Our study supports the prognostic role of the GNRI in DLBCL patients. Lower GNRI was associated with worse PFS and OS. Notably, patients who did not meet any of the two criteria for sarcopenia had a favorable prognosis regardless of GNRI score, with the exception of those with major GNRI risk scores, while all patients who met both criteria for sarcopenia had an unfavorable prognosis even in cases of no GNRI risk. In contrast, for patients who met only one of the criteria for sarcopenia, disease prognoses were determined based on GNRI score. These findings suggest that the combined use of GNRI and sarcopenia may improve the predictability of each factor in DLBCL patients.
A previous Japanese study showed that the GNRI score could identify patients with poorer prognosis among those with high-intermediate to high NCCN-IPI [18]. In contrast, a Chinese study found that while there was a marginal difference in OS by univariate analysis, GNRI score was not an independent prognostic factor for OS in multivariate analysis [19]. Given the differences in patient populations and inclusion criteria it is difficult to compare the results of our study directly with those of previous studies; however, there were considerable differences in patient characteristics between studies. The patients in the Chinese study were younger (mean age, 55 years) than those in both the Japanese study and this investigation (median ages, 68 and 64 years, respectively). The proportions of patients with low to low-intermediate NCCN-IPI were 80%, 46%, and 46.5% in the Chinese, Japanese, and current studies, respectively. In subgroup analyses, the GNRI score could not identify patients with a worse prognosis among those with low to low-intermediate NCCN-IPI in any these studies, whereas there was a significant association between GNRI score and OS among those with high-intermediate to high NCCN-IPI in both the Japanese and current studies. These findings may explain why the prognostic value of GNRI was differently reported in the literature [18, 19] and suggests that the GNRI alone can be a prognostic factor only in DLBCL patients with higher NCCN-IPI.
There is debate about which single parameter for cancer cachexia is most appropriate to predict the prognosis of DLBCL patients. Large database cohort studies reported that patients with low to normal BMI had shorter survival times relative to overweight or obese patients [23, 24], while subset analysis from a phase III trial failed to prove the prognostic role of BMI [25]. Sarcopenia, as determined by CT imaging, has been proposed as an independent prognostic factor in several studies [15, 22, 26, 27]. However, other studies found that the prognostic value of sarcopenia was limited in elderly and male patients [28, 29]. There are also contradictory reports regarding the prognostic role of hypoalbuminemia with various cut-off points [13, 16, 30].
Essentially, multifactorial elements are intricately linked to cancer cachexia. Muscle wasting and atrophy, which are key features in cancer cachexia, are mediated by tumor-derived factors such as proteolysis-inducing factor involving nuclear factor-κB pathway [31, 32]. Tumor-driven inflammatory cytokines are responsible for the development of cancer cachexia by inducing alterations in protein metabolism, as well as by activation of apoptosis and inhibition of regeneration of muscle mass [33]. White adipose tissue browning and lipolysis promoted by tumor-derived cytokines and hormones mediates adipose tissue and muscle wasting through molecular crosstalk between adipose and different tissues [34]. Myostatin expression and activity are enhanced in experimental cancer cachexia, with inhibition sufficient to reduce muscle loss [35, 36]. Furthermore, an international consensus has suggested that the staging criteria of cancer cachexia consist of various clinical factors, including weight loss, BMI, sarcopenia, systemic inflammation, anorexia, response to anticancer therapy, and performance status [9]. Therefore, the cachexia risk of our study, which reflects body weight, sarcopenia, and systemic inflammation may be a better surrogate marker for evaluating the severity of cancer cachexia compared with other single parameters. Cachexia risk was a predictor of treatment response, treatment-related toxicity, and survival in DLBCL. Given the intolerance to R-CHOP treatment observed in patients with high cachexia risk, dose adjustment may be considered in this group. However, chemotherapy dose adjustment resulted in a remarkable decrease of CR rate in the patients with high cachexia risk, whereas there was little effect in those with low cachexia risk. This suggests that a novel therapeutic strategy may be warranted in patients with high cachexia risk.
Our study has several limitations. First, the retrospective, non-randomized study design with a relatively small sample size makes it difficult to determine whether the differences in patients’ characteristics between the HCR and LCR groups were caused by potential selection bias or by essential differences between the two groups. In this regard, cachexia risk may be a significant confounding variable. To reduce this potential bias, all consecutive patients who were treated with the same treatment modality were included in this study. Furthermore, the prognostic value of cachexia risk was still significant after adjustment for important covariates and in stratified analysis by the NCCN-IPI. Second, laboratory biomarkers for cachexia and systemic inflammation other than serum albumin were not assessed in our study. Although serum albumin, one of the representative markers for systemic inflammation [37], was used to define cachexia risk in this study, the absence of a biomarker that better reflects the muscle wasting process may weaken the relevance of our risk model for cancer cachexia. To overcome these pitfalls, a prospectively designed study including various biomarkers for cancer cachexia is needed to validate our findings.
Taken together, the data presented here raise the possibility of the GNRI score as a prognostic factor in DLBCL. In addition, we found that the combined risk model including GNRI and sarcopenia could better predict patient prognosis relative to GNRI alone. These findings emphasize the complexity of cancer cachexia and suggest a close relationship between cachexia, systemic inflammation, and DLBCL.