GAC with clinical T1N0 can be treated by either endoscopic therapy or surgery without any adjunctive therapy, while advanced localized GAC should undergo either preoperative therapy or surgery first followed by adjuvant chemotherapy. Adequate lymphadenectomy (D2 dissection) is desired with gastrectomy ¹⁰ . If the depth of invasion suggests eusT1a, then endoscopic resection is preferred according to the Japanese guidelines ¹⁰ . When surgery is performed first, postoperative adjuvant chemotherapy should be considered on the basis of the pathological stage or quality of surgery.

Postoperative adjuvant chemotherapy is the most common strategy in East Asia. The ACTS-GC trial, a phase III trial in Japan, showed that postoperative adjuvant therapy with S-1 for 12 months improved overall survival (OS) (5-year OS 72% versus 61%, hazard rate [HR] 0.67, 95% confidence interval [CI] 0.52–0.82) and relapse-free survival (RFS) (5-year RFS 65% versus 53%, HR 0.65, 95% CI 0.54–0.79) in GAC patients with stage II/III who underwent D2 gastrectomy ¹¹ . The CLASSIC trial, a phase III trial performed in South Korea, China, and Taiwan, showed that adjuvant capecitabine plus oxaliplatin combination given for 6 months after D2 gastrectomy improved OS (5-year OS 78% versus 69%, HR 0.66, 95% CI 0.51–0.85) and RFS (5-year RFS 68% versus 53%, HR 0.58, 95% CI 0.47–0.72) ¹² . These trials are the basis for recommending postoperative chemotherapy after optimal surgery. More intensive regimens have no advantage for postoperative chemotherapy ^{13, 14} . Recently, the JACCRO GC-07 trial suggests that S-1 plus docetaxel has significant advantage over S-1 alone and it has become the standard of care ¹⁵ .

Efficacy of postoperative chemoradiation was shown in the INT-0116 trial; compared with the surgery-alone group, the postoperative chemoradiotherapy group had longer OS (median OS 27 versus 36 months, HR 1.35, 95% CI 1.09–1.66) and RFS (median RFS 19 versus 30 months, HR 1.52, 95% CI 1.23–1.86) ^{16, 17} . However, the quality of surgery was suboptimum in patients enrolled in this trial; D0, D1, and D2 lymph node dissection rates were 54%, 36%, and 10%, respectively. Importantly, the ARTIST and CRITICS trials demonstrated the lack of efficacy of postoperative chemoradiation after D2 or D1+ nodal dissection ^{18, 19} . Therefore, postoperative chemoradiation is not useful if optimal or near-optimal surgery is performed.

The MAGIC trial provided evidence that perioperative chemotherapy (three preoperative and three postoperative cycles of epirubicin, cisplatin, and fluorouracil [ECF]) for resectable GAC could improve cure rates (5-year OS 23% versus 36%, HR 0.75, 95% CI 0.60–0.93) ²⁰ . However, a follow-up analysis of the MAGIC results has not yet been presented. The results overall were suboptimal. Moreover, epirubicin is not necessary ²¹ . The FNCLCC/FFCD trial showed that surgery plus preoperative cisplatin and fluorouracil (FP) improved OS compared with surgery alone (5-year OS 24% versus 38%, HR 0.69, 95% CI 0.50–0.95) ²² . Recently, the MRC-OEO5 trial compared two cycles of FP and four cycles of ECF as perioperative chemotherapy and showed that the two regimens had similar OS (3-year rate 42% versus 39%, HR 0.92, 95% CI 0.79–1.08) ²³ . These results suggest that the addition of epirubicin and longer duration of chemotherapy do not provide any advantage.

The FLOT4 trial compared perioperative (predominantly preoperative) chemotherapy with docetaxel, oxaliplatin, and fluorouracil/leucovorin (FLOT) and ECF/ECX ^{24, 25} . A total of 716 patients were randomly assigned to the ECF/ECX group (n = 360) or the FLOT group (n = 356). FLOT improved median RFS (30 versus 18 months, HR 0.75, p = 0.001) and median OS (50 versus 35 months, HR 0.77, p = 0.012) compared with ECF/ECX. A total of 50% of patients in the FLOT group completed the planned postoperative treatments, while 37% of patients in ECF/ECX completed them. Perioperative complications were similar across the two groups and unacceptably high ^{24, 25} . Grade 3–4 diarrhea, infections, sensory disorder, and neutropenia were more frequent in the FLOT group, while vomiting, nausea, thromboembolism, and anemia were more frequent in the ECF/ECX group. FLOT should not be recommended to every patient. FLOT might be more suitable for only very fit patients with GAC rather than patients with esophageal or gastroesophageal junction adenocarcinoma where chemoradiation followed by surgery is the preferred strategy. FLOT resulted in 90-day mortality of 5% and grade 3–4 diarrhea (10%) and infections (18%). The follow-up in this trial is also short, and it is likely that, with further follow-up, the differences between the two regimens will narrow. If the 5-year OS rate of FLOT is about 40%, then it would not be a major advance.

The efficacy of targeted therapy for perioperative treatment was assessed. The ST03 trial evaluated whether adding bevacizumab to perioperative chemotherapy could improve survival ²⁶ . A total of 1,063 patients were randomly assigned to receive chemotherapy alone (n = 533) or chemotherapy plus bevacizumab (n = 530), and prognosis was similar in the two groups (3-year OS 50.3% versus 48.1%, HR 1.08, 95% CI 0.91–1.29, p = 0.36) ²⁶ . Moreover, anastomotic leak was more frequent in the chemotherapy-plus-bevacizumab group ²⁶ . To date, there is no place for targeted therapy in the preoperative setting. PETRARCA/FLOT6 (ClinicalTrials.gov Identifier: NCT02581462) is evaluating the efficacy of adding herceptin/pertuzumab to perioperative chemotherapy for human epidermal growth factor receptor 2 (HER2)-positive GAC. RAMSES/FLOT7 (ClinicalTrials.gov Identifier: NCT02661971) is evaluating the efficacy of adding ramucirumab (VEGFR2 monoclonal antibody) for HER2-negative GAC.

For esophageal or gastroesophageal junction adenocarcinoma, the CROSS trial documented benefit of preoperative chemoradiation ²⁷ . For GAC, preoperative chemoradiation is an attractive but non-standard option. Several trials have evaluated the efficacy of preoperative chemoradiation ^{28– 31} , and a retrospective study at the M.D. Anderson Cancer Center showed excellent prognosis of GAC patients who had preoperative chemoradiation ³² . To date, phase III trials assessing the value of preoperative chemoradiation in GAC are ongoing. The TOPGEAR trial is assessing the efficacy of adding preoperative radiation to the MAGIC trial regimen ³³ . The CRITICS-II trial is comparing three arms: preoperative chemotherapy followed by surgery, preoperative chemotherapy and subsequent chemoradiation followed by surgery, and preoperative chemoradiation followed by surgery (ClinicalTrials.gov Identifier: NCT02931890). Result of these trials are expected.

The recommended first-line therapy for patients with HER2-negative GAC is a two-drug combination of oxaliplatin (preferred) or cisplatin plus 5-FU or capecitabine. For HER2-positive GAC, the ToGA study showed that trastuzumab should be added to the first-line cytotoxic therapy ³⁴ . Irinotecan or taxane should be considered when platinum-based chemotherapy cannot be tolerated in the first-line setting ^{35– 37} . DCF (docetaxel, cisplatin, and 5-FU) provides marginal OS advantage but its toxicity is significant ^{38, 39} . Therefore, a modified regimen has been evaluated ^{40– 43} . Modified DCF is still one of the options in selected cases, but the routine use of a taxane in the first-line setting should be avoided. Also, ECF is not recommended for metastatic GAC ⁴⁴ .

In the second-line therapy setting, ramucirumab is the only molecular-targeted drug with a confirmed but marginal survival benefit (as a single agent) in a global phase III trial. The RAINBOW study showed that OS as a result of ramucirumab plus paclitaxel was significantly longer than in the placebo-plus-paclitaxel group (median OS 9.6 versus 7.4 months) and thus this regimen is the preferred choice for second-line therapy ⁴⁵ . Ramucirumab monotherapy is not recommended ^{46, 47} . Docetaxel, irinotecan, and paclitaxel have significantly prolonged OS compared with best supportive care (BSC) ^{48– 50} . However, the data on these molecules are based on a few small trials.

There is no established therapy for peritoneal carcinomatosis, which is a common site of metastases in patients with advanced GAC. Systemic chemotherapy or BSC is often recommended ²⁸ . Recently, intraperitoneal chemotherapy or hyperthermic intraperitoneal chemoperfusion (HIPEC) has been assessed as a potential therapy for peritoneal metastases. PHOENIX-GC, a Japanese phase III trial, compared S-1 in combination with intravenous/intraperitoneal paclitaxel and S-1 in combination with intravenous cisplatin (standard therapy in Japan) ⁵¹ . Primary results showed that there was no significant difference between the two groups (median OS 17.7 and 15.2 months), suggesting lack of superiority of the intraperitoneal chemotherapy ⁵¹ . A subgroup appeared to benefit, but this would only be hypothesis generating ^{52, 53} . Also, perioperative intraperitoneal chemotherapy is being assessed ^{54, 55} . The CY-PHOENIX trial showed that intraperitoneal chemotherapy for CY-positive patients (no other site of metastasis) resulted in conversion to CY-negative in 36 patients (94.7%) and 84.2% of the 1-year OS rate ⁵⁵ . Neoadjuvant laparoscopic HIPEC (mitomycin C 30 mg and cisplatin 200 mg) was assessed in a phase II study, which showed that seven patients (37%) had negative peritoneal cytology after HIPEC ⁵⁶ .

HER2 is the only molecularly targeted drug accepted in first-line therapy. The ToGA study showed that adding trastuzumab in chemotherapy for patients with HER2-positive GAC resulted in longer median OS (13.8 versus 11.1 months, HR 0.74, 95% CI 0.60–0.91) ³⁴ . However, the subsequent analysis of the ToGA study showed that all of these differences have decreased by 40%. HER2 status should be tested according to guidance in all candidates for HER2-targeted therapy ⁵⁷ . The randomized phase III GATSBY trial compared T-DM1, conjugated with trastuzumab and emtansine, and taxane in patients with HER2-positive GAC in a second-line setting ⁵⁸ . However, T-DM1 failed to show benefit in OS (median OS 7.9 versus 8.6 months, HR 1.15, 95% CI 0.87–1.51) ⁵⁸ .

TRIO-013/LOGiC, a randomized phase III trial, assessed lapatinib, a dual inhibitor of HER2 and epidermal growth factor receptor (EGFR) ⁵⁹ . However, the addition of lapatinib to chemotherapy did not show benefit in OS for patients with HER2-positive GAC in the first-line setting (median OS 12.2 versus 10.5 months, HR 0.91, 95% CI 0.73–1.12) ⁵⁹ . TyTAN, a randomized phase III study, also showed that lapatinib was not effective for patients with HER2-positive GAC in the second-line setting (median OS 11.0 versus 8.9 months, HR 0.84, 95% CI 0.64–1.11) ⁶⁰ .

EGFR-targeted therapies showed no benefit in any trials for GAC. Two trials—the EXPAND trial and the REAL3 study—failed to show a benefit from the addition of EGFR inhibitor to first-line chemotherapy ^{61, 62} . To date, EGFR inhibitor is evaluated in selected patients who have EGFR overexpression with immunohistochemistry (IHC) 2+ or 3+ (ClinicalTrials.gov Identifier: NCT01813253).

Ramucirumab is marginally effective for second-line therapy based on REGARD and performs somewhat better, as shown in the RAINBOW study ^{45, 46} . The RAINFALL trial assessed whether adding ramucirumab to first-line standard chemotherapy could be effective ⁶³ . A total of 645 patients were randomly assigned to two groups—ramucirumab plus chemotherapy (n = 326) or placebo plus chemotherapy (n = 319)—and adding ramucirumab marginally prolonged progression-free survival (PFS) (median PFS 5.7 versus 5.4 months, HR 0.75, 95% CI 0.61–0.94) but not OS (median OS 11.2 versus 10.7 months, HR 0.96, 95% CI 0.80–1.16) ⁶³ . These results strongly suggest that ramucirumab should not be used as first-line therapy. Bevacizumab, in the first-line setting, was found to have no efficacy when added to chemotherapy in the AVAGAST study or perioperative chemotherapy in the STO3 study ^{26, 64} .

The phase III RILOMET-1 trial compared ECX with and without MET-positive gastroesophageal adenocarcinoma, but this trial was stopped early because of a higher number of deaths in the rilotumumab group; the median OS was 8.8 months in the rilotumumab group (n = 304) compared with 10.7 months in the placebo group (n = 304) ⁶⁵ . The MET gastric trial assessed adding onartuzumab to FOLFOX in HER2-negative GAC but did not significantly benefit OS (median OS 11.0 versus 11.3 months, HR 0.82, 95% CI 0.59–1.15) ⁶⁶ . These results suggest that MET inhibitors are not particularly efficacious in most patients with MET-positive GAC.

The GRANITE-1 study compared everolimus, an inhibitor of mammalian target of rapamycin (mTOR), and placebo in patients with GAC previously treated with standard care ⁶⁷ . Unfortunately, everolimus showed no benefit for OS (median OS 5.9 versus 4.3 months, HR 0.90, 95% CI 0.75–1.08) ⁶⁷ . The PADPAC study assessed adding everolimus to paclitaxel in the second-line setting but did not show significant benefit; the median OS was 6.1 months in the everolimus group (n = 150) compared with 5.1 months in the placebo group (n = 150) ⁶⁸ .

Fibroblast growth factor receptor (FGFR) inhibitor failed to show a benefit in a phase II trial, the SHINE study ⁶⁹ . GAC patients with FGFR2 amplification were randomly assigned to an AZD4547 (an FGFR inhibitor) group (n = 41) or a paclitaxel group (n = 30) and showed similar median PFS: 1.8 months in the AZD4547 group and 3.5 months in the paclitaxel group ⁶⁹ .

Poly (ADP-ribose) polymerase (PARP) inhibitor is potentially useful in cancers with a deficiency in the repair of double-strand breaks ⁷⁰ . In patients with GAC, a phase II trial showed that lower/absent expression level of ATM, which has a key role in activating DNA damage response to double-strand breaks, seemed to benefit from PARP inhibitor ⁷¹ . However, the phase III GOLD study did not show an advantage from the addition of olaparib ⁷² . A total of 643 patients were assigned to the olaparib-plus-paclitaxel arm (n = 263) or the placebo-plus-paclitaxel arm (n = 262), and the median OS values in the olaparib and placebo groups were 8.8 and 6.9 months, respectively (not significant) ⁷² . Even in the ATM-negative population, olaparib was not effective ⁷² .

Cancer stem cells are often resistant to therapy that is non-specific and thus can be targeted to overcome resistance ⁷³ . Thus far, few clinical trials have assessed the effect of inhibiting stemness-related pathways, such as the Hedgehog and signal transducer and activator of transcription 3 (STAT3) pathway. Vismodegib, an inhibitor of Hedgehog signal by binding smoothened (SMO), was assessed in a phase II randomized study, but adding vismodegib to FOLFOX did not prolong PFS (11.5 versus 9.3 months, p = 0.34) ⁷⁴ . Moreover, the BRIGHTER study showed that adding napabucasin, a STAT3 inhibitor, to paclitaxel did not prolong OS ⁷⁵ . The trial did not enrich patients by expression of stem cell markers in the tumor cells. It may be that selected patients should be assessed ⁷⁶ . STAT3 inhibitors may be better suited for immune modulation as well.

The isoform 2 of the tight junction molecule claudin-18 (CLDN18.2) is an important component of the tight cell junctions and overexpressed significantly in GAC ⁷⁷ . IMAB362, an inhibitor of CLDN18.2, was assessed in clinical trials. In a randomized phase II trial, IMAB362 plus epirubicin, oxaliplatin, and capecitabine (EOX) significantly prolonged PFS (median 5.7 versus 7.9 months, HR 0.5, 95% CI 0.35–0.78) ⁷⁸ . Moreover, the benefit of IMAB362 was more pronounced in the patients with CLDN18.2 high-expression (70% labeling index) GAC (PFS 6.1 versus 9.1 months, HR 0.46; OS 9.3 versus 16.6 months, HR 0.44) ⁷⁸ . A phase III trial is ongoing.

Matrix metalloproteinase 9 (MMP9) has a key role in extracellular matrix remodeling and angiogenesis, and its inhibition in combination with FOLFOX led to potential improvement of prognosis in a phase I study ⁷⁹ . A phase III trial assessing potential benefit of MMP9 inhibition has completed accrual (ClinicalTrials.gov Identifier: NCT02545504) ⁸⁰ .