In this study, we investigated the prognostic factors for young and elderly patients with recurrent and unresectable GC. In the Y group, poor PS was an independent factor for a short OS, whereas NLR was the independent prognostic factor in the E group.
Among oncologists, PS is considered an important tool to determine the general conditions of patients. Almost all physicians refer to the PS when considering whether a patient is eligible for treatment or participation in clinical trials. Poor PS has been reported as a predictor of a poor clinical outcome, such as increased adverse events and decreased treatment efficacy in patients receiving chemotherapy [15–17]. However, PS may not be sufficient when assessing the elderly for chemotherapy  due to the high heterogeneity of this population, including medical history, organ function, and nutritional status . Indeed, in this study, NLR rather than PS was an independent prognostic factor for OS of the E group.
NLR is useful for predicting clinical outcomes in several studies including both young and elderly patients with cancer [13, 20, 21]. A high NLR indicates an increased neutrophil count and/or a decreased lymphocyte count, as well as relative lymphopenia. The relationship between NLR and prognosis of patients with cancer is still not well understood. However, both neutrophils and lymphocytes are considered to be related to cancer prognosis. First, neutrophils play significant roles in the process of cancer progression, including tumor initiation, growth, proliferation, or metastatic stage [22, 23]. Furthermore, neutrophilia inhibits the cytotoxic activity of lymphocytes such as T cells and natural killer cells and facilitates the extravasation of tumor cells . By contrast, lymphocytes play an important role in the immune system against cancers. Peripheral low lymphocyte counts have been associated with a poor outcome in various cancers [25–27] and were related with lymphatic invasion and recurrence of lung cancer . Lymphocytes play a crucial role in the anti-tumor immune response. Accordingly, reduction in the lymphocyte count reduces the anti-tumor effect of the immune system, resulting in accelerated tumor occurrence and development .
This background clearly demonstrates an association of NLR with the cancer immune environment; consequently, NLR is considered to be related with the clinical outcome. The present finding that NLR may be a specific prognostic factor of OS in elderly patients with GC indicates that NLR might not only reflect the tumor environment but may also be associated with host frailty [30, 31]. Frailty is a syndrome affecting physiologic reserves across multiple organ systems and has been identified as a poor prognostic factor in geriatric oncology [32, 33].
Nishijima et al.  reported a significant association between NLR and frailty only in the geriatric population, which supports our findings. Nevertheless, the mechanism underlying the association between NLR and frailty is uncertain. Gilmore et al.  reported that chronic low-grade inflammation may be involved in the relationship between NLR and frailty.
In summary, our results suggest that NLR is a useful prognostic factor by reflecting not only the tumor immune environment but also frailty in elderly patients with cancer. By contrast, PS emerged as a useful marker for predicting survival in younger patients, which may not be a significant prognostic factor in elderly patients because of its reduced reliability for this population, as suggested in previous reports.
The present study had various limitations. First, as the data were collected from a single center, we could not entirely avoid selection bias associated with patients and chemotherapy regimens. Second, we could not completely exclude the possibility of complications such as subclinical infection that might affect the prognostic values of biomarkers. This limitation might more strongly affect the results in the E group due to the greater complexity of their comorbidities. Third, the most significant limitation of this study was the low number of patients investigated; thus, validation will be required via prospective studies with a larger cohort. The relatively small number of patients and events in our cohort did not allow for comprehensive multivariable analyses and precluded making definitive conclusions. However, the correlation of high NLR and poor prognosis of the E group was highly statistically significant and also seems to be clinically meaningful.