (95%) between two central laboratories when evaluating results of FISH assays, whereas the concordance between local and central labora‐ tories was 87%. Expression of HER2 was tested using the FDA‐approved IHC test HercepTest (Dako Biotechnology). Comparison of local laboratory and central laboratory HER2 testing using IHC assay for the assessment of HER2 status in patients assigned to TRIO‐013/LOGiC trial showed that the concordance rate was less than 50%. Comparison of agreement between IHC and FISH assays in central laboratories showed 88% overall agreement for cases from the commercially obtained upper gastrointestinal carcinomas and 91% for the TRIO‐013/LOGiC cohort. Additional analyses confirmed the findings of Hecht and colleagues [45] that progres‐ sion‐free survival as well as overall survival was significantly higher in selected groups of patients, such as Asian patients and younger patients [45, 86]. These findings correlated well with the fact that these patients had higher levels of HER2 gene amplification. Interestingly, other studies also reported similar outcomes in GC patients with high HER2 amplification sta‐ tus when treated with monoclonal antibody trastuzumab [87, 88]. These findings pointed out clinically important aspect, which could underlie the discrepancy between studies and clini‐ cal trials, evaluating the benefit of anti‐HER2 targeted chemotherapies. First, HER2 expres‐ sion patterns differ between GC and breast cancer and furthermore, in GC the expression patterns are frequently heterogeneous [44, 46, 86, 88]. The optimal cutoff for selecting patients with GC who would benefit from addition of lapatinib to chemotherapy should be evaluated in further studies; however, at present, the results indicated that the cutoff value, based on FISH assays, could be the ratios 5.01–10.0 and >10.0 [86]. Second, it was also recognized that other alterations could affect the treatment with lapatinib. For example, it was established that in trastuzumab‐resistant breast and esophagogastric cancers, MET amplification could contribute to intrinsic or treatment‐acquired resistance to trastuzumab [89, 90]. Studies of breast and lung cancers have indicated that overexpression of other tyrosine kinases, includ‐ ing IGF‐1R, other members of HER family, and EphA2, could lead to development of resis‐ tance mechanisms against anti‐HER2 drugs, by bypassing anti‐HER2 inhibition of MAPK and PI3K/Akt signaling pathways [91, 92]. Therefore, additional studies, focusing on molecular biomarkers for selection of eligible patients for anti‐HER2 therapy, could improve the efficacy and safety of small molecular HER2 inhibitors as well as the safety of anti‐HER2 antibodies. Several other small molecule inhibitors, which have been approved for use in treatment of other cancers, are being tested in clinical studies. For example, sunitinib, which inhibits cel‐ lular signaling by targeting PDGFRs and VEGFRs, has been evaluated as safe for treatment of GC patients in a few Phase I studies; however, Phase II studies have not confirmed its efficacy and benefit [93–96]. The safety and benefit of apatinib, which selectively inhibits VEGFR2, have been shown in Phase II and Phase III clinical trials [97, 98]. However, recent reports from other studies have raised concerns regarding the toxicity of apatinib, since it has shown toxic‐

94 Gastric Cancer

ity in previous studies on patients with metastatic triple‐negative breast cancer [99].

In recent years, only two targeted therapeutics, trastuzumab and ramucirumab, have been approved in Western countries for treatment of advanced GC, which is less than the num‐ ber of approved biological drugs for use in other common cancers. Several reasons could be responsible for that. First, although the explosion of knowledge on molecular mechanisms involved in human diseases has led to novel perspectives in medical treatment and diagnos‐ tic procedures, it appears that the enormous amount of molecular and biological informa‐ tion and the complexity of bioinformatic approaches, used to decipher the experimental data, in reality impede the transition from basic research to clinical applications. Second, clinical trials as well as basic research, utilizing novel high‐throughput techniques, revealed great heterogeneity among populations. The consequences of interracial differences are particu‐ larly evident in the field of developing novel small‐molecule drugs and antibodies. Genetic background in populations appears to account for unequal effectiveness and different safety profiles of targeted therapies in different population [100]. In addition, intratumor heteroge‐ neity found within individuals further complicates the development of effective drugs. There is common consensus that novel molecular determinants should be investigated in order to establish genetic profiles, which would enable the identification of the patient subpopula‐ tions, in which the treatment with targeted anti‐cancer agents would be most effective and beneficial. The first milestone in this process involves determination of different genetic land‐ scapes of GCs across the world, followed by tight collaborations between researchers, health‐ care practitioners and pharmaceutical companies. In addition, bioinformatic exploitation of biomedical data collected in databases and utilization and aggregation of already available research data from clinical studies and basic research could provide additional opportunities to identify disease‐specific genetic profiles and establish suitable prognosis prediction models, which could guide personalized treatment management.
