Nomogram to predict overall survival of patients receiving radical gastrectomy and incomplete peri-operative adjuvant chemotherapy for stage II/III gastric cancer: a retrospective bi-center cohort study

Although several nomograms have been established to forecast recurrence or the survival probability of GC patients [8, 10‐11], none have focused on patients who received incomplete PAC (defined as 1–5 cycles in the present study). Moreover, whether the exact cycles of PAC was independently related to long-term outcomes has not previously been clarified. In this retrospective study of 1,070 consecutive patients with stage II/III GC who underwent radical gastrectomy and incomplete PAC in two tertiary hospitals in China, we confirmed that the number of chemotherapy cycles, along with BMI, total gastrectomy and stage III disease were significantly linked with prognosis. Thereafter, for the first time, a nomogram was constructed to predict survival probability, exploring the influence of the number of chemotherapy cycles in these patients. Further analyses established that the nomogram had satisfactory calibration and good discrimination to predict the 1-, 3- and 5-year survival probabilities. It seemed to be more consistent with clinical practice than the eighth version of TNM staging system.

PAC has been established as the most effective adjuvant management for LAGC patients [3‐4], but it is common to encounter patients who could not complete all of the planned dose intensity. In one of our previous studies [16], we reported that among the entire 2,510 patients with LAGC, 546 patients (21.8%) underwent no AC and another 1,044 patients (41.6%) received 1 to 5 cycles of PAC. Only 920 patients (36.7%) underwent ≥ 6 cycles of PAC. Not surprisingly, patients with adequate PAC (≥ 6 cycles) had significantly better prognosis. It was echoed by Lu and colleagues [9], who reported that 425 patients (25.6%) among a cohort of 1,657 did not receive even one cycle of AC in a retrospective multicenter study. Further analyses confirmed that AC significantly decreased recurrence but the exact cycles of AC were not reported. More importantly, in the well-known CLASSIC study [14], 174 of 520 patients (33.5%) with stage II-IIIB GC could not complete the planned 8 cycles of adjuvant oxaliplatin and capecitabine therapy. Post hoc analysis identified that patients who received at least 6 cycles had significantly prolonged survival than those with low dose intensity. Another of our previous studies also confirmed that patients who were given ≥ 6 cycles of PAC had significantly better cancer-specific survival (CSS) [15]. Thus, patients undergoing 1–5 cycles were defined to have incomplete PAC in the present study. As shown in Table 2, the prognosis was comparable in patients undergoing 1, 2 or 3 cycles of PAC, but significantly worse than those who underwent 4 or 5 cycles of PAC. Consistence with previous studies, our findings confirmed the relationship between dose intensity and prognosis, and thus, emphasized the importance of completing as many PAC cycles as possible in stage II/III GC patients [9, 14‐16].

The inclusion criteria, investigated variables and conclusions differed significantly in predictive models for GC patients. Liu et al. [8] established a nomogram to predict CSS in 688 stage II or III patients who had received more than 4 cycles of AC after resection. They concluded that inflammatory, nutritional and tumor markers, tumor location and the stage were significant predictors for CSS. The C-index of their nomogram based on these factors was 0.714, which was higher than the TNM stage (0.630, P < 0.001). Park and colleagues [10] developed a new staging system and a nomogram for advanced GC patients without adjuvant managements. They confirmed that post-operative morbidity, age and American Society of Anesthesiologists (ASA) score were independent predictors for OS. However, it should be noted that only 185 patients were included in the training group. In another retrospective study that included 639 stage I to III GC patients (except T1a) who underwent ≥ 2 cycles of AC within 2 months following their operations, single chemotherapy regimens were identified to be associated with a poorer prognosis compared to multiple chemotherapy regimens [11]. Unfortunately, the exact number of cycles of AC was not given. In addition, there has been increasing evidence favoring PAC instead of AC for the treatment of locally advanced GC, even in Eastern countries [7, 19]. It is noteworthy that almost all previous similar studies excluded those involving the administration of NAC, which inevitably harmed the generalizability of the conclusions [8‐9, 11].

The prognostic factors and recurrence patterns might be different among patients given different dose intensities. In the study presented by Kanda et al. [21], 70 pairs of stage II/III patients were analyzed after propensity score matching. The authors identified that stage III, pT4, vessel invasion, total gastrectomy and carcinoembryonic antigen levels ≥ 5 ng/mL were independent predictors for long term survival in patients given no AC, whereas only a macroscopic tumor size ≥ 5 cm was significantly associated with survival in patients who received adjuvant S-1. In our previous studies, we have confirmed that complete PAC (≥ 6 cycles) could cancel out the adverse influence of a low prognostic nutritional index, low BMI, peri-operative blood transfusion and/or infections on survival in stage II or III GC patients [15‐17]. Given that PAC could impact recurrence patterns and modify the predictive factors for survival in GC patients [22‐23], we must cite the previous conclusions with caution and a new nomogram was clearly needed in patients treated incompletely with PAC.

Consistent with our previous findings, we also revealed that a poor nutritional status, late TNM stage and total gastrectomy were significantly related to poor prognosis [8‐9, 11, 16‐17, 22]. In addition, the exact dose intensity was confirmed to be an independent predictor of prognosis for the first time. Previous studies revealed that poor patient condition (age ≥ 65 years, ASA score ≥ 3), a poor nutritional and immune status (serum albumin level < 35 g/L, prognostic nutritional index < 43.9, BMI < 20.3 kg/m²), body weight loss and post-operative infection complications adversely affected compliance with AC [15, 24‐27]. The possible explanations included that poor patient condition and nutritional status increased chemotherapy-related adverse events [28]. Post-discharge oral nutritional supplements significantly decreased weight loss and chemotherapy modifications after 3 months of invention [29]. On the other hand, NAC was confirmed to be a protective factor for patients to complete PAC, especially in those at a high risk of experiencing post-operative complications [16, 30]. Taken together, the findings suggest that in order to increase the completeness of PAC and improve prognosis, performing NAC and nutritional intervention are feasible strategies. But further prospective studies are still needed.

The present study had several limitations. First, it was a retrospective study and some important variables, such as tumour markers, the type of chemotherapy regimen, the exact time to initiate treatment, the exact number of patients having dose de-escalation, changing chemotherapy regimen or were put on single agent treatment was not collected and analyzed, which might also act as confounders and impact the reliability of our conclusions. Second, given that it was a long study period over 10 years, several chemotherapy regimens combinations were used in our institutions, such as S-1 alone, ECF, FLOT, XELOX and SOX [17]. The convenience and safety of different regimens might also impact the completion of PAC. Third, the median follow-up of 30 months of the entire cohort seemed not long enough to analyze later relapse and the deaths of patients. Fourth, for the 102 patients (9.5%) undergoing NAC, the pre-treatment clinical TNM stage was used, which might be inconsistent with the pathological TNM stage. Fifth, given that only a small proportion of patients underwent laparoscopic surgery, whether the conclusions were applicable to these patients still needs further investigation. Last but not least, our findings still need external validation, especially in patients from Western countries, where NAC is recommended as standard treatment for LAGC and the management strategy differed significantly from our institutions [3, 30].