Predictive Ability of Chinese Visceral Adipose Index for Gastric Intestinal Metaplasia in Korean Adults

Abstract

Introduction

Gastric intestinal metaplasia (GIM) is a precancerous lesion of gastric cancer. Despite increasing interest for association between obesity and gastric cancer, there is still ongoing controversy for effect of obesity on gastric cancer. Therefore, it is important to discover obesity index with reliable predictability for GIM.

Methods

Study participants were 85,986 Koreans who were free of GIM at endoscopy. They were categorized by quartile levels of body mass index (BMI), waist circumference (WC), and Chinese visceral adipose index (CVAI). Cox proportional hazard assumption was used to evaluate hazard ratio (HR) and 95% confidence interval (CI) for incident GIM (adjusted HR [95% CI]) according to quartile levels of BMI, WC, and CVAI for about 8 years of follow-up. Using the time ROC package, ROC and area under curve (AUC) were calculated for each quartile group of BMI, WC, and CVAI at 2, 4, 6, and 8 years.

Results

In men, BMI ≥ quartile 3, WC ≥ quartile 2, and CVAI ≥ quartile 2 were more significantly associated with incident GIM, compared with quartile 1 groups (reference). In women, while BMI and WC did not show the significant association with GIM, CVAI was solely associated with the increased risk of GIM. Time-dependent ROC and AUC analyses indicated that CVAI was superior to BMI and WC in predicting GIM among both men and women. The predictive ability of CVAI was more prominent in women than men.

Conclusion

CVAI was an effective predictor for GIM, surpassing the predictive ability of BMI and WC.

Introduction

Although the incidence and mortality of gastric cancer has declined globally over the decades, gastric cancer is still responsible for the fifth most common malignancy and the fourth leading cause of cancer-related deaths in 2020 [1]. In particular, East Asia has an exceptionally high incidence of gastric cancer, accounting for 60% of newly developed gastric cancers [2]. South Korea has the highest incidence of gastric cancer, which is approximately 10 times higher than that of the US whites [3].

Multidirectional approaches have been used to determine the pathophysiology and risk factors of gastric cancer. The major pathogenesis of gastric cancer is the progression from normal gastric mucosa to invasive gastric carcinoma through precancerous lesions [4]. Gastric intestinal metaplasia (GIM) is an intermediate precancerous gastric lesion in the gastric cancer cascade of chronic gastritis, atrophic gastritis, intestinal metaplasia (IM), dysplasia, and adenocarcinoma [5]. Thus, it is clinically important to investigate the risk factors for GIM to prevent gastric cancer.

Obesity is known to be a risk factor for diverse cancers of the digestive system, including gastric cardia cancer [6, 7]. However, it remains unclear whether obesity is independently associated with the overall risk of gastric cancer. To ascertain the association between obesity and gastric cancer, it is necessary to examine the effect of obesity on precancerous lesions for gastric cancer. However, observational studies have presented heterogeneous results for the association between obesity indices and precancerous lesions, such as atrophic gastritis and GIM [8–12].

The Chinese visceral adiposity index (CVAI) was developed to precisely assess visceral adipose tissue (VAT) in Chinese adults [13]. Evidence has shown that CVAI is superior to other surrogate markers of obesity in predicting cardiometabolic diseases, including prediabetes, diabetes, atherosclerotic disease, diabetic kidney disease, and hypertension, among Asians [14–16]. However, no study has investigated the association between CVAI and the risk of precancerous lesions of gastric cancer.

To evaluate the predictive ability of CVAI for GIM, we quantified the risk of GIM in relation to the quartile levels of CVAI in 85,986 Korean adults who periodically underwent endoscopy as an item of health checkup. In addition, we compared the predictive ability of GIM among surrogate markers of obesity, including the CVAI, body mass index (BMI), and waist circumference (WC).

Methods

Study Participants

The data used in this retrospective study were obtained from the Kangbuk Samsung Health Study (KSHS) cohort. The KSHS cohort was based on health data of employees and their spouses who underwent health checkups annually or every 2 years in accordance with Korea’s Industrial Safety and Health Law. The baseline study participants were 177,151 adults who visited the Total Healthcare Center, Kangbuk Samsung Hospital, between March 2011 and December 2012 and underwent health checkups, including gastroscopy. For participants who underwent several health checkups during the baseline study period, the first visit was used as the reference.

Among the 177,151 study participants in the baseline study, 66,228 participants were excluded because they met one or more of the following exclusion criteria: (1) 178 participants had missing values in the variables necessary to calculate BMI, CVAI, or WC; (2) 60,770 participants had one or more missing values in covariates (e.g., alcohol intake, history of hypertension, and marriage); (3) 2,746 participants taking antihyperlipidemic medication; (4) 92 participants had a history of subtotal or total gastrectomy; and (5) 2,442 participants had a history of cancer, including gastric cancer. A total of 5,555 participants who did not meet the exclusion criteria, but already had GIM, were excluded from the study. The remaining 105,368 participants were included in the baseline study. Among these participants, 85,986 attended the Total Healthcare Center, Kangbuk Samsung Hospital, from January 2013 to December 2019 and underwent gastric endoscopy. Consequently, 85,986 participants were included in the final analysis.

Clinical, Anthropometric, and Biochemical Data

Data pertaining to medical history, current medication usage, socioeconomic status (e.g., education), and health-related behaviors (e.g., exercise) were obtained using a self-administered questionnaire. Anthropometric measurements including weight, WC, and height were obtained for all participants. Smoking patterns were classified into four categories: never, former, current, and no response. Never smoker is an adult who has never smoked or has smoked fewer than 100 cigarettes in their lifetime. A former smoker is an adult who has smoked at least 100 cigarettes in their lifetime but has quit smoking. Physical activity was assessed using the Korean-validated version of the International Physical Activity Questionnaire (IPAQ) short form (SF) and categorized into three levels (low, moderate, and high) in accordance with the guidelines established by the IPAQ core group (http://www.ipaq.ki.se). Hypertension was defined as a prior diagnosis of hypertension, current use of antihypertensive medication, or measured blood pressure of ≥140/90 mm Hg in accordance with the guidelines of the Korean Hypertension Society [17]. Blood pressure was measured three times in a sitting position after a 5-min rest at a minimum of 30-s intervals by a trained nurse using an automated device (53000-E2; Welch Allyn, USA). Diabetes mellitus (DM) was defined as one of the following conditions; fasting glucose ≥126 mg/dL, hemoglobin A1c ≥6.5%, current use of oral hypoglycemic agents and/or insulin, and a prior diagnosis of DM [18].

Total cholesterol and triglyceride levels were quantified using enzymatic colorimetric assays. Low-density lipoprotein cholesterol (LDL cholesterol) was determined employing a homogeneous enzymatic colorimetric test, while high-density lipoprotein cholesterol (HDL cholesterol) was assessed using a selective inhibition method (Advia 1650 Autoanalyzer, Bayer Diagnostics; Leverkusen, Germany).

BMI was calculated by dividing weight (kg) by height (m²). Obesity was defined as BMI ≥25 kg/m² according to the International Obesity Task Force recommendations [19]. WC was measured at the narrowest point between the lower border of the rib cage and iliac crest during minimal respiration in the erect position. Abdominal obesity was defined according to criteria established by the Korean Society for the Study of Obesity (WC ≥90 cm for men and ≥85 cm for women) [20]. CVAI was calculated using the following equation [13]:

Men: CVAI = −267.93 + 0.68 × age + 0.03 × BMI + 4.00 × WC + 22.00 × Log₁₀TG − 16.32 × HDL-cholesterol.

Women: CVAI = −187.32 + 1.71 × age + 4.23 × BMI + 1.12 × WC + 39.76 × Log₁₀TG − 11.66 × HDL-cholesterol.

TG and HDL cholesterol levels were calculated in mmol/L.

Endoscopic Data Collection

Experienced and certified endoscopists conducted endoscopic examinations using a conventional white light endoscope (GIF H260; Olympus Medical Systems). Endoscopists systematically examined and digitally recorded images of the esophagus, stomach, and duodenum. Any abnormal findings observed during the endoscopic procedure were described by noting their location, size, and characteristics. GIM was diagnosed based on endoscopic findings in accordance with the established standard criteria [21]. Endoscopic findings indicative of GIM are characterized by white plaque-like elevations in the antrum and corpus [21]. Examination for Helicobacter pylori (H. pylori) infection was performed only when deemed necessary according to the Korean guidelines, specifically in cases of suspected gastric cancer, mucosa-associated lymphoid tissue lymphoma, and peptic ulcer disease in the stomach and duodenum [22]. Comprehensive descriptions of the study population and data collection methodologies have been previously published [16].

Statistical Analysis

Considering the differences in anthropometric measurements, visceral obesity, and prevalence of GIM between the sexes, analyses were performed separately for men and women. In the baseline study from 2011 to 2012, we calculated the mean values with standard deviations for men, women, and all participants. To assess the differences between sexes, a t test was used for continuous variables (e.g., BMI), whereas a chi-square test was used for categorical variables (e.g., current smoking rate).

To analyze the association between BMI, WC, CVAI, and the risk of developing GIM in longitudinal analyses, study participants were categorized according to quartiles of baseline BMI, WC, and CVAI. The Cox proportional hazards model was used to calculate the unadjusted and multivariable-adjusted hazard ratios (HRs) for GIM and their 95% confidence intervals (95% CIs) in each quartile group (adjusted HR [95% CI]). The covariates of the multivariable model were selected from the factors that could potentially influence obesity and the development of GIM. Multiple covariates included age, sex, physical activity, smoking, alcohol intake (g/day), glycemic status, hypertension, marital status, and education level. To assess multicollinearity between variables, we analyzed the variance inflation factor, and it was confirmed that no variables exhibited a variance inflation factor greater than 10. The proportional hazards assumption was verified using log-log plots. The incidence of cases and the incidence density (incidence cases per 1,000 person-years) of GIM, and mean age at incident GIM were calculated for each group. In addition, the dynamic changes in BMI, WC, and CVAI during the follow-up period were categorized into four groups: increased, stable, mildly decreased, and moderately decreased. Among the study participants, 65,980 individuals (42,398 males and 23,582 females) who underwent gastroscopy in the final phase of the study (2018–2019) and had no missing values for weight, WC, triglycerides, and HDL cholesterol were included in the dynamic change analysis. Both unadjusted and multivariate-adjusted HRs with 95% CIs were calculated, and the incidence density, case numbers, percentage, and person-years were obtained. The mean values of the covariates in the four dynamic change groups are also presented in the online supplementary Table (for all online suppl. material, see https://doi.org/10.1159/000551835).

To compare the predictive power of GIM for BMI, WC, and CVAI over time, we conducted time-dependent receiver operating characteristic (ROC) analysis. Using the time ROC package, ROC and area under curve (AUC) were calculated for BMI, WC, and CVAI at 2, 4, 6, and 8 years, respectively. Due to computational limitations, the analysis was performed on a randomly selected 20% sample of men and women. Randomized sample selection was performed three times using the R sample function, and a moderate result (dataset 1) was selected. Sensitivity analyses to assess the robustness of the randomized sample selection were performed using dataset 2 and dataset 3, both of which were generated by the R sample function.

We also compared the predictive capabilities of WC and CVAI with BMI (reference) and obtained p value. All statistical analyses were executed using R 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria), and a two-sided p value <0.05 (two-sided) was considered statistically significant across all analyses.

Results

The baseline characteristics of the study participants are presented in Table 1. The overall characteristics of the study participants were featured by preponderance of men (63.2%, n = 54,353) and relatively young age (39.1 ± 6.7 years). In the baseline study, women had more favorable clinical characteristics, including lower levels of current smoking, alcohol intake, and prevalence of hypertension and diabetes, than men. During the follow-up period, 6,556 participants (7.6%) had GIM, which was more common in men (9.7%) than in women (4.0%).

Table 1.

Baseline characteristics of the study participants according to gender

Characteristics	Overall	Women	Men	p value
Number	85,986	31,633	54,353	​
Age, years	39.1±6.7	38.6±6.5	39.4±6.8	<0.001
WC, cm	82.2±9.0	76.0±7.8	85.8±7.7	<0.001
BMI, kg/m2	23.3±3.2	21.6±2.9	24.4±2.9	< 0.001
CVAI	64.2±42.1	31.9±30.8	83.0±36.0	<0.001
Obesity, %	28.8	11.5	38.9	<0.001
Abdominal obesity, %	22.0	12.3	27.6	<0.001
Triglyceride, mmol/L	1.3±0.9	0.9±0.5	1.5±1.0	<0.001
HDL cholesterol, mmol/L	1.5±0.4	1.7±0.4	1.4±0.3	<0.001
Average alcohol use, g/day	15.3±22.6	6.0±12.4	20.7±25.2	<0.001
Current smoker, %	23.9	2.0	36.7	<0.001
High PA, %	17.3	17.1	17.4	<0.001
High education, %	73.0	62.9	78.8	<0.001
Married, %	86.6	89.1	85.1	<0.001
HTN, %	9.6	3.6	13.1	<0.001
Glycemic status, %	​	​	​	<0.001
DM	3.2	1.5	4.2	​
Prediabetes	46.3	42.5	48.5	​
Normal	50.5	56.1	47.3	​
Incidence of IM, n (%)	6,556 (7.6)	1,266 (4.0)	5,290 (9.7)	<0.001

Continuous variables are expressed as mean (±SD), and categorical variables are expressed as number (percentage [%]).

BMI, body mass index; PA, physical activity; HTN, hypertension; DM, diabetes mellitus; CVAI, Chinese visceral adiposity index; IM, intestinal metaplasia.

Table 2 indicates the HR and 95% CI for GIM according to the quartile levels of BMI, WC, and CVAI in men. Compared to the first quartile (reference), higher quartiles of BMI, WC, and CVAI were associated with a modest increase in the HR for GIM. Multivariable-adjusted HR and 95% CI showed that BMI ≥ quartile 3 was more significantly associated with GIM than quartile 1 (quartile 1: reference, quartile 2: 1.03 [0.95–1.12], quartile 3: 1.20 [1.11–1.30], and quartile 4: 1.14 [1.05–1.24]). In WC, groups with quartile level ≥2 had a higher risk of GIM, compared with quartile 1 (quartile 1: reference, quartile 2: 1.13 [1.04–1.22], quartile 3: 1.12 [1.04–1.22], and quartile 4: 1.14 [1.05–1.23]). This finding was identically observed in association between CVAI and the risk for GIM (quartile 1: reference, quartile 2: 1.10 [1.02–1.20], quartile 3: 1.15 [1.06–1.24], and quartile 4: 1.16 [1.07–1.26]).

Table 2.

HRs and 95% CIs for intestinal metaplasia according to the quartile groups of BMI, WC, and CVAI in men

​	Quartile 1	Quartile 2	Quartile 3	Quartile 4
BMI, n	14,157	13,654	13,329	13,213
Range of BMI, kg/m2	≤22.5	22.6–24.2	24.3–26.1	≥26.2
Unadjusted HR	1.00 (reference)	1.08 (1.00–1.17)	1.30 (1.21–1.40)	1.17 (1.09–1.27)
Multivariable-adjusted HR	1.00 (reference)	1.03 (0.95–1.12)	1.20 (1.11–1.30)	1.14 (1.05–1.24)
Incidence density/person-year	14.1/86,654	15.3/82,973	18.4/80,037	16.6/80,117
Incidence cases, n (%)	1,219 (8.6)	1,270 (9.3)	1,470 (11.0)	1,331 (10.1)
Age of GIM diagnosis	44.9±6.9	45.7±6.8	46.0±6.7	45.4±6.3
WC, n	13,747	13,938	13,384	13,284
Range of WC, cm	≤80.7	80.8–85.5	85.6–90.5	≥90.6
Unadjusted HR	1.00 (reference)	1.21 (1.12–1.31)	1.26 (1.17–1.37)	1.26 (1.16–1.36)
Multivariable-adjusted HR	1.00 (reference)	1.13 (1.04–1.22)	1.12 (1.04–1.22)	1.14 (1.05–1.23)
Incidence density/person-year	13.6/84,348	16.4/84,814	17.2/80,672	17.1/79,946
Incidence cases, n (%)	1,144 (8.3)	1,394 (10.0)	1,388 (10.4)	1,364 (10.3)
Age of GIM diagnosis	44.7±6.8	45.6±6.7	46.0±6.7	45.7±6.6
CVAI, n	13,589	13,588	13,588	13,588
Range of CVAI	≤58.7	58.7–82.5	82.5–106.2	≥106.2
Unadjusted HR	1.00 (reference)	1.30 (1.19–1.41)	1.50 (1.39–1.63)	1.63 (1.51–1.77)
Multivariable-adjusted HR	1.00 (reference)	1.10 (1.02–1.20)	1.15 (1.06–1.24)	1.16 (1.07–1.26)
Incidence density/person-year	11.8/84,035	15.4/83,389	17.8/81,790	19.3/80,566
Incidence cases, n (%)	993 (7.3)	1,287 (9.5)	1,459 (10.7)	1,551 (11.4)
Age of GIM diagnosis	43.5±6.3	45.3±6.5	46.4±6.7	46.8±6.9

Adjusted for age, sex, physical activity, smoking, alcohol intake (g/day), glycemic status, hypertension, marital status, and education level.

HR, hazard ratios; CIs, confidence intervals.

However, women failed to show a significant association between quartile levels of BMI and the risk of GIM (quartile 1: reference, quartile 2: 0.97 [0.82–1.15], quartile 3: 1.05 [0.89–1.24], and quartile 4: 1.09 [0.93–1.29]) (Table 3). In addition, although the quartile 2 group of WC presented an increase in adjusted HR and 95% CI for GIM (1.24 [1.05–1.48]), there was no specific pattern of relationship between quartile levels of WC and the risk for GIM (quartile 1: reference, quartile 2: 1.24 [1.05–1.48], quartile 3: 1.06 [0.90–1.26], and quartile 4: 1.14 [0.96–1.35]). In contrast with BMI and WC, CVAI ≥ quartile 2 had a higher risk for GIM, compared with quartile 1 (quartile 1: reference, quartile 2: 1.31 [1.07–1.60], quartile 3: 1.34 [1.10–1.64], and quartile 4: 1.26 [1.02–1.56]).

Table 3.

HRs and 95% CIs for IM according to the quartile groups of BMI, WC, and CVAI in women

​	Quartile 1	Quartile 2	Quartile 3	Quartile 4
BMI, n	8,340	7,705	7,740	7,848
Range of BMI, kg/m2	≤19.6	19.7–21.1	21.2–23.0	≥23.1
Unadjusted HR	1.00 (reference)	1.16 (0.98–1.38)	1.46 (1.24–1.72)	1.80 (1.54–2.11)
Multivariable-adjusted HR	1.00 (reference)	0.97 (0.82–1.15)	1.05 (0.89–1.24)	1.09 (0.93–1.29)
Incidence density/person-year	5.0/50,225	5.8/46,063	7.3/45,777	9.0/45,655
Incidence cases, n (%)	251 (3.0)	268 (3.5)	335 (4.3)	412 (5.3)
Age of GIM diagnosis	42.4±5.5	44.0±6.1	45.2±6.5	46.6±7.1
WC, n	8,009	7,837	7,940	7,847
Range of WC, cm	≤70.5	70.6–75.0	75.1–80.4	≥80.5
Unadjusted HR	1.00 (reference)	1.43 (1.21–1.70)	1.39 (1.17–1.65)	1.81 (1.54–2.14)
Multivariable-adjusted HR	1.00 (reference)	1.24 (1.05–1.48)	1.06 (0.90–1.26)	1.14 (0.96–1.35)
Incidence density/person-year	4.7/47,573	6.9/46,750	6.7/47,366	8.8/46,030
Incidence cases, n (%)	224 (2.8)	321 (4.1)	318 (4.0)	403 (5.1)
Age of GIM diagnosis	42.5±5.9	43.9±6.0	45.1±6.3	46.4±7.1
CVAI, n	7,909	7,908	7,908	7,908
Range of CVAI	≤9.9	9.9–27.6	27.7–49.7	≥49.7
Unadjusted HR	1.00 (reference)	1.73 (1.42–2.11)	2.30 (1.90–2.80)	3.39 (2.83–4.07)
Multivariable-adjusted HR	1.00 (reference)	1.31 (1.07–1.60)	1.34 (1.10–1.64)	1.26 (1.02–1.56)
Incidence density/person-year	3.2/47,465	5.7/48,291	7.5/47,444	10.9/44,519
Incidence cases, n (%)	151 (1.9)	273 (3.5)	357 (4.5)	485 (6.1)
Age of GIM diagnosis	40.0±4.4	43.2±4.6	45.6±5.3	49.2±7.5

Adjusted for age, sex, physical activity, smoking, alcohol intake (g/day), glycemic status, hypertension, marital status, and education level.

HR, hazard ratios; CIs, confidence intervals.

The time-dependent ROC and AUC analyses for BMI (reference), WC, and CVAI are presented in Table 4. Time-dependent ROC and AUC analyses demonstrated that CVAI had higher AUC values at 2, 4, 6, and 8 years of follow-up, compared with BMI and WC. In particular, AUC values of CVAI are more distinctly increased in women (AUC [95% CI]: 0.642 [0.570–0.715] at 2 years, 0.653 [0.609–0.697] at 4 years, 0.635 [0.595–0.675] at 6 years, and 0.641 [0.596–0.687] at 8 years of follow-up) than men (AUC [95% CI]: 0.570 [0.529–0.611] at 2 years, 0.574 [0.550–0.599] at 4 years, 0.548 [0.527–0.570] at 6 years, and 0.545 [0.519–0.572] at 8 years of follow-up). In ROC graphs (online suppl. Fig. 1, 2), the AUC of WC did not show a statistically significant difference compared with the AUC of BMI. However, the AUC of CVAI was significantly higher than those of BMI and WC in each biannual follow-up period.

Table 4.

Time-dependent AUC with 95% CI for BMI, WC, and CVAI in 2-year, 4-year, 6-year, 8-year follow-up time

​	BMI	WC	CVAI
Men
2 year	0.523 (0.482–0.563)	0.522 (0.481–0.563)	0.570 (0.529–0.611)***
4 year	0.530 (0.506–0.555)	0.534 (0.509–0.559)	0.574 (0.550–0.599)***
6 year	0.512 (0.491–0.533)	0.512 (0.490–0.533)	0.548 (0.527–0.570)***
8 year	0.516 (0.490–0.543)	0.522 (0.495–0.549)	0.545 (0.519–0.572)***
Women
2 year	0.597 (0.522–0.673)	0.568 (0.489–0.647)	0.642 (0.570–0.715)*
4 year	0.601 (0.552–0.649)	0.588 (0.539–0.638)	0.653 (0.609–0.697)***
6 year	0.583 (0.541–0.625)	0.568 (0.526–0.611)	0.635 (0.595–0.675)***
8 year	0.607 (0.559–0.654)	0.581 (0.533–0.628)*	0.641 (0.596–0.687)***

Reference: BMI.

AUC, Area Under the Curve.

*p < 0.05 (p value).

***p < 0.001 (p value).

Online supplementary Tables 1 and 2 present the unadjusted and multivariable-adjusted HRs with 95% CI in BMI, WC, and CVAI dynamic change group. Prior to covariate adjustment, men exhibited proportionally decreased HRs for GIM across the BMI, WC, and CVAI dynamic change groups. However, the multivariable-adjusted HRs for both men and women, as well as the unadjusted HRs for women, showed statistically nonsignificant or marginally significant results (summarized in online suppl. Fig. 3). Online supplementary Table 3 highlights a significant increase in BMI, WC, and CVAI, along with other adverse clinical characteristics such as old age and a high prevalence of hypertension and diabetes in the increased group (group 1). Finally, online supplementary Tables 4 and 5 detail the time-dependent ROC and AUC analyses for BMI (reference), WC, and CVAI in dataset 2 and dataset 3. The analyses of two randomly selected datasets showed a similar pattern with Table 4.

Discussion

In the present study, we longitudinally analyzed the risk of GIM in relation to obesity indices, including BMI, WC, and CVAI, among 85,986 Koreans (54,353 men and 31,633 women) of working age. In the analysis for men, BMI ≥ quartile 3, WC ≥ quartile 2, and CVAI ≥ quartile 2 had a higher risk for GIM, compared with quartile 1 groups. Our analysis indicated that an increase in obesity indices is positively associated with the risk of GIM in men. This result is in line with those of previous studies indicating the adverse effects of obesity on precancerous lesions of gastric cancer among Koreans. Cross-sectional studies in Korean populations have demonstrated that obesity is associated with GIM [11], gastric dysplasia [23, 24], and early gastric cancer [23]. A cohort study from our group showed that a BMI >30 kg/m² was significantly associated with the risk of incident GIM (1.48 [1.20–1.83]) [25]. Meta-analyses also concluded that obesity and excess body weight increased the risk of gastric cancer [7, 26]. These reports support the role of obesity as a risk factor for precancerous lesions in the development of gastric cancer. However, inconsistent results have raised doubts regarding this association. In the aforementioned meta-analyses, stratified analysis demonstrated that the adverse effects of obesity on gastric cancer were present only in non-Asians [7, 26]. A study of US citizens showed that BMI and WC were not associated with the risk of non-cardia GIM among 409 GIM cases and 1,748 controls [27]. Thus, it is still debatable whether obesity increases the risk of gastric cancer across ethnicities. Moreover, studies in the Japanese have suggested that BMI is inversely associated with the risk of atrophic gastritis and GIM [8, 9]. These inconsistent results require further evidence to discover more precise and specific surrogate markers of obesity for predicting precancerous lesions of gastric cancer.

Our study provides epidemiological evidence that fulfills this requirement. In the present study, time-dependent ROC and AUC analyses showed that CVAI had higher AUC values at 2, 4, 6, and 8 years of follow-up in both men and women, compared with BMI and WC. This finding suggests that CVAI is superior to BMI and WC for predicting GIM. A plausible explanation for this finding is the superiority of CVAI to assess VAT and its associated morbidity [13–16]. In particular, increased VAT has been suggested to be involved in the pathophysiology of GIM. VAT acts as an active endocrine organ that secretes various cytokines including adiponectin and leptin. Increased VAT is associated with dysregulated adipokine secretion, including altered leptin signaling [28, 29] and reduced adiponectin [30, 31]. Leptin potently affects the development of GIM and gastric cancer cell proliferation via activation of leptin signaling pathway [28, 29]. Laboratory evidence suggests that adiponectin can exert inhibitory effects on the growth and proliferation of gastric cancer cells [32]. The inhibitory role of adiponectin on gastric cancer is supported by a result that low plasma adiponectin levels were associated with an increased risk of gastric cancer [33]. Moreover, it is presumed that the VAT has a more powerful function in Asians. Asians have 3–5 percent higher total body fat than white Europeans, even with the same BMI [34]. In particular, Asians tend to have relatively greater levels of VAT than other ethnic groups, which predispose Asians to cardiometabolic diseases even at the same BMI [35, 36]. Thus, precise assessment of VAT by CVAI may allow for a greater predictive ability for GIM.

The predictive ability of CVAI for GIM is more evident in women than in men. In our analysis of women, while BMI and WC had no significant relationship with GIM, CVAI was associated solely with the risk of GIM. Time-dependent ROC and AUC analyses showed that women had a greater AUC value at 2, 4, 6, and 8 years of follow-up than men. These findings indicate that CVAI is a more useful marker of the risk of GIM in women than in men. The predominant predictability of CVAI in women has been similarly observed in previous studies on Asians [16, 37, 38]. This feature of CVAI may be derived from sex differences in fat deposition. Asian women are characterized by greater abdominal fat deposition [39], which can lead to a higher risk of diseases related to VAT in women [16, 37, 38]. In particular, it is presumed that the effect of VAT on GIM is greater in women than in men. Compared with women, men tend to have more unfavorable health behaviors such as smoking, alcohol intake, and dietary habits. Female gender is more resistant to gastric cancer and precancerous lesions than are men [40]. Studies have demonstrated that estrogen has a protective function against gastric cancer and precancerous lesions, including GIM [41, 42]. The interaction of behavioral and hormonal factors can contribute to a lower risk of incident GIM in women. Thus, VAT has a relatively greater impact on GIM in women with fewer risk factors and higher protective properties for GIM. In contrast, men are exposed to more risk factors for GIM, which may surpass the adverse effects of VAT on GIM. Considering the lower risk of GIM in women, our results indicate the potential usefulness of CVAI in screening high-risk groups for GIM in women.

Our study presents that CVAI is a stronger predictive ability for GIM than other adiposity-related indices. Nevertheless, it should be noted that the AUC values of CVAI for predicting GIM remained in the modest range in both men and women. These findings suggest that the role of CVAI may be limited to a supportive component rather than serving as a primary screening tool for GIM. Previous studies have demonstrated that noninvasive serologic biomarkers, such as the serum pepsinogen I/II ratio and gastrin levels, can be used to stratify the risk of atrophic gastritis and GIM [43, 44]. Although these biomarkers were not available in the present study, combining serologic markers, reflecting gastric mucosal status with CVAI may improve the precision of GIM risk prediction by integrating systemic metabolic burden with local gastric mucosal changes. Future studies incorporating both visceral adiposity-based indices and noninvasive gastric biomarkers are warranted.

The present study had several limitations. First, GIM was defined solely based on endoscopic findings, without histologic confirmation. Although histopathology is considered the gold standard, systematic biopsy of endoscopically suspected GIM is not routinely performed during screening endoscopy in Korea. Consequently, information on histological subtype, extent, and operative link on GIM staging was unavailable. MAPS III guidelines indicated that these histological characteristics have important prognostic implications [45]. Thus, the absence of histological data may have limited more refined risk stratification and influenced the magnitude of observed associations. Second, a detailed evaluation of the various factors that may influence the development of GIM was challenging. In particular, socioeconomic status, dietary factors, and family history of gastric cancer are known to affect the risk of GIM. However, information on these variables was collected as optional items in the health checkup questionnaires, resulting in substantial missing data. Therefore, we acknowledge that adjustment for these potential confounding factors was limited. Third, information on H. pylori infection was not available in the present study. This represents an important limitation, as H. pylori infection is the most dominant and well-established risk factor for GIM. However, the Korean endoscopic guidelines recommend H. pylori examination only in specific cases, not including GIM [23]. Consequently, we were unable to account for H. pylori infection in our multivariable analyses. Future studies incorporating H. pylori infection status may allow a more precise assessment of the independent predictive value of CVAI beyond this major etiologic factor.

In conclusion, our analysis showed that CVAI was positively associated with the risk of GIM in relatively young and apparently healthy Koreans. CVAI was superior to BMI and WC in predicting GIM. This feature of CVAI may be attributable to its accuracy in assessing VAT. Further studies are needed to prove whether the increased risk of GIM predicted by CVAI is linked to the increased risk of gastric cancer.

Acknowledgments

This study was based on medical data collected and arranged by Kangbuk Samsung Cohort Study (KSCS). Therefore, this study could be done by virtue of the labor of all staff working in KSCS and Total Healthcare Center, Kangbuk Samsung Hospital.

Statement of Ethics

This study was approved by the Institutional Review Board (IRB) of Kangbuk Samsung Hospital (IRB no. KBSMC 2022-08-041). The requirement for informed consent was waived by the Ethics Committee because of the retrospective nature of the study.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This study was not supported by any sponsor or funder.

Author Contributions

Conception, design, and drafting of the article: Sung Keun Park, Yeongu Chung, and Ju Young Jung. Data analysis and interpretation: Sung Keun Park, Yeongu Chung, Ju Young Jung, and Chang-Mo Oh. Critical revision of the article for important intellectual content: Sung Keun Park, Yeongu Chung, Ju Young Jung, and Chang-Mo Oh. Final approval of the article: all authors.

Funding Statement

This study was not supported by any sponsor or funder.

Data Availability Statement

The data that support the findings of this study are not publicly available due to restricted data which were used under license for the current study but are available from the corresponding author upon reasonable request and with permission from the Kangbuk Samsung Cohort Study.