Geriatric Nutrition Risk Index as a Predictive Marker of Tolerability and Vulnerability to Chemotherapy in Patients With Colorectal Cancer

Purpose: Malnutrition increases chemotherapy toxicity and impairs quality of life. Previous studies have shown an association between nutrition assessment tools, such as the Geriatric Nutrition Risk Index (GNRI), and adverse events of chemotherapy. However, none are specic to colorectal cancer (CRC). Therefore, we evaluated this association by investigating adverse events needing treatment (AENT) in patients with CRC. Methods: We retrospectively classied 147 patients with CRC into the risk group (GNRI<98, 85 patients) and the no-risk group (GNRI ≥ 98, 62 patients). We dened AENT as infection requiring antibiotics administration, grade ≥ 2 leukocytopenia, feasible neutropenia, anemia and thrombocytopenia requiring transfusion, grade ≥ 3 diarrhea, and acute organ failure. Results: We compared the two groups regarding AENT, antibiotic use, admission treatment for AENT, and mortality. Fifty-two (61.2%) and twenty-eight (45.1%) patients in the risk group and no-risk group, respectively, experienced AENT (odds ratio [OR], 0.948; [95% condence interval, 0.919–0.978]; area under the curve of the receiver-operating characteristic curve, 0.694; [0.608–0.780]). Those in the risk group had increased antibiotic use (OR, 0.945; [0.912–0.979]) and mortality (OR, 0.845; [0.765–0.932]). AENT and performance status were not associated, while GNRI score and chemotherapy toxicity were inversely associated. Conclusion: GNRI can predict a patient’s tolerability for cytotoxic chemotherapy. Prospective studies should validate GNRI and nutrition support benets.


Introduction
Malnutrition is common among patients with unresectable cancer [1,2]. About 40-65% of patients with colorectal cancer (CRC) are diagnosed with malnutrition [3]. Moreover, malnutrition increases infection, hospitalization, and medical expenses and is associated with poor prognosis in patients with cancer [4,5]. If cancer treatment is ineffective, the disease progression worsens the patient's nutrition status, and cancer-related in ammation and metabolic changes increase their risk of infection and mortality [6,7].
However, cytotoxic chemotherapy may be overtreatment for malnourished patients due to the severe adverse events arising as a result of chemotherapy. If cytotoxic chemotherapy is excessive, adverse events worsen nutrition status and impair quality of life in patients with cancer [8,9]. Therefore, we considered the adverse events associated with cytotoxic chemotherapy to indicate malnutrition and tolerance to chemotherapy [6,[10][11][12]. Thus, in patients with cancer, nutrition assessment and support are essential [1].
However, these nutrition assessment tools based on extensive questionnaires are primarily subjective and have the risk of memory bias [22,23]. Therefore, PG-SGA and MNA-SF are di cult for geriatric patients and uncooperative patients. The NRI comprises objective calculations based on basal body weight, height, and serum albumin levels. However, basal body weight is uncertain, which may cause memory bias, and most patients with cancer experience weight loss before diagnosis [24]. Furthermore, NRI calculation is impossible without records of the patient's weight. A previous study reported the association between NRI and chemotherapy toxicity [24]. In contrast, other studies reported that NRI might not predict chemotherapy-induced adverse events [3,25]. Thus, the association between NRI and chemotherapy toxicity remains controversial. Among the four validated nutrition assessment tools, GNRI is the only accurate and objective tool.
GNRI is a simple and objective tool requiring less time to perform and aids in undertaking patient nutrition risk assessment [1]. Physicians can calculate GNRI in clinical practice without changing the clinical routine because patient's body weight, height, and blood test parameters are routinely measured before cytotoxic chemotherapy. Bouillanne et al. developed GNRI as a tool to assess older patients' nutrition risk primarily [6,22,26]. However, recent studies showed that GNRI score can assist with the identi cation of nutrition risk and prediction of tolerability, vulnerability, and prognosis in patients with cancer [10,11,19,23,27], chronic diseases [28,29], and who undergo hemodialysis [30]. Although GNRI is a wellestablished nutrition assessment tool for predicting adverse events associated with chemotherapy in various cancers, this tool's utility in patients with CRC remains unexplored.
This study aimed to assess the association between GNRI and tolerability, vulnerability to cytotoxic chemotherapy in patients with CRC by investigating the occurrence of chemotherapy-related adverse events.

Patients
We reviewed the medical records of patients with CRC visiting the Department of Medical Oncology at Hirosaki University Hospital from April 2008 to March 2020.
We found 214 medical records of patients with CRC. The eligibility criteria were as follows: chemotherapy undertaken at the Department of Medical Oncology, Hirosaki University Hospital; Eastern Cooperative Oncology Group Performance Status (PS) ≤2; stable condition without severe comorbidities; CRC diagnosed histologically; data available on height, weight, and serum albumin level at the rst chemotherapy session; and follow-up data for 12 weeks after chemotherapy initiation. The exclusion criteria to reduce biases were the presence of grade ≥2 leukocytopenia, neutropenia, or receiving granulocyte colony-stimulating factor (G-CSF) in the rst chemotherapy session at our hospital. Because, on the colorectal cancer chemotherapy, G-CSF administration is not routine generally. Moreover, G-CSF administration will modify the progress of bone marrow suppression. We thought that if the patient with cancer has leukocytopenia or neutropenia in the rst chemotherapy session, the medical oncologist will generally hesitate the cytotoxic chemotherapy. Besides, patients with cancer showing leukopenia or neutropenia before cytotoxic chemotherapy have a high risk of infection compared to not showing leukopenia or neutropenia. Sixty-seven patients were excluded based on these criteria. Finally, we enrolled 147 patients in this study (Figure1).
We assessed the nutritional risk in patients with CRC using GNRI at the rst chemotherapy session for each patient. We calculated GNRI from body weight and serum albumin levels using the following equation: GNRI = 1.489 × serum albumin (g/dL) + 41.7 × (body weight/ideal body weight). Ideal body weight was de ned as height (m) 2 × 22; this formula was derived from the patient's height and a body mass index (BMI) of 22 kg/m 2 . We classi ed patients into two groups according to previously published thresholds [22]: the no-risk group (GNRI ≥98) and the risk group (GNRI <98). In the original GNRI equation, Bouillanne et al. calculated the ideal body weight using the Lorentz formula. However, we calculated ideal body weight from height and a BMI of 22 because a previous study validated the BMI formula [31]. A previous report showed that each formula's GNRI scores varied little from the calculated values [30]. We de ned adverse events needing treatment (AENT) as infection used antibiotics; grade ≥2 leukocytopenia or neutropenia leading to the postponement of chemotherapy; febrile neutropenia; anemia and thrombocytopenia requiring transfusion; grade ≥3 diarrhea; acute liver failure; and acute kidney injury [3]. We considered AENT to indicate tolerability and vulnerability to chemotherapy [1,10,11,19]. Therefore, we compared the two groups in terms of AENT, antibiotics use, admission treatment for adverse events, and mortality.

Statistical analysis
We examined the association of GNRI, AENT, antibiotics use, admission treatment for adverse events, and mortality. We used multivariable logistic regression analysis to control for potentially confounding factors such as age, sex, PS, liver metastasis, chemotherapy protocol, and dose-adjustment (dose reduction). We performed all statistical analyses using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria) [32]. More precisely, it is a modi ed version of R commander designed to add statistical functions frequently used in biostatistics.

Characteristics
The median GNRI was 95.05 (range, 66.26 -139.01). Of the 147 patients, we classi ed 85 and 62 patients in the risk and no-risk groups, respectively. Table 1 shows patients' characteristics from the GNRI Classi cation. In the risk group, more patients had liver metastasis than in the no-risk group. Age, sex, primary cancer site, PS, chemotherapy protocol, chemotherapy line, dose-adjustment (dose reduction), height, weight, and albumin were not signi cantly different between the two groups. Patients in the risk group had more AENT and antibiotic use, as well as increased mortality than patients in the no-risk group.

AENT and GNRI
The risk group patients tended to have more AENT than the no-risk group patients. In the risk group, 52 patients had AENT (61.2%), while in the no-risk group, 28 patients had AENT (45.1%). After applying logistic regression analysis, we observed a signi cant association between GNRI and AENT (odds ratio [OR], 0.948; 95% CI, 0.919-0.978; and the area under the curve of the receiver-operating characteristic curve(AUC of ROC) was 0.694; 95% CI, 0.608-0.78). (Table 2.).

Admission treatment
In the risk group, 20 patients received admission treatment (23.5%), while in the no-risk group, eight patients received admission treatment (12.9 %). However, we did not observe a signi cant association between GNRI and admission treatment (OR, 0.978; 95% CI, 0.936-1.02; p=0.319) in the multivariable logistic regression analysis.

Mortality
Seven patients died during the study period. Four patients died due to infection (cholangitis, pneumonia, and sepsis), one due to cancer-associated thrombosis, and two due to primary cancer. In the multivariable logistic regression analysis, we observed a signi cant association between GNRI and mortality (OR, 0.845; 95% CI, 0.765-0.932; AUC of ROC, 0.881; 95% CI, 0.77-0.994). Additionally, we observed a two-factor model PS 2 and Low GNRI were signi cantly associated with mortality (AUC of ROC, 0.966; 95% CI, 0.916-1]).

Discussion
Malnutrition increases the toxicity of chemotherapy and impairs quality of life of patients with cancer [33]. Our current ndings suggest an association between nutrition risk classi ed by GNRI, and AENT in the early period, antibiotics use, and mortality in patients with unresectable CRC receiving cytotoxic chemotherapy. We calculated GNRI using three factors: height, body weight, and serum albumin levels.
Therefore, GNRI is an accurate, objective, and time-saving tool for nutrition assessment [1,10]. More patients in the risk group (GNRI <98) had AENT than in the no-risk group(GNRI ≥98). Furthermore, AENT may increase due to infection and malnutrition, which results in higher treatment costs for the patients in the risk group because of the need for antibiotic treatment and other supportive care forms.
Previous reports validated the association between malnutrition and the patient's tolerability and vulnerability to cytotoxic chemotherapy due to severe chemotherapy-associated adverse events [1,8]. However, in patients with CRC, the effect of body weight, BMI, or hypoalbuminemia alone on tolerability and vulnerability has remained controversial [25, 33-35]. The body weight of patients with CRC changes easily due to decreased food intake, ascites, pleural effusion, and dehydration [36]. Similarly, serum albumin levels are easily in uenced by in ammation and dehydration [37]. Therefore, the association between BMI or albumin alone and the nutrition risk of patients with cancer may be controversial [4,34].
In the GNRI calculation, body weight and serum albumin levels are inversely proportional. In other words, if the patient's body weight increases due to ascites or pleural effusion, the patient's serum albumin levels decrease by increasing body uid volume unless intra-vascular dehydration occurs. Conversely, if the amount of body uid decreases, the patient's body weight decreases, but the patient's serum albumin levels increase due to high blood concentration. However, our ndings showed no correlation between body weight and albumin. Thus, GNRI calculation may re ect nutrition status and risk complementarity, although BMI, body weight, or albumin alone cannot re ect nutrition status and risk comprehensively because of unknown risk factors.
NRI is composed of basal weight and serum albumin levels. NRI seems useful for identifying malnutrition and nutrition risk. However, baseline body weight is uncertain because malnutrition occurs before cancer diagnosis. In addition, it is di cult to exclude the patient's memory bias. Therefore, NRI may not comprehensively re ect nutrition status and risk [3]. In contrast, the body weight used in GNRI is obtained at just one point in time. Thus, GNRI may help identify nutrition status and risk before chemotherapy, enabling prediction concerning the tolerability to chemotherapy.
Malnutrition increases the risk of infection and hospitalization in patients with cancer [4]. Moderate or severe infection disease is critical in some cases. Prediction of patients' immunocompromised status through a novel method can help patients exercise preventive measures, e.g., avoiding crowds, following hand wash practices, wearing a mask on the mouth, following a diet avoiding fresh vegetables and fermented food. Furthermore, we need to assess for the presence of comorbidities that lead to severe infection, such as diabetes mellitus, chronic obstructive pulmonary disease, and asthma. Therefore, nutrition risk assessment is essential in patients with cancer.
We consider that nutrition risk assessment can be used similarly to Geriatric Assessment (GA) to discover (diagnose) a missing problem, predict adverse events of treatment and disease prognosis, and determine treatment strategy [38]. Importantly, the diagnosis of unrecognized problems underlying malnutrition has bene ts [39]. Patients with nutritional risk have metabolic, cardiovascular, and respiratory comorbidities and potential psychosocial and economic problems [40][41][42]. Thus, supportive care for physical comorbidities and psychosocial and economic di culties is important in patients with cancer. GNRI is useful for assessing nutrition risk in patients with cancer, and the GNRI assessment enables the discovery of problems associated with malnutrition. Recognizing comorbidities and malnutrition-associated problems in patients with cancer will enhance supportive care [2]. Supportive care enhancement will reduce antibiotics use, admission treatment for adverse events, and mortality-associated excessive chemotherapy. Therefore, enhanced supportive care can reduce medical expenses [13,43].
Another aspect of essential supportive care for patients with cancer involves identifying the patients requiring supportive care and what type of care they may bene t from. Our ndings support the utilization of GNRI in predicting AENT. Recent studies reported that nutrition support might reduce adverse events associated with toxic chemotherapy [44]. Support for comorbidities that causes malnutrition is also important. However, nutrition intervention for all patients with cancer may not be practical and increase medical expenses [26,41]. Therefore, before chemotherapy initiation, the GNRI nutrition risk assessment will help physicians identify patients who may bene t from nutrition support and comorbidity screening.
If the patient with cancer has a GNRI lower than 98, we should introduce nutrition support and intervention practically as a part of the cancer treatment strategy. Nutrition support and intervention comprise nutrition education and counseling, exercise with rehabilitation therapists, and life support from medical social workers and care workers. However, the effect of nutrition support based on GNRI remains unexplored on patients with CRC of tolerance and vulnerability to chemotherapy. Thus, a prospective study is needed to validate the bene t of nutrition support for chemotherapy in patients with CRC.
We did not observe any association between GNRI and admission. Admission treatment is complicated. When the patients determine whether to be admitted for treatment, disease severity is not the only determinant. Many patients and families may have many other determinants, such as nancial conditions, family situations, and insurance issues. Therefore, nutrition risk assessment alone may not re ect all essential issues. To expand upon the current ndings, we intend to delve into issues concerning admission treatment.
The present study has some limitations. First, the present study enrolled only patients covered by a single center. Thus, it is undeniable there was bias regarding diet and exercise habits as regional culture, climate, and economy in uence patients' lifestyle. Second, most of the patient's chemotherapy protocol was an oxaliplatin-based protocol. Japanese clinical physicians more commonly use an oxaliplatin-based protocol than an irinotecan-based protocol for the rst-line chemotherapy [45]. Thus, we might have some selection bias. Finally, regarding treatment strategy, the chemotherapy protocol was determined by internal meetings within the department. We determined the treatment strategy based on the Japanese Society for Cancer of the Colon and Rectum guidelines and considering the patient's age, PS, organ dysfunction status, and patient's wishes. Although our strategy included more than one physician's opinion, the inherent aspects of single-center studies may cause a degree of selection bias in the treatment strategy. Thus, a multi-center study is needed to resolve the biases.
In conclusion, in patients with CRC receiving chemotherapy, our study results indicate an association between nutrition risk classi cation by GNRI and AENT, antibiotic use, and mortality. GNRI may be a useful screening tool predicting tolerability and vulnerability to chemotherapy. Further prospective research is needed to validate nutrition support based on nutrition status and risk classi cation by GNRI in patients with CRC.  FOLFOX: modi ed FOLFOX(mFOLFOX) †A signi cant difference was observed for "Liver Metastasis" and "Performance Status 2" between the risk group and the no-risk groups.
All p-values were obtained from the Fisher exact test.