In the study, we evaluated elderly GC patients integrated with clinicopathological, nutritional and surgical variables comprehensively. 34.8% had a high prevalence of high nutritional risk (NRS score ≥ 5), which was associated with over 2-fold increase in hazards of 5-year TSS in univariate analysis. And similar findings were shown with other nutritional variables including PNI score, preoperative anemia, albumin and prealbumin levels. Multivariate analysis by Cox model demonstrated preoperative BMI ≤ 20.6 kg/m2 and postoperative ICU admission were independently predictors of poor 5-year TSS (HR = 2.297; 95%CI, 1.362–3.874, p = 0.002; HR = 1.965; 95%CI, 1.171-3.300, p = 0.011, respectively).
Recent studies demonstrated that low BMI predicted increased surgical morbidity and mortality [29–31]. A study of 510 GC patient reported by Kim et al  showed that preoperative BMI ≤ 18.5 kg/m2 was an independent poor prognostic factor of TSS in the stage I/II group (HR = 13.521; 95%CI, 1.186-154.197; P < 0.05) and an independent risk factor of postoperative complications in the stage III/IV subgroup (HR = 17.158; 95%CI, 1.383–212.940; P < 0.05). A recent meta-analysis  including 12 studies showed GC patients with BMI < 18.5 had higher risk of surgical complications and poorer survival. Controversially, some studies had different results [32, 33]. Lee et al demonstrated that there was no significance of 5-year TSS between preoperative low and high BMI groups (BMI < 18.5 vs BMI ≥ 25.0, p = 0.297), and significances between normal and high BMI groups (BMI 18.5–24.9 vs BMI ≥ 25.0, p = 0.001) in a study of 1909 GC patients . Besides, they found that low BMI had significant lower overall survival (OS) compared to high-normal BMI (23.0-24.9). Interestingly, results from this study implied that traditional BMI cut-off point is not as much helpful as we expected and better BMI cut-off point probably lie between low-normal BMI (18.5–22.9). Ejaz et al reported of 775 GC patients that BMI was not related to postoperative complications or TSS or OS, but patients with underweight (BMI < 18.5 kg/m2) with preoperative low albumin (< 35 g/L) had worse OS . However, baseline albumin level in low BMI group was significantly lower and small proportion (7.1%) of their cohort was classified as low BMI, indicating selection bias of surgical GC patients may exist as data was retrospectively collected.
BMI, as an easily available index, is universally used as an indicator of nutrition assessment in clinical practice and BMI < 18.5 kg/m2 is the most commonly used cut-off point as underweight or malnutrition. However, no consensus about cut-off value of BMI for the elderly is achieved, and better BMI cut-off point is yet to be set up in elderly surgical patients with GC. For the first time, we verified preoperative BMI ≤ 20.6 kg/m2 as an independent predictor and significantly associated with poor 5-year TSS in the elderly GC patients after gastrectomy. Since more clinicians realized that novel low BMI cut-off point might be more suitable and useful in elderly patients, Dutch definition  of low BMI for patients over 65 years is < 20 kg/m2 and European Society of Clinical Nutrition and Metabolism definition  of low BMI is < 22 kg/m2 for patients ≥ 70 years. Novel low BMI cut-off points were validated as valuable predictors in general hospital population . Currently, very limited studies focused on this topic, and further study may validate our newly found BMI cut-off point in larger and prospective elderly GC patients with gastrectomy and may be extrapolated to other cancers.
Nutritional variables other than BMI in the present study showed predictive value of TSS in the univariate analysis. Preoperative weight loss > 5%, NRS2002 > 5 [37, 38], PNI < 45 , preoperative anemia, low albumin (< 35 g/L) [40, 41] and low prealbumin levels  are reported in several studies to be related to surgical morbidities and/or cancer survival individually or as variables of a model. Preoperative weight loss and anemia as main clinical symptoms of GC are closely related to low BMI. Intrinsic connections or crossover concepts among BMI with NRS2002, albumin with PNI, and albumin with prealbumin are probably confounders which should eliminated by multivariate analysis methodologically. Besides, BMI probably acts as reflection of general fitness of these elderly GC patients given the fact that cancer is a chronic and consumptive disease commonly ending up with cachexia in terminal phase. Consequently, it is understandable that BMI is the only nutritional variable as risk factor of 5-year TSS in multivariate analysis.
The Charlson comorbidity index (CCI) has been proved numerously with good reliability, excellent correlation with mortality and progression-free survival (PFS) outcomes, particularly to account for the effect of age . CCI is a common tool in geriatric units and helpful in assessment of concurrent diseases in elderly patients. Postoperative ICU admission often results from cardiac, pulmonary, hepatic and/or renal dysfunctions, which is common in elderly patients with more morbid illnesses. ICU admission might be closely related to complex comorbidities in elderly GC patients and as an independent predictor of TSS (HR = 1.965; 95%CI, 1.171-3.300). Also, we found TNM staging (stage II, HR = 5.56, 95%CI, 1.59–19.42; stage III, HR = 16.20, 95%CI, 4.99–52.59, p < 0.01) were associated with TSS significantly as widely known.
Elderly GC patients have higher nutritional risk, more comorbidities and increased mortality. When it comes to surgery for patients with early or locally advanced GC patients, multiple factors result in high surgical morbidities and mortality. However, chronological age alone is not an independent factor to withhold curative or palliative treatment from elderly GC patients, and cofactors including physical and physiological functions, mental and even socioeconomic status have to be considered . Thereafter, several tools are recommended to assess elderly GC patients, such as comorbidity indices (CCI), nutrition indices (NRS2002) and prognostic indices for survival (PNI). Recently, GO2 trial  including 514 elderly and/or frail patients with advanced gastroesophageal (GO) cancer had preliminary results, showing that 60% dose of standard regimens had non-inferiority PFS (HR = 1.10; 95%CI, 0.90–1.33) compared to standard regimens (Oxaliplatin 130 mg/m2 d1, Capecitabine 625 mg/m2 bd d1-21, q21d), as well as less toxicities and better overall treatment utility (OTU), a novel patient-centered endpoint . Previously, results of 321GO demonstrated favoring a doublet regimen over single agent or triplet chemotherapy in 55 elderly or frail advanced GO patients as a randomized phase II and feasibility study . And both implied a common refrain in some cancer treatment is “less may be more”, which probably enlighten us whether elderly and/or frail GC patients with surgical indication could have a shrink conversion from current D2 lymphadenectomy as the standard and curative approach, or could the adjuvant chemotherapy regimens be simple or reduced dose.
There are some limitations in the present study. Firstly, the retrospective nature of the study design lowers the quality of data and sample size is relatively small. Secondly, we did not collect surgical morbidity data here and QoL was not assessed postoperatively, although we recognized the significance of it to elderly GC patients. What’s more, it is a single center study from China, we should be cautious when generalizing to populations whether in other countries or with other cancers. To the best of our knowledge, it is for the first time that an exact BMI cut-off value (< 20.6 kg/m2) was determined as risk factor of poor 5-year TSS in elderly GC patients. Future studies may have further comprehensive assessment of elderly surgical patients and probably use novel tools, such as OUT [43, 44].