Basal metabolic rate by FAO/WHO/UNU as a prognostic factor for survival in patients with gastric cancer: A cohort study

Abstract

The basal metabolic rate (BMR) is a crucial indicator of the body’s energy expenditure at rest and is essential for understanding metabolic needs. This retrospective study evaluated the prognostic significance of BMR in 521 predominantly Asian patients with stage I–III gastric cancer who underwent curative-intent resection. BMR was calculated using the Food and Agriculture Organization/World Health Organization/United Nations University (FWU BMR) equation. Multivariate Cox regression analysis identified FWU BMR as a significant predictor of overall survival (OS) (P < .001). Fractional polynomial modeling revealed a linear relationship between FWU BMR and OS, with higher values correlating with lower mortality risk. The FWU model, which included FWU BMR along with other clinical variables, showed superior predictive performance (C-index: 0.815, iAUC: 0.775) compared to that of the same model lacking BMR. Additionally, although the differences were not statistically significant, the FWU model also outperformed those using the BMR derived from alternative equations, including the Harris–Benedict equation. The nomogram, based on the FWU model, demonstrated good calibration. These findings suggest that the FWU BMR is a valuable prognostic factor in patients with gastric cancer post-resection, enhancing predictive accuracy and aiding in personalized post-surgical care. However, further validation in diverse populations is required.

1. Introduction

Gastrectomy is the primary curative treatment for patients with stage I–III gastric cancer. Despite this intervention, patients remain at a risk of relapse and mortality. The Tumor-Node-Metastasis (TNM) staging system is widely used to predict the prognosis of patients with cancer. However, owing to tumor heterogeneity, treatment responses, and other factors, patients with the same TNM stage may have different prognoses. Moreover, TNM staging only considers tumor invasion, nodal involvement, and metastases, without accounting for other crucial factors that affect prognosis, such as patient age, comorbidities, and molecular tumor characteristics.^[1,2] Therefore, novel biomarkers are needed to improve the prognostic accuracy in patients with gastric cancer.

Malnutrition is prevalent among individuals with upper gastrointestinal cancer and is associated with shorter survival and poor quality of life. To effectively prevent or treat malnutrition, nutritional interventions must ensure appropriate energy provision to meet the daily metabolic demands. The basal metabolic rate (BMR) represents the essential daily energy expenditure required to sustain vital bodily functions. The concept of basal metabolism emphasizes the need to conduct experiments under strictly standardized conditions. Traditionally, this was achieved by measuring the minimum rate of heat production to ensure that the measurements are free from the effects of food consumption and extreme physical environments. However, consistently imposing such stringent conditions has proven impractical.^[3]

In 1918, Harris and Benedict developed equations to predict BMR using rigorous statistical methods and introduced biometric principles including age, sex, height, and body weight.^[4] Despite the tendency to overestimate the BMR, these equations have remained popular and widely used in North America owing to their simplicity (Table 1).^[3] Similarly, the 1985 Food and Agriculture Organization (FAO), World Health Organization (WHO), and United Nations University (UNU) equations also consider age, sex, height, and body weight, underscoring the importance of BMR for estimating energy needs (Table 1). However, this equation tends to overestimate BMR.^[5,6] Despite these limitations, the FAO/WHO/UNU (FWU) equations are widely accepted.^[3]

Table 1.

Equations for formulating basal metabolic rates

Equation	Sex	Formula
FAO/WHO/UNU	Men	Age 10–18	: (16.6 × W in kg) + (77 × H in meter) + 572
		Age 18–30	: (15.4 × W in kg) − (27 × H in meter) + 717
		Age 30–60	: (11.3 × W in kg) + (16 × H in meter) + 901
		Age > 60	: (8.8 × W in kg) + (1128 × H in meter) − 1071
	Women	Age 10–18	: (7.4 × W in kg) + (482 × H in meter) + 217
		Age 18–30	: (13.3 × W in kg) + (334 × H in meter) + 35
		Age 30–60	: (8.7 × W in kg) − (25 × H in meter) + 865
		Age > 60	: (9.2 × W in kg) + (637 × H in meter) − 302
Harris–Benedict	Men	(13.752 × W in kg) + (5.003 × H in cm) − (6.755 × age in years) + 66.473
	Women	(9.563 × W in kg) + (1.850 × H in cm) − (4.676 × age in years) + 655.096
Roza et al.	Men	(13.397 × W in kg) + (4.799 × H in cm) − (5.677 × age in years) + 88.362
	Women	(9.247 × W in kg) + (3.098 × H in cm) − (4.330 × age in years) + 447.593
Mifflin et al.	Men	(10 × W in kg) + (6.25 × H in cm) − (5 × age in years) + 5
	Women	(10 × W in kg) + (6.25 × H in cm) − (5 × age in years) − 161

FAO = Food and Agriculture Organization, H = height, UNU = United Nations University, W = weight, WHO = World Health Organization.

Both the Harris–Benedict (HB) and FWU equations, which are based on anthropometric or demographic data, are commonly used as cost-effective methods for determining the BMR. However, these equations were formulated on studies involving healthy subjects, potentially rendering them inaccurate for patients with cancer.^[7–9]

Recently, Huang et al found that a lower BMR (<1149 Kcal/d) calculated using the HB equation is linked to higher postoperative complications and poorer overall survival (OS) in patients with stage I–III gastric cancer.^[10] This suggests that while the accuracy of the HB equation compared to the BMR measured under strict conditions is unclear, the BMR calculated using the HB equation remains a significant prognostic factor for OS.

In the present study, we evaluated the prognostic value of FWU BMR in a cohort predominantly composed of Asians patients (approximately 96%) diagnosed with stage I–III gastric cancer. Unlike previous studies, we did not categorize BMR to avoid potential bias and facilitate its straightforward application in future research.

2. Methods

2.1. Study population

We conducted a retrospective study of patients who underwent curative-intent gastrectomy at Kyung Hee University Hospital in Gangdong between October 2006 and December 2018. The study included patients who met the following criteria: diagnosis of primary gastric cancer based on the Lauren classification of gastric cancer,^[11] stage I–III gastric cancer based on the 8th edition of the American Joint Committee on Cancer system,^[12] and microscopically negative resection margins. Exclusion criteria encompassed patients with concurrent or prior malignancies within the past 5 years, prior anticancer treatment, active infection, or autoimmune diseases, and those lacking available hemograms or serum albumin data. The Institutional Review Board of Kyung Hee University Hospital at Gangdong approved this study (IRB 2024-06-003), and the requirement for informed consent was waived because of the retrospective nature of the study.

2.2. Baseline clinical characteristics

We retrieved data from the electronic medical records. All data were collected within 7 days before surgical resection, and if multiple values were available, the result closest to the operation date was selected. The data included demographic parameters (age, sex, body weight, height), along with clinicopathological parameters including the American Society of Anesthesiologists Physical Status (ASA-PS) classification;^[13] tumor location; TNM stage; Lauren et al.’s histological classification;^[11] lymphatic, vascular, and perineural invasion; type of gastrectomy (TOG), hemograms (white blood cell count, hemoglobin concentration, mean corpuscular volume [MCV], and platelet count); and serum albumin level. Additionally, several combinations of prognostic factors were calculated, such as the neutrophil/lymphocyte ratio (NLR), platelet/lymphocyte ratio (PLR), lymphocyte/monocyte ratio (LMR), and Systemic Immune Inflammation Index (SII). The NLR was calculated by dividing the absolute neutrophil count (ANC) by the absolute lymphocyte count (ALC). The LMR was calculated by dividing the ALC by the absolute monocyte count (AMC). The PLR was calculated by dividing the platelet count by the ALC. The SII was calculated using the following formula: SII = (ANC × platelet count)/ALC.^[14–17]

2.3. Measurement of BMR using FAO/WHO/UNU (FWU) equations

For males aged 10 to 18 years, the BMR is calculated as (16.6 × weight in kg) + (77 × height in meters) + 572. Between ages 18 to 30 years, it is (15.4 × weight in kg) − (27 × height in meters) + 717. For men aged 30 to 60 years, the formula is (11.3 × weight in kg) + (16 × height in meters) + 901. Beyond 60 years, the BMR is calculated as (8.8 × weight in kg) + (1128 × height in meters) − 1071.

For females aged 10 to 18 years, the BMR is calculated as (7.4 × weight in kg) + (482 × height in meters) + 217. Between ages 18 to 30 years, it is (13.3 × weight in kg) + (334 × height in meters) + 35. For women aged 30 to 60 years, the formula is (8.7 × weight in kg) − (25 × height in meters) + 865. Beyond 60 years, the BMR is calculated as (9.2 × weight in kg) + (637 × height in meters) − 302 (Table 1).^[6]

2.4. Statistical analysis

The OS was calculated as the time from the date of gastrectomy to the date of death from any cause. Continuous variables remained uncategorized to avoid potential biases, except in cases where widely accepted consensus on specific cutoff values for certain variables existed. Continuous variables were presented as medians with interquartile ranges (IQRs). Nonparametric tests, such as the Mann–Whitney U test, were used for inter-group comparisons of variables. Correlations among clinical variables were assessed using a Spearman correlation analysis.

Cox regression analysis was used to calculate the hazard ratios (HRs) for various variables. Multivariate Cox regression analysis was conducted on variables deemed significant (P < .05) in univariate Cox regression analysis. Variables that did not meet the assumption of proportional hazards were excluded from multivariate analysis. Multicollinearity among variables was assessed using the variance inflation factor (VIF). Fractional polynomial curves were applied to model the relationships between significant continuous variables and log relative hazards, providing a detailed visualization of these associations.

The discriminative capacity of the models was evaluated using several metrics, including the concordance index (C-index), integrated area under the curve (iAUC), and continuous net reclassification improvement (cNRI). Inter-model differences in the C-index were assessed employing 1000 bootstrap re-samples. Additionally, C-indices for OS were plotted over a 10-year period using bootstrapping with 1000 re-samples. Differences in the iAUC over 10 years were evaluated using permutation tests with 1000 re-samples. cNRI analyses were conducted to compare the OS between the models at 36- and 60-months post-surgery. Nomograms were developed using established models to predict the OS, and calibration curves were utilized to internally validate the nomograms with 1000 bootstrap re-samples to avoid overfitting. All P values were two-sided, and statistical significance was defined as P < .05. The statistical analyses were conducted using MedCalc^® (version 22.021) and R packages (r-project.org).

3. Results

3.1. Patients’ clinical characteristics

Among the initial 528 patients evaluated for eligibility, three individuals with concurrent malignant tumors, two with microscopic residual disease, and two with stage IV disease were excluded. Consequently, this study comprised 521 patients with gastric cancer, predominantly Asians (96.3%, n = 502), with a smaller proportion of Caucasians (3.6%, n = 19). The median age of the patients was 61 years (IQR: 52–70 years). Regarding the ASA-PS score, 48 patients had a score of I; 414, score of II; and 59, score of III. The most frequent tumor location was the lower region (54.5%, n = 284), followed by the middle (33.8%, n = 176), upper (9.8%, n = 51), and diffuse (1.9%, n = 10) regions. The distribution of the cancer stages was as follows: stage I, 61.2% (n = 319); stage II, 18.6% (n = 97); and stage III, 20.2% (n = 105). Most patients underwent partial gastrectomy (79.1%, n = 412), while 20.9% (n = 109) underwent total gastrectomy (Table 2).

Table 2.

Patient characteristics: demographic, clinicopathological, and laboratory data

Variables	Median (IQR) or n (%)
Age, years	61 (52; 70)
Sex
Men	352 (67.6)
Women	169 (32.4)
ASA-PS
I	48 (9.2)
II	414 (79.5)
III	59 (11.3
Site of tumor
Upper	51 (9.8)
Middle	176 (33.8)
Lower	284 (54.5)
Diffuse	10 (1.9)
TNM stage
I	319 (61.2)
II	97 (18.6)
III	105 (20.2)
Histology (Lauren)
Intestinal	262 (50.3)
Others	259 (49.7)
Lymphatic invasion
No	343 (65.8)
Yes	178 (34.2)
Vascular invasion
No	492 (94.4)
Yes	29 (5.6)
Perineural invasion
No	469 (90.0)
Yes	52 (10.0)
TOG
Partial	412 (79.1)
Total	109 (20.9)
WBC, per μL	6500 (5390; 7900)
Hemoglobin, g/dL	13.1 (11.3; 14.2)
MCV, fL	92.0 (88.2; 95.4)
Platelet, × 103/μL	240 (206; 283)
NLR	1.9 (1.4; 2.7)
PLR	124.4 (97.4; 161.6)
LMR	4.2 (3.2; 5.4)
SII	468.9 (308.3; 717.1)
Albumin, g/dL	4.1 (3.9; 4.3)

ASA-PS = American Society of Anesthesiologists Physical Status, IQR = interquartile range, LMR = lymphocyte/monocyte ratio, MCV = mean corpuscular volume, NLR = neutrophil/lymphocyte ratio, PLR = platelet/lymphocyte ratio, SII = systemic immune-inflammation index, TNM = Tumor-Node-Metastasis, TOG = type of gastrectomy, WBC = white blood cell.

3.2. Associations of FWU BMR with clinical and laboratory variables

The median BMR, calculated using the FWU equation, was 1380 Kcal/day (IQR: 1251–1629 Kcal/day). The Mann–Whitney U test revealed significant differences in BMR values between groups for categorical variables such as sex and anemia (both P < .001). No significant differences in the BMR values were observed for ASA-PS, TNM stage, histology, lymphatic invasion, vascular invasion, or perineural invasion (Table 3).

Table 3.

Difference in FWU BMR across variable groups

Variables	n (%)	FWU BMR Median (IQR)	P value
Sex
Men	352 (67.6)	1520 (1333; 1687)	<.001
Women	169 (32.4)	1258 (1140; 1329)
ASA-PS
I–II	462 (88.7)	1386 (1247; 1632)	.171
III	59 (11.3)	1331 (1273; 1440)
TNM stage
I–II	416 (79.9)	1380 (1261; 1629)	.900
III	105 (20.1)	1387 (1249; 1622)
Histology
Intestinal	262 (50.3)	1367 (1229; 1588)	.053
Others	259 (49.7)	1390 (1283; 1651)
Lymphatic invasion
No	343 (65.8)	1373 (1270; 1623)	.709
Yes	178 (34.2)	1389 (1225; 1636)
Vascular invasion
No	492 (94.4)	1383 (1263; 1630)	.210
Yes	29 (5.6)	1364 (1187; 1451)
Perineural invasion
No	469 (90.0)	1373 (1256; 1619)	.313
Yes	52 (10.0)	1476 (1250; 1670)
Anemia
Present	194 (37.2)	1333 (1211; 1472)	<.001
Absent	327 (62.8)	1438 (1288; 1668)

The Mann–Whitney U test was used for inter-group comparisons of variables.

ASA-PS = American Society of Anesthesiologists Physical Status, BMR = basal metabolic rate, FWU = the Food and Agriculture Organization/World Health Organization/United Nations University, IQR = interquartile range, TNM = Tumor-Node-Metastasis, TOG = type of gastrectomy.

Spearman correlation analysis revealed a moderate negative correlation between FWU BMR and age (r = −0.56, P < .001). In contrast, FWU BMR showed weak or minimal correlations with other clinical or laboratory variables, including tumor size (r = −0.07, P = .136), WBC (r = 0.14, P = .001), ANC (r = 0.12, P = .005), AMC (r = 0.11, P = .009), ALC (r = 0.11, P = .010), hemoglobin level (r = 0.39, P < .001), MCV (r = −0.10, P = .022), platelet count (r = 0.04, P = .376), NLR (r = 0.02, P = .671), LMR (r = −0.00, P = .946), PLR (r = −0.07, P = .107), SII (r = 0.04, P = .309), and serum albumin level (r = 0.21, P < .001).

3.3. Cox regression of the risk factors of OS

The median follow-up duration was 104.2 months (IQR: 54.2–140.0 months). The 5-year OS rates were 92.3%, 76.6%, and 46.5% for stages I, II, and III, respectively. Univariate Cox regression analysis revealed significant associations between OS and several variables including age, sex, ASA-PS, TNM stage, lymphatic invasion, vascular invasion, perineural invasion, TOG, hemoglobin level, MCV, NLR, PLR, LMR, SII, serum albumin level, and FWU BMR. However, age, lymphatic invasion, and perineural invasion did not meet the assumption of proportional hazards and were thus excluded from further analysis (Table 4).

Table 4.

Univariate and multivariate cox regression analysis for predicting overall survival

Covariate*	Univariate HR (95% CI)	P value	Multivariate HR (95% CI, FWU model)	P value
Age, years†	1.065 (1.040; 1.081)	<.001
Sex (women vs men)	0.617 (0.424; 0.897)	.012	0.463 (0.298; 0.720)	<.001
ASA-PS	2.133 (1.509; 3.013)	<.001	1.584 (1.103; 2.274)	.013
Stage (III vs I–II)	4.684 (3.377; 6.497)	<.001	2.675 (1.867; 3.831)	<.001
Histology (intestinal vs others)	1.052 (0.763; 1.451)	.757
Lymphatic invasion (yes vs no)†	2.829 (2.047; 3.909)	<.001
Vascular invasion (yes vs no)	3.695 (2.276; 5.998)	<.001	1.752 (1.032; 2.972)	.038
Perineural invasion (yes vs no)†	2.028 (1.288; 3.194)	.002
TOG (total vs partial)	2.252 (1.599; 3.172)	<.001	1.637 (1.150; 2.331)	.006
WBC, per μL	1.000 (1.000; 1.000)	.391
Hemoglobin, g/dL	0.774 (0.729; 0.822)	<.001	0.905 (0.841; 0.974)	.008
MCV, fL	0.972 (0.954; 0.991)	.005
Platelet, ×103/μL	1.002 (1.000; 1.004)	.134
NLR	1.163 (1.096; 1.234)	<.001
PLR	1.006 (1.004; 1.008)	<.001
LMR	0.838 (0.749; 0.937)	.002
SII	1.000 (1.000; 1.000)	<.001
Albumin, g/dL	0.185 (0.139; 0.247)	<.001	0.388 (0.266; 0.565)	<.001
BMR (FWU)	0.999 (0.998; 0.999)	<.001	0.999 (0.998; 0.999)	<.001

ASA-PS = American Society of Anesthesiologists Physical Status, CI = confidence interval, FWU = the Food and Agriculture Organization/World Health Organization/United Nations University, HR = hazard ratio, LMR = lymphocyte/monocyte ratio, MCV = mean corpuscular volume, NLR = neutrophil/lymphocyte ratio, PLR = platelet/lymphocyte ratio, SII = systemic immune-inflammation index, TNM = Tumor-Node-Metastasis, TOG = type of gastrectomy, WBC = white blood cell.

^*The right-hand values in parentheses are the reference values.

^†Not compatible with the assumption of proportional hazards.

Multivariate Cox regression analysis identified sex (hazard ratio [HR] 0.463; P < .001), ASA-PS (HR 1.584; P = .013), TNM stage (HR 2.675; P < .001), vascular invasion (HR 1.752; P = .038), TOG (HR 1.637; P = .006), hemoglobin level (HR 0.905; P = .008), serum albumin level (HR 0.388; P < .001), and FWU BMR (HR 0.999; P < .001) as significant determinants of OS (Table 4). The VIFs for the covariates were 1.37 for sex, 1.14 for ASA-PS, 1.17 for TNM stage, 1.11 for vascular invasion, 1.06 for TOG, 1.48 for hemoglobin level, 1.42 for albumin level, and 1.41 for FWU BMR, demonstrating no collinearity between the covariates. The FWU model was established by incorporating the following significant variables: sex, ASA-PS score, TNM stage, vascular invasion, TOG, hemoglobin level, serum albumin level, and FWU BMR.

Fractional polynomial analysis of the log relative hazard revealed a continuous inverse association between each variable and the risk of death. For hemoglobin, a linear inverse relationship was observed, with higher levels associated with a reduced risk of death. Similarly, serum albumin showed a continuous inverse association, where higher levels were linked to a lower risk of mortality. For FWU BMR, a linear inverse association was also noted, indicating that higher FWU BMR values correlated with a reduced risk of death (Fig. 1).

Figure 1.

Fractional polynomial models of log relative hazard for hemoglobin level, serum albumin level, and FWU BMR with 95% confidence intervals. BMR = basal metabolic rate, FWU = the Food and Agriculture Organization/World Health Organization/United Nations University.

3.4. Model comparison between FWU model and baseline or intermediate models

When comparing the FWU model with the baseline model, which relied solely on the TNM stage, the FWU model’s C-index was significantly higher at 0.815 than the baseline model’s C-index of 0.664 (P < .001). Similarly, the iAUC of the FWU model was 0.775, significantly surpassing the baseline model’s iAUC of 0.646 (P < .001) (Table 5). Over a 10-year period, the C-index of the FWU model remained consistently higher than that of the baseline model, as demonstrated by bootstrap cross-validation with 1000 resamples (Fig. 2). Furthermore, cNRI analysis indicated a significant improvement in the FWU model compared to the baseline model at both 36 and 60 months after surgery (both P < .001) (Table 6).

Table 5.

Comparisons of FWU model with baseline and intermediate models for predicting survival outcomes using C-index and iAUC

Model comparison	Model	C-index (SE)	iAUC (SE)	P value (C-index)	P value (iAUC)
FWU vs Baseline	FWU	0.815 (0.017)	0.775 (0.016)	<.001	<.001
	Baseline	0.664 (0.020)	0.646 (0.010)
FWU vs Intermediate	FWU	0.815 (0.017)	0.775 (0.016)	.010	.002
	Intermediate	0.799 (0.019)	0.761 (0.017)

The FWU model incorporates sex, American Society of Anesthesiologists Physical Status, stage, vascular invasion, type of gastrectomy, hemoglobin level, serum albumin level, and FWU BMR. The intermediate model incorporated the same covariates as the FWU model, except for the FWU BMR. The baseline model relied solely on the stage.

C-index = concordance index, FWU = the Food and Agriculture Organization/World Health Organization/United Nations University, iAUC = integrated area under the curve, SE = standard error.

Figure 2.

Concordance indices of models for survival outcomes over a 10-year period.

Table 6.

Improvement in survival outcome predictions at specific time points using cNRI

Model comparison	Metrics (mo)	Difference (95% CI)	P value
FWU vs Baseline	cNRI (36)	0.317 (0.190; 0.449)	<.001
	cNRI (60)	0.328 (0.227; 0.440)	<.001
FWU vs Intermediate	cNRI (36)	0.221 (0.002; 0.326)	.048
	cNRI (60)	0.272 (0.040; 0.377)	.032

The FWU model incorporates sex, American Society of Anesthesiologists Physical Status, stage, vascular invasion, type of gastrectomy, hemoglobin level, serum albumin level, and FWU BMR. The intermediate model incorporated the same covariates as the FWU model, except for the FWU BMR. The baseline model relied solely on the stage.

CI = confidence interval, cNRI = continuous net reclassification improvement, FWU = the Food and Agriculture Organization/World Health Organization/United Nations University.

When comparing the FWU model with the intermediate model, which included the same covariates as those of the FWU model (except for the FWU BMR), the FWU model demonstrated superior predictive performance, with a C-index of 0.815, compared to the intermediate model’s C-index of 0.799 (P = .010). The iAUC was also significantly higher for the FWU model at 0.775 compared with 0.761 for the intermediate model (P = .002) (Table 5). When the C-indices for OS were plotted over a 10-year period, the FWU model consistently outperformed the intermediate model (Fig. 2). The cNRI analysis showed a significant improvement in the FWU model over the intermediate model at 36- and 60-months post-surgery (P = .048 and P = .032, respectively) (Table 6).

Therefore, the inclusion of the FWU BMR in the prognostic model significantly enhanced the predictive accuracy for OS in patients with gastric cancer undergoing curative-intent resection.

3.5. Model comparison between FWU model and other models such as the HB model, the Roza model, and the Mifflin model

To evaluate the prognostic impact of the FWU model, we established three additional models: HB, Roza, and Mifflin. Each of these models incorporated the same covariates as those of the FWU model (sex, ASA-PS, TNM stage, TOG, hemoglobin level, and serum albumin level) but excluded vascular invasion and included BMR calculated using different equations. Specifically, the HB model used the HB equation;^[4] the Roza model, the Roza et al equation;^[18] and the Mifflin model, the Mifflin et al equation (Tables 1 and 7).^[19]

Table 7.

Predictive models for overall survival using different BMR equations

Covariate*	Harris–Benedict model		Roza model		Mifflin model
	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value
Sex (women vs men)	0.509 (0.335; 0.775)	.002	0.474 (0.305; 0.736)	<.001	0.366 (0.217; 0.619)	<.001
ASA-PS	1.558 (1.093; 2.223)	.014	1.572 (1.103; 2.242)	.013	1.578 (1.107; 2.250)	.012
Stage (III vs I–II)	2.905 (2.043; 4.129)	<.001	2.889 (2.031; 4.119)	<.001	2.900 (2.039; 4.127)	<.001
TOG (total vs partial)	1.694 (1.191; 2.408)	.003	1.691 (1.189; 2.406)	.004	1.691 (1.188; 2.407)	.004
Hemoglobin, g/dL	0.917 (0.850; 0.988)	.023	0.915 (0.849; 0.986)	.020	0.915 (0.849; 0.986)	.020
Albumin, g/dL	0.383 (0.262; 0.560)	<.001	0.381 (0.261; 0.556)	<.001	0.378 (0.259; 0.551)	<.001
BMR (Harris–Benedict)	0.998 (0.997; 0.999)	<.001	–	–	–	–
BMR (Roza)	–	–	0.998 (0.997; 0.999)	<.001	–	–
BMR (Mifflin)	–	–	–	–	0.998 (0.998; 0.999)	<.001

ASA-PS = American Society of Anesthesiologists Physical Status, BMR = basal metabolic rate, CI = confidence interval, HR = hazard ratio, TOG = type of gastrectomy.

^*The right-hand values in parentheses are the reference values.

The C-index values of the HB, Roza, and Mifflin models were 0.813, 0.812, and 0.811, respectively. The FWU model had a high C-index of 0.815, although the differences between the FWU model and the other 3 models were not statistically significant. Similarly, the iAUC for the HB, Roza, and Mifflin models were 0.772, 0.771, and 0.772, respectively. The iAUC of the FWU model was higher, at 0.775; however, the differences were not statistically significant (Table 8).

Table 8.

Comparisons of survival outcome prediction models using C-index and iAUC

Model comparison	C-index (FWU vs Other)	P value	iAUC (FWU vs Other)	P value
FWU vs Harris–Benedict	0.815 vs 0.813	.813	0.775 vs 0.772	.539
FWU vs Riza	0.815 vs 0.812	.813	0.775 vs 0.771	.322
FWU vs Mifflin	0.815 vs 0.811	.490	0.775 vs 0.772	.326

C-index = concordance index, FWU = the Food and Agriculture Organization/World Health Organization/United Nations University, iAUC = integrated area under the curve.

3.6. Establishment and validation of nomogram

A nomogram for predicting OS was developed based on the FWU model (Fig. 3), incorporating variables such as sex, ASA-PS, TNM stage, vascular invasion, TOG, hemoglobin level, serum albumin level, and FWU BMR. Calibration curves demonstrated good agreement between the predicted and actual survival outcomes (Fig. 4).

Figure 3.

Nomogram predicting 3-year and 5-year overall survival according to the FWU model. FWU = the Food and Agriculture Organization/World Health Organization/United Nations University, TOG = type of gastrectomy, VI = vascular invasion.

Figure 4.

Calibration curve analysis of the FWU model for 3-year (A) and 5-year (B) Overall survival. FWU = the Food and Agriculture Organization/World Health Organization/United Nations University

4. Discussion

FWU BMR was identified as a significant predictor of OS in our multivariate Cox regression analysis, with fractional polynomial modeling demonstrating a linear relationship between higher FWU BMR values and a reduced risk of death. The predictive capability of the FWU model, which incorporates FWU BMR into the intermediate model, exceeded that of both the intermediate model and the baseline model based solely on TNM staging. Moreover, the superior performance of the FWU model, as highlighted by comparisons with other BMR predictive models, underscores its clinical relevance and utility in optimizing patient management.

The BMR denotes the fundamental daily energy expenditure required to maintain essential bodily functions. In the early 20th century, BMR measurements were primarily utilized for diagnosing and evaluating the therapeutic effectiveness of thyroid disorders, including hypo- and hyperthyroidism.^[3] Subsequently, BMR has served as a marker for energy expenditure. Achieving precise BMR measurements requires controlled conditions to minimize external influences, such as temperature variations, physical activity, and the effects of food or medications. Nevertheless, consistently imposing such stringent conditions has proven to be impractical.^[3]

In the early 20th century, Harris and Benedict conducted a meticulous biometric analysis, resulting in the publication of “A Biometric Study of Basal Metabolism in Man.”^[4] They developed equations to predict the BMR using rigorous statistical methods, marking a significant departure from previous approaches. Their study introduced biometric principles into BMR analysis. Despite their contributions, the HB equations tended to overestimate the BMR, particularly among young women.^[3,20] Nonetheless, their simplicity has ensured their enduring popularity, and clinicians in North America continue to widely utilize the HB equations.

Schofield et al conducted an extensive literature review to support the FWU Expert Consultation on “Energy and Protein Requirements,” laying the groundwork for predictive equations for both sexes across various age groups. These equations form the basis of the 1985 FWU document, “Energy and Protein Requirements.”^[5,6] While emphasizing the importance of using BMR to estimate human energy needs, Schofield’s equations often overestimate the BMR. The dataset used included 47% Italian subjects with a higher BMR per kilogram, contributing to this overestimation and limiting the global applicability of the equations. Despite these limitations, these equations remain in use because of their historical significance and widespread acceptance.^[3,17]

Based on anthropometric and demographic data, both the HB and FWU equations are commonly used as cost-effective alternatives for estimating the BMR (Table 1). However, these were developed based on studies of healthy subjects and may not accurately predict BMR in cancer cohorts.^[7] In Ozorio et al.’s study involving patients with advanced gastrointestinal cancer (stages III–IV), a moderate agreement was found between resting energy expenditure (REE) measured through calorimetry and BMR estimated using the HB equation (ICC = 0.520).^[8] Conversely, a study by de Souza et al in patients with gastrointestinal or head and neck cancer (97% in stages III–IV) found poor agreement in patients aged > 70 years (ICC = 0.466).^[9] Given that patients with gastric cancer constituted only approximately 20% of these cohorts, mostly in advanced stages, the concordance between REE and the HB equation in patients with gastric cancer remains uncertain, especially for those with locally advanced disease. Further research is required to determine the applicability of this equation in patients with gastric cancer.

The disparity in the BMR values between patients with gastric cancer and healthy controls remains unclear. However, some studies suggest that REE in patients with gastric cancer is equal to or exceeds that in normal controls.^[7] In individuals without malignancy, elevated BMR is often considered a negative prognostic indicator and is correlated with increased susceptibility to malignancies, cardiovascular diseases, and reduced lifespan.^[21–26] Conversely, in individuals with malignancies, a decreased BMR is often regarded as a negative prognostic factor. Huang et al recently reported that patients with stage I–III gastric cancer and a lower BMR (as calculated using the HB equation) are more susceptible to postoperative complications and experience poorer OS outcomes.^[10] Similarly, in the present study involving patients with stage I–III gastric cancer, predominantly of Asian descent (approximately 96%), a lower BMR (as calculated using the FWU equation) was associated with poorer OS outcomes. Therefore, BMR can be used as a prognostic marker of survival outcomes.

The mechanisms underlying the predictive power of BMR in forecasting survival outcomes among patients with gastric cancer are likely complex. The BMR represents the energy required to maintain basic physiological functions at rest. Higher BMRs may indicate better metabolic and nutritional status, which are essential for immune function, wound healing, and recovery post-surgery.^[10,27,28] An elevated BMR suggests better overall health and resilience to stressors, including surgery. Interestingly, in our study, BMR showed only weak correlations with nutritional parameters, including serum albumin (r = 0.21, P < .001), prognostic nutritional index (r = 0.22, P < .001), and geriatric nutritional risk index (r = 0.29, P < .001), which differs from findings in previous research.^[26,27] Additionally, BMR is closely linked to muscle mass and lean body mass. Patients with higher muscle mass typically exhibit higher BMRs, indicating a healthier body composition.^[29] Muscle wasting, which is common in advanced stages of cancer, is associated with poor prognosis and survival outcomes.

The clinical application of the FWU BMR includes several key aspects. First, BMR, which reflects energy expenditure at rest, serves as a surrogate marker for nutritional assessment, helping identify patients who could benefit from nutritional interventions to improve recovery and survival outcomes. Second, by identifying patients with lower BMRs at a higher risk of poor outcomes, clinicians can implement targeted interventions such as personalized nutritional support, intensified monitoring, and proactive management. Future research should refine these equations or develop new ones to enhance predictive accuracy across diverse populations.

In addition to the FWU BMR, several other prognostic factors for OS were identified, including sex, ASA-PS, TNM stage, vascular invasion, TOG, hemoglobin level, and serum albumin level. The TNM staging system, widely acknowledged as the gold standard for cancer prognostication,^[30–33] has been consistently emphasized for its prognostic significance in gastric cancer survival.^[34] Multivariate Cox regression analysis confirmed that the TNM stage was a significant determinant of OS, even after adjusting for other covariates. Total gastrectomy has historically been associated with higher rates of 30-day morbidity and readmission owing to nutritional challenges and remains a significant predictor of OS.^[33,35,36] Anemia, arising from various causes including nutritional deficiencies and chronic disease, was identified as an independent prognostic factor for OS, with our study corroborating prior findings.^[37] Low serum albumin levels, indicative of poor nutritional status and systemic inflammation, were associated with an increased risk of postoperative complications and poor survival outcomes.^[38–42] They remained a significant predictor of OS in our study.

For continuous variables such as hemoglobin level, serum albumin level, and FWU BMR, all of which were significant in the multivariate Cox regression, we employed fractional polynomial curves to explore their associations with log relative hazard. The analysis revealed a continuous inverse association between each variable and the risk of death, underscoring their potential prognostic importance for survival outcomes.

In this study, the FWU model, which integrates FWU BMR with other clinical variables, demonstrated superior predictive performance compared with models excluding BMR, underscoring its clinical relevance in gastric cancer prognostication. The c-index and iAUC of the FWU model were higher than those of the models that used alternative BMR equations (such as HB, Roza et al, and Mifflin et al), although these differences were not statistically significant. A nomogram for OS prediction was developed based on the FWU model, and the calibration curves indicated good agreement between the predicted and actual survival outcomes.

This study has multiple strengths. Notably, it is the first to investigate the prognostic value of FWU BMR in patients with gastric cancer undergoing curative-intent resection, offering new insights into its clinical utility. To ensure the robustness and applicability of our findings, we employed rigorous statistical methods, including 1000 bootstrap iterations, enhancing the model’s generalizability. By maintaining FWU BMR as a continuous variable, we preserved statistical power and reduced overfitting, supporting applicability across diverse patient groups. Additionally, fractional polynomial modeling provided a nuanced view of the linear association between FWU BMR and OS, revealing patterns that may not be immediately apparent in tabular data. We also integrated FWU BMR into a novel prognostic model, demonstrating superior predictive accuracy over existing equations like HB, Roza, and Mifflin. This approach adds precision to survival predictions in patients with gastric cancer, assisting clinicians in making informed, individualized management decisions. The development of a nomogram based on the FWU model provides an accessible, cost-effective tool for clinical use, as FWU BMR is routinely measured preoperatively. Despite inherent limitations, the broad acceptance and established usage of FWU equations lend historical context and credibility to this study, which may significantly impact tailored postsurgical care and follow-up strategies for patients with gastric cancer.

However, caution is warranted when interpreting the results owing to several limitations. While the FWU equations, based on anthropometric and demographic data, are commonly used as expedient and cost-effective alternatives for estimating BMR, it is important to note that these equations were derived from studies involving healthy subjects and may not accurately predict BMR in cancer cohorts.^[7] In addition, there is a possibility of overestimation of the BMR equations. Further studies comparing the FWU equations with calorimetry are necessary to evaluate their concordance in patients with gastric cancer.

In conclusion, FWU BMR was identified as a significant predictor of OS in our multivariate Cox regression analysis, demonstrating a linear relationship between higher FWU BMR values and reduced mortality risk. The FWU model, which incorporates FWU BMR, outperformed both the intermediate model and the baseline TNM staging model. Its superior predictive performance underscores its clinical relevance in patient management. Future research should refine BMR assessment methods and validate the model with independent cohorts to enhance prognostic accuracy and generalizability across diverse populations.

Author contributions

Conceptualization: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Data curation: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Formal analysis: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Funding acquisition: Soomin An, Wankyu Eo, Sookyung Lee.

Investigation: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Methodology: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Project administration: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Resources: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Software: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Supervision: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Validation: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Visualization: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Writing – original draft: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.

Writing – review & editing: Soomin An, Wankyu Eo, Seol Bin Kim, Sookyung Lee.