Phase II trial of S-1 plus leucovorin in patients with advanced gastric cancer and clinical prediction by S-1 pharmacogenetic pathway

To our knowledge, the current study is the first one-arm and the second phase II trial to evaluate efficacy and toxicity of S-1/LV chemotherapy in AGC patients, which is the first to predict outcomes using S-1 pharmacogenetic pathway approach, on the basis that LV is but a cofactor entering the 5-FU metabolism. The enrollment was between July 2011 and July 2012, which was similar to between October 2011 and December 2012 of that randomized phase II study of S-1/LV versus SOL versus SP for AGC [5].

We reported S-1/LV regimen yielded promising ORR, PFS, TTF, and OS, without combination with platinum, taxane, irinotecan, or trastuzumab as first-line treatment. In previous phase II studies of S-1 monotherapy in AGC and previous S-1 monotherapy arms of phase III JCOG9912 study, SPIRITS study, SC-101 study, GC0301/TOP-002 study, and START study, conventional dose for 4 weeks followed by 2-week rest was usually given [18‐22]. The 2-week S-1/LV here was generally more effective, with a less dose intensity than S-1 monotherapy in these studies. LV can enhance the efficacy of 5-FU by maintaining the plasma 5-FU concentration [23]. Similarly with the previous trial of S-1/LV in mCRC [4], S-1/LV regimen here also demonstrated better efficacy and safety compared with previously reported UFT/LV in AGC [3, 24]. S-1/LV also showed more potential than 5-FU/LV plus oxaliplatin that we reported in terms of efficacy and safety [25].

To reduce the duration of effective drug concentration and keeping an appropriate rest period could be effective method of improving safety. So we modified the schedule as 1-week S-1/LV, followed by 1-week rest period, the same as S-1/LV group in that randomized phase II study. The efficacy S-1/LV here was quite comparable with S-1/LV group in that randomized phase II study, in terms of ORR (41.0 vs. 43%), PFS (4.13 vs. 4.2 m), and TTF (3.70 vs. 4.1 m) [5]. Fewer patients in our study received the second-line treatment (53.9 vs. 77%), which helped explain the relatively not so promising OS. Another reason was that there were 5 patients discontinuing treatment before progression not due to adverse events. Meanwhile, S-1/LV in this study saw very satisfactory safety. The frequencies of each type of AEs were generally lower than those in the phase II trial of mCRC and comparable with those in the randomized phase II trial of AGC [4, 5]. Encouragingly, grade 4 AEs occurred in 1 patient (2.6% neutropenia), similarly for the randomized study (2% grade 4 neutropenia and 2% leucopenia). Each type of the grade 3 AEs occurred in only 2.6–7.7% patients, similar to 2–13% for the randomized study [5]. In our study, significantly fewer patients experienced dose reduction, or delayed courses [4], compared to patients with S-1 monotherapy in previous phase III trials [19, 26]. In our study, gastrointestinal reactions were more common than myelosuppression both in total (89.7 vs. 74.4%) and in grade 3–4 (15.4 vs. 10.3%), and dose reductions or interruptions were chiefly due to diarrhea, stomatitis, anorexia, and vomiting. We also observed dose-limiting toxicity was shifted from hematological to gastrointestinal when S-1 was administered with LV [4]. Probably the capacity of oral intake and gastrointestinal tolerability would be the important indications for this regimen.

Japan Gastric Cancer Association guideline states that S-1 alone could be considered for patient who is not suitable for S-1 plus cisplatin therapy [27]. Previous studies showed S-1 monotherapy could be a reasonable option in the treatment of elderly patients. In our study, the ORR, PFS, OS were sound for patients of age >70 (47.1%, 6.00, 11.40 m) without difference compared with those of age ≤70 (36.4%, 4.00, 11.20 m) and seemed better than previously reported S-1 monotherapy (ORR 14.3–26.3%) and UFT/LV (ORR 22%) for elderly AGC patients [28‐30]. The oral convenience makes the S-1/LV regimen extremely useful clinically, especially for elderly patients.

Genotyping the peripheral blood for CYP2A6 polymorphism, quantifying plasmic protein expression of DPD, TS, TP, and OPRT with ELISA, is more optimal, convenient, quicker than evaluating genes, mRNAs, or proteins in tumor tissue, especially when 5-FU plasmic pharmacokinetics were to be integrated at the same time. Thus, there may be more clinical accessibility and prospects. What is more, the cutoff levels were determined by standard statistic analysis, not the simple median or mean. The plasmic expression of DPD in this study was consistent with previously reported with ELISA; however, no report for plasmic OPRT, TP, or TS was available [7]. The frequencies of CYP2A6 alleles in this population were compatible with other Asian population [31‐33]. We especially examined the CYP2A6*13, because its function in Asian patients was unclear. The 0% of CYP2A6*13 again proved it is rare in Asia.

In clinical setting, numerous studies have reported a low TS, low TP, or a high OPRT expression contributed to a high sensitivity to UFT, S-1, or S-1-based treatment in gastric cancer patients, with or without influencing the PFS or OS [34‐38]. There were also evidences that low TS was a predictor of high response for AGC patients with 5-FU/LV, or 5-FU/LV plus cisplatin/oxaliplatin [10, 39, 40]. However, most studies failed to find the DPD expression related to either response, PFS, or OS and less studies did demonstrate high DPD mRNA was predictor of poor OS, not ORR or PFS in AGC [35, 41, 42]. These clinical findings reflect theoretical roles of TS and OPRT. We found high OPRT/TS alone significantly predicted responding. The resultant high OPRT/TS here revealed preferential use of the OPRT pathway versus TS pathway during 5-FU metabolism. In humans, the preferential use of the OPRT pathway was revealed to correlate with a higher sensitivity to 5-FU [43]. Ichikawa et al. and Tanemura et al. [13, 14] both reported the combination of high OPRT and low TS was more predictive of responders to S-1 or S-1 based chemotherapy in gastric cancer patients than either alone, while the low TP was not. Apart from the above PCR or ELISA methods, the quantitative double-fluorescence immunohistochemistry method, reported by Hashiguchi K, was used to access the protein expressions and their ratios quantitatively and found a significant correlation between OPRT/TS, OPRT/DPD, or OPRT/(TS + DPD) and response to S-1 in the AGC patients, among which OPRT/TS showed the strongest correlation with the clinical response [15]. These three studies generally agree with our finding.

In this current study, low baseline plasmic DPD was not related to response; however, it was related to long OS, compatible with previous study that high intratumoral DPD mRNA was predictor of poor OS, not ORR or PFS [41]. Previous studies reported clinical response to S-1 in gastric cancer was not influenced by intratumoral DPD expression [42]. It can be explained that S-1 has antitumor activity even in tumor with high expression of DPD because of the inhibition of DPD by CDHP [44]. Even though it had no prediction of response, low DPD did relate to long OS. Firstly, S-1 enabled high 5-FU concentrations to be maintained in blood for long periods of time by inhibiting of DPD and 5-FU maintenance was a reason of long survival. Secondly, low levels of intratumoral DPD have been generally shown to predict long survival in gastric cancer patients treated with 5-FU-based chemotherapy [45]. Many patients here received second-line chemotherapy comprising of 5-FU with platinum or other drugs.

We found plasmic OPRT (P = 0.024), TS (P = 0.044) expression significantly inversely, while DPD (P = 0.073), TP (P = 0.080) borderline inversely correlated with AUC_0–24h of 5-FU. The number of CYP2A6 gene variants (P = 0.889) did not correlate with AUC_0–24h. This was consistent with that CYP2A6 gene correlated with tegafur pharmacokinetics, but not with 5-FU pharmacokinetics [16]. AUC_0–24h of 5-FU, not OPRT/TS, was predictive of PFS revealed that other factors may also influence PFS by AUC_0–24h of 5-FU. We did not demonstrate the number of CYP2A6 gene variants correlated with efficacy. Although some studies demonstrated patients having fewer CYP2A6 variants had better PFS in AGC patients with S-1 plus cisplatin, or S-1 plus docetaxel [6, 46], divergences on the relation between CYP2A6 genetic polymorphisms and response existed for both gastric and colorectal cancer patients with S-1 or S-1-based chemotherapy [47, 48]. In our study, second-line treatment excluded S-1 and tegafur. CYP2A6 converts enzymatically tegafur, the effector molecule of S-1, to 5-FU, and this role produces no meaning on second-line treatment.

That high OPRT predicted grade 3–4 AEs as well as affected response can be theoretically understood, and in animal models, oteracil in S-1 was found to inhibit the OPRT by 70% in the small intestine; however, the inhibition was limited to 0–20% in tumor regions without affecting the antitumor activity of 5-FU. Besides, high OPRT/DPD, OPRT/TP, OPRT/TS, OPRT/DPD + TP + TS, OPRT/DPD + TS, OPRT/TP + TS, and OPRT/DPD + TP were all associated with grade 3–4 AEs, among which, high OPRT/DPD exhibited the highest accuracy. Cui et al. [7] reported lower baseline plasmic DPD correlated with higher grade of toxicities in AGC patients with S-1 plus docetaxel by ELISA. Further studies were warranted to decide whether OPRT or OPRT/DPD better predicts grade 3–4 AEs. The literature shows severe diarrhea was dose-limiting toxicity for S-1 in Caucasians and severe neutropenia in East Asians perhaps due to CYP2A6 gene polymorphism [17], while this study did not see its relation to severe diarrhea, neutropenia, or total grade 3–4 AEs and no relation of CYP2A6 gene polymorphism to AUC_0–24h or C_max of 5-FU here helped explain.