Association of high-risk comorbidity with overall survival among patients with gastric cancer and its sex-specific differences in China: a retrospective observational cohort study

In this hospital-based cohort study, the findings of this study confirm that comorbidities have a significant impact on the OS of Chinese patients with GC, and more importantly, ACCI-measured comorbidities were independent prognostic factors of long-term mortality. Our results highlight the utility of the ACCI as an indicator for OS among patients with GC. To improve the survival of patients with GC who undergo radical gastrectomy treatment, assessing ACCI might be a feasible and useful option for establishing a reasonable treatment strategy for gastric cancer.

Several prospective and retrospective studies have reported the association between the comorbidities/multimorbidities and survival outcomes among patients with GC, but the conclusions have remained inconsistent, some have shown a significant association [20, 24‐27], whereas some have not [18, 28]. In a cohort study of 488 Japanese patients with endoscopic submucosal dissection (ESD) for GC after a 5-year follow-up, Tanoue et al. showed that patients with severe comorbidities, defined by the ASA-PS classification, had significantly shorter survival than their counterparts with nonsevere comorbidities (5-year OS rate, 79.1 vs. 87.7%; P < 0.01); furthermore, severe comorbidities were significantly associated with a 2.56-fold increased mortality risk after adjustment for confounders [24]. Consistent findings were confirmed by another three Japanese cohort studies [12, 20, 29], and an inverse association of high-risk comorbidity with the prognosis was observed, with a nearly 8-fold increased risk of mortality, even after adjustment for confounders in both elderly and nonelderly patients [20]. Likely, Cao et al., using a hospital-based cohort of 639 Chinese elderly patients with early gastric adenocarcinoma (EGAC), found an impaired 5-year OS and increased hazard of mortality risk in elderly patients who had CCI ≥ 2 when compared to their counterparts with CCI < 2 (79.7% compared to 94.4%) [25], which was in agreement with results from another Chinese cohort study [30], supporting that the presence of comorbidities is a clinically significant prognostic factor of overall survival. Notably, the authors also showed that age was an important independent risk factor for impaired OS in elderly patients with EGAC, which is consistent with our study in which we used the ACCI to evaluate comorbidities. Similarly, two recent cohort studies that both included Korean elderly early GC patients demonstrated that patients in the high-risk comorbidity group were associated with poorer OS after a long-term follow-up, with a nearly twice higher risk of mortality compared with their counterparts in the low-risk comorbidity group, which suggested its role as independent prognostic factor affecting the survival of elderly GC patients [31, 32]. These consistent and significant results were also reported an a perspective of Western population [26, 33, 34]. For instance, from a recent Swedish cohort including 2154 GC patients, Asplund et al. found that the risk of all-cause mortality was 1.63-fold higher among GC patients with high-risk comorbidity than those with low-risk over a 3-year follow-up after adjusting for demographic, and histopathological confounders [33]. Lombardi et al. analysed a Western multicenter data, which comprised 20 Italian centres and 717 patients with advanced GC, and showed that a positive assocation of high comorbidity burden with greater mortality risk for advanced GC patients who underwent both open and laparoscopic gastrectomy after a median follow-up of 40 months (HR, 1.35; 95% CI, 1.01–1.81) [26]. Similarly, a nationwide study, which comprised 11,196 French patients undergoing esophageal or gastric cancer surgery, also confirmed that patients in the high risk comorbidity group had 3.86-fold increased risk of 30-day postoperative mortality, when compared with their counterparts in the low risk group [27]. Moreover, due to largely limited literature on young GC patients, De et al., using a large cohort of 70,084 American patients, focused specifically on young adults GC patients aged ≤ 40 years and demonstrated that comorbidity was still a significant and strong predictor of 5-year overall survival for this young subgroup of patients [34]. The findings from our hospital-based study added robust evidence to the literature, and confirms a significant increase in 3-year mortality risk among Chinese GC patients with high-risk comorbidities, with a 70% higher risk than their peers with low-risk comorbidities.

While a higher comorbidity burden was found predictive of increased morbidity among patients with GC, irrespective of age groups; however, there are still controversial and conflicting findings with nonsignificant associations observed in Asian [28, 35], Western [36] and Brazilian [37] patients with GC. An early retrospective study including 85 Japanese patients with early GC aged ≥ 85 years revealed no significant difference in 5-year OS rates between patients with ASA-PS-based high and low comorbidities after a median 39-month follow-up, although a significant favourable OS rate was observed for patients without comorbidities when compared with their counterparts with comorbidities [28]. Yang et al. et Shen et al. also showed that despite a higher CCI in elderly Chinese GC patients, comorbidities had not been identified as a prognostic factor for short- and long-term survival [35, 38]. Consistently, a recent Spanish study using data from 591 GC patients found that comorbidity was not associated with postoperative mortality and other short-term inhospital mortalities in the multivariable analysis [36]. Similar nonsignificant results were also observed in two Brazilian cohort studies that high-risk comorbidity defined by CCI was not associated with recurrence and mortality risk [37, 39]. The potential explanation for this discrepancy may be due to the small sample size or short follow-up. Besides, some studies failed to take into account relevant confounders, or dated population-based studies where common comorbidities, such as cardiovascular diseases, were underreported.

Recently, many studies have expanded the ACCI to appraise short- and long-term outcomes in a variety of cancer conditions and reported a promising prognostic role in laryngopharyngeal cancer [19], esophageal cancer [14], prostate cancer [40] and vulvar cancer [15], as well as other severe medical disorders [41, 42]. This study confirms the significant prognostic effect of the ACCI on OS in Chinese patients with GC, in line with previous studies among various populations [12, 18, 30]. In a retrospective study including 122 patients with esophageal cancer after a median follow-up of 6 years, Aoyama et al. found that the OS and RFS rates at 5 years after surgery were 54.4% and 43.6%, respectively, in the low-ACCI group but decreased to 29.2% and 21.3%, respectively, in the high-ACCI group, with adjusted HRs of 1.932 and 2.24, respectively [14]. Koseki et al. and Maezawa et al. demonstrated that the ACCI had a significant prognostic impact on both short- and long-term outcomes among GC patients undergoing curative gastrectomy after a 5-year follow-up [12, 17]. Consistently, to reduce confounding bias, Lin et al. used the propensity score matching method and showed that high-risk comorbidities defined by the ACCI were positively associated with a 1.386-fold increased risk of mortality, indicating that the presence of high-risk comorbidities was an independent risk factor for OS [30].

Moreover, our findings demonstrated that the ACCI score, but not the CCI score, was significantly associated with OS among patients with GC, highlighting the superior clinical influence of the ACCI over the CCI as the comorbidity risk scoring system. Lin et al. reported the ability of the ACCI to predict survival in GC patients after radical gastrectomy, whereas the CCI was not prognostic factor for OS during a median follow-up of 4 years [30]. Similarly, a population-based study, which included 4508 lung cancer patients, also found that the ACCI had better discrimination and predictive accuracy for prognosis in operated lung cancer patients compared with the CCI and ECI, suggesting a merit of widespread applicability for ACCI [18]. Consistently, another retrospective study of 198 patients with laryngopharyngeal cancer revealed a better prognostic effect of ACCI than CCI after a follow-up of 5 years [19]. To some extent, this indicates the importance of age-related physiology for cancer treatment outcomes, and adjusting age to CCI is a reasonable way to evaluate the role of comorbidities.

In the abovementioned studies about the clinical impact of the ACCI on the prognosis in patients with malignant tumours, the way to choose the appropriate ACCI cut-off value was quite discrepant. Some explored the cut-off value on the basis of previous and similar studies that evaluated all potential endpoints [17, 43], while some used a statistical modelling approach [12, 30]. Koseki et al. [12] and Aoyama et al. [14] established the optimal cut-off value according to survival analysis, with the most significant results at ACCI = 4, while Dias-Santos et al. obtained the cut-off point based on the ROC curve method [16]. Inspired by Lin et al.’s study [30] and others, the present study used the X-tile program to determine the ACCI cut-off value from a long-term survival prospective. X-tile software, developed by Robert Camp et al. [22], is a commonly used program that is specific for biomarker cut-off optimization within a time-dependent framework. Thus, we obtained the best cut-off value of the ACCI at 2 and 4 and then divided the whole cohort into a low-risk (ACCI = 0–2), moderate-risk (ACCI = 3–4) and high-risk comorbidity group (ACCI≥5). In addition, a sensitivity analysis was conducted by using the 4-group ACCI classification provided by Yang et al. [18] and similar results were found; we thus combined two moderate groups of ACCI at 3 and 4 into a single moderate group (results not shown). Although there is not yet a consensus on the definitive ACCI cut-off value due to differences among cancer types, patients’ background characteristics, and sample size, a systematic review of the literature suggested that a cut-off value of 3–5 might be useful for the ACCI [14]. Future studies are warranted to clarify the optimum cut-off value for the ACCI.

The present study demonstrated that the ACCI was an independent prognostic factor for OS in patients with GC. One possible explanation is that the ACCI might be associated with postoperative surgical complications. Indeed, recent studies have observed a significant impact of comorbidities on postoperative complications in gastric cancer [17]. For instance, the incidence of surgical complications in the high-ACCI group was more than 1.5 times than in the low-ACCI group among patients with GC [17, 30]. Consistent results have also been observed in other types of malignancies [14, 43, 44]. Kahl et al. showed that, among patients with advanced epithelial ovarian cancer, high ACCI scores were significantly associated with more than a 3.72-fold higher risk of severe postoperative complications than low ACCI scores [43]. Furthermore, Aoyama et al. reported that the incidence of postoperative complications was 27% in the low-ACCI group and 54% in the high-ACCI group [14]. The onset of postoperative complications is associated with decreased survival or an increased risk of disease recurrence in various types of cancers [45‐47]. Thus, careful attention should be devoted to the possible development of surgical complications in patients with high ACCI scores when undergoing curative gastrectomy or other surgical treatments for gastric cancer.

The current study further demonstrated sex differences in effect of high-risk comorbidity on overall survival among Chinese GC patients, with a significant association among male patients with GC. To the best of our knowledge, there are few studies with specific focus on this sex-specific effect, and most of previous studies included sex as a relevant confounder. Despite this, consistent results with ours were also reported. Of note, two recent cohort studies found a significant higher mortality risk in Japanese GC men than their female counterparts over a follow-up of 5 years (HR, 1.32, 95% CI 1.04–1.66 and 1.29, 1.06–1.58, respectively) after adjustment for comorbidity and other confounders [12, 17]. A recent study, which used a real-world data from a Western population-based perspective with large sample size, showed the similar results [48]. More precisely, Peltrini et al. investigated changes and perioperative mortality over a 6-year period among 14,512 Italian GC patients who underwent gastrectomies, and they demonstrated that sex was independent risk factors for 90-day mortality, with a 1.14-times increased risk of mortality in men than women [48]. Consistent results of higher mortality risk in male GC patients were also reported in other studies using Asian and European data [6, 24, 26] despite lack of significant differences found. Although the reasons underlying these sex differences are still not justified, but seem complex and multifactorial, a higher proportion of unhealthy lifestyle among Chinese men than women may contribute to this. Indeed, men may have a higher prevalence of gastric cancer risk factors, such as poor hygiene, lower consumption of fruits or vegetables, and higher tobacco consumption and alcohol intake, while women are more attentive to health-related information and behaviors [49]. In the present study, the proportion of smokers and alcohol intake was 45.19% and 37.78% among men, respectively, which were substantially higher than that among women (1 and 3 out 164 female patients). These discrepancies could potentially affect or increase severity of their comorbid conditions, which increases the risk of detecting gastric cancer in late stages and, thus, the mortality risk.

The present study has several limitations that should be mentioned. First, despite being solid in terms of the follow-up period, the present study was retrospective in nature and might suffer from some uncontrolled selection bias. Second, our data were derived from a single institution; further validation with additional datasets is warranted. Third, the present study did not have a large sample size which may lead to a reduced statistical power. In addition, the comorbidity scores were assembled retrospectively and may be underestimated because of missing data. However, in terms of capturing comorbidities, inpatient data and main and contributing diagnoses were included, assuring the sound validity of the ACCI. These abovementioned concerns will be considered in further studies.