In the present study, we examined the correlation of GRIm score with POCs and prognosis in elderly patients with CC. Age, comorbidities, clinical T/N/M status, and POCs were higher in patients with a high GRIm score. Multivariate analysis for POCs identified high GRIm score (OR, 4.146; 95%CI, 2.536–6.777; p < 0.001) as an independent predictive factor. The median follow-up period was 42 months (range, 1–160 months). Five-year RFS (GRIm-H, 79.3% vs. GRIm-L, 92.7%; p = 0.001), OS (76.0% vs. 92.2%, p = 0.001), and non-CSS (84.7% vs. 94.6%, p = 0.003) were significantly lower in the GRIm-H group. Multivariate analysis showed that high GRIm was an independent predictive factor for OS (HR, 2.875; 95%CI, 1.451–5.698; p = 0.002) and non-CSS (HR, 3.650; 95%CI, 1.612–8.265; p = 0.001).
GRIm score was first reported as a prognostic indicator for patients undergoing immunotherapy [13]. The three components of the score (NLR, LDH, and albumin) reflect patients’ nutritional status, systemic inflammation, and immune status [13]. Previous studies have reported that GRIm score correlated with tumor stage, tumor size, BMI, and performance status [15, 16, 20]. In the present study, GRIm score was closely correlated with patient characteristics and tumor conditions, including advanced age, comorbidities, and T/N/M status, in agreement with the previous reports.
No previous study has examined the relationship between GRIm score and POCs, although correlations with POCs have been reported for its components [9, 17–19]. The present results showed that GRIm (p < 0.001), NLR (POCs, 2.793 vs. non-POCs, 2.327; <0.001), LDH (182 vs. 160, p = 0.042), and albumin (3.4 vs. 4.0, p < 0.001) were significantly correlated with POCs (Suppl. Table 1). Furthermore, GRIm score was an independent predictive factor for POCs in multivariate analysis. ROC curve analysis for POCs of each parameter revealed that the AUC for GRIm (0.679) was higher than those for LDH (0.520), NLR (0.606), and albumin (0.641). This finding suggests that GRIm score could be a more useful parameter for predicting POCs compared with its components, for the following reason. We considered that the present cohort of elderly patients was heterogenous in terms of comorbidities and performance status [9]; therefore, the simple parameters of NLR, albumin, and LDH might be influenced by various conditions. As GRIm score is a combined indicator, it might reflect an overall value that is more accurate in elderly patients.
Previous studies have reported that POCs were correlated with poor prognosis [5, 6]. In elderly patients, poor prognosis has been based on early mortality in the first postoperative years, which is due mainly to POCs [21]. The present study found no relation of POCs with poor prognosis, probably for the reason of the lower incidence of POCs and of early mortality in the present study, due to recent improvements in surgical technique and perioperative management.
GRIm score has reported as a prognostic factor for various cancers [14–16]. Feng and colleagues examined the prognostic value of GRIm score in esophageal cancer [14], and found that patients with high GRIm had poor CSS compared to those with low GRIm (10.3% vs. 35.0%, p < 0.001). In the survival analysis of Tian and colleagues regarding the prognostic role and predictive performance of GRIm score for adult patients with colorectal cancer [15] patients with a high GRIm score had poor disease-free survival (DFS) (p < 0.0001) and poor OS (p < 0.0001). Multivariate analysis also indicated that high GRIm score was a potent predictive factor for poor DFS (HR, 1.743; 95%CI, 1.188–2.558; p = 0.0045) and poor OS (HR, 1.622; 95%CI, 1.118–2.355; p = 0.0109). They also reported that because of the strong correlation with outcomes, the combination index (GRIm score) was a more powerful index than single indices such as NLR, LDH, or albumin. In contrast to those results, the present study did not find a relationship between GRIm score and cancer prognosis, possibly for the reason that our study included only elderly patients with CC and there was a relatively high incidence of non-cancer cause of death.
In general, the relative risk of non-cancer cause of death in patients with colorectal cancer is reported to be higher than that in the age-matched general population [7]. In 2016, the incidence of non-cancer cause of death in patients with colorectal cancer was 40.0% [7]. The incidence of non-cancer cause of death in colorectal cancer patients increases gradually with advancing age, and has been reported as 0.9% in those aged ≤ 49 years, 22.9% in those aged 50–74 years, and 76.0% in those aged ≥ 75 years [7].
The most common non-cancer cause of death in colorectal cancer patients have been reported as heart disease (36.7%), followed by chronic obstructive pulmonary disease (7.5%) and cerebrovascular disease (7.5%) [7]. In the present cohort of patients aged ≥ 75 years with CC, the incidence of non-cancer cause of death was 72.2% and most deaths were due to heart or pulmonary disease, in agreement with the previous study.
Our findings additionally revealed that GRIm score was a useful predictor of non-cancer cause of death.
The present finding that GRIm score was an independent predictive factor for both POCs and non-cancer cause of death leads us to propose the regarding performing CC surgery in elderly patients on the basis of GRIm score. First, high GRIm patients were more likely to have malnutrition or poor preoperative general condition. Indeed, preoperative stabilization of the general condition is known to improve short- and long-term outcomes [21]. Preoperative management with such as nutrition therapy might improve GRIm score, and thus reduce unexpected POCs and in-hospital mortality. Second, the relative risk of non-cancer cause of death in patients with CC is reported to be high within one year after surgery [7], which might indicate the effect of chemotherapy on cancer survivorship due to prolonged cancer-specific survival. However, some patients die from chemotherapy-related side effects such as heart and lung disease, and some from suicide [22–25]. The indications for chemotherapy in elderly patients with CC who have a high GRIm score should be determined carefully based on their general condition. Third, several guidelines recommend that because 10% of patients with early colorectal cancer are likely to have regional lymph node metastases, such patients who receive endoscopic resection and have proven non-curative factors (such as depth of submucosal invasion > 1000 um or vascular invasion) require additional colectomy with lymph node dissection [26–28]. To date, there is no established consensus regarding whether additional surgery is reasonable and effective in elderly patients. The indication for additional surgery in elderly patients with high GRIm should be considered carefully in terms of its risks and benefits.
There were several limitations in this study. First, this was a retrospective, single-center study which include only a limited number of patients. In addition, we use an arbitrary cutoff of the score for patient grouping. Large scale, multicenter study is needed to solve these problems. Second, the choice of whether or not to perform adjuvant chemotherapy and the selection of chemo-regimen was at the discretion of the surgeon.