Published results of randomized controlled trials of second-line therapy versus best supportive care or placebo.
Gastric cancer (GC) is the fourth (in men) and fifth (in women) most commonly newly diagnosed cancer worldwide and the third (in men) and fifth (in women) leading cause of cancer death . Radical resection remains the only way cure, but cure rates following surgery for resectable GC in the West lag behind those in Japan. In addition, many GC patients will suffer recurrence even after curative resection in approximately 50% of cases , which eventually results in a still poor 5-year survival rate of <30% both in the USA  and in Europe . Although the 5-year survival rate in Japan is much better than that in the West, there is a global consensus that advanced GC patients need further anticancer therapy.
Currently, five classes of cytotoxic agents (fluoropyrimidines, platinums, taxanes, topoisomerase inhibitors, and anthracyclines) and new molecular targeting agents have been used in GC. Recent randomized controlled trials (RCTs) in the field of treatment for advanced GC have established first-line chemotherapy, although the approach and management for advanced GC varies from region to region so that there is no worldwide consensus on this matter. The most promising regimens are ECF (epirubicin, cisplatin, and 5-fluorouracil) in Europe , DCF (docetaxel, cisplatin, and 5-fluorouracil) in the USA (V325 trial) , and SP (S-1 and cisplatin) in Japan (SPIRITS trial) . Strikingly, however, the global RCT conducted in the USA (FLAGS trial) failed to exhibit any survival advantage of the SP regimen over the FP (5-fluorouracil plus cisplatin) regimen . Very recently, trastuzumab combined with cisplatin and fluoropyrimidines has been found to be active for human epidermal growth factor receptor-2 (HER2 or ERBB2)-positive GC (ToGA trial) . Although these regimens are active, the treatment progress is painfully slow . The median time to progression (TTP) or progression free survival (PFS) was 7.4 months by ECF , 5.6 months by DCF , 6.0 months by Japanese style SP , 4.8 months by global style SP , and 6.7 months even by adding trastuzumab , suggesting that many advanced GC patients experience failure after first-line chemotherapy.
Interestingly, the median TTP or PFS of these regimens are almost equivalent (5.6-7.4 months) [5-7] while median overall survival times (OS) differ between the studies (Figure 1). The 13 months of median OS by the Japanese SP regimen  was apparently longer than those (8.7-9.2 months) by ECF and DCF regimens [5,6] (Figure 1). In addition, even by the use of the SP regimen, the post-progression survival –the survival length difference between median PFS and median OS– differed considerably, being 7 months in the SPIRITS trial  and 3.8 months in the FLAGS trial  (Figure 1). Such inter-trial differences in post-progression survival are partly attributable to different proportions of subsequent second-line therapy after the failure of first-line chemotherapy, being over 70% in the SPIRITS trial while 30-45% in other trials [6,8,9]. In this regard, patients who retain good performance status at the time of first-line treatment failure are candidates for second-line therapy. Currently, however, no standard regimens for any second-line therapy hitherto determined suggest an urgent need for the establishment of second-line therapy. Reflecting this urgency, clinical research concerning second-line therapy has been recently commenced. In the era of post first-line chemotherapy, this chapter reviews the next research issues deserving of focus in the field of advanced GC treatment.
2. Rationale for second-line therapy
The significance of second-line therapy after disease progression by first-line chemotherapy has been evaluated by several randomized comparisons of survival times between second-line therapy and best supportive care (BSC) or placebos. Various kinds of agents have been investigated in the RCTs of second-line therapy, such as chemotherapy drugs [10-12] and molecular targeting agents [13-15]. Some of these trials have demonstrated significantly improved survival times by second-line therapies as compared with BSC or a placebo (Table 1).
The AIO conducted a RCT to compare irinotecan (21 patients) with BSC (19 patients) . Although this study was closed prematurely due to poor patient accrual, irinotecan could significantly (p=0.012) prolong median OS (4.0 months) as compared to those by BSC (2.4 months). The second RCT conducted in Korea  assigned patients in a ratio of 2:1 to chemotherapy (docetaxel or irinotecan) plus BSC (133 patients) or BSC alone (69 patients). Choice of chemotherapy drugs was left to the discretion of the investigators. Median OS of the chemotherapy arm (5.3 months) was significantly (p=0.007) longer than those of the BSC arm (3.8 months). This survival difference may be partly ascribed to the significantly (p=0.02) greater number of patients receiving further chemotherapy (third-line chemotherapy) in the chemotherapy arm (40%) than in the BSC arm (22%) because median OS was longer (8.0 months) for patients who received subsequent treatment than those who did not (3.76 months), regardless of treatment arm. Of note, the lack of any difference in median OS between docetaxel use (5.2 months) and irinotecan use (6.5 months) implies that both agents are equally active, although bias for patient allocation can not be denied. In the third RCT conducted in the UK , 168 patients were allocated to either docetaxel plus active symptom control or active symptom control alone (84 patients each). The median OS in the docetaxel group (5.2 months) was significantly longer (p=0.01) than that in the active symptom control alone group (3.6 months). Measurements of disease specific quality of life showed benefits for docetaxel in reducing dysphagia and abdominal pain.
The efficacy of molecular targeting agents as a second-line setting has been investigated by three other RCTs [13-15]. First, a GRANITE-1 trial  randomly assigned 656 patients at a 2:1 schedule either to everolimus (n=439) or a matching placebo (n=217). Significantly longer median PFS by everolimus (1.7 months, p<0.0001) than that by placebo (1.4 months) did not reflect a difference in median OS (5.4 months in the everolimus arm versus 4.3 months in the placebo arm, p=0.12). Such a discrepancy, namely a significant PFS difference and nonsignificant OS difference, could be ascribed to the fact of the similar or even longer median OS (4.3 months) in the placebo arm in this study as compared to those (2.4-3.8 months) of BSC arms in other second-line RCTs [10-12]. Therefore, one should remember that the longer survival length in the control arm might diminish any statistical significance even when the experimental regimen is effective. Subsequently, a global REGARD trial  was able for the first time to demonstrate that ramucirumab, a monoclonal antibody of vascular endothelial growth factor receptor (VEGFR)-2, was significantly superior to placebo with regard to median OS (5.2 months versus 3.8 months, p=0.047) and median PFS (2.1 months versus 1.3 months, p<0.0001), respectively. In addition, more patients receiving ramucirumab experienced stable or improved QOL than those in the placebo arm. Finally, apatinib, a tyrosine-inhibitor agent targeting VEGFR with an anticipated anti-angiogenesis effect, has been now tested in a clinical trial (NCT 01512745) . This study aims to determine whether apatinib can improve PFS or OS compared to placebos.
Subsequently, a metaanalysis has very recently elucidated a 36% reduction in the risk of death (p<0.0001) by second-line therapy , suggesting that second-line therapy is effective and should be considered in some segments of GC patients refractory to first-line therapy.
3. Seeking the best regimen of second-line therapy
Several randomized trials have been released to explore the optimal combination and doses for second-line therapy using the currently active agents or those deemed active. The agents investigated by these trials include chemotherapeutic drugs [17-20], a molecular targeting agent , and their combinations [22-26].
The efficacy of chemotherapy as a second-line setting has been investigated in several RCTs. In the TCOG GI-0801 trial , median PFS was significantly longer (3.8 months, n=64) by biweekly irinotecan plus cisplatin than by irinotecan alone (2.8 months, n=63), but median OS did not differ between the arms (10.7 months versus 10.1 months). Notably, the proportion of patients receiving third-line therapy was the same (75%) in both groups. The results of another randomized trial (ECRIN) have been reported in the abstract form , in which irinotecan plus cisplatin (n=84) was compared with irinotecan alone (n=84) in patients refractory to an S-1 containing regimen. WJOG 4007  is a RCT comparing weekly paclitaxel (n=108) with biweekly irinotecan (n=111) in GC patients refractory to fluoropyrimidines plus cisplatin. Although there are no statistically significant differences between weekly paclitaxel and biweekly irinotecan for median OS (9.5 months vs 8.4 months, p=0.38), median PFS (3.6 months vs 2.3 months, p=0.33), or response rate (20.9% vs 13.6%, p=0.24), the median OS of both arms are equally longer than those of the previous second-line studies (Table 1, Figure 2). Several explanations are possible for the longer median OS in this study. One explanation is the lower proportion of patients with severe peritoneal metastasis in this study (25.6%) as compared with those in the AIO  (45%) and Kang  (45%) studies. Another explanation is that the proportion of patients receiving third-line therapy in this study was substantial, being 89.8% in the paclitaxel arm and 72.1% in the irinotecan arm. Interestingly, the third-line therapy was an irinotecan-containing regimen in 75% of patients of the paclitaxel arm and a taxane-containing regimen in 60% of the irinotecan arm. Including later lines, 81% patients in the paclitaxel arm received irinotecan and 68% of patients in the irinotecan arm received paclitaxel. The more prolonged median OS in both arms of the WJOG 4007 than those of the previous studies implies that both irinotecan and paclitaxel in second- and further-line settings could potentially contribute to prolonged survival. The ongoing fourth RCT (JACCRO GC05)  has been comparing irinotecan plus S-1 with irinotecan alone to GC patients refractory to S-1. Since the standard regimen in adjuvant and advanced settings in Japan is respectively S-1 and S-1 plus cisplatin, this study will provide one answer to the clinically important question of whether the prolonged use of S-1 is effective even to the S-1 resistant patients whom the clinicians face.
|Thuss-Patience ||Iriontecan||250-350mg/m2, q3w||21||19||4||2.4||2.5||ND|
|Kang ||Iriontecan or Docetaxel||150mg/m2, q2w or 60mg/m2, q3w||133||69||5.3||3.8||ND||ND|
|COUGAR-02 ||Docetaxel||75mg/m2, q3w||84||84||5.2||3.6||ND||ND|
|REGARD ||Ramucirumab||8mg/kg, q2w||238||117||5.2||3.8||2.1||1.3|
The RCT comparing a molecular targeting agent with a chemotherapeutic agent has been ongoing. HER2-positive advanced GC patients have been allocated to either trastuzumab emtansine or taxane (GATSBY, NCT01641939) .
A combination of molecular targeting agents and chemotherapy agents has been investigated in several RCTs. First, paclitaxel with or without lapatinib, which binds to the intracellular tyrosine kinase domains of ERBB1 and ERBB2 (HER2), was compared in HER2-positive advanced GC (TYTAN) . Neither median OS nor median PFS were prolonged by lapatinib plus paclitaxel (n=132) as compared to paclitaxel alone (n=129). Second, the results of the RAINBOW trial [23,24], which compared ramucirumab plus paclitaxel with placebo plus paclitaxel, have been very recently reported in the abstract form. Sixty hundred and sixty-five GC patients, who were refractory to cisplatin and fluoropyrimidine-containing regimens, were randomized in a 1:1 ratio to receive paclitaxel plus ramucirumab or paclitaxel plus a placebo. The paclitaxel and ramucirumab arm showed a significant benefit in median OS (HR=0.807, p=0.0169) and median PFS (HR=0.635, p<0.0001). The subsequent geographical region-stratified analysis elucidated that the advantage of median PFS and median OS was more evident in Western patients than in Japanese patients (Figure 2). Such a regional difference could partly explained by a higher use of post-discontinuation treatment in Japan (75% of patients) than in Western countries (36-38% of patients), which may eventually mask any potential OS benefit of ramucirumab plus paclitaxel in Japan. Two other RCT trials [25,26] are now ongoing. Paclitaxel with or without everolimus has been compared in patients with advanced GC refractory to fluoropyrimidine containing regimens . Patients will be randomized in a 1:1 ratio for a total of 240 patients per treatment arm. Nimotuzumab plus irinotecan versus irinotecan monotherapy as second-line therapy has been investigated in patients with ERBB1 overexpressed advanced GC . This is a Japan and Korea collaborative RCT allocating patients at 1:1 ratio for a total of approximately 400 patients per arm.
The results of some trials have been published; however, at present, trials in which the primary endpoint has been met have been unfortunately very limited. The positive results hitherto obtained include a prolonged PFS by a combination of irinotecan and cisplatin after the failure of S-1 based first-line chemotherapy, and prolonged PFS and OS by ramucirumab in combination with weekly paclitaxel. The results of other studies, especially using molecular targeting agents, are awaited.
4. Consideration of adverse events
When considering second-line therapy, clinicians must be aware that, by definition, patients receiving second-line therapy are not chemonaive and experience treatment failure; thus, they are more likely to be resistant to therapy than first-line treatment. Furthermore, patient performance status is more likely to deteriorate due to a recurrent cancer burden and treatment against it. Indeed, widely ranging grades of gastrointestinal symptoms-related adverse events occurred in a substantial proportion of patients in the BSC arms of clinical trials, suggesting a poor quality of life even by the BSC. In addition, a more intensive regimen aiming at a prolonged OS may cause a higher likelihood of severe adverse events that may result in treatment discontinuation and limitation. For example, the DCF regimen caused 82% of grade 3-4 neutropenia and 29% of febrile neutropenia. Therefore, balancing efficacy and adverse events should be taken into account when developing second-line therapy. The development of more intensive while less toxic second-line regimens could be positioned as a breakthrough marker of GC treatment, which can then ultimately achieve further survival prolongation without treatment withdrawal.
In this context, another direction for research is to establish less toxic regimens with preserved anticancer activity. As described earlier, the conventional first-line treatment comprises 5-fluorouracil and emetogenic/nephrotoxic cisplatin. Therefore, the challenges for a less toxic regimen are the substitution of oxaliplatin for nephrotoxic cisplatin [27,28] or substitution of oral capecitabine for intravenous 5-fluorouracil . Such substitutions aim to reduce toxicity and maintain organ function, thus avoiding treatment withdrawal and prolonged treatment duration.
Fluorouracil, leucovorin, and oxaliplatin (FLO) were associated with significantly less anemia, nausea, vomiting, fatigue, renal toxicity, and serious adverse events related to the treatment as compared with fluorouracil, leucovorin, and cisplatin (FLP) . In addition, there was a better trend toward improved median PFS by FLO than FLP. Similarly, as compared to SP, S-1 plus oxaliplatin (SOX) resulted in a reduced risk of grade 3/4 neutropenia (41.5% vs 19.5%) and febrile neutropenia (6.9% vs 0.9%) while median PFS was similar between the two regimens (5.4 months versus 5.5 months) . These findings further support the possibility that the substitution of oxaliplatin for cisplatin reduced toxicity while maintaining efficacy. Accordingly, length of hospital stay per cycle of SOX regimen (0.85 day) was significantly shorter than that for the SP regimen (6 days) . Furthermore, capecitabine and cisplatin (XP) versus 5-fluorouracil and cisplatin (FP) proved similar median OS (10.5 months versus 9.3 months), suggesting that the inconvenience of infusional 5-fluorouracil could be replaced by oral capecitabine, thus realizing a more simplified dosing schedule . Whether this concept can be applied to second-line therapy warrants further investigation.
5. Future perspectives
With the development of a new generation of cytotoxic agents and molecular targeting agents, the research field of GC treatment appears to be transitioning into a new era with a focus on the targetable molecules in GC. Several molecular targeting agents are currently being evaluated both in first- and second-line settings. The pending results of ongoing clinical trials motivates researchers to continue the challenge of establishing the best second-line regimens or less toxic combinations for the treatment of advanced GC.
Clarifying the genetic alterations of targeted molecules and their interactions.
Unfortunately, in sharp contrast to colorectal cancer in which there has been a significant achievement in treatment by the use of molecular targeting agents, only one agent (ramucirumab) has been currently proved to be an agent for second-line therapy presumably due to the genetic complexity and molecular heterogeneity of the disease.
Recent findings have highlighted the mechanisms of the actions of molecular targeting agents or patterns of expression of targetable molecules in several tumors. For example, genetic alterations that are critical for cell growth occur with considerably different frequencies, being 60%, 33%, and 32% in cancers of the pancreas, biliary tract, and colon, respectively . Furthermore, certain combinations of targets are expressed in a mutually exclusive or co-occurring manner in the same tumor. A mutually exclusive fashion would be a case being KRAS and BRAF in colorectal cancer, or ERBB1 and KRAS in lung cancer . The co-occurring expression of targets is a case of being ERBB2 and PIK3CA in breast cancer . With regard to GC, although KRAS mutations are initially recognized as an infrequent phenomenon in GC , the alterations and amplifications of other genes --both known or previously unreported-- have been subsequently found in a substantial number of GC patients in a mutually exclusive manner [34,35] or co-amplified manner . These findings create a challenge in the treatment regimen with either a use of each blockade of each targetable molecule or a use of dual- or pan-inhibitors of kinases . In addition, elucidation of the roles of each domain in ERBB2 that plays a role in resistance to anti-HER2 therapy further provides the theoretical basis to modify therapeutic strategies to circumvent this resistance . These advances may expand therapeutic options, thereby making larger proportions of GC patients possible candidates for molecular targeting therapies than previously appreciated. Accordingly, there may be a desperate need for multiple gene profiling which could help establish rational molecular criteria for patient inclusion and exclusion in clinical trials. Patient selection by gene profiling may allow better patient recruitment for those most likely to respond to targeted therapies.
Positive and negative interactions between chemotherapy and molecular targeting therapies.
It should be noted that the recent clinical trials of first-line molecular targeting therapies regrettably resulted in some negative results [39-41] (Figure 2). The addition of cetuximab to cisplatin plus capecitabine had a similar median PFS (4.4 months) and median OS (9.4 months) compared to cisplatin plus capecitabine alone (5.6 months and 10.7 months, respectively)  (EXPAND). Surprisingly, the addition of panitumumab to epirubicin, oxaliplatin, and capecitabine (EOC) resulted in a significantly shorter (p=0.013) median OS (8.8 months) compared to EOC (11.3 months) (REAL-3) . Furthermore, lapatinib in combination with capecitabine and oxaliplatin could not show any significant survival benefit (LoGiC) . These negative results lead to speculation that there may be ideal combinations or compatibility between chemotherapy and molecular targeting therapy. In addition, even the addition of panitumumab to EOC achieved only similar median OS to those of the WCOG 4007 trial (Figure 2) [19,40]. Therefore, a certain combination, such as anti-EGFR antibody and capecitabine or oxaliplatin, may be ineffective or even interfere with the other. This speculation can be supported by the observation that addition of cetuximab to oxaliplatin plus fluoropyrimidine-based chemotherapy failed to demonstrate survival benefits for advanced colorectal cancer .
Need to evaluate the updated molecular profile.
Another obstacle for the progression of second-line treatment lies in the possibility that the molecular profile of GC is likely to change under the stress of treatment. Generally, lower response rates and shorter survival times in patients receiving second-line therapy than chemonaive patients may partly be explained by such molecular changes that lead to a resistance to therapy. Ideally, the on-demand tissue samples from tumor sites currently of interest are needed to assess the updated genetic and molecular patterns and to predict whether the planned second-line therapy is really effective. However, direct tumor tissue sampling and subsequent gene analysis are often hampered by virtually inaccesible tumor localization and by overly small sample volumes to perform gene analysis. It is necessary to discover novel markers which can be alternatives for those obtained only by direct tumor tissue sampling as well as to improvise new methods to assess them in order to select the right patients for the right second-line regimen.
Motivated by the first promising results of trastuzumab use in the ToGA trial, several molecular targeting therapies have been challenged in clinical trials. Continuous efforts should be necessary to clarify the mutation and amplification of targeted molecules, and novel methods for their genetic profiling should probably become part of clinical routines. Ultimately, GC patients harboring unique genetic profiles of targeted molecules should be allocated into specific, suitable trials. Targeted therapies tailored to individual genetic profiles maximize treatment efficacy because this allows the recruitment of selected, most suitable patients rather than unselected ones .
Against the background of survival advantages of second-line therapy over BSC in GC patients refractory to first-line treatment, efforts to establish the most effective regimens have been just begun. Parallel to the development of molecular targeting agents, the investigated regimens comprise doublet or triplet chemotherapeutic agents, molecular targeting agents, and their combinations. In addition, the minimization of adverse events should be taken into account to avoid treatment discontinuation. There is a desperate need to explore genetic mutations of targeted molecules and the interactions between them, to establish novel methods to assess them, to clarify the positive or negative interactions between chemotherapy and molecular targeting agents, and to find regimens in which adverse events are least likely to occur for avoiding treatment discontinuation. These could help determine rational molecular criteria for patient inclusion and exclusion in clinical trials, realize the most efficient patient recruitment for those most likely to respond to therapies, and accelerate the establishment of the most effective second-line therapy that could achieve greater survival prolongation.