Modified scoring model incorporating waist-hip ratio for predicting advanced colorectal neoplasia

Abstract

BACKGROUND

Designing a feasible risk prediction model for advanced colorectal neoplasia (ACN) can enhance colonoscopy screening efficiency. Abdominal obesity is associated with colorectal cancer development.

AIM

To propose and evaluate a modified scoring model incorporating waist-hip ratio for the prediction of ACN.

METHODS

A total of 6483 patients who underwent their first screening or diagnostic colonoscopy in our center between 2020 and 2023 were recruited, in which 4592 were in the derivation cohort and 1891 formed a validation cohort. Multivariate logistic regression was used to investigate the risk factors of ACN in the derivation cohort based on endoscopic findings, and a new scoring model for ACN prediction was developed. The discriminatory capability of the scoring model was validated by the validation cohort.

RESULTS

Age, male gender, smoking, and wait-to-hip ratio were identified as independent risk factors for ACN, and a 7-point scoring model was developed. The prevalence of ACN was 3.3%, 9.3% and 18.5% in participants with scores of 0-2 [low risk (LR)], 3–4 [moderate risk (MR)], and 5–7 [high risk (HR)], respectively, in the derivation cohort. With the scoring model, 49.9%, 38.4%, and 11.7% of patients in the validation cohort were categorized as LR, MR, and HR, respectively. The corresponding prevalence rates of ACN were 5.0%, 10.3%, and 17.6%, respectively. The C-statistic of the new scoring model was 0.66, which was higher than that of the Asia-Pacific Colorectal Screening model (0.63).

CONCLUSION

A modified scoring model incorporating waist-hip ratio has an improved predictive performance in the prediction of ACN.

Key Words: Advanced colorectal neoplasia; Abdominal obesity; Colorectal cancer screening; Risk scoring model; Metabolic syndrome

Core Tip: Designing a feasible risk prediction model for advanced colorectal neoplasia (ACN) can greatly improve the efficiency of colonoscopy screening. Abdominal obesity is associated with development of colorectal cancer. We proposed and evaluated a modified scoring model incorporating waist-hip ratio for the prediction of ACN in a Chinese population. Results showed that the modified scoring model had an improved predictive performance than the Asia-Pacific Colorectal Screening model in the prediction of ACN.

INTRODUCTION

Colorectal cancer (CRC) ranks as the third most prevalent cancer diagnosis globally and represents the second primary contributor to cancer-associated fatalities[1]. It is estimated that 550000 new CRC cases and 290000 deaths occurred in China in 2020, making it the second most common cancer and the fifth leading cause of cancer-related mortality[2]. The neoplastic progression primarily originates from benign colorectal adenomatous polyps through the adenoma-carcinoma progression pathway, with advanced adenomas characterized by diameters exceeding 10 mm, presence of high-grade dysplastic changes, or villous architectural components[3]. Current preventive strategies emphasize timely identification and excision of colorectal adenomatous lesions, particularly advanced colorectal adenomas (CRAs). Clinical investigations demonstrate 53%-72% reduction in CRC incidence rates coupled with notable mortality decline following systematic endoscopic screening and polyp removal interventions[4]. While international protocols recommend initial screening colonoscopy for individuals beyond 50 years, this invasive diagnostic modality demands significant healthcare resources and carries potential procedural risks, contributing to limited global implementation-particularly within developing nations. Implementing risk stratification strategies for screening candidates could optimize resource allocation by identifying high-risk individuals requiring prioritized endoscopic evaluation.

The Asia-Pacific Consensus Recommendations[5] advocate a risk-stratified methodology utilizing the Asia-Pacific Colorectal Screening (APCS) criteria[6]. This scoring framework evaluates variables including age, gender, smoking history, and familial CRC background as predictors for advanced colorectal neoplasia (ACN). A revised APCS scoring model developed in Hong Kong integrates body mass index (BMI) as an additional parameter, demonstrating reasonable accuracy in identifying individuals at risk for ACN[7,8]. Emerging evidence highlights abdominal obesity metrics-particularly waist circumference and waist-to-hip ratio-as superior predictive indicators for oncological risk compared to generalized obesity measures. The pathophysiological mechanism involves metabolically active visceral adipose tissue (VAT), which secretes numerous bioactive adipocytokines implicated in carcinogenesis[9-11].

In the present study, we developed a modified scoring model that incorporates waist-hip ratio as a parameter for predicting ACN in a Chinese population. The discriminatory performance of the new scoring system was evaluated in a validation cohort and compared with that of the APCS[6] and the BMI-modified APCS model[8].

MATERIALS AND METHODS

Study design and population

Patients aged ≥ 18 years old who underwent first-time colonoscopy for standard indications or screening at the University of Hong Kong-Shenzhen Hospital between July 1, 2020 and Decmber 31, 2023 were recruited. Patients with a history of colectomy, poor bowel preparation, incomplete colonoscopy, inflammatory bowel disease, or other chronic bowel inflammation and those who refused to participate in the study were excluded. The study inclusion and exclusion criteria are shown in Figure 1.

Figure 1  Flowchart illustrating the inclusion and exclusion criteria for study participants.

Measurements and definitions

Research assistants conducted in-person interviews to gather comprehensive patient information encompassing medical histories, coexisting health conditions, substance use patterns, hereditary cancer risks, prescribed medications, exercise routines, and symptomatic presentations. Body indices, including weight, height, waist circumference, and hip circumference, were measured and recorded.

Current smoking was classified as consuming a minimum of one cigarette daily within the past year[12,13]. Alcohol intake exceeding 140 grams weekly was categorized as regular consumption[12,13]. Physical activity involving cycling, swimming, or brisk walking for over 30 minutes daily was considered regular exercise when performed three or more times weekly[12,13]. Aspirin utilization was identified as daily intake of aspirin for a minimum of three consecutive months within the past year[12,13]. A positive family history of CRC was determined by having one or more first-degree relatives diagnosed with the condition, regardless of age[12,13].

Height assessment was conducted with a wall-mounted stadiometer, while body mass quantification employed a hospital-calibrated digital weighing instrument. Anthropometric measurements included waist circumference determination at the mid-point between the lowest rib margin and iliac crest using non-elastic measuring tape, with subjects maintaining erect posture. Concurrently, hip circumference evaluation was performed at the maximal gluteal protrusion inferior to the iliac crest, maintaining horizontal tape alignment during upright stance[12,13].

Colonoscopy and histopathological examination

Colonoscopy procedures were conducted either directly by qualified endoscopists or under their professional guidance. In alignment with established colonoscopy quality guidelines, practitioners maintained a minimum withdrawal time exceeding six minutes during examinations. All identified colonic polyps underwent removal procedures compliant with internationally accepted surgical protocols. Excised tissue samples received initial pathological analysis from a certified specialist, followed by comprehensive reassessment conducted by a senior pathologist.

Derivation and validation cohorts

The derivation cohort enrolled patients who underwent first-time colonoscopy at our center between July 1, 2020 and December 31, 2022. Participants in the derivation cohort were divided into ACN and non-ACN groups based on their colonoscopy findings. Whereas validation cohort enrolled patients who underwent first-time colonoscopy at our center between January 1, 2023 and December 31, 2023.

Development and validation of the risk scoring model for the prediction of ACN

In the derivation cohort, associations between ACN and clinical parameters were examined through Pearson's χ² analysis. Evaluated variables encompassed demographic characteristics (age, gender), lifestyle factors (regular exercise, smoking, alcohol consumption), medical history components [family history of CRC, hypertension, diabetes, hyperlipidemia, non-steroidal anti-inflammatory drug (NSAID) usage, cholecystectomy], clinical symptoms (abdominal discomfort, altered bowel patterns, rectal bleeding), and anthropometric measurements [BMI, wait-to-hip ratio (WHR), and waist circumference]. For identification of independent predictors, multivariable logistic regression was employed on variables showing preliminary associations (P < 0.05) in univariate screening. All analyses utilized two-tailed tests with statistical significance threshold set at α = 0.05. The predictive scoring system incorporated regression coefficients derived from multivariable analysis, where each risk factor's weight was calculated by doubling the β-coefficient before rounding to whole numbers. Individual risk scores represented cumulative sums of weighted component values.

Within the validation cohort, subjects were stratified into three distinct risk categories according to their assessment scores: (1) Low risk (LR); (2) Moderate risk (MR); and (3) High risk (HR). The occurrence rates of ACN across these stratified groups were systematically analyzed. Model calibration was assessed through the Hosmer-Lemeshow goodness-of-fit test, with P-values exceeding 0.05 suggesting adequate calibration between predicted probabilities and actual outcomes. Comparative analysis of predictive performance was conducted using C-statistics and receiver operating characteristic curves, benchmarking the novel scoring system against both the conventional APCS framework and its modified iteration. All comparative evaluations were replicated within the validation cohort to ensure methodological consistency.

Statistical analysis

Statistical Package for Social Science (version 23.0; Chicago, IL, United States) was used for data analysis. The relationships between ACN and the clinical factors of the derivation cohort were examined by univariate analyses using the χ² test. A BMI ≥ 23 kg/m² was chosen as the cut-off for BMI, based on the World Health Organization (WHO) definition of being overweight for Asian populations. All variables with initial P values < 0.05 in the univariate analysis were analyzed in a binary logistic regression model to identify independent risk factors for ACN. Two-sided P values < 0.05 were considered statistically significant.

Ethical considerations

Written consents were obtained from the participants for research data collection and study purposes with identities kept confidential. The Ethics Committee of the University of Hong Kong-Shenzhen Hospital approved this study in July 2019 (No. [2019]228).

RESULTS

Participant characteristics

The derivation cohort enrolled 4722 patients who underwent first-time colonoscopy in our center between July 2020 and December 2022. Patients with past colorectal surgery (n = 56), incomplete colonoscopy (n = 25), poor bowel preparation (n = 8), lost or incomplete data (n = 34), inflammatory bowel disease, or other chronic bowel inflammation (n = 7) were excluded. Finally, 4592 patients were included in the derivation cohort. The validation cohort enrolled 1958 patients who underwent first-time colonoscopy in our center between January 2023 and December 2023. Patients with past colorectal surgery (n = 45), incomplete colonoscopy (n = 11), poor bowel preparation (n = 3), lost or incomplete data (n = 5), inflammatory bowel disease, or other chronic bowel inflammation (n = 3) were excluded. Finally, 1891 patients were included in the validation cohort.

Compared with the derivation cohort, the validation cohort was older (47.1 years ± 12.4 years vs 45.8 years ± 12.4 years), and the waist circumference, WHR, and prevalence of hypertension and hyperlipemia were higher in the validation cohort, probably due to the older age of the validation cohort (P < 0.05) (Table 1). Mean value of BMI were similar in the derivation and validation cohorts (23.4 kg/m² ± 3.3 kg/m²vs 23.3 kg/m² ± 3.3 kg/m²). The incidences of CRA (34.3% vs 32.0%), ACN (8.5% vs 8.1%), and CRC (1.1% vs 1.3%) were not significantly different between the two cohorts (P > 0.05) (Table 1).

Table 1 Characteristics of the derivation and validation cohorts, n (%).

	Derivation cohort (n = 4592)	Validation cohort (n = 1891)	P value
Age (years), mean ± SD	45.8 ± 12.4	47.1 ± 12.4	0.000
Sex, male	2458 (53.5)	1035 (54.7)	0.381
Regular excise	2090 (45.6)	670 (35.4)	0.000
Smoking	809 (17.6)	378 (20.0)	0.029
Alcohol consumption	606 (13.2)	352 (18.6)	0.000
Use of non-steroidal anti-inflammatory drugs	105 (2.3)	89 (4.7)	0.000
Hypertension	542 (11.8)	276 (14.6)	0.003
Diabetes mellitus	222 (4.9)	104 (5.5)	0.289
Hyperlipemia	410 (8.9)	266 (14.1)	0.000
Colorectal cancer family history	347 (7.6)	136 (7.2)	0.640
Post-cholecystectomy	130 (2.9)	34 (1.8)	0.015
Rectal bleeding	199 (4.3)	90 (4.8)	0.499
Change of bowel habit	275 (6.0)	120 (6.3)	0.585
Abdominal pain	731 (15.9)	354 (18.7)	0.105
Body mess index (kg/m2), mean ± SD	23.3 ± 3.3	23.4 ± 3.3	0.264
Waist circumference (cm), mean ± SD	82.5 ± 10.3	83.4 ± 10.8	0.003
Waist-hip ratio, mean ± SD	0.885 ± 0.79	0.891 ± 0.89	0.014
Normal/diverticulum/colitis/neuroendocrine neoplasms/hyperplastic polyps	3124 (68.0)	1241 (65.6)	0.061
Advanced colorectal neoplasia	372 (8.1)	161 (8.5)	0.582
Cancer	58 (1.3)	20 (1.1)	0.490

Independent risk factors of ACN in the derivation cohort

In the univariate analysis, male gender, older age, smoking, regular intake of NSAIDs, hypertension, diabetes mellitus, hyperlipidemia, BMI > 23 kg/m², large waist circumference, and high WHR were associated with a higher ACN prevalence, while alcohol consumption, family history of CRC, excise, change of bowel habit, abdominal pain and post-cholecystectomy were not associated with the prevalence of ACN (P > 0.05) (Table 2). Multivariate regression analysis revealed that male gender, age, smoking, and increased WHR were independent risk factors for the prevalence of ACN (P < 0.05). In the multivariate regression analysis, WHR, but not BMI, was independently associated with the prevalence of ACN. Compared with females, males had an adjusted odds ratio (AOR) of ACN incidence of 1.49 (95%CI: 1.14–1.93). Compared with participants aged < 40 years, participants aged 40–49 years, 50–59 years, 60–69 years, and ≥ 70 years had AORs of ACN incidence of 3.11 (95%CI: 2.17–4.44), 3.93 (95%CI: 2.71–5.71), 5.56 (95%CI: 3.70–8.35), and 6.10 (95%CI: 3.64–10.2), respectively. Smokers had an AOR of 1.50 (95%CI: 1.14–1.98) for ACN compared with non-smokers, and participants with high WHR had an AOR of 1.46 (95%CI: 1.11–1.92) for ACN development, compared with participants with normal WHR.

Table 2 Relationships between clinical features and presence of advanced colorectal neoplasia in derivation cohort, n (%).

Clinical feature	Total	Advanced colorectal neoplasia
		Univariate analysis				Multivariate analysis
		Present (n = 372)	Absent (n = 4220)	OR (95%CI)	P value	OR (95%CI)	P value
Sex
Female	2134	134 (6.3)	2000 (93.7)	1 (Reference)		1 (Reference)
Male	2458	238 (9.7)	2220 (90.3)	1.54 (1.26-1.89)	0.000	1.49 (1.14-1.93)	0.003
Age (years)
< 40	1636	46 (2.8)	1590 (97.2)	1 (Reference)		1 (Reference)
40–49	1315	111 (8.4)	1204 (91.6)	3.00 (2.15-4.20)		3.11 (2.17-4.44)
50–59	934	101 (10.8)	833 (89.2)	3.85 (2.74-5.41)		3.93 (2.71-5.71)
60–69	519	81 (15.6)	438 (84.4)	5.56 (3.92-7.87)		5.56 (3.70-8.35)
≥ 70	188	33 (17.6)	155 (84.4)	6.25 (4.10-9.52)	0.000	6.10 (3.64-10.2)	0.000
Regular exercise
No	2491	193 (7.7)	2298 (92.3)	1 (Reference)
Yes	2091	177 (8.5)	1914 (91.5)	1.09 (0.90-1.33)	0.373
Smoking
No	3783	275 (7.3)	3508 (92.7)	1 (Reference)		1 (Reference)
Yes	809	97 (12.0)	712 (88.0)	1.65 (1.33-2.05)	0.000	1.50 (1.14-1.98)	0.004
Alcohol drinking
No	3986	314 (7.9)	3673 (92.1)	1 (Reference)
Current or past drinker	606	58 (9.6)	548 (90.4)	1.21 (0.93-1.59)	0.158
Regular intake of non-steroidal anti-inflammatory drugs
No	4437	350 (7.9)	4087 (92.1)	1 (Reference)		1 (Reference)
Yes	105	18 (17.0)	87 (83.0)	2.16 (1.340-3.32)	0.001	1.11 (0.64-1.93)	0.707
Hypertension
No	4050	296 (7.3)	3754 (92.7)	1 (Reference)		1 (Reference)
Yes	542	76 (14.0)	466 (86.0)	1.92 (1.51-2.43)	0.000	1.13 (0.83-1.54)	0.423
Diabetes mellitus
No	4370	342 (7.8)	4028 (92.2)	1 (Reference)		1 (Reference)
Yes	222	30 (13.5)	192 (86.5)	1.72 (1.21-2.43)	0.003	0.92 (0.60-1.42)	0.718
Hyperlipemia
No	4182	324 (7.7)	3858 (92.3)	1 (Reference)		1 (Reference)
Yes	410	48 (11.7)	362 (88.3)	1.51 (1.13-2.01)	0.005	1.03 (0.74-1.45)	0.848
Colorectal cancer family history
No	4239	344 (8.1)	3895 (91.9)	1 (Reference)
Yes	347	28 (8.1)	319 (91.9)	1.01 (0.69-1.44)	0.976
Post cholecystectomy
No	4462	357 (8.0)	4105 (92.0)	1 (Reference)
Yes	130	15 (11.5)	115 (88.5)	1.43 (0.88-2.33)	0.154
Rectal bleeding
No	4393	349 (7.9)	4044 (92.1)	1 (Reference)		1 (Reference)
Yes	199	23 (11.6)	176 (88.4)	1.51 (0.97-2.37)	0.068	1.55 (0.97-2.45)	0.064
Change of bowel habit
No	4317	348 (8.1)	3969 (91.9)	1 (Reference)
Yes	275	24 (8.7)	251 (91.3)	1.09 (0.71-1.68)	0.695
Abdominal pain
No	3861	320 (8.3)	3541 (91.7)	1 (Reference)
Yes	731	52 (7.1)	679 (92.9)	0.85 (0.63-1.15)	0.286
Body mass index (kg/m²)
≤ 23	1961	121 (6.2)	1840 (93.8)	1 (Reference)		1 (Reference)
> 23	2621	249 (9.5)	2372 (90.5)	1.54 (1.25-1.90)	0.000	1.10 (0.84-1.44)	0.500
Waist circumference (cm)
Normal	2865	203 (7.1)	2662 (92.9)	1 (Reference)		1 (Reference)
Higher (> 90, males; > 80, females)	1727	169 (9.8)	1558 (90.2)	1.38 (1.14-1.68)	0.001	0.90 (0.68-1.20)	0.487
Waist-hip ratio
Normal	2957	190 (6.4)	2767 (93.6)	1 (Reference)		1 (Reference)
Higher (> 0.9, males; > 0.85, females)	1635	182 (11.1)	1453 (88.9)	1.73 (1.43-2.11)	0.000	1.46 (1.11-1.92)	0.006

OR: Odds ratio.

Development of the risk score

An 7-point scoring model for the prediction of ACN was developed based on the beta coefficient for the four independent risk factors identified in the multivariate analysis, including male sex (score: 1), age < 40 years (score: 0), age of 40–49 years (score: 2), age of 50–59 years (score: 3), age of 60–69 years (score: 3), age > 70 years (score: 4), smoking (score: 1), and high WHR (score: 1) (Table 3). The prevalence of ACN increased from 2.5% to 26.3% in the order of calculated risk scores from 0 to 7 in the derivation cohort (Table 4). A score of 2 indicated a significantly lower prevalence of ACN than the overall prevalence (4.6% vs 8.1%), and a score of 5 indicated a significantly higher prevalence of ACN than the overall prevalence (17.4% vs 8.1%) in the derivation cohort. Therefore, scores ≤ 2 were assigned as LR, scores 5 were assigned as HR, and a score of 3 or 4 had a prevalence of ACN close to the overall prevalence (9.3% vs 8.1%) and was therefore assigned as MR. The prevalence of ACN in the derivation cohort was 3.3%, 9.3 %, and 18.5% in the LR, MR, and HR groups, respectively (Table 4).

Table 3 Development of a scoring model to predict advanced colorectal neoplasia.

Clinical features	Odds ratio (95%CI)	Beta coefficient	Risk score	P value
Sex
Female	1 (Reference)	0	0
Male	1.49 (1.14-1.93)	0.396	1	0.003
Age (years)
< 40	1 (Reference)	0	0
40–49	3.11 (2.17-4.44)	1.134	2
50–59	3.93 (2.71-5.71)	1.370	3
60–69	5.56 (3.70-8.35)	1.715	3
≥ 70	6.10 (3.64-10.2)	1.807	4	0.000
Smoking
No	1 (Reference)	0	0
Yes	1.50 (1.14-1.98)	0.406	1	0.004
Waist-hip ratio
Normal	1 (Reference)	0	0
Higher (> 0.9, males; > 0.85, females)	1.46 (1.11-1.92)	0.378	1	0.006

Table 4 Prevalence of advanced colorectal neoplasia according to scores and subgroups in the derivation cohort.

Score	Proportion of individuals with ACN	95%CI	Subgroups	Proportion of individuals with ACN	95%CI
0	2.5% (13/530)	1.3–4.2	Low-risk group (0–2)	3.3% (63/1938)	2.4-4.0
1	2.6% (19/730)	1.6–4.0
2	4.6% (31/678)	3.1–6.4
3	7.3% (69/944)	5.7–9.2	Moderate-risk group (3–4)	9.3% (185/1983)	8.3-10.9
4	11.2% (116/1039)	9.3–13.2
5	17.4% (84/482)	14.1–21.1	High-risk group (5–7)	18.5% (124/671)	15.6-21.6
6	20.6% (35/170)	14.8–27.5
7	26.3% (5/19)	9.1–51.2
Total	8.1% (372/4592)	7.3–8.9

ACN: Advanced colorectal neoplasia.

Validity of the newly developed score model and comparison with APCS and BMI-modified APCS model

In the validation cohort, 49.9%, 38.4 %, and 17.6% of patients were classified as LR, MR, and HR, respectively. The prevalences of ACN in the LR, MR, and HR groups of the validation cohort were 5.1%, 10.4%, and 16.2%, respectively. Compared with the LR group, the MR and HR groups had higher risks of ACN (AOR = 2.03 and AOR = 3.16, respectively) (Table 5). The P value of the Hosmer-Lemeshow goodness-of-fit statistic assessing the reliability of the validation set was 0.71, which implied a close match between the predicted and real risks. When the APCS[6] and modified APCS models[8] were applied to predict the risk of ACN in the validation cohort, the C-statistics of the two models were 0.63 and 0.62, respectively, which were both lower than the C-statistic of the newly developed model in this study (C-statistic = 0.66) (Table 6).

Table 5 Prevalence of advanced colorectal neoplasia by risk tier, n (%).

Risk tier (risk score)	Derivation cohort		Validation cohort			Validation cohort (male)			Validation cohort (female)
	Number of subjects	ACN	Number of subjects	ACN	Relative risk (95%CI)	Number of subjects	ACN	Relative risk (95%CI)	Number of subjects	Advanced colorectal neoplasia	Relative risk (95%CI)
Low risk (0–2)	1938 (42.2)	63 (3.3)	943 (49.9)	47 (5.0)	1 (Reference)	356 (34.4)	16 (4.5)	1 (Reference)	587 (68.6)	31 (5.3)	1 (Reference)
Moderate risk (3–4)	1983 (43.2)	185 (9.3)	727 (38.4)	75 (10.3)	2.07 (1.46-2.94)	489 (47.2)	42 (8.6)	1.91 (1.09-3.34)	238 (27.8)	33 (13.9)	2.62 (1.65-4.18)
High risk (5–7)	671 (14.6)	124 (18.5)	221 (11.7)	39 (17.6)	3.55 (2.38-5.26)	190 (18.4)	32 (16.8)	3.75 (2.11-6.67)	31 (3.6)	7 (22.6)	4.27 (2.05-8.93)
Total	4592	372 (8.1)	1891	161 (8.5)		1035 (54.7)	90 (8.7)		856 (45.3)	71 (8.3)

ACN: Advanced colorectal neoplasia.

Table 6 Validity of the newly developed score model and comparison with Asia-Pacific Colorectal Screening model and modified Asia-Pacific Colorectal Screening model, n (%).

Validation cohort (risk score model of this study) (age, male, smoking, and high wait-to-hip ratio)			Validation cohort (according to APCS model) (age, male, smoking, family history of CRC)			Validation cohort (according to modified APCS model) (age, male, smoking, family history of CRC and body mass index )
Risk tier (risk score)	ACN	Relative risk (95%CI)	Risk tier (risk score)	ACN	Relative risk (95%CI)	Risk tier (risk score)	Advanced colorectal neoplasia	Relative risk (95%CI)
LR (0-2)	47 (5.0)	1	LR (0-1)	46 (5.1)	1	LR (0)	15 (4.9)	1
MR (3-4)	75 (10.3)	2.07 (1.46-2.94)	MR (2-3)	81 (10.4)	2.03 (1.48-2.87)	MR (1–2)	86 (7.7)	1.58 (0.93-2.70)
HR (5-7)	39 (17.6)	3.55 (2.38-5.26)	HR (4-7)	34 (16.2)	3.16 (2.09-4.81)	HR (3–6)	60 (12.9)	2.65 (1.53-4.57)
Hosmer-lemeshow goodness-of-fit statistic	0.71		Hosmer-lemeshow goodness-of-fit statistic	0.59		Hosmer-lemeshow goodness-of-fit statistic	0.27
C-statistic	0.66 ± 0.02		C-statistic	0.63 ± 0.02		C-statistic	0.62 ± 0.02

APCS: Asia-Pacific Colorectal Screening model; CRC: Colorectal cancer; HR: High risk; LR: Low risk; MR: Moderate risk.

DISCUSSION

This single-center cross-sectional study clarified the prevalence and risk factors of ACN by examining a Chinese population and building a scoring model to predict the risk of ACN. Male sex, age, smoking, and WHR were found to be independent risk factors for ACN and were incorporated as parameters in the new scoring model for predicting ACN. A 7-point scoring model was built, in which male gender, smoking, and high WHR were all weighted as 1 point, while age had the largest weight, reaching 4 points when the patient was 70 years old. This model indicates that age plays the most important role in ACN development. Male gender and smoking are two common risk factors for most human malignancies and associated with the development of ACN and have been included as parameters in many risk models for ACN, such as the APCS and BMI-modified APCS models. Alcohol consumption is a suspected risk factor for CRC. However, it was not associated with ACN in this study, which is consistent with the results of other previous studies conducted in Asian countries, indicating that the association between alcohol consumption and the development of CRC requires further study[7,8]. The discriminative ability of the new risk model was assessed in the validation cohort and compared with that of the APCS and BMI-modified APCS models. The Hosmer-Lemeshow goodness-of-fit and C-statistics of the new model were 0.71 and 0.66, respectively, which were better than those of the APCS model (0.59 and 0.63, respectively) and the BMI-modified APCS model (0.27 and 0.62, respectively), indicating that the new model has a better predictive value for ACN risk.

The WHO defines overweight or obesity as excessive fat accumulation in the body, which may have negative effects on general health[13]. Obesity is associated with cancer incidence and even mortality and is a risk factor for many different types of cancer, including CRC[14,15]. BMI is the most commonly used indicator of obesity and has been shown to be a risk factor for CRC[16]. Several risk scoring models of ACN had included BMI as a scoring parameter. There are two types of obesity based on the distribution of body fat: (1) Peripheral; and (2) Central. Peripheral obesity refers to excessive adipose tissue that accumulates mainly in the trunk, whereas central obesity (also named abdominal obesity) refers to excessive adipose tissue that accumulates in the abdominal cavity, including the adipose tissue of the omentum or mesentery[17,18]. The carcinogenic effects of obesity are suspected to be associated with adipose tissue, particularly VAT. Visceral fat is an endocrine organ that produces a variety of proteins, hormones, and cytokines that are referred to as adipokines, including adiponectin, leptin, interleukin-6, and tumor necrosis factor alpha[19]. The underlying mechanism of the positive association between central obesity and CRA may be an imbalance in the adipokines produced by visceral fat. An imbalance in adipokines contributes to carcinogenesis and promotes the development of various kinds of tumors[20]. The WHR is a simple clinical indicator of abdominal obesity. Some studies have demonstrated that WHR may be a better indicator than BMI for predicting CRC risk[9-11]. In this study, we investigated the link between BMI, WHR, metabolic syndrome (including hypertension, diabetes, and hyperlipidemia), and the incidence of ACN. Multivariate analysis revealed that WHR, but not BMI, was an independent risk factor for ACN.

Approximately 35% of patients with CRC have a family history of CRC, and the underlying mechanisms may be genetic factors, co-exposure, or a combination of both[21]. The family history of CRC is generally considered as a HR factor for the development of CRC. A study reported that individuals with two or more false discovery rates (FDRs) with CRC or individuals with one FDR diagnosed with CRC before the age of 50 years had a significantly increased risk of CRC, in whom the relative risk of developing CRC was almost three times that in the general population[22]. However, the same study suggested that only 5%–7% of patients with CRC are affected by a clearly defined inherited CRC syndrome[23]. In the current study, the family history of CRC was not significantly different between the ACN and non-ACN groups on univariate analysis in the derivation cohort (P = 0.98). The underlying mechanism remains unclear. One possible reason may be that the family history provided by the participants was inaccurate. Health awareness and medical services in last century in China was not well developed and therefore the past medical history of our previous generations were not accurately documented. This situation is common in China, suggesting that attention should be paid to inaccuracy and lack of family history data when assessing the risk of CRC in Chinese individuals. However, the accuracy of the family histories of Chinese people will improve with the economic and medical care being developed. In this study, the scoring model we developed did not involve family history data but still had a good predictive value, as evaluated in the validation cohort.

The participants in this study were younger than those in most previous studies in the world because only people aged > 50 years can participate in screening programs in most countries. However, with the plenty supply of medical resources, colonoscopy for screening purposes can be performed on demand in China, with self-financed manner. Many young people request a screening colonoscopy because of concerns about CRC, regardless of whether they have symptoms. The results showed that 2.8% of people aged < 40 years were diagnosed with ACN, and 8.4% of people aged 40–49 years had ACN, which was close to the detection rate in people aged 50–59 years (10.8%). Therefore, people aged 40-50 years should not ignore the risk of ACN. The beginning age for colonoscopy screening in the current Chinese guidelines is 40 years[24].

Most predictive risk score models for ACN are based on large-scale colonoscopy screening programs. Currently, there is no standard colonoscopy screening program in China. The subjects of this study were patients who visited the hospital for first-time colonoscopy, and some had symptoms such as rectal bleeding, abdominal pain, or changes in bowel habits. However, the statistical analysis showed that these symptoms were not directly related to the ACN detection rate, similar to the findings of a previous study in Australia[25]. This may because that these symptoms are non-specific. For example, rectal bleeding may be related to anal diseases, such as hemorrhoids, anal fissures, and enteritis, especially in young people. Changes in abdominal pain or bowel habits may be related to acute and chronic enteritis, intestinal dysfunction, or dietary changes. The proposed stratification tool shows potential for implementation in community-based screening programs and gastroenterology referrals across China. By prioritizing high-risk cohorts for endoscopic evaluation, this approach could facilitate timely detection and management of ACN, ultimately improving patient prognoses through early therapeutic interventions.

While the current risk assessment framework demonstrates notable progress in evaluating ACN susceptibility, certain limitations persist. A critical concern involves insufficient inclusion of younger demographics potentially susceptible to ACN development. Given that CRC manifestations in younger populations frequently occur without traditional risk indicators, subsequent iterations could investigate supplementary biomarkers-potentially genomic or proteomic signatures-to enhance detection accuracy for this subgroup. The study cohort comprised self-referred screening participants rather than conventional asymptomatic populations, which introduces potential selection bias concerns despite comparative analysis revealing no significant disparities in symptom presentation between ACN and control groups. Furthermore, the model's derivation from Chinese demographic data leaves its generalizability across diverse ethnic groups requiring verification through external validation studies involving multi-ethnic cohorts. Subsequent research directions should prioritize longitudinal investigations monitoring clinical trajectories across stratified risk categories, thereby strengthening prognostic reliability and practical implementation efficacy. The precision of forthcoming model adaptations could be substantially improved through systematic integration of novel biomarker discoveries and real-world clinical outcome data.

CONCLUSION

In conclusion, we developed a modified scoring model incorporating waist-hip ratio for the prediction of ACN in a Chinese population. Comparative analysis revealed the refined model demonstrated superior predictive accuracy compared to the APCS system and BMI-modified APCS system during validation testing. Future investigations involving extensive multi-institutional longitudinal studies are warranted to evaluate the clinical applicability of this novel stratification tool.