Efficacy and safety for Apatinib treatment in advanced gastric cancer: a real world study

Apatinib has been proved to be effective and safe among patients in gastric cancer in Phase II and III Trials. We aimed to evaluate its efficacy and safety in real world practice, and to explore factors associated with efficacy. Between January 2015 and February 2017, totally 36 patients with advanced gastric adenocarcinoma or adenocarcinoma of gastroesophageal junction (GEJ) were enrolled and followed up retrospectively after failing at least two lines of systemic therapy. The mPFS was 2.65 months (95%CI 1.66–3.54), and mOS was 5.8 months (95%CI 4.77–6.83). Two patients achieved partial response, and nineteen achieved stable disease. The disease control rate (DCR) was 58.3%, and objective response rate (ORR) was 5.6%. Common grade adverse events were hypertension (38.9%), proteinuria (36.1%), and neutropenia (33.3%). And the most common adverse events over grade 3 were hand-foot syndrome (8.3%), anemia (5.6%), and diarrhea (5.6%). No treatment-related death was documented during the drug administration. Exploratory analyses indicated patients treated with antiangiogenic therapy previously were more likely to benefit from apatinib.

All patients had received previous treatment, including but not limited to gastrostomy, chemotherapy, radiotherapy and targeted therapy. Twenty-four patients (66.7%) hadn't undergone any surgery while 10 patients (27.8%) had radical surgery and 2 patients (5.6%) had palliative surgery. Five patients (13.9%) had radiotherapy due to positive surgical margin. Most patients had received doublet or triplet chemotherapy in the first and second line therapy and the most important chemotherapeutics were Platinum, Taxanes and Fluorouracil. All 8 patients with positive Her2 status had received anti-Her2 therapy in the first or second line according to guidelines. Nine patients (25%) had undergone antiangiogenic therapy in first or second line therapy involving bevacizumab and apatinib. In these 9 patients' previous treatment, bevacizumab was combined with therapy including XELOX, DOF and Everolimus, while apatinib was applied alone.
Complete clinical and pathologic characteristics at the initiation of apatinib therapy are shown in Table 1, and the previous treatment details are provided in Tables 2 and 3.

Characteristics
No. % As shown in Table 4, seven factors including age group, prior gastrostomy, liver metastasis, peritoneal metastasis, previous targeted therapy (anti-Her2 and antiangiogenic) and combination therapy, were brought into our exploratory analysis in consideration of clinical practice and previous study. Cox regression model showed a significant association between apatinib therapy PFS and prior antiangiogenic therapy (adjusted HR = 2.935, 95%CI 1.047-8.229, P = 0.041). Besides that, factors considered as potential markers associated with the efficiency like liver metastasis, peritoneal metastasis and number of metastasis didn't make any differences in our study, probably because of small sample size. Data of univariate analysis and multivariate analysis were shown in Table 4.
These findings are consistent with those from previous clinical trials. No treatment-related death was documented during the drug administration. All adverse events are listed in Table 5.    Table 4. Exploratory analysis of factors to predict PFS of Apatinib. (Antiangiogenic therapy included bevacizumab or apatinib therapy).
Our study is a real-world observation of the efficacy and safety of the apatinib therapy for patients with advanced gastric cancer or advanced GEJ cancer. In this study of the third or more line treatment for gastric cancer, apatinib therapy led to a median PFS of 2.65 months (95%CI 1.66-3.54), a median OS of 5.8 months (95%CI 4.77-6.83), an ORR of 5.56% and a DCR of 58.33%. Common adverse events were hypertension (38.9%), proteinuria (36.1%), and neutropenia (33.3%). And the most common adverse events over grade 3 were hand-foot syndrome (8.3%), anemia (5.6%), and diarrhea (5.6%). As shown in Table 6, our apatinib therapy seems to be nearly efficient and safe compared with phase II and III trials 11,12 . However, this is a real world study, an observational study, which mingled with some complicated factors. There is still some dissimilitude between them. On one hand, patients' performance status before apatinib in this study was much worse than in clinic trails. Five   Table 6. Comparison with previous studies. (Survival and response data in Phase II Trial were from intent-totreat (ITT) patients. In Phase III Trial, survival data were from the full analysis set (FAS) patients, and response data were assessed by investigators).
patients ECOG PS was 2/3 (13.9%) in our study, while all patients in previous clinic trails scored 0 or 1. In our study, 83.3% of the patients had more than two metastatic lesions, while the corresponding rate in phase III trial was 21%. Only 33.3% the patients in present study had received surgery, involving radical and palliative surgery at the same time, while the rate in phase III trial was 69.3%. Besides that, patients in our study were heavily pretreated in consideration of the three patients who had received over four lines chemotherapy. These discrepancies illustrate patients in real world perform worse and have more visceral metastases and higher tumor burden, and highlight the gap in the baseline of apatinib treatments between the randomized controlled trials and real-world treatment. On the other hand, dose titration was more flexibly in our study although initial apatinib dosage of 500 mg once daily in our study was situated between the two clinic trials, which could be increased to 750 mg or decreased to 250 mg. Combination with other therapy was also allowed according to their actual performance status in our study, which wasn't covered in trials. We believed it was these rectifications in the treatment method that we had obtained similar efficacy results with previous trials, even if patients were heavily pretreated and performed worse. Moreover, these rectifications, especially dose up-regulation and combination chemotherapy, didn't increase the incidence of adverse events, which meant a well-behaved tolerance of patients. Preclinical data demonstrated that vascular endothelial growth factor (VEGF) was continuously expressed during oncogenesis, tumor growth and metastasis, by facilitating tumor angiogenesis, and prolonged exposure to VEGF inhibitors could delay tumor growth and even maintain tumor regression [21][22][23] . Continuous angiogenic blockade strategy has been evaluated in the clinical settings and been proved to benefit patients with metastatic colorectal cancer [24][25][26] , but there is still no evidences directly supporting this concept in gastric cancer. In our exploratory analysis, multivariate analysis indicated prior antiangiogenic therapy was an independent factor associated with PFS of apatinib therapy. Nine patients had undergone antiangiogenic therapy in prior treatment, involving apatinib and bevacizumab. As shown in Table 3, four patients received bevacizumab combined with chemotherapy or targeted therapy regularly, and 5 received apatinib alone in second or third line before they were brought into our study. After disease progression from previous treatment, they continued antiangiogenic therapy with apatinib and benefited from it, whether a small-molecule TKI selectively targeting VEGFR-2 (apatinib) or a monoclonal antibody targeting VEGF ligand (bevacizumab) was used before. Bevacizumab as a continuous therapy had been discussed in the ML18147 trial and the BEBYP trial, in which bevacizumab was continued or reintroduced after the first progression of bevacizumab interruption, and bevacizumab plus chemotherapy significantly prolonged patients OS and PFS (OS in ML18147, B + C 11.2 mo vs. C 9.8 mo, HR = 0.81, 95%CI 0.69-0.94, P = 0.0062; PFS in BEBYP, B + C 6.8 mo vs. C 5.0 mo, HR = 0.70, 95%CI 0.52-0.95, P = 0.010) 24,27 . But this strategy still needs more exploration and discussion.
Antiangiogenic therapies, no matter monoclonal antibodies or tyrosine kinase inhibitors, were usually combined with chemotherapy, because of the poor efficiency used alone. Preclinical models demonstrated that sustained monoclonal antibody antiangiogenic treatment could create or remodel an environment suitable for normalization of the stable vascular endothelial cells, leading to increased tumor uptake of chemotherapy, which could be a possible explanation for the beneficial effect of this combination therapy 28,29 . However, the combined therapy strategy involving antiangiogenic and chemo therapy in gastric cancer has long been controversies. Most evidences were inclined to support the counterview. Monoclonal antibodies such as bevacizumab or ramucirumab, neither used alone nor combined with chemotherapy could significantly improve patients' survival. For instance, in AVATAR study, ST03 study, REGARD study and RAINBOW study, addition of bevacizumab or ramucirumab didn't show any advantages 8,[30][31][32] , and the same results also occurred in TKIs, like orantinib, pazopanib and sorafenib [33][34][35][36] . Only two researches showed that monotherapy of antiangiogenic TKIs could significantly prolong the PFS as the primary endpoint, including apatinib (Phase III, in 3rd line, apatinib 2.6 mo vs. placebo 1.8 mo, HR = 0.44, 95%CI 0.54-0.94, P < 0.001) and regorafenib (Phase II, in 2nd or 3rd line, regorafenib 2.6 mo vs. placebo 0.9 mo, HR = 0.40, 95%CI 0.28-0.59, P < 0.001) 11,37 . In our study, fourteen patients added other systemic and local treatment, including chemotherapy, radiotherapy and ablation to apatinib therapy. However the combination didn't significantly prolong patients' PFS compared to apatinib, which was ascribed to differences in tumor biology and subsequent treatment between the first and second line settings, or limitations of study designs and patient selection. We also noticed the five patients in Table 3 who progressed from apatinib monotherapy but benefited from the combination of apatinib plus chemotherapy, and we attributed this to the enhancement of apatinib to conventional chemotherapy. In preclinical data, apatinib significantly increased the intracellular accumulation of rhodamine 123 and doxorubicin in cells by down-regulating the expression of P-glycoprotein (P-gp, ABCB1) or breast cancer resistance protein (BCRP, ABCG2), then reversed the multidrug resistance (MDR) and significantly enhanced the cytotoxicity substrate drugs 38,39 . Further studies remain needed.
Consequently, rigorous randomized controlled trials are needed to establish whether apatinib with an alternative chemotherapy regimen can benefit patients more than apatinib monotherapy, whether apatinib with or without an alternative chemotherapy regimen can benefit patients who have progressed after previous antiangiogenic therapy, or whether antiangiogenic therapy history can predict the advantaged population of apatinib therapy.
In spite of the inspiring improvement of PFS and DCR, apatinib has also exposed patients to its toxicity. It draws more and more attention of physicians and patients before considering the administration of apatinib. Hypertension, proteinuria and hand-foot syndrome are the most common AEs in antiangiogenic therapy 7,8,40 . In our present study, the safety profile was almost consistent with previous clinical trials, except that the incidence of certain AE hand-foot syndrome was a little lower 11,12 . Dealing with hand-foot syndrome has always been a tough thing. We attribute the lower incidence of hand-foot syndrome to prophylactic treatment, including moisturization for hand and foot, herbal medicine immersion and supplementation of multi-vitamin. Hypertension, proteinuria were treated in accordance with standard principles. In addition, hematologic toxicities related to apatinib in our study included leukopenia and neutropenia, while common non-hematologic toxicities involved elevated transaminase, hyperbilirubinemia, nausea and vomiting and diarrhea. Given that 66.7% of our patients not received any surgery, the incidence of bleeding events had not increased obviously compared to clinical trials.
It indicates that apatinib does not increase risks associated to antiangiogenic therapy and can be tolerated by patients with heavy tumor burden of primary lesion. Besides that, a puzzle emerged in the present study, how to define the bleeding. For instance, should we categorize positive result of fecal occult blood as hemorrhage in digestive tract? On one side, primary lesion itself of advanced gastric cancer could lead a positive result, which might lead to an overestimated incidence. On the other side, in clinic work, fecal occult blood is usually not monitored as frequently as other laboratory examination such as blood routine examination. Only when patients noticed the red change of their fecal color, or when they experienced abdominal pain or diarrhea symptom, this test would be monitored regularly. Therefore, the incidence was highly likely to be underestimated. On the whole, from the observations in our study and previous trials, we can see that the AEs of apatinib are manageable, based on physician awareness and patient education.
This study offers first-hand efficacy and safety data of apatinib in the real world, which are informative for physicians and patients. Secondly, exploratory analysis provides several clues for selection of patients who are more likely to benefit from apatinib. Thirdly, safety analysis of this study helps gain better knowledge and familiarization with possible side effects and how to deal with them. However, this study was subject to limitations of its retrospective observational methodology, including potential missing data, possible information bias, small sample size and lack of control group. Moreover, quality of life was not formerly assessed, which could have provided more comprehensive information on apatinib toxicities. Furthermore, potential crowds remained to be crystallized.
Taken together, the efficacy and safety profiles of apatinib in this study were similar to previous clinic trials. Heavily pretreated advanced gastric cancer patients can tolerate and benefit from apatinib therapy, which makes apatinib therapy a promising option. And the specific application strategies need further exploration.

Methods
Patients. This is a retrospective real world study approved by the ethics committee of Chinese PLA General Hospital (PLAGH). Patients with advanced gastric adenocarcinoma or adenocarcinoma of the gastroesophageal junction who had progressed or relapsed after undergoing at least two lines of systemic therapy in accordance with the recommendations and guidelines of NCCN (National Comprehensive Cancer Network) in Oncology Department of the Chinese PLAGH between January 2015 and February 2017 were included. Data were obtained from patients' medical history. Demographic and clinical characteristics and previous treatment were evaluated in all patients.
Treatment. Apatinib therapy was initiated from an oral administration dosage of 500 mg once a day, 4 weeks for a cycle, which could be adjusted according to patients' actual performance status ranging from 750 mg to 250 mg once daily. The daily dosage could be decreased to 250 mg due to patients' severe adverse events. The chemotherapy or targeted therapy combined with apatinib was hinged on physicians' determination based on patients' situations. Informed consent was reviewed and signed by the patients or their legal guardian before apatinib therapy.
Only patients who had finished at least one cycle apatinib therapy and evaluated the efficacy were included in this study.
Efficacy and safety. The efficacy of apatinib was evaluated including progression-free survival (PFS), overall survival (OS), objective response rate (ORR), disease control rate (DCR). PFS was defined as time from initiation of apatinib to disease progression or death, whichever occurred first. OS was defined as the duration from the time of treatment initiation to the time of death of any cause or the last follow-up time. Tumor responses were evaluated by computed tomography (CT), magnetic resonance imaging (MRI), bone scan and physical examination every cycle until disease progression, categorized as complete response (CR), partial response (PR), stable disease (SD) and progressive disease (PD) which were confirmed by physicians according to Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 criteria. ORR was estimated as the percentage of CRs and PRs. DCR was considered as the percentage of CRs, PRs, and SDs.
All adverse events (AEs) were reviewed and determined from patients' medical history and laboratory examination results or from telephone follow-up according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0.
Statistical analysis. PFS and OS and their corresponding 95% confidence intervals (CIs) were estimated by Kaplan-Meier method. The hazard ratios (HRs) and corresponding 95% confidence intervals (95% CIs) were estimated using the Cox's proportional hazards regression model.
Quantitative variables are presented as median (range) or number of patients (percentage). Exploratory analysis for potential factors to predict PFS of Apatinib was a two-step process, consisting of univariate analysis and multivariate analysis. Univariate analyses were performed with the Log-rank test. Factors included in the univariate analysis were age (<median = 58 vs. ≥58), prior gastrostomy (yes vs. no), liver metastasis (yes vs. no), peritoneal metastasis (yes vs. no), previous anti-Her2 therapy (yes vs. no), previous antiangiogenic therapy (yes vs. no) and combination therapy (apatinib only vs. combined with chemo or other targeted therapy). Multivariate analyses were performed with the Cox's proportional hazards regression model based on results of the univariate analyses. All crude and adjusted HRs and 95% CIs were estimated. All statistical analyses were two sided. HR < 1 implied a lower risk of progression for patients. A statistical significance cutoff of p = 0.05 was used to retain the variables in the final model.
Responses and AEs were both aggregated in the form of frequency counts and percentages. The ORR and DCR analyses were based on frequency counts.
All the statistical analyses were performed using SPSS for Windows (version 21; IBM ® , Armonk, NY).