Impact of malnutrition on cancer recurrence, colorectal cancer‐specific death, and non‐colorectal cancer‐related death in patients with colorectal cancer who underwent curative surgery

This study aimed to examine how malnutrition, as reflected by the Geriatric Nutritional Risk Index (GNRI), is associated with colorectal cancer (CRC) recurrence and cause of death.

Malnutrition is a common hallmark of cancer patients that adversely affects several aspects of cancer treatment and outcomes. 1Cancerrelated malnutrition is a common comorbidity in colorectal cancer (CRC) patients, experienced by approximately 30%, and leads to reduced treatment intensity, increased treatment toxicities, impaired quality of life, and worse oncologic outcomes. 2,3Thus, nutritional screening and intervention are important aspects of cancer treatment.Recently, the Geriatric Nutritional Risk Index (GNRI), a comprehensive malnutrition index that integrates height, body weight, and serum albumin levels, has been proposed for use in the evaluation of nutritional risk. 4][10][11] This is not surprising because GNRI was originally developed to assess the risk of morbidity and mortality in elderly hospitalized patients. 4,12Given improvements in CRC survival due to medical advances, patients may now live long enough after a CRC diagnosis such that non-CRC-related death may impact their survival. 13,14Non-cancer causes of death are major health threats for CRC survivors in the United States, and the mortality rate of noncancer death among CRC patients is more than twice the rate of that in the general population. 15Thus, evaluating the risks of non-CRCrelated death together with CRC-specific death is crucial for making decisions about cancer treatment and postoperative surveillance.
It remains unclear if and how malnutrition affects cancer recurrence, cancer-specific death, and non-cancer-related death in CRC patients.To this end, the present study aimed to examine the association of malnutrition, as assessed by GNRI, with cancer recurrence and cause of death in CRC patients.

| Study population
Patients were selected from a database of Japanese CRC patients who underwent curative surgery between 2011 and 2017 at the Department of Surgery at Himeji Medical Center in Japan.Inclusion criteria were consecutive patients with colorectal adenocarcinoma who underwent curative surgery.Exclusion criteria were patients with Stage 0 or Stage IV CRC, those who underwent palliative surgery, and those with insufficient data.The present study population included both elderly and young patients.This study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the Institutional Review Board (IRB) of Himeji Medical Center (IRB code, 2020-11).Owing to the retrospective nature of the study, written informed consent could not be obtained from our participants.However, the participants were guaranteed the right to withdraw from the study through an opt-out procedure that was posted on the hospital's website.

| Data collection, GNRI, and cutoff thresholds
The following data were obtained from medical records: age, sex, body mass index (BMI), American Society of Anesthesiologist Physical Status (ASA-PS), Charlson Comorbidity Index (CCI), primary tumor site (colon or rectum), TNM stage (UICC 8th), tumor differentiation ("differentiated" defined as tubular and papillary adenocarcinomas, and "others" defined as poorly differentiated, mucinous, and signetring cell carcinomas), pretreatment carcinoembryonic antigen (CEA) levels ( > 5 or ≤5 ng/mL), use of adjuvant chemotherapy (Yes or No), and the presence or absence of postoperative complications (Clavien-Dindo grade ≥III).Laboratory data were obtained at the first visit.GNRI was calculated using the following formula: GNRI = 1.489 × serum albumin concentration (g/L) + 41.7 × present body weight/ ideal body weight (kg). 4Ideal body weight was defined as the value calculated from height and a BMI of 22 (given its previously reported validity), rather than the value calculated with the Lorentz formula used in the original GNRI equation. 16Prognostic nutritional index (PNI) was calculated using the following formula: serum albumin concentration (g/L) + 0.005 × absolute count of lymphocytes. 17

GNRI has different cutoff values for different diseases.
Referencing a previous study, which provided GNRI cutoff values for predicting OS in CRC, we set an optimal cut-off value of 98 for predicting OS of CRC patients (low risk: >98, high risk: ≤98). 6,18In this study, we defined a patient as malnourished if his/her GNRI was ≤98.
For PNI, a receiver operating characteristic (ROC) curve analysis was performed with survival at 5 years after surgery as the outcome.
We set a cutoff value of 42.9 for predicting OS of CRC patients and defined a patient as malnourished if his/her PNI was ≤42.9.

| Statistical analysis
Pearson's chi-square test was used to compare categorical variables, and the Wilcoxon rank sum test was used to compare continuous variables between groups.Continuous variables are presented as median (interquartile range).OS was defined as the interval between the date of surgery and the date of all-cause death, and surviving patients were censored at the end of the follow-up period.
Recurrence-free survival (RFS) was defined as the interval between the date of surgery and the date of first recurrence or death from any cause.The Kaplan-Meier method was used to estimate OS and RFS.Follow-up data were collected prospectively until an event occurred, or until the data cutoff date of May 31, 2022.Differences in survival were assessed with the log-rank test.Multivariable Cox proportional hazards regression models were subsequently fitted to evaluate factors independently associated with survival.Since a statistically nonsignificant factor can also be a confounder, covariates included in the multivariable analysis were selected according to prior knowledge based on previous studies (age, sex, ASA-PS, CCI, primary tumor site, TNM stage, tumor differentiation, pretreatment CEA levels, and adjuvant chemotherapy).
Colorectal cancer-specific death and non-CRC-related death are competing events.If several competing events occur, they can lead to inaccurate conclusions, that is, it can lead to an overestimation of the investigated events.To compare the incidence of CRC-specific death with that of non-CRC-related death for each group over time, we analyzed the cumulative incidence function in patients with CRC according to GNRI groups.Gray's test was used to evaluate differences between groups. 19,20The Fine-Gray subdistribution hazard model was used to evaluate the relationship between the GNRI group and non-CRC-related death adjusted for potential confounding covariates, including age, ASA-PS, CCI, TNM stage, and adjuvant chemotherapy.Results are expressed as subdistribution hazard ratios (sdHRs) and 95% confidence intervals (CIs) corrected for CRC-specific death as a competing event.
To clarify the relationship between malnutrition and recurrence or cause of death over time, we used the hazard function.The Kaplan-Meier method is used to estimate the survival function, which conveys information about cumulative probabilities of events at time t for the entire cohort.2][23][24][25] The recurrence hazard and peak recurrence time for each group were calculated using the RFS hazard function, with recurrence or death from any cause defined as an event.In addition, hazard function curves were created with recurrence only or each cause of death (all-cause, CRC-specific, or non-CRC related) as the endpoint.
Units of measure for hazard rates are events per month.Hazard rates were estimated using the kernel-smoothing method. 26< 0.05 was considered statistically significant, and 95% CIs were calculated.All statistical analyses were performed using the JMP14 software program (SAS Institute Japan Ltd.) or R version 4.1.2.

| Study cohort characteristics
A total of 827 CRC patients who underwent surgery at our hospital were identified.Of these, 601 (361 males and 240 females) were included in the analysis, excluding 47 with Stage 0 CRC, 122 with Stage IV CRC, 10 who underwent palliative surgery, and 47 with insufficient data (albumin; n = 45, BMI; n = 2) (Figure 1).The median survival time was 58.9 months, with 116 deaths, 485 censored patients, and 109 patients with recurrence.The number of RFS events (i.e., recurrence and death from any cause) was 169.
Of the 601 analyzed patients, 180 (29.9%) had a low GNRI (GNRI ≤ 98), and 421 (70.1%) had a high GNRI (GNRI > 98).Table 1 summarizes clinicopathologic features of the low and high GNRI groups.Sex, CCI, primary tumor site, pathological N category, preoperative CEA levels, and use of adjuvant chemotherapy did not significantly differ between the two groups (p = 0.140, p = 0.866, p = 0.174, p = 0.108, p = 0.069, and p = 0.455, respectively).Patients in the low GNRI group were older (p < 0.001) and had a lower BMI (p < 0.001), higher ASA-PS (p = 0.003), larger tumor size (p < 0.001), higher rate of neoadjuvant treatment (p < 0.001), higher pathological T stage (p < 0.001), more advanced TNM stage (p < 0.001), and a higher proportion of tumor differentiation categorized as "others" (p = 0.026) compared to the high GNRI group.The low GNRI group had a higher rate of complications of Clavien-Dindo grade III or higher (p = 0.049) compared with the high GNRI group.A low GNRI was significantly correlated with a worse level of PNI (p < 0.001).

| Relationships between long-term outcomes and each marker
Figure 2 shows the prognostic impact of GNRI on OS and RFS.
Five-year OS rates were 67.9% for the low GNRI group and 88.8% for the high GNRI group (p < 0.001) (Figure 2A).Five-year RFS rates were 61.8% for the low GNRI group and 78.9% for the high GNRI group (p < 0.001) (Figure 2B).
Given the association between GNRI and several clinicopathologic features, we generated OS curves for subgroups of patients according to age and TNM stage.There was a clear separation between the OS curves for patients in the two GNRI groups regardless of age (p < 0.001, respectively) (Supporting Information: Figure 1).For all TNM stages, the OS curves were separated between the two GNRI groups (Stage I; p < 0.001, Stage II; p < 0.001, Stage III; p = 0.004) (Figure 3).

| Factors affecting RFS and OS in CRC patients
The multivariable analysis using Cox proportional hazards regression models revealed that age ≥65 years (hazard ratio (HR) range, 1.03-    3).

| Cause of death and competing risk analysis
The (Table 4).| 323 time and an overlap in the risk of recurrence for both groups after 3 years.
The hazard curve of all-cause death for the low GNRI group increased from 0 to nearly 2 years after surgery, plateaued until nearly 4 years after surgery, and then decreased slightly.
The hazard curve of CRC-specific death for the high GNRI group was relatively flat, whereas that for the low GNRI group mildly increased from 0 to nearly 3 years after surgery, T A B L E 3 Univariable and multivariable analyses of overall survival.than the high GNRI group, which is nearly consistent with the RFS hazard curve for the total cohort.

RFS hazard functions by adjuvant chemotherapy in patients with
Stage III CRC are shown in Figure 6B.The risk of recurrence and death in patients without adjuvant chemotherapy was highest the year after surgery, with a rapid decrease after that, but increased consistently over time compared to patients with adjuvant chemotherapy.
3.6 | Hazard functions for recurrence, CRC-specific death, and non-CRC-related death, according to the PNI group Hazard functions for each endpoint according to the PNI group are shown in Figure 7. Similar to the recurrence hazard curves for each GNRI group, those for the low PNI group approached the hazard for the high PNI group after 3 years.The hazard curve of CRC-specific death for the low PNI group increased slightly from 0 to nearly 3 years after surgery, peaked at 32 months (peak rate: 0.007), and then decreased through year 5.The hazard curve of non-CRC-related Cumulative incidence of CRC-specific death and competing non-CRC-related death after curative surgery in the high GNRI group (n = 421) and low GNRI group (n = 180).CRC, colorectal cancer; GNRI, Geriatric Nutritional Risk Index.
T A B L E 4 Association between each variable and non-CRCrelated death using Fine-Gray subdistribution hazard model.

| DISCUSSION
The present study demonstrated that GNRI has a clear prognostic value for CRC patients who underwent curative surgery.OS curves were stratified into two groups by GNRI level, and the prognostic value was independent of the pathological stage.Moreover, the multivariable analysis revealed that low GNRI was strongly associated with RFS and OS after adjusting for known key clinical factors.
The competing risk analysis and conventional survival analysis we conducted took into account the impact of non-CRC-related death.The competing risk analysis revealed that patients with low GNRI had a higher rate not only of CRC-specific death but also non-CRC-related death compared to patients with high GNRI.We also examined the impact of malnutrition over time using several endpoints.The results of the hazard function analysis suggest that the hazard function curve of recurrence for patients with low GNRI had an earlier and higher peak than that for high GNR; however, the curves overlapped after approximately 3 years.In contrast, the hazard function with non-CRC-related death as the endpoint showed an upward curve for 5 years in the low GNRI group.These findings were consistent even when patients were grouped by PNI level.
There are several possible explanations for why malnutrition affects cancer recurrence.First, malnutrition reduces the rate and intensity of postoperative adjuvant chemotherapy.For instance, the low GNRI group had a significantly lower rate of chemotherapy than the high GNRI group in Stage III CRC patients (50.7% vs. 65.5%,p = 0.030).The risk of recurrence in Stage III CRC patients without adjuvant chemotherapy was greater than that in those with adjuvant chemotherapy for 5 years after surgery.Second, the low GNRI group had a higher incidence of serious postoperative complications than the high GNRI group.Some studies have reported serious complications and severe inflammation as risk factors for postoperative cancer recurrence. 23,27Further investigation will be necessary to clarify whether interventions for malnutrition can contribute to a decrease in postoperative complications and an increase in the rate of adjuvant chemotherapy, leading to improved oncological outcomes.
Another important finding of the present study is that patients with low preoperative GNRI were at a higher risk of non-CRC-related death than patients with high preoperative GNRI in the postoperative years.Although several studies have reported an increased rate of non-CRC-related death in CRC patients in the postoperative years, [13][14][15] to the best of our knowledge, this is the first report of To date, reports using the hazard function in CRC patients have shown that the RFS hazard function curve in Stages II and III CRC peaks at about 1 year after surgery, and the risk decreases thereafter. 21,22,24,25Moreover, the peak of the hazard function curve is much more prominent for patients with Stage III CRC than those with Stage II CRC.The hazard of all-cause death increases from baseline (0 years) to nearly 2 years after surgery, plateaus until nearly 4 years after surgery, and then decreases slightly to year 5. 22,25 In Stage III CRC patients without adjuvant chemotherapy, the absolute risk of recurrence and death is greater than in patients with adjuvant chemotherapy over time. 24,25Findings from these reports are fairly consistent with those from the present study.In the present study, we found that the risk of recurrence is comparable between the low and high GNRI groups after about 3 years, whereas the RFS hazard is consistently higher in the low GNRI group than in the high GNRI group due to the increased risk of non-CRC-related death.While intensive follow-up for 3 years is still appropriate for detecting the recurrence of cancer, it is also important to understand that the risk of non-CRC-related death increases over time in CRC patients with low GNRI.
This study has some limitations.First, although our study cohort consisted of consecutive patients, there may have been selection bias due to the retrospective design.Second, the effects of perioperative changes in nutritional status on survival were not evaluated due to the lack of relevant data.In a recent study, tumor resection was found to alter the nutritional state of patients, and postoperative nutritional status in the non-cancer-bearing state more accurately predicted outcomes in patients with Stage III CRC. 28,29Further studies will be needed to verify these important findings.Third, potential confounders such as KRAS, BRAF, and MSI status were not included in the analyses.

| CONCLUSIONS
The impact of malnutrition on the risk of recurrence, CRC-specific death, and non-CRC-related death varies considerably over time.It is important to consider preoperative nutritional status along with the cancer stage when developing strategies to improve outcomes for CRC patients.
2.27), ASA-PS ≥ 3 (HR range, 1.60-3.68),rectal cancer (HR range, F I G U R E 1 CONSORT diagram for patient selection.After excluding patients with Stage 0 and Stage IV colorectal cancer (n = 47, n = 122), those who underwent palliative surgery (n = 10), and those with insufficient data (albumin; n = 45, BMI; n = 2) from the initial group of recruited patients with colorectal cancer (n = 827), the final number of analyzed patients was 601.T A B L E 1 Patient characteristics and associations of GNRI with clinicopathological factors and inflammatory markers.

F
I G U R E 2 (A) Overall survival curve for 601 colorectal cancer patients who underwent curative surgery, stratified according to high or low Geriatric Nutritional Risk Index (GNRI); 421 patients with a high GNRI and 180 patients with a low GNRI were analyzed.(B) Recurrence-free survival curve for 601 colorectal cancer patients who underwent curative surgery, stratified according to high or low GNRI.stage, tumor differentiation, pretreatment CEA levels, and adjuvant chemotherapy), a low GNRI was independently associated with reduced OS (HR range, 1.72-3.71)(Table cause of death was CRC-specific in 57 patients and non-CRCrelated in 59 patients.Causes of non-CRC-related death were other malignant diseases (n = 30), pneumonia (n = 8), cardiovascular disease (n = 6), sepsis (n = 3), liver failure (n = 2), and sudden death (n = 10).Of the other malignant diseases, 10 were diagnosed at the time of CRC diagnosis, and 20 developed after CRC surgery.Of the 175 Stage I CRC patients, 23 died during follow-up.All of these deaths were non-CRC-related.Of the 201 Stage II CRC patients, 10 (4.98%) died from CRC, and 23 (11.4%) died from other causes.Of the 225 patients with Stage III CRC, 47 (20.9%) died from CRC, and 13 (5.78%)died from other causes.Cumulative incidence function curves for CRC-specific death and non-CRC-related death according to GNRI groups are shown in Figure4.In the high GNRI group, the 5-year cumulative rate was 7.1% for CRC-specific death and 4.1% for non-CRC-related death.In the low GNRI group, the 5-year cumulative rate was 18.6% for CRC-specific death and 13.5% for non-CRC-related death.Both CRC-specific death and non-CRC-related death were significantly higher in the low GNRI group compared with the high GNRI group (p < 0.001, respectively).A Fine-Gray subdistribution hazard model for non-CRCrelated death was constructed.The low GNRI group was an independent risk factor for non-CRC-related death adjusted for potential confounding covariates, including age, ASA-PS, CCI, TNM stage, and adjuvant chemotherapy (sdHR range 1.67-5.27)

3. 5 |
Hazard functions for recurrence, CRC-specific death, and non-CRC-related death according to the GNRI group Hazard functions (representing the instantaneous risk of the event at each time-point) for each endpoint according to GNRI groups F I G U R E 3 Overall survival curve for 601 colorectal cancer patients stratified according to Geriatric Nutritional Risk Index (GNRI) group by stage.(A) Stage I, (B) Stage II, and (C) Stage III.are shown in Figure5.The RFS hazard curve (recurrence and all-cause death as endpoints) for the low GNRI group had approximately twice the recurrence hazard of the high GNRI group throughout the 5-year follow-up period and peaked at 8.2 months (peak hazard rate: 0.012).In contrast, the hazard function with recurrence only as an endpoint showed that the low GNRI group had a higher and earlier peak in hazard function (peak rate: 0.011, peak month: 3.14), followed by a decrease over T A B L E 2 Univariable and multivariable analyses of recurrence-free survival.
peaked at 34 months (peak rate: 0.005), and then decreased to year 5.The hazard curve of non-CRC-related death for the high GNRI group was also flat.In contrast, for the low GNRI group showed an upward curve for 5 years.RFS hazard functions by CRC stage are shown in Figure6A.The RFS hazard curve for the low GNRI group at Stage I CRC showed an upward curve over 5 years, which is consistent with the hazard curve for non-CRC-related death in the low GNRI group.On the other hand, the RFS hazard curve for the low GNRI group at Stage III CRC had a higher and earlier peak (peak rate: 0.019, peak month: 8.52)

F I G U R E 5
Hazard functions for recurrence stratified according to GNRI group by each endpoint.Units of measure for hazard rates are events per month.(A) Recurrence and death hazard, (B) Recurrence hazard, (C) All-cause death hazard, (D) CRC-specific death hazard, (E) Non-CRC-related death hazard.CRC, colorectal cancer; GNRI, Geriatric Nutritional Risk Index.death for the high PNI group was flat, whereas that for the low PNI group showed an upward curve for 5 years.

F I G U R E 6
Hazard functions for recurrence and death by (A) pathological stage and (B) adjuvant chemotherapy in Stage III CRC.Units of measure for hazard rates are events per month.CRC, colorectal cancer; GNRI, Geriatric Nutritional Risk Index.an increased risk of non-CRC-related death in CRC patients with malnutrition.This finding has at least two important clinical implications.First, the risk of non-CRC-related death should be considered when determining the indication for surgery and adjuvant therapy in patients with low GNRI.Second, interventions to improve the nutritional status of patients with low GNRI might help prevent death from other causes, although future studies are warranted to clarify this possibility.
Notwithstanding the limitations above, the present study, which targeted a cohort of CRC patients who underwent curative surgery, demonstrated for the first time that malnutrition has a different F I G U R E 7 Hazard functions for recurrence stratified by PNI group and by each endpoint.Units of measure for hazard rates are events per month.(A) Recurrence and death hazard, (B) Recurrence hazard, (C) CRC-specific death hazard, (d) Non-CRC-related death hazard.CRC, colorectal cancer; PNI, prognostic nutritional index.effect over time when endpoints are considered separately by using the hazard function.Identifying the major causes of death among CRC patients is essential for counseling patients regarding their care and survivorship.Our findings provide valuable information regarding healthcare prioritization during CRC survivorship.