The first 5 pages, shown as text so search engines and readers can find this document.
Page 1
HDL-C presents a suggestive
negative moderating trend for
the association between total
cholesterol and colorectal
polyps: model construction and
bootstrap internal assessment
of a predictive nomogram
Lichao Gao, Yujia Liu, Xiaojia Lv and Zhengli Dou*
The Fourth Affiliated Hospital of Anhui Medical University, Hefei, China
Background: Colorectal polyps are classic precancerous lesions of colorectal
cancer, closely linked to lipid metabolic disorders. Elevated total cholesterol does
not consistently raise polyp risk, implying interactive effects among lipoprotein
fractions. This study aimed to explore whether HDL-C exhibits a moderating trend
in the TC-polyp association and construct a predictive nomogram with bootstrap
internal performance evaluation.
Methods: A retrospective single-center case-control study enrolling 411
hospitalized colonoscopy patients. Our pre-defined primary multivariate logistic
regression was adjusted for age and standardized objective baseline indicators
(gender, BMI category, Helicobacter pylori, GSP, CST4). Normal weight was set as
BMI reference; restricted cubic splines were applied to test BMI nonlinearity. Data
on smoking, alcohol intake, diet, physical activity and family history of colorectal
polyps were incompletely recorded in inpatient medical records and thus
excluded from regression adjustment. Sensitivity analyses substituting non-
HDL-C for TC were performed to eliminate compositional coupling bias
between TC and HDL-C. Subgroup analyses stratified by polyp pathological
subtypes were conducted. Nomogram discrimination and calibration were
evaluated via ROC curves and 500 bootstrap resampling on the full cohort
without train-test splitting. Random forest was used to rank variable
importance; such rankings cannot independently corroborate the TC × HDL-C
interaction term.
Results: After adjustment for age and baseline clinical covariates, we detected a
suggestive inverse TC × HDL-C interaction (OR = 0.787, 95% CI: 0.626–0.994,
P = 0.040). Sensitivity analysis with non-HDL-C reproduced this consistent
inverse trend (P = 0.023). Restricted cubic spline modeling indicated no
nonlinear correlation between continuous BMI and polyp risk (P for non-
linearity=0.348). The internally assessed nomogram achieved an AUC of 0.834
(95% CI: 0.789–0.879) with satisfactory bootstrap calibration. Consistent inverse
moderating trends were observed across adenomatous and hyperplastic polyp
subgroups at the descriptive level.
Conclusion: We identified a suggestive inverse TC–HDL-C moderating trend
after covariate adjustment, which remained robust in non-HDL-C sensitivity
Frontiers in Endocrinology frontiersin.org01
OPEN ACCESS
EDITED BY
Esteban Orenes-Piñero,
University of Murcia, Spain
REVIEWED BY
Chen Yuhua,
Lanzhou University, China
Carlo Petruzzellis,
ARNAS Garibaldi, Italy
*CORRESPONDENCE
Zhengli Dou
ch2004640011@fy.ahmu.edu.cn
RECEIVED 30 April 2026
REVISED 05 July 2026
ACCEPTED 08 July 2026
PUBLISHED 22 July 2026
CITATION
Gao L, Liu Y, Lv X and Dou Z (2026)
HDL-C presents a suggestive negative
moderating trend for the association
between total cholesterol and colorectal
polyps: model construction and
bootstrap internal assessment
of a predictive nomogram.
Front. Endocrinol. 17:1864105.
doi: 10.3389/fendo.2026.1864105
COPYRIGHT
© 2026 Gao, Liu, Lv and Dou. This is an
open-access article distributed under the
terms of the Creative Commons
Attribution License (CC BY). The use,
distribution or reproduction in other
forums is permitted, provided the
original author(s) and the copyright
owner(s) are credited and that the
original publication in this journal is
cited, in accordance with accepted
academic practice. No use, distribution
or reproduction is permitted which does
not comply with these terms.
TYPE Original Research
PUBLISHED 22 July 2026
DOI 10.3389/fendo.2026.1864105
Page 2
analysis. HDL-C exhibited no independent main effect on colorectal polyps, yet it
exerted a suggestive negative moderating trend that attenuated the positive
association between total cholesterol and colorectal polyps. The internally
assessed nomogram provides preliminary reference for individualized
colorectal polyp risk stratification. Observations are constrained by inpatient-
only recruitment and incomplete lifestyle covariate adjustment; further large
community-based cohorts are needed to replicate these findings.
KEYWORDS
cholesterol, colorectal polyps, high-density lipoprotein cholesterol, nomogram, risk
prediction, suggestive moderating trend
Introduction
Colorectal polyps are widely recognized as the most critical
precancerous lesions of colorectal cancer (1, 2), and their pathogenesis
is closely associated with systemic metabolic disorders (3–5). As one of
the most common malignant tumors threatening human health,
colorectal cancer has shown a marked increase in incidence in recent
years (6), with a trend toward younger onset (7). Over 85% of cases
arise from adenomatous polyps through the classic “polyp-adenoma-
carcinoma” sequence (8, 9). Given the pivotal role of polyps in
colorectal carcinogenesis, identifying risk factors for their formation
is essential for early cancer prevention (10).
Among these, abnormal lipid metabolism has emerged as a key
area of investigation (11). Hypercholesterolemia is widely consid-
ered a major metabolic risk factor that promotes intestinal mucosal
hyperplasia, chronic inflammation, and abnormal epithelial prolif-
eration, thereby facilitating polyp development (12). However,
clinical observations indicate that not all individuals with elevated
cholesterol levels develop colorectal polyps (13), and the incidence
of polyps varies considerably among populations with similar
cholesterol levels (14). This suggests that the relationship between
cholesterol and colorectal polyps is not a simple linear one, but
rather modulated by complex interactions among multiple lipid
components (15). Consequently, a single lipid parameter is insuf-
ficient to accurately reflect the metabolic risk associated with
colorectal polyps (16).
High-density lipoprotein cholesterol (HDL-C) is widely recog-
nized as a core “protective lipoprotein” essential for maintaining
lipid metabolic homeostasis (17). It is involved in multiple physi-
ological processes, including cholesterol reverse transport—which
removes excess cholesterol from peripheral tissues—as well as
antioxidant and anti-inflammatory responses (18). Accumulating
evidence has demonstrated that HDL exerts a significant protective
role in the pathogenesis and progression of cardiovascular diseases
and various malignancies (19), and its potential protective effect on
the intestinal mucosa has also been suggested in relevant studies
(20). However, direct, high-quality clinical evidence remains lacking
regarding whether HDL-C modulates the association between
cholesterol and colorectal polyps, as well as the specific mode and
intensity of this regulatory effect (21).
Against this background, the present study analyzed colorectal
polyps using clinical data collected from patients who underwent
routine colonoscopy, with rigorous statistical methods. This study
aimed first to systematically characterize the association between
serum cholesterol levels and the occurrence of colorectal polyps.
Second, we aimed to explore whether HDL-C exerts a moderating
trend on the cholesterol-polyp association and quantify potential
modulating intensity via interaction term modeling. Third, the
study aimed to construct a predictive nomogram and perform
internal assessment of its predictive performance to facilitate
individualized risk stratification in clinical practice.
It should be noted that all participants in this study were
hospitalized patients receiving colonoscopy for clinical indications,
which may introduce selection bias and alter serum metabolic
biomarker levels. In addition, adenomatous and hyperplastic
polyps arise from distinct biological proliferation pathways; we
further conducted stratified subgroup analyses to explore whether
lipid regulation differs across polyp pathological subtypes.
Materials and methods
Study design and population
This was a retrospective case–control study conducted at the
Department of Gastroenterology, the Fourth Affiliated Hospital of
Anhui Medical University. The study group consisted of 310
patients hospitalized between January 1, 2021, and January 30,
2025, who underwent colonoscopy and were diagnosed with colo-
rectal polyps. The control group included 101 patients hospitalized
during the same period who underwent colonoscopy with normal
findings. The study group comprised 219 males and 91 females,
with a mean age of 57.42 ± 10.08 years. The control group included
30 males and 71 females, with a mean age of 61.04 ± 13.99 years.
Abbreviations: ALT, Alanine aminotransferase; AST, Aspartate aminotransfer-
ase; AUC, Area under the curve; BMI, Body mass index; CA125, Cancer antigen
125; CA153, Cancer antigen 153; CA199, Cancer antigen 199; CA724, Cancer
antigen 724; CEA, Carcinoembryonic antigen; CI, Confidence interval; Cr,
Creatinine; CST4, Cysteine protease inhibitor 4; FBG, Fasting blood glucose;
GSP, Glycated serum protein; HDL-C, High-density lipoprotein cholesterol; Hp,
Helicobacter pylori; LDL-C, Low-density lipoprotein cholesterol; OR, Odds ratio;
ROC, Receiver operating characteristic; TBil, Total bilirubin; TBA, Total bile
acids; TC, Total cholesterol; TG, Triglycerides; UA, Uric acid; VIF, Variance
inflation factor.
Gao et al. 10.3389/fendo.2026.1864105
Frontiers in Endocrinology frontiersin.org02
Page 3
Inclusion criteria: first-time colonoscopy during hospitalization
with no history of polypectomy, no use of hypoglycemic or lipid-
lowering therapy within the past six months, and no prior
Helicobacter pylori eradication treatment. Pathological diagnosis
of colorectal polyps was required to be confirmed independently by
two associate chief physicians or above through blinded review.
Complete clinical records were necessary, including metabolic
indicators, H. pylori status, endoscopy findings, and pathol-
ogy reports.
Exclusion criteria: presence of non-sporadic polyposis syn-
dromes, such as inflammatory bowel disease, intestinal tuberculosis,
familial adenomatous polyposis, or Peutz–Jeghers syndrome; severe
impairment of cardiac, cerebral, hepatic, or renal function; or use of
medications within the preceding six months that could alter lipid
metabolism (e.g., statins, fibrates) or exert chemopreventive effects
on colorectal polyps (e.g., aspirin or other nonsteroidal anti-
inflammatory drugs).
All enrolled patients were hospitalized individuals undergoing
colonoscopy due to gastrointestinal complaints rather than com-
munity screening crowds, which may lead to selection bias and
interfere with the stability of serum lipid indicators.
Data collection
General clinical data: Gender, age, body mass index (BMI)
classification (underweight, normal weight, overweight, obese),
history of hypertension, history of diabetes, and Helicobacter
pylori infection status (positive or negative).
Metabolic indicators: Total cholesterol (TC), triglycerides (TG),
low−density lipoprotein cholesterol (LDL−C), high−density lipo-
protein cholesterol (HDL−C), fasting blood glucose (FBG), glycated
serum protein (GSP), uric acid (UA), creatinine (Cr), urea, aspartate
aminotransferase/alanine aminotransferase ratio (AST/ALT), total
bilirubin (TBil), and total bile acids (TBA).
Tumor markers: carcinoembryonic antigen (CEA), cancer an-
tigen 125 (CA125), cancer antigen 153 (CA153), cancer antigen 199
(CA199), cancer antigen 724 (CA724), cysteine protease inhibitor
4 (CST4).
Polyp characteristics: Polyp location was classified as left colon
(descending colon, sigmoid colon, and rectum), right colon (cecum
and ascending colon), transverse colon (including hepatic and
splenic flexures), or multiple (polyps present in two or more
segments). Polyp number was categorized as solitary or multiple
(two or more polyps). Histologically, polyps were classified as
hyperplastic or adenomatous based on pathology reports.
Statistical analysis
Data were analyzed with SPSS 25.0 and R 4.2.1. Categorical
variables were compared by chi-square test; continuous variables
were assessed for normality with t-test or Wilcoxon rank-sum test
accordingly. Multicollinearity was excluded by VIF < 5 threshold.
Age is a well-documented confounder for both lipid levels and
colorectal polyps, so age adjustment was pre-specified in our
analytical framework from study design. Considering age is a
well-established confounder of both lipid metabolism and colorectal
polyp risk in existing epidemiological studies, age was forced into
the multivariate regression regardless of univariate P-values to
eliminate residual age-related confounding. The primary multivar-
iate logistic regression was adjusted for age and complete objective
baseline covariates extracted from electronic medical records:
gender, BMI category, Helicobacter pylori infection, GSP and
CST4. Normal weight, male sex, and H. pylori-negative status
were set as reference categories for categorical variables.
Restricted cubic splines were applied to test the nonlinear relation-
ship between continuous BMI and polyp risk.
Smoking, alcohol, diet and activity data were collected through
clinical inquiry, yet records lacked unified standards and contained
frequent missing values. To avoid missing-data bias, these lifestyle
variables were not incorporated into regression. To test interaction
robustness and exclude compositional coupling bias, sensitivity
analyses were performed using non-HDL-C, LDL-C, and the TC/
HDL-C ratio in place of TC. Subgroup analyses by polyp patho-
logical subtype were performed.
The TC × HDL-C interaction term was reverse-coded only for
nomogram scoring without altering model results. Model discrim-
ination was measured by AUC, and calibration was evaluated via
500 bootstrap repetitions (internal assessment). A random forest
model with 500 decision trees ranked predictors by mean decrease
in Gini index. Variable importance rankings only reflect predictive
weight within this inpatient cohort and cannot independently
confirm the presence of the TC × HDL-C moderating trend.
Model development and internal
assessment
Based on the independent predictors identified in the multivar-
iate moderation analysis, a nomogram was constructed using the
rms package in R (version 4.2.1). Each predictor was assigned a
weighted score proportional to its regression coefficient, and the
total score corresponded to the predicted probability of colorectal
polyp presence. Model discrimination was assessed by receiver
operating characteristic (ROC) curves and quantified by the area
under the curve (AUC) with 95% confidence intervals. Calibration
was evaluated using bootstrap-corrected calibration curves with 500
resampling repetitions. To further characterize relative variable
predictive weight, a random forest model (random Forest package)
was trained with 500 trees, and the mean decrease in Gini index was
used to rank predictors.
Results
Baseline data
In univariate comparison, the control group had a significantly
higher mean age than the polyp group (61.04 ± 13.99 vs. 57.42 ±
10.08 years, P = 0.018). Additionally, the two groups differed
significantly in gender, BMI classification, hypertension prevalence,
and H. pylori infection rate (all P < 0.05), while diabetes prevalence
showed no significant difference (P > 0.05) (Table 1). This
Gao et al. 10.3389/fendo.2026.1864105
Frontiers in Endocrinology frontiersin.org03
Page 4
counterintuitive finding is likely multifactorial. First, the marked
sex imbalance between groups—predominantly male in the polyp
group (70.6%) versus predominantly female in the control group
(70.3%)—introduced substantial confounding, as female sex exhib-
ited a strong protective effect in multivariate analysis (OR = 0.14).
Second, the inpatient nature of the cohort may have amplified this
disparity: control subjects were older hospitalized patients under-
going colonoscopy for various indications, whereas the polyp group
included relatively younger individuals with milder complaints.
After adjusting for sex and other covariates, age lost statistical
significance (OR = 0.989, P = 0.387), confirming that the crude age
difference was driven by confounding and selection bias rather than
a genuine inverse age–polyp association.
Univariate analysis of key metabolic and
tumor markers
Univariate analysis revealed significant differences between the
polyp and control groups in total cholesterol, triglycerides, LDL-C,
fasting blood glucose, uric acid, creatinine, AST/ALT ratio (all P <
0.05), as well as in tumor markers CA125 and CA724 (both P <
0.01). Among lipid parameters, total cholesterol was significantly
higher in the polyp group (4.82 ± 0.91 vs. 4.43 ± 1.12 mmol/L,
P = 0.001), whereas HDL-C did not differ significantly between
groups (P = 0.495). Glycated serum protein (GSP) and CST4
showed borderline associations (P = 0.156 and P = 0.064, respec-
tively) and were retained for multivariate analysis based on clinical
relevance. Complete univariate comparisons of metabolic indicators
and tumor markers are provided in Supplementary Tables 1, S2,
respectively. Stratified analyses grouped by polyp location, polyp
number and pathological subtype are displayed in Supplementary
Tables 3-S5.
Multicollinearity assessment
Multicollinearity analysis revealed severe collinearity between
LDL-C and total cholesterol, as well as between AST, ALT, and the
AST/ALT ratio (VIF > 5). To avoid model bias, LDL-C, AST, and
ALT were excluded from subsequent multivariate analyses.
Multivariate moderation analysis and
independent predictors
In our pre-specified multivariate logistic regression fully ad-
justed for age and all available baseline clinical covariates, a
suggestive inverse moderating trend between standardized TC
and HDL-C was observed (OR = 0.787, 95% CI: 0.626–0.994,
P = 0.040, Figure 1). Elevated standardized TC independently
increased the risk of colorectal polyps (OR = 1.570, 95% CI:
1.160–2.160, P = 0.004), whereas HDL-C exhibited no independent
predictive effect after full covariate adjustment (OR = 1.150, 95% CI:
0.840–1.580, P = 0.394).
After full covariate adjustment, the main effect of HDL-C was
statistically non-significant (OR = 1.150, P = 0.394), indicating
isolated HDL-C levels could not independently predict polyp
occurrence. However, the TC × HDL-C interaction term reached
statistical significance (OR=0.787, P = 0.040), demonstrating HDL-
C only functions by modifying the pro-polyp effect of total
TABLE 1 Comparison of baseline characteristics between the two groups [n(%)].
Index Study group Control group Statistic P
Age (years) 57.42 ± 10.08 61.04 ± 13.99 -2.405 0.018
Gender 53.475 <0.001
Male 219 (70.6%) 30 (29.7%)
Female 91 (29.4%) 71 (70.3%)
Hypertension 5.879 0.015
Yes 131 (42.3%) 29 (28.7%)
No 179 (57.7%) 72 (71.3%)
Diabetes mellitus 1.578 0.209
Yes 46 (14.8%) 10 (9.9%)
No 264 (85.2%) 91 (90.1%)
Helicobacter pylori 5.889 0.015
Positive 90 (29.0%) 17 (16.8%)
Negative 220 (71.0%) 84 (83.2%)
BMI classification 41.788 <0.001
Underweight 8 (2.6%) 19 (18.8%)
Normal weight 138 (44.5%) 52 (51.5%)
Overweight 120 (38.7%) 26 (25.7%)
Obesity 44 (14.2%) 4 (4.0%)
Although the control group had a significantly higher mean age than the polyp group in univariate comparison (P = 0.018), age lost independent predictive significance after adjusting for gender,
BMI category, H. pylori status, GSP and CST4 in the multivariate model (P = 0.387). The crude intergroup age disparity was mediated by confounding clinical covariates rather than reflecting a
causal age-related polyp risk.
Gao et al. 10.3389/fendo.2026.1864105
Frontiers in Endocrinology frontiersin.org04
Page 5
cholesterol. Specifically, elevated HDL-C buffered risk increments
driven by high TC, while HDL-C concentrations made no difference
in polyp risk among individuals with normal TC levels. The
univariate comparison (Supplementary Table 1) further supported
this pattern: baseline HDL-C levels were comparable between polyp
and control groups (P = 0.495), so intergroup polyp disparities
could not be attributed to raw HDL-C concentrations alone.
Age was mandatorily incorporated into the regression per our
pre-planned analytical framework, despite lacking independent
predictive power (OR = 0.989 per year of age, 95% CI: 0.966–
1.010, P = 0.387). Despite the higher baseline age observed in
controls, age showed no independent association with colorectal
polyps after comprehensive covariate adjustment, suggesting the
crude age gap between groups was driven by residual confounding
rather than an inherent age-related polyp risk gradient.
For categorical covariates, normal weight, male sex and Hp-
negative patients served as unified reference groups. Compared with
normal-weight individuals, obesity was associated with significantly
higher polyp odds (OR = 3.720, 95% CI: 1.250–14.000, P = 0.030).
Overweight showed no statistical correlation with polyp risk.
Underweight was associated with lower polyp odds (OR = 0.120,
95% CI: 0.038–0.341, P<0.001), which further supports a positive
correlation between adiposity and colorectal neoplasia. Restricted
cubic spline analysis further verified the absence of a significant
nonlinear association between continuous BMI and polyp incidence
(P for non-linearity=0.348, Figure 2). Other independent risk
predictors identified in this fully adjusted model included female
sex (reference: male), which exerted a protective effect against
polyps (OR=0.140, 95%CI: 0.070–0.240, P<0.01). Positive H.
pylori status (reference: Hp-negative) was an independent risk
factor for colorectal polyps (OR = 2.330, 95%CI: 1.200–4.740,
P=0.016), elevated GSP (OR=1.010, 95%CI: 1.000–1.020, P=0.016)
and elevated CST4 (OR=1.010, 95%CI: 1.001–1.030, P = 0.011). Full
regression outputs are summarized in Table 2 and visualized in
Figure 3. The total points correspond to an estimated probability of
colorectal polyp presence, ranging from <10% to >90%. (Figure 4).
FIGURE 1
Suggestive negative moderating trend of HDL-C on the association between standardized TC and colorectal polyp risk. The risk gradient of TC was
flattened at high HDL-C levels.
FIGURE 2
Restricted cubic spline curve for continuous BMI; P for non-linearity=0.348.
Gao et al. 10.3389/fendo.2026.1864105
Frontiers in Endocrinology frontiersin.org05