Conversion surgery for stage IV gastric cancer: a multicenter retrospective study

Although stage IV GC includes a mixture of tumor types with various metastatic characteristics and biology, recent advances in the use of chemotherapy have made CS a feasible and promising option for this disease. However, there is a paucity of data regarding the long-term outcomes and prognostic factors that enable the selection of stage IV GC patients who might be good candidates for CS. Our multicenter retrospective study demonstrated that IR stage IV GC (category 1) and R0 resection were significant predictors of favorable OS after CS. Among patients with initially UR disease, those with peritoneal dissemination alone (category 3), which is commonly uncurable and usually not considered for CS, had a favorable prognosis only after CS with R0 resection. Thus, the results of the present study may prove helpful in selecting patients with this highly heterogeneous disease who might respond to CS.

We found that the 3-year OS rate after CS was significantly better for the IR group (78.3%) than the UR group. Our definition of IR corresponded to category 1 defined by Yoshida et al., and included PAN No. 16a2 and/or 16b1 metastasis, a solitary liver metastasis of < 5 cm, and CY1 [13]. Kinoshita et al. reported that patients with potentially resectable disease, such as PAN No. 16a2 and/or 16b1 metastasis or fewer than three peripheral liver metastases, had significantly better prognosis than patients with initially UR disease [10]. However, CY1 differs from IR PAN or liver metastasis because it indicates the presence of microscopically unresectable disease. Although S-1 adjuvant chemotherapy following R1 resection due to CY1 has a relatively good outcome and prognosis, some studies reported favorable OS for patients with CY1 who underwent R0 resection after chemotherapy-induced conversion to a cytologically negative status (CY0) [11, 18‐22]. In our study, all patients in the IR group with CY1 achieved CY0 and underwent R0 resection after chemotherapy, which contributed to the favorable prognosis in this group. The median time from the initiation of chemotherapy to CS in CY1 patients was 103 days, which was significantly shorter than duration of UR patients (data not shown). Yoshida et al. showed that the median duration of preoperative chemotherapy was 84 days for P0CY1 and R0 resection was achieved in 73.8% of patients with P0CY1 disease [11]. In CY1 GC, R0 resection rate may be higher even after a shorter duration of chemotherapy compared to UR. IR include stage IV GCs which are resectable before chemotherapy and which are potentially resectable after a shorter duration of chemotherapy compared to UR. Thus, we suggest that patients with IR are possible to be treated differently from those in UR.

We achieved a R0 resection rate of 79.7%, which is comparable to the range of 49.5–100% following CS for stage IV GC obtained in earlier studies [4, 5, 7, 10, 23]. The previous reports also showed that patients who achieved R0 resection had significantly better OS rates than patients who did not and found that R0 resection was an independent favorable prognostic factor for OS [4, 9]. Whether R1 resection is an acceptable outcome for CS is another issue. R1 resection is often observed in stage IV GC patients with CY1; indeed, this was true for 10 of the 11 patients who achieved R1 resection in the present study. The 5-year OS rate for patients with CY1 has been shown to be > 20% [18]. In our study, all of the patients with R1 resection due to CY1 were in the UR group and the 3- and 5-year OS rates were 20% and 0%, respectively, despite having received additional chemotherapy after CS (data not shown). The reason for this poor prognosis compared with the results of the previous report may be the presence of chemoresistant or poorly susceptible residual cancer cells after chemotherapy. These cancer cells are difficult to eliminate even with postoperative chemotherapy and may result in peritoneal dissemination. Surgeons should therefore confirm that cytology is negative before CS, especially in patients with UR disease. Another common reason for R1 resection is positive microscopic resection, and Solaini et al. reported that this was the only risk factor for recurrence after CS [4]. Therefore, R0 resection may be of great importance to observe improved survival in patients undergoing CS for stage IV GC.

Peritoneal dissemination is the most common noncurative factor in GC and is included in categories 3 and 4 stage IV GC. Patients with tumors in these categories are expected to have a poor prognosis even when R0 resection is accomplished [24]. However, some studies have shown favorable survival outcomes in category 3 patients who underwent CS [4, 5, 7]. For example, Leonardo et al. reported a 3-year OS rate of > 35% [4], Yamaguchi et al. reported an MST of 22.0 months for category 3 patients with CS and an MST of 33.3 months for those achieving R0 [5], and Chen et al. reported an MST for category 3 patients of 43.6 months [7]. In the present study, we observed an MST of 43.0 months and 3- and 5-year OS rates of 65.5% and 32.3%, respectively, for category 3 patients who underwent R0 resection, which is consistent with the results of previous reports. Fukuchi et al. reported that the combination of several noncurative factors, such as T4b, P1, H1, M1, and CY1, but not each factor individually, was associated with poor survival of patients who underwent CS [9]. These findings suggest that, although category 3 patients have a low R0 resection rate compared with category 1 or 2 patients, they are still good candidates for CS if R0 resection is possible.

In category 4, the conversion rate and the incidence of R0 resection was low, and the OS in the CS patients was significantly shorter than compared with the other categories [5, 7]. In the present study, the incidence of R0 resection was the same for category 4 and category 3 patients, but the 3-year OS rate was 0% for category 4 patients, even if R0 resection was achieved. CS is considered to be less effective for category 4 patients than for the other categories. However, Yoshida et al. demonstrated that in category 4 the 5-year OS was more than 50% and the prognosis was significantly longer in patients who underwent R0 resection than in those who underwent R1 or R2 resection, so R0 resection for category 4 is controversial [11].

The optimal timing of CS and the ideal number of chemotherapy courses before surgery in stage IV GC patients is controversial. In general, the best timing for CS is considered to be attainment of a partial or complete response to chemotherapy after a regimen of 4–6 courses on a 3–5-week cycle. This is based on the results of several studies showing that the median progression-free survival after each regimen for stage IV GC is about 6 months [25‐27]. In the present study, the median number of chemotherapy cycles was 4 and the median time from initiation of chemotherapy to CS was 4.6 months, both of which were comparable to the results of previous studies [9, 28].

In this study, most of the patients received triplet regimens consisted of FU, platinum, and taxane or doublet regimens consisted of FU, and platinum. The phase III trail JCOG1013 did not demonstrate the superiority of docetaxel, cisplatin and S-1 compared with cisplatin and S-1 in chemotherapy-naive patients with advanced gastric cancer [29]. Chen GM et al. reported that the OS of the patients who underwent CS was not different between the patients who received triplet regimen and the patient who received doublet regimen [7]. Therefore, the difference between triplet regimen and doublet regimen is not expected to have a significant impact on OS. However, it is possible that a variety of regimens lead to differences in conversion rates. Yamaguchi T et al. reported that the incidence of conversion to CY0P0 in the patients with CY1 or P1a were 51.4% in the patients with docetaxel, cisplatin, and S-1 regimen and 37.4% in CS regimen [20]. Although the high chemotherapy response contributes to a survival benefit [30], high histological response did not improve survival in this study. In other CS studies, chemotherapy response before CS and histological response are also not one of prognostic factors [7‐10]. This result supports the negative association between histological response and survival in the present study. Further research should be needed to reveal conversion rates by regimens and a potential association between chemotherapy response and survival.

This study has some limitations. First, this is a retrospective study with small sample size. Lage-scale multicenter randomized trails are needed to explore the role of CS and to determine the best treatment strategy for CS. Second, it was a multicenter retrospective study, and the indication and timing of CS differed between institutions. However, the treatment strategy for patients with stage IV GC who were given chemotherapy was to convert to surgery if R0 resection was possible, which was common to all participant institutions. Thus, there was little bias in patient indication of CS for stage IV GC. Third, the chemotherapy regimens were not unified, so the real impact of the regimen on CS outcomes was not clear. Forth, the study population included only patients who underwent CS, and whether CS had a survival benefit compared with chemotherapy alone is thus unclear. The MST after 1st line chemotherapy trails and CS is approximately 15 months [25‐27]. The MST in the present study was much longer than 15 months, even for patients with UR or R1/R2. Moreover, most patients who underwent CS had histologically viable cancer cells in their resected specimens, and CS may have contributed to their improved survival by removing these cells. Nevertheless, we believe that our findings of this study provide useful information for the consideration of treatment strategies for stage IV GC.