Incidence and risk of remnant gastric cancer after gastrectomy for gastric cancer: a population-based study from the SEER database

This SEER-based study revealed that patients subjected to partial gastrectomy for GC had a 7.70-fold increased risk of developing GC than the U.S. general population and the 20-year cumulative incidence was 1.88%. As far as we know, this is the first U.S. population-based study on RGC following GC gastrectomy.

The incidence of RGC varies widely across previous studies (Table 2). A Japanese nationwide survey reported the precise incidence of metachronous RGC was 2.94% after 5-year follow-up [16], while a meta-analysis (mostly Asian studies) exclusively on RGCs after GC gastrectomy revealed a lower cumulative incidence of 1.2% [7]. Moreover, Mak et al. reported a pooled prevalence of 2.6% through meta-analysis, but the value is higher in European populations (5.8%) than that in Chinese populations (1.6%) [3]. In our study, we estimated the cumulative incidence of RGC after gastrectomy for GC in the U.S. population to be 0.57% at 5 years and 1.88% at 20 years after FPGC diagnosis. These discrepancies could be explained by several reasons. Different criterion was used to identify RGC by each study, while we followed the SEER criteria to define MPs which were considered more stringent [17, 18]. Moreover, a 1-year latency was used to exclude synchronous RGCs which might be missed at FPGC diagnosis. We also calculated the cumulative incidence with competing risk methods, which account for the high death rate of GC and produce lower estimates than traditional Kaplan-Meier methods [19]. Thus, we provided a reliable estimate of the incidence of RGC after gastrectomy for GC in the U.S. population, which was similar to that of the Asian populations.

Male sex, older age, depth of invasion, intestinal histologic type, and macroscopic type have been confirmed as independent risk factors for RGC after GC gastrectomy in previous studies [9‐13], with which most of our findings were accordant. Intriguingly, we also identified widowed or single marital status, and lower income as independent risk factors (Fig. 3), which have not been reported before. Similar association was observed in the development of de novo GCs [27, 28]. Living with a partner improves overall well-being while widowed or single people might be at higher risk of bad lifestyle [29]. Similarly, lower income discourages the adoption of healthier lifestyle choices [30]. Patients with these factors are more likely to encounter carcinogens for GC such as tobacco and alcohol. This phenomenon may also be explained by that individuals with worse socioeconomic status have less chance to identify and treat precancerous lesions since they might be less adherent to GC surveillance [31]. Special attention should be directed to these vulnerable groups in post-gastrectomy management to prevent RGC.

PG and pylorus-preserving gastrectomy (PPG) are speculated to be risk factors for RGC because preserving the distal stomach or pylorus reduces bile reflux and increases the possibility of H. pylori infection [9, 11, 13]. Contrarily, we found no association between RGC development and tumor site or surgical mode. Although patients with tumor of NOS site had significantly higher risk (HR = 1.53, Fig. 3), it should not be considered as an independent risk factor given the uninformative nature. In fact, the role of bile reflux-H. pylori relation in RGC development remains at debate [5, 32]. Although Roux-en-Y reduces biliopancreatic reflux compared to Billroth-I reconstruction [33], the incidence of RGC following Roux-en-Y was reported significantly lower, questioning the hypothesis [16]. This inconsistency may also be attributed to SEER database’s incomplete information on tumor site and gastrectomy. In fact, large proportions of patients (59.2%) received unknown mode of partial gastrectomy in this study and specific type of reconstruction was not recorded in SEER database.

Another possible explanation could be the lower H. pylori infection rate in the U.S. than that in Asian countries, especially for urban areas where most of our cases (91.4%) lived [34]. Several factors contributed to this low prevalence, including strain difference [35], cost-effective vaccination [36], and high socioeconomic states [37]. A meta-analysis reported that H. pylori is prevalent in 35% U.S. population while the prevalence is over 50% in Japan, Korea and Italy, where most previous RGC studies took place [38]. Furthermore, this study included only patients of the last two decades, during which the infection rate decreased from 35.9 to 18.4% in the U.S [39]. Further studies are required to clarify the role of H. pylori, tumor site and surgical mode in RGC development.

We found that female patients were at significantly lower risk of developing RGC (HR = 0.67, Fig. 3), which is consistent with previous studies [9‐11]. However, in comparison with the general population, female had higher SIR than male in all subgroups (Fig. 4), indicating a more prominent increased risk than male. This discrepancy might be owing to the higher incidence of de novo GC in male than that in female [1, 7, 40], potentially diluting the increased risk for RGC in male patients. Similar phenomenon has been observed in second primary lung cancer [41] and colorectal cancer [42], both known as male-predominant cancer.

It is intuitive that earlier year of diagnosis was associated with a significantly higher risk of RGC because the incidence and risk increased for each year (Figs. 2 and 4B). Nevertheless, our analysis also revealed that age and invasion depth increased the risk of RGC (Fig. 3). Older age has been previously reported as an independent risk factor for metachronous multiple cancers including RGC [9, 42, 43, 44], though this seemed counterintuitive because younger patients with FPGC have more time to develop RGC. Indeed, we did observe a higher SIR in younger patients in comparison with the general population (Fig. 4A). This phenomenon could be explained by the multicentric carcinogenesis theory of SPGCs [9, 45]. RGCs originate from noncancerous mucosa of the gastric remnant which is constantly influenced by the carcinogenic factors. The older the patient is, the longer time has passed to the progression of multicentric carcinogenesis. This time-dependent manner could also explain the increased RGC risk with deeper depth of invasion [6, 8, 9]. Also, the incidence of overall GC is higher in older people, which thus dilutes the SIR of RGC, similar to other second primary cancers [42, 43, 44].

In our analysis, patients underwent gastrectomy for GC were at significantly higher risk of developing RGC than the general population across all strata (SIR > 4.5, Supplementary Table S4). Taking together the time-dependently increasing SIR and cumulative incidence of RGC (Figs. 2 and 4C), our findings suggested that patients underwent gastrectomy for GC should receive lifelong endoscopic surveillance. Current guidelines require a follow-up for the first 5 years after surgery, but individualized follow-up plans beyond that period was also encouraged to detect RGCs [46]. Thus, risk factors found in our study could help tailoring surveillance strategies for individual patients.

There were some limitations to this study that should be considered when interpreting the results. Firstly, our analyses used the latencies from FPGC diagnosis instead of those from gastrectomy because the date of surgery is unknown in SEER database. Secondly, recurrences and metastases might be recorded mistakenly in the database, which could overestimate the rate of RGCs. However, this bias might be tiny because SEER follows criteria for primary tumors strictly [47]. Moreover, the generalizability of our study might be limited because SEER areas are more urban. This cohort had a higher proportion of T1 patients (27.4%), and a lower percentage of White patients (46.1%) than that of non-SEER areas [48]. Lastly, several important factors in GC carcinogenesis such as H. pylori infection, smoking, alcohol use, comorbidities, reconstruction methods etc., were not included in our analyses because SEER database does not collect these data. Given the difficulty to conduct a randomized controlled trial on RGC, multi-center studies using prospectively collected data from a finer scale would be needed in the future to compensate for the limitations of this database study.