Dose–response effect of statins on colorectal cancer risk in IBD: a nationwide cohort study

Abstract

Background

Patients with inflammatory bowel disease (IBD) face an elevated risk of colorectal cancer (CRC). Statins have demonstrated potential anticancer properties, but evidence in IBD populations remains limited.

Objective

To evaluate the association between sustained statin use and the risk of CRC in patients with IBD, and to assess dose–response relationships and statin type-specific effects.

Design

Nationwide, retrospective cohort study using time-dependent Cox proportional hazards models and Fine–Gray competing risk models.

Results

Among 6,120 propensity score–matched adults with IBD (2,040 statin users and 4,080 non-statin lipid-lowering agent users), statin use was associated with a 65% reduction in CRC risk compared with users of non-statin lipid-lowering agents (adjusted hazard ratio [aHR] 0.35; 95% CI, 0.24–0.52; p<0.0001). A clear dose–response relationship was observed, with the highest quartile of cumulative exposure associated with an aHR of 0.08. Rosuvastatin and simvastatin conferred the strongest protective effects. Statin use also correlated with lower all-cause mortality (aHR 0.42; 95% CI, 0.35–0.51).

Conclusion

In this nationwide observational cohort, sustained statin use was associated with a significantly lower risk of CRC and mortality. While the observed dose–response and statin-specific patterns suggest a potential chemopreventive signal, these findings are hypothesis-generating. Confirmation in randomized or pragmatic trials is required before these results can translate into clinical practice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-026-15970-y.

Summary box

What is already known on this topic

Patients with inflammatory bowel disease (IBD) are at increased risk of colorectal cancer (CRC). Statins have demonstrated chemopreventive potential against CRC in the general population; however, evidence in IBD populations remains limited and inconsistent, with few large-scale studies evaluating dose–response relationships.

What this study adds

With the important limitation of potential residual confounding inherent to claims-based observational studies, this nationwide cohort study provides hypothesis-generating evidence of a dose–response association between statin use and reduced colorectal cancer risk among patients with inflammatory bowel disease. However, given the observational nature of the study, these findings should be interpreted as hypothesis-generating rather than establishing causality.

How this study might affect research, practice or policy

These findings do not support the immediate initiation of statins solely for CRC prevention but provide a strong rationale for designing prospective randomized or pragmatic trials. While statins should continue to be prescribed based on established cardiovascular indications, their potential role in CRC risk reduction serves as a proof-of-concept for future IBD-focused prevention research.

Introduction

Colorectal cancer (CRC) ranks as the third most commonly diagnosed cancer and the second leading cause of cancer-related death worldwide [1, 2]. In Taiwan, CRC has remained one of the top three cancers by incidence and mortality for over a decade, with more than 16,000 new cases and approximately 6,000 deaths reported annually [3–5]. Among high-risk groups, patients with inflammatory bowel disease (IBD)—including ulcerative colitis (UC) and Crohn’s disease (CD)—face an even greater threat. Studies have shown that individuals with longstanding IBD are at increased risk for CRC, and once diagnosed, these patients tend to experience poorer prognosis compared to those in the general population [6–8]. This makes cancer prevention in the IBD population an urgent clinical and public health priority.

Over the past two decades, various pharmacologic agents have been investigated for their potential chemopreventive effects against CRC [9]. Among them, statins—widely used for hyperlipidemia and cardiovascular risk reduction—have attracted attention due to their pleiotropic properties, including anti-inflammatory and antiproliferative effects [10]. However, findings from epidemiologic studies on statin use and CRC risk have been inconsistent, with some cohort and case-control studies demonstrating a significant protective effect, while others found no association or only marginal benefit [11–14]. These discrepancies may stem from differences in study design, population heterogeneity, statin type or dose, duration of follow-up, and residual confounding [11–15]. Moreover, few studies have adequately addressed time-varying exposure, cumulative dose, or medication adherence, limiting the interpretability and generalizability of the results. In Taiwan, throughout the study period (2008–2019), lipid-lowering therapy was guided by cardiovascular risk stratification rather than calendar-time policy shifts [16, 17]. Statins were the mandated first-line therapy across risk strata, whereas non-statin agents (e.g., fibrates or bile acid sequestrants) had distinct indications unrelated to BMI or smoking status [18, 19]. This prescribing framework reduces the likelihood that temporal guideline changes or baseline cardiometabolic characteristics systematically altered first-line drug choice, thereby supporting the validity of our active-comparator design.

Despite these challenges, the biological plausibility and recurring positive signals in the literature continue to support a potential protective role of statins against CRC [20]. Given that IBD patients represent a population at particularly high risk [6–8], it is both timely and necessary to investigate whether statins might offer meaningful protection against CRC in this vulnerable group. Furthermore, unanswered questions remain regarding which statin agents, and at what intensity or cumulative dose, are most effective. Addressing these questions could offer valuable guidance for future preventive strategies, particularly in tailoring long-term care for patients with IBD.

Methods

Study population

The Taiwan National Health Insurance Research Database (NHIRD), covering nearly 99% of the population since 1995, offers validated, population-representative data on diagnoses, prescriptions, and hospitalizations. Its large sample size, longitudinal design, and standardized ICD coding enable robust analyses of chronic disease patterns and treatment outcomes [21–23]. This study used de-identified NHIRD data with approval from the Tzu Chi Medical Foundation Research Ethics Committee, waiving the need for informed consent. Given these constraints, we designed this nationwide cohort study to approximate a target-trial comparison as closely as possible within claims data. Specifically, we used an active-comparator approach by comparing sustained statin users with sustained users of non-statin lipid-lowering agents, rather than untreated individuals, to improve comparability with respect to the indication for lipid-lowering therapy, medication-taking behavior, and general healthcare utilization. Non-statin lipid-lowering agents (e.g., fibrates, ezetimibe, bile acid sequestrants) were chosen as a pragmatic active comparator because they reflect clinician-recognized need for lipid management and involve similar prescribing and follow-up intensity. Crucially, prior meta-analyses and randomized trials indicate that these non-statin agents exert a neutral effect on colorectal cancer risk [24–28], unlike the potential chemopreventive effects of statins. This design was intended to reduce (though not eliminate) healthy user/healthy adherer bias compared with an untreated comparator. To further mitigate residual confounding in an IBD setting, we additionally incorporated claims-based proxies for IBD phenotype/treatment and surveillance colonoscopy utilization into the propensity score model. These variables were included to improve baseline balance related to treatment assignment.

Study timeline and variable definitions

Index date (cohort entry) was defined as the date of the first prescription of either a statin or a non-statin lipid-lowering agent after IBD diagnosis; this was also the timepoint for propensity score matching. Baseline covariates were ascertained during the 12 months prior to and including the index date. Time-varying measures (exposure status and cumulative dose) were updated during follow-up as described below.

Inclusion and exclusion criteria

We included adults aged ≥ 20 years with newly diagnosed inflammatory bowel disease (IBD) between January 1, 2008, and December 31, 2019, with follow-up through December 31, 2022. IBD was identified using validated ICD-9/10 codes for Crohn’s disease and ulcerative colitis, requiring ≥ 1 inpatient or ≥ 2 outpatient claims within 12 months. A total of 241,590 patients were identified before matching; 2,955 received statins, and 21,204 received non-statin lipid-lowering agents. Patients were assigned to the statin group if they had a medication possession ratio (MPR) > 80% for a single statin and < 20% for others [29], and received ≥ 28 cumulative defined daily doses (cDDDs) per year. The cDDD was calculated annually to reflect changes in prescriptions over time. Cumulative exposure was treated as time-updated during follow-up. Daily statin intensity was defined as average DDD per day and categorized as low-intensity (< 1 DDD/day) or high-intensity (≥ 1 DDD/day). The active comparator group comprised users of non-statin lipid-lowering agents with MPR > 80% and minimal statin exposure (< 28 cDDDs during follow-up), to improve comparability in medication-taking behavior and healthcare utilization.

We excluded patients with missing demographic data, prior cancer, or CRC diagnosed or death occurring within the first year. Additional exclusions included ESRD on dialysis, autoimmune diseases (e.g., SLE, RA), HIV, prior organ transplant, and those with mixed exposure (≥ 28 cDDDs of both statin and non-statin agents). All exclusion criteria were applied prior to cohort entry (i.e., before the index date). Therefore, patients who actively switched between statin and non-statin therapies were excluded by design to maintain a strict active-comparator contrast. Individuals meeting any exclusion criteria were removed at baseline and did not contribute any person-time to follow-up. A detailed flowchart (Fig. 1) illustrates the number of patients excluded at each step leading to the final matched cohort. Propensity score matching (PSM) was performed using a 1:2 nearest-neighbor algorithm without replacement, to balance the probability of treatment assignment based on baseline characteristics, including demographics, IBD duration, IBD subtype (ulcerative colitis, Crohn’s disease, and indeterminate colitis), claims-based proxies of IBD treatment/severity (e.g., 5-ASA, systemic corticosteroids, immunomodulators, biologics, and IBD-related hospitalization), and surveillance colonoscopy utilization, comorbidities, lifestyle factors (ICD-code proxies for smoking, alcohol-related diseases, and physical inactivity; Table 1), and concurrent medications. All matching variables were defined at baseline (12-month look-back prior to and including the index date).

Fig. 1.

Study Flowchart of Patient Selection and Exclusion

Table 1.

Baseline characteristics of patients with inflammatory bowel disease receiving statins versus non-statin lipid-lowering therapy after propensity score matching

	Non-Statins Use		Statins Use		ASMD
	N = 4080		N = 2040
	N	%	N	%
Age (mean ± SD)	51.76 ± 18.13		53.26 ± 14.04		0.0120
Age-median (IQR, Q1, Q3)-years-old	53.00 (39.00,65.00)		57.00 (50.00,67.00)
Age group-years-old					0.0720
20–50	1,729	42.4%	933	45.7%
51–60	990	24.3%	482	23.6%
61–70	734	18.0%	343	16.8%
> 70	627	15.4%	282	13.8%
Sex					0.0457
Female	2,089	51.2%	1,091	53.5%
Male	1,991	48.8%	949	46.5%
Income levels (NTD)					0.1090
Low income/Financial dependent	878	21.5%	530	26.0%
<20,000	1,213	29.7%	566	27.8%
2001–30,000	1,030	25.3%	488	23.9%
30,001–45,000	620	15.2%	308	15.1%
>45,000	339	8.3%	148	7.3%
Urbanization					0.0015
Rural	1,191	29.2%	597	29.3%
Urban	2,889	70.8%	1,443	70.7%
Duration of Inflammatory Bowel Disease (median IQR), years
Mean (SD) follow-up year	2.05 ± 0.56		2.08 ± 1.44		0.0221
Median (IQR, Q1, Q3) follow-up year	1.73 (1.55,2.24)		1.79 (1.06,2.88)
Duration of Inflammatory Bowel Disease, years					0.0018
0 ~ 1	2,436	59.7%	200	58.8%
> 1	1,644	40.3%	840	41.2%
Coexisting comorbidities
Diabetes	383	9.4%	259	12.7%	0.1057
Hypertension	1,218	29.9%	633	31.0%	0.0256
Obesity	38	0.9%	17	0.8%	0.0107
Benign neoplasm of colon	58	1.4%	28	1.4%	0.0043
COPD	751	18.4%	375	18.4%	0.0008
Depression	249	6.1%	134	6.6%	0.0193
Anxiety	585	14.3%	297	14.6%	0.0063
Family History
Family history of malignant neoplasm of gastrointestinal tract	8	0.2%	4	0.2%	0.0000
Lifestyle Factors
Smoking	24	0.6%	16	0.8%	0.0230
Alcohol-Related Diseases	88	2.2%	37	1.8%	0.0251
Lack of physical exercise	11	0.3%	9	0.4%	0.0117
Medication Use
Contraceptives	284	7.0%	150	7.4%	0.0151
NSAIDs	1,452	35.6%	721	35.3%	0.0052
Antibiotics	604	14.8%	302	14.8%	0.0000
Charlson Comorbidity Index
Mean (SD)	0.58 ± 1.08		0.55 ± 0.97		0.0034
Median (Q1-Q3)	0.00 (0.00,1.00)		0.00 (0.00,1.00)
Charlson Comorbidity Index					0.0015
0	2,843	69.7%	1,420	69.6%
≥ 1	1,237	30.3%	620	30.4%
Statins
Statins, cDDD
Non-Statins use	4,080	100.0%	0	0.0%
Q1	0	0.0%	569	27.9%
Q2	0	0.0%	545	26.7%
Q3	0	0.0%	478	23.4%
Q4	0	0.0%	448	22.0%
Statins, Daily intensity (DDD)
Mean (SD)	0		0.71 ± 0.32
Median (IQR)	0		0.66 (0.50,0.90)
Daily intensity (DDD)
< 1	0	0.0%	1,619	79.4%
≥ 1	0	0.0%	421	20.6%
Statins
Non-Statins use	4,080	100.0%	0	0.0%
Lipophilic statins
Atorvastatin	0	0.0%	750	36.8%
Lovastatin	0	0.0%	116	5.7%
Simvastatin	0	0.0%	417	20.4%
Fluvastatin	0	0.0%	165	8.1%
Pitavastatin	0	0.0%	26	1.3%
Hydrophilic statins
Rosuvastatin	0	0.0%	402	19.7%
Pravastatin	0	0.0%	164	8.0%
Statins
Non-Statins use	4,080	100.0%	0	0.0%
Lipophilic statins	0	0.0%	1,505	73.8%
Hydrophilic statins	0	0.0%	535	26.2%
Mean (SD) follow-up year	8.74 ± 4.15		9.65 ± 3.91		0.0262
Median (IQR-Q1-Q3) follow-up year	9.13 (5.50,12.13)		10.24 (6.77,12.88)

Abbreviations: ASMD Absolute Standardized Mean Difference, N Number, NTD New Taiwan Dollars, SD Standard Deviation, IQR Interquartile Range, Q1, Q3 First and Third Quartiles, DDD Defined Daily Dose, cDDD Cumulative Defined Daily Dose

Defined daily dose (DDD) and cumulative dose calculations

Defined daily doses (DDDs) for individual statins were based on WHO standards: atorvastatin (20 mg), rosuvastatin (10 mg), simvastatin (30 mg), pravastatin (30 mg), lovastatin (45 mg), fluvastatin (60 mg), and pitavastatin (2 mg) [30]. Cumulative DDDs (cDDDs) were calculated annually as:

Statin exposure and cDDD were treated as time-varying during follow-up, and users were stratified into quartiles (Q1–Q4) by total cDDD and categorized as low-intensity (< 1 DDD/day) or high-intensity (≥ 1 DDD/day) based on daily cDDD (total cDDD / total follow-up days). These standardized dose metrics supported dose–response analyses.

Primary endpoints

The primary endpoint was incident CRC diagnosed after cohort entry, identified using validated ICD-9-CM (153.x, 154.0, 154.1, 154.8) and ICD-10-CM (C18.x, C19, C20) codes. CRC cases were confirmed through linkage with the Taiwan Cancer Registry Database, which includes histologically verified diagnoses. Patients with a prior history of CRC were excluded. Each patient was counted once based on the date of first CRC diagnosis. Due to inconsistent staging data, only the initial confirmed event was used.

The secondary endpoint was all-cause mortality, identified from inpatient discharge records and the national death registry. Time-to-event was defined from the index date to CRC diagnosis, death, or end of follow-up (December 31, 2022), whichever occurred first.

Statistical analysis

After propensity score matching, a time-dependent Cox proportional hazards model was used to assess the association between statin use and CRC risk in patients with IBD, with statin exposure modeled as a time-varying covariate updated quarterly. Post-matching regression models were applied to improve precision and to accommodate time-varying exposure, rather than to correct an inadequate propensity score. Exposure was defined according to an “as-treated” time-varying approach. Specifically, time-varying exposure status and cumulative dose were updated quarterly based on active prescriptions. If a patient had a prescription gap of ≥ 90 days, their status for that interval was dynamically reclassified as “unexposed” (statin = 0). Thus, “unexposed” person-time represented periods of statin discontinuation within the statin cohort. Consequently, if an outcome (CRC or death) occurred during a discontinuation interval, it was statistically attributed to the unexposed person-time. This dynamic exposure allocation aligns person-time with active medication use and prevents immortal time misclassification. Follow-up spanned from the index date to CRC diagnosis, death, or December 31, 2022. In our analytic strategy, variable selection for the PS model prioritized true confounders and baseline prognostic factors associated with CRC risk [31]. PSM was utilized to achieve prognostic balance between the exposure groups at baseline, ensuring that the statin and non-statin cohorts shared a comparable baseline risk profile. The subsequent multivariable Cox regression was applied not to ‘correct’ an inadequate propensity score, but to improve estimation precision and rigorously account for time-varying exposure dynamics in a pharmacoepidemiologic setting.

Within the matched cohort, four hierarchical models were constructed:

(1) unadjusted; (2) adjusted for age and sex; (3) further adjusted for socioeconomic and clinical factors (e.g., income, urbanization, IBD duration, comorbidities, family history); and (4) additionally adjusted for behavioral and medication-related variables, including smoking, alcohol-related disease, physical inactivity, NSAIDs, aspirin, metformin, antibiotics, and contraceptives, as well as claims-based proxies for key IBD therapies/severity and surveillance colonoscopy utilization. The Charlson Comorbidity Index (CCI) was defined using validated ICD-9-CM and ICD-10-CM diagnosis codes from inpatient and outpatient claims. CCI was assessed at baseline, utilizing the 12-month look-back period strictly prior to the index date, to represent pre-existing comorbidity burden. Covariates in the PS model were baseline-defined, whereas exposure and cDDD were treated as time-varying during follow-up. We additionally conducted prespecified subgroup analyses by IBD subtype (ulcerative colitis, Crohn’s disease, and indeterminate colitis/IBD-unclassified when identifiable), reported in Supplemental Table 1. To evaluate potential surveillance bias, we conducted a sensitivity analysis stratifying patients by colonoscopy surveillance frequency using claims procedure codes, defining regular surveillance as colonoscopy performed every 1–2 years and irregular surveillance as less frequent or none (Supplemental Table 3).

To address competing mortality risks, Fine and Gray subdistribution hazard models were used with all-cause death specified a priori as the competing event. The Fine and Gray model estimates the subdistribution hazard ratio (sHR), reflecting the effect of covariates on the cumulative incidence function (CIF), rather than an etiologic cause-specific hazard. Statin exposure was also stratified by cumulative defined daily dose (cDDD; quartiles) and by daily intensity (< 1 vs. ≥ 1 DDD/day). Subgroup analyses assessed individual statin agents and solubility class (lipophilic vs. hydrophilic). For statin-agent subgroup analyses, strata with zero CRC events were reported as “not estimable,” and corresponding P values were not presented.

Incidence rates and incidence rate ratios per 10,000 person-years were calculated via Poisson regression. CIF curves based on the Aalen–Johansen estimator were generated to visualize the cumulative probability of CRC while accounting for the competing risk of all-cause death (Fig. 2). Standard Kaplan–Meier curves were used exclusively for overall survival, not for CRC incidence. Restricted cubic splines modeled nonlinear dose–response relationships.

Fig. 2.

Cumulative Incidence Function (CIF) Curves for Colorectal Cancer Incidence in Patients With Inflammatory Bowel Disease: Comparison of Statin Versus Non-Statin Therapy

All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC), with two-sided P values < 0.05 considered statistically significant.

Results

Baseline characteristics by statin use

Before matching, 241,590 eligible patients were identified. After applying baseline exclusion criteria prior to cohort entry, the remaining patients were included in the matching process, as detailed in Fig. 1. After PSM, a total of 2,040 patients receiving statins were matched to 4,080 users of non-statin lipid-lowering agents (Table 1). The two groups were well balanced across age, sex, income level, urbanization, IBD duration, Charlson Comorbidity Index (CCI), coexisting comorbidities, and lifestyle factors, with all absolute standardized mean differences below 0.1. The mean age was slightly higher in statin users (53.3 ± 14.0 years) compared with the non-statin group (51.8 ± 18.1 years). Women comprised 53.5% of the statin group and 51.2% of the non-statin group. The median duration of IBD was similar between groups. Comorbidity burden, including diabetes, hypertension, and COPD, was comparable. Most statin users received lipophilic statins (73.8%), with atorvastatin (36.8%) and simvastatin (20.4%) being the most common agents. The median cumulative statin exposure was 0.66 DDD/day, with 20.6% receiving ≥ 1 DDD/day. Because individuals with mixed exposures were explicitly excluded at baseline, the proportion of study participants who switched between the statin and non-statin exposure groups during the follow-up period was 0%.

Adjusted risks of colorectal cancer and all-cause mortality

Statin use was associated with a significantly lower relative risk of CRC compared with users of non-statin lipid-lowering agents (Table 2). In the fully adjusted time-varying Cox model (Model 4), statin use was associated with a 65% reduction in CRC relative risk (aHR, 0.35; 95% CI, 0.24 to 0.52; P < 0.0001). In the Fine and Gray competing risk model, statin use was associated with a significantly lower subdistribution hazard of CRC (sHR, 0.39; 95% CI, 0.26 to 0.57; P < 0.0001), accounting for all-cause death as a competing event. Statin use was also associated with significantly lower all-cause mortality (aHR, 0.42; 95% CI, 0.35 to 0.51; P < 0.0001). CIF curves (Fig. 2) demonstrated a lower cumulative incidence of CRC in statin users while simultaneously accounting for the competing risk of all-cause death (capturing only deaths that occurred prior to CRC). Kaplan–Meier curves were used only to depict overall survival (capturing all-cause deaths at any time) (Fig. 3).

Table 2.

Adjusted Hazard Ratios for Colorectal Cancer Incidence and All-Cause Mortality in Patients With Inflammatory Bowel Disease: Multivariable Cox Regression Comparing Statin and Non-Statin Lipid-Lowering Therapy

	Multivariable Cox Regression			Competing Risk Analysis for Colorectal Cancer Incidence
	Adjusted HRa (95% CI)		P-value	Adjusted HRb (95% CI)		P-value
Colorectal Cancer
Model 1	0.36	(0.24, 0.53)	< 0.0001	0.38	(0.26, 0.56)	< 0.0001
Model 2	0.38	(0.25, 0.56)	< 0.0001	0.41	(0.28, 0.6)	< 0.0001
Model 3	0.38	(0.25, 0.56)	< 0.0001	0.41	(0.28, 0.6)	< 0.0001
Model 4	0.35	(0.24, 0.52)	< 0.0001	0.39	(0.26, 0.57)	< 0.0001
All-Cause Mortality
Model 1	0.41	(0.34, 0.48)	< 0.0001	-	-	-
Model 2	0.43	(0.36, 0.51)	< 0.0001	-	-	-
Model 3	0.42	(0.35, 0.51)	< 0.0001	-	-	-
Model 4	0.42	(0.35, 0.51)	< 0.0001	-	-	-

Abbreviations: HR Hazard Ratio, CI Confidence Interval

^aA time-varying Cox proportional hazards model was used to account for changes in statin use over time. The model was adjusted progressively as follows: Model 1 included statin use alone (univariable); Model 2 was adjusted for age and sex; Model 3 was further adjusted for age, sex, income level, urbanization, duration of inflammatory bowel disease, coexisting comorbidities, and family history of gastrointestinal malignancy; and Model 4 included additional adjustment for lifestyle factors and medication use

^bThe Fine and Gray subdistribution hazard model was used to estimate the risk of colorectal cancer while accounting for all-cause mortality (death) as a competing risk

Fig. 3.

Kaplan–Meier Curves for Overall Survival in Patients With Inflammatory Bowel Disease Receiving Statin Versus Non-Statin Therapy

Effect consistency across subgroups

Figure 4 illustrates the results of prespecified subgroup analyses stratified by demographic and clinical characteristics. The association between statin use and a lower relative risk of colorectal cancer was consistent across all examined subgroups, including age (< 50 vs. ≥50 years), sex, comorbidity burden (CCI = 0 vs. ≥1), and the presence or absence of diabetes and hypertension. In each subgroup, statin users demonstrated a significantly lower adjusted hazard ratio for CRC compared with the non-statin group, and there was no evidence of significant effect modification across strata. In addition, the association was directionally consistent across IBD subtypes and in sensitivity analyses incorporating available claims-based proxies for key concomitant IBD therapies, aspirin/metformin use, and surveillance colonoscopy utilization, including stratification by colonoscopy surveillance frequency (regular vs. irregular; Supplemental Table 3) (Supplemental Tables 1–3).

Fig. 4.

Subgroup Analysis of the Association Between Statin Use and Risk of Colorectal Cancer in Patients With Inflammatory Bowel Disease. Abbreviations: aHR (Adjusted Hazard Ratio), CI (Confidence Interval), CCI (Charlson Comorbidity Index)

Notably, among patients without any baseline comorbidities (CCI = 0), statin use remained strongly associated with a lower risk of colorectal cancer, which reduces (but does not eliminate) concerns that the association is explained solely by baseline comorbidity burden.

Dose–response, statin type, and competing risk models

A clear dose–response relationship was observed. Compared with the non-statin group, patients in the highest quartile of cumulative statin dose (cDDD, Q4) had the lowest CRC relative risk (aHR, 0.08; 95% CI, 0.02 to 0.34; P for trend < 0.0001), with similar trends across quartiles (Table 3, Supplemental Fig. 1). Daily statin intensity ≥ 1 DDD/day was associated with a lower CRC relative risk (aHR, 0.28; 95% CI, 0.10 to 0.75) compared with lower intensity (Supplemental Fig. 2). When stratified by statin solubility, both hydrophilic (aHR, 0.24; 95% CI, 0.10 to 0.57) and lipophilic statins (aHR, 0.39; 95% CI, 0.26 to 0.60) were associated with lower CRC risk compared with non-statin therapy (Supplemental Fig. 4). Among individual statins, rosuvastatin (aHR, 0.18; 95% CI, 0.06 to 0.57) and simvastatin (aHR, 0.33; 95% CI, 0.14 to 0.75) showed the strongest protective effects showed the largest observed inverse associations (Supplemental Fig. 3). However, several statin-specific subgroups had sparse or zero CRC events (e.g., pitavastatin), yielding non-estimable risk estimates due to zero events in those strata; therefore, comparisons across individual statins should be interpreted cautiously.

Table 3.

Hazard ratios for colorectal cancer in patients with inflammatory bowel disease: comparison of statin and non-statin lipid-lowering therapies

Variable	Crude HR (95% CI)	P-value	aHRa (95% CI)	P-value	aHRb (95% CI)	P-value
Non-Statins use (Ref)	1.00	-	1.00	-	1.00	-
Statins use (Total)	0.36 (0.24, 0.53)	< 0.0001	0.35 (0.24, 0.52)	< 0.0001	0.39 (0.26, 0.57)	< 0.0001
Individual Statin Agents
Hydrophilic statins
Pravastatin	0.29 (0.07, 1.18)	0.0843	0.30 (0.07, 1.21)	0.0903	0.34 (0.08, 1.36)	0.1264
Rosuvastatin	0.18 (0.06, 0.57)	0.0036	0.18 (0.06, 0.57)	0.0036	0.20 (0.06, 0.62)	0.0051
Lipophilic statins
Pitavastatin	NA	NA	NA	NA	NA	NA
Fluvastatin	0.42 (0.13, 1.32)	0.1380	0.39 (0.12, 1.25)	0.1127	0.42 (0.13, 1.32)	0.1366
Atorvastatin	0.45 (0.26, 0.77)	0.0041	0.45 (0.26, 0.78)	0.0044	0.49 (0.29, 0.84)	0.0096
Lovastatin	0.57 (0.18, 1.78)	0.3326	0.48 (0.15, 1.51)	0.2077	0.53 (0.17, 1.65)	0.2749
Simvastatin	0.33 (0.15, 0.75)	0.0082	0.33 (0.14, 0.75)	0.0080	0.36 (0.16, 0.82)	0.0148
Statin Solubility (Summary)
Hydrophilic statins	0.23 (0.09, 0.56)	0.0012	0.24 (0.10, 0.57)	0.0015	0.26 (0.11, 0.62)	0.0027
Lipophilic statins	0.40 (0.27, 0.62)	< 0.0001	0.39 (0.26, 0.60)	< 0.0001	0.43 (0.29, 0.65)	< 0.0001
Cumulative Dose (cDDD)
Non-Statins use (Ref)	1.00	-	1.00	-	1.00	-
Q1	0.49 (0.26, 0.91)	0.0230	0.53 (0.29, 0.99)	0.0447	0.56 (0.30, 1.04)	0.0659
Q2	0.49 (0.26, 0.91)	0.0232	0.51 (0.28, 0.95)	0.0342	0.56 (0.30, 1.02)	0.0572
Q3	0.35 (0.16, 0.74)	0.0061	0.31 (0.15, 0.68)	0.0030	0.34 (0.16, 0.73)	0.0057
Q4	0.10 (0.02, 0.38)	0.0009	0.08 (0.02, 0.34)	0.0005	0.10 (0.02, 0.39)	0.0011
P for trend			< 0.0001		< 0.0001
Daily Intensity (DDD)
Non-Statins use (Ref)	1.00	-	1.00	-	1.00	-
< 1 DDD	0.39 (0.26, 0.59)	< 0.0001	0.37 (0.24, 0.56)	< 0.0001	0.41 (0.27, 0.61)	< 0.0001
≥ 1 DDD	0.24 (0.09, 0.65)	0.0052	0.28 (0.10, 0.75)	0.0120	0.29 (0.11, 0.77)	0.0132
P for trend			< 0.0001		< 0.0001

Abbreviations: CI Confidence interval, aHR Adjusted hazard ratio, HR hazard ratio, cDDDs cumulative defined daily doses, DDD defined daily doses, ref reference group, Q quartiles, NA not estimable (zero events)

^aA time-varying Cox proportional hazards model was employed, treating statin use as a dynamic exposure. The model was adjusted for multiple covariates, including age, sex, income level, degree of urbanization, duration of inflammatory bowel disease, coexisting comorbidities, family history of gastrointestinal malignancy, lifestyle factors, and medication use

^bThe Fine and Gray subdistribution hazard model was used to estimate the risk of colorectal cancer while accounting for death as a competing risk

Restricted cubic spline regression revealed a nonlinear inverse association between both cumulative and daily statin exposure and the risk of colorectal cancer. As shown in Supplemental Figs. 5 and 6, higher cDDD and greater average daily intensity (DDD/day) were each associated with a progressively lower hazard of colorectal cancer. However, the magnitude of risk reduction began to attenuate at higher levels of exposure, suggesting a plateau effect in the dose–response curve.

Incidence rates and risk ratios for colorectal cancer

During a median follow-up of 10.2 years in statin users and 9.1 years in the non-statin group, the incidence of colorectal cancer was 15.7 per 10,000 person-years in the statin group versus 45.9 in the non-statin group (Table 4). The incidence rate ratio (IRR) was 0.34 (95% CI, 0.23 to 0.51; P < 0.0001). Rosuvastatin users had the lowest incidence rate (8.0 per 10,000 person-years; IRR, 0.18; 95% CI, 0.06 to 0.55), followed by simvastatin (IRR, 0.32) and atorvastatin (IRR, 0.43). The IRRs decreased progressively across statin cDDD quartiles and daily DDD intensity categories, with the highest statin dose group (cDDD Q4) demonstrating an IRR of 0.09 (95% CI, 0.02 to 0.36; P = 0.0007).

Table 4.

Incidence rates and incidence rate ratios of colorectal cancer in patients with inflammatory bowel disease: comparison of statin and non-statin lipid-lowering therapies

	Events	Person-years	IR (per 10,000 person-year)	IRR	95% CI for IRR	P *
Statins use
Non-Statins use	133	28,997.2	45.87	Ref.
Statins use	31	19,689.0	15.74	0.34	(0.23, 0.51)	< 0.0001
Statins use
Non-Statins use	133	28,997.2	45.87	Ref.
Hydrophilic statins
Rosuvastatin	3	3,736.0	8.03	0.18	(0.06, 0.55)	0.0028
Pravastatin	2	1,563.4	12.79	0.28	(0.07, 1.13)	0.0731
Lipophilic statins
Atorvastatin	14	7,176.9	19.51	0.43	(0.25, 0.74)	0.0023
Lovastatin	3	1,221.6	24.56	0.54	(0.17, 1.68)	0.2846
Simvastatin	6	4,146.3	14.47	0.32	(0.14, 0.71)	0.0057
Fluvastatin	3	1,629.0	18.42	0.40	(0.13, 1.26)	0.1181
Pitavastatin	NA	NA	NA	NA	NA	NA
Statins use
Non-Statins use	133	28,997.2	45.87	Ref.
Lipophilic statins	26	14,700.6	17.69	0.39	(0.25, 0.59)	< 0.0001
Hydrophilic statins	5	4,988.4	10.02	0.22	(0.09, 0.53)	0.0008
Statins use (cDDD)
Non-Statins use	133	28,997.2	45.87	Ref.
cDDD, Q1	11	5,067.7	21.71	0.47	(0.26, 0.88)	0.0171
cDDD, Q2	11	5,092.9	21.60	0.47	(0.25, 0.87)	0.0164
cDDD, Q3	7	4,638.5	15.09	0.33	(0.15, 0.7)	0.0041
cDDD, Q4	2	4,889.9	4.09	0.09	(0.02, 0.36)	0.0007
Statins use (daily density, DDD)
Non-Statins use	133	28,997.2	45.87	Ref.
< 1	27	15,956.5	16.92	0.37	(0.24, 0.56)	< 0.0001
≥ 1	4	3,732.5	10.72	0.23	(0.09, 0.63)	0.0042

Abbreviations: CI Confidence interval, aHR Adjusted hazard ratio, HR hazard ratio, cDDDs cumulative defined daily doses, DDD defined daily doses, ref. reference group, Q quartiles, IR Incidence Rate, IRR Incidence Rate Ratio

^*The IRR was calculated using Poisson regression models

Discussion

In this nationwide IBD cohort, sustained statin use was significantly associated with a lower relative risk of CRC (aHR, 0.35; 95% CI, 0.24–0.52; P < 0.0001; Table 2). The 10-year cumulative CRC incidence of approximately 5.0% in our non-statin cohort aligns well with contemporary meta-analyses [32, 33]. While the CRC risk in the first decade following an initial IBD diagnosis is typically low (< 1%), our study cohort represents an older population initiating statin therapy. Consequently, our 10-year follow-up period captures a later window in their overall IBD disease course, making a 5.0% cumulative incidence epidemiologically consistent with the risk expected in patients with longstanding IBD (e.g., > 20 years of disease duration) [32, 33]. This association was consistent across time-varying Cox models and competing risk analyses, and it persisted even among patients without baseline comorbidities (CCI = 0), which reduces (but does not eliminate) concerns about confounding by indication and healthy user/healthy adherer bias. We acknowledge that the consistency of risk reduction across both CRC and all-cause mortality could reflect residual healthy-user bias; however, the observed dose–response gradient and the distinct heterogeneity among statin types (e.g., rosuvastatin vs. pravastatin) argue against a uniform selection effect. Furthermore, from a competing-risk perspective, reduced mortality typically increases the cumulative risk of cancer by prolonging the time-at-risk for diagnosis. The fact that statins reduced CRC incidence despite significantly improving survival (thereby retaining more high-risk elderly patients in the cohort) suggests that the observed association is not readily explained by selective survival alone. Nevertheless, unmeasured health-seeking behaviors—particularly surveillance practices—may still differ between groups; therefore, our findings should be interpreted as associative rather than causal. To assess potential surveillance bias, we conducted a claims-based sensitivity analysis stratified by colonoscopy surveillance frequency (regular every 1–2 years vs. irregular/none). The inverse association remained directionally consistent across strata (Supplemental Table 3), although residual surveillance bias cannot be excluded. We chose an active comparator (non-statin lipid-lowering therapy) rather than untreated IBD patients because untreated comparators are typically more susceptible to confounding by healthcare utilization and preventive behaviors (including surveillance intensity), which can amplify healthy-user bias. Non-statin lipid-lowering agents (e.g., fibrates, ezetimibe, bile acid sequestrants) were selected as a pragmatic comparator because prior meta-analyses and randomized trials indicate that these agents exert a neutral effect on colorectal cancer risk [24–28], serving as an ideal “cancer-neutral” reference. Their users share similar medication-refill behaviors and healthcare utilization patterns, thereby improving comparability and allowing us to better isolate the specific chemopreventive signal of statins. Under Taiwan’s risk-stratified reimbursement framework during 2008–2019, statins were universally recommended as first-line therapy for hypercholesterolemia, while non-statin agents were prescribed for specific indications (e.g., severe hypertriglyceridemia or bile acid malabsorption in IBD) [17]. Baseline factors such as BMI or smoking influenced treatment intensity targets but did not typically determine the choice between statin and non-statin agents [18, 19]. This structure minimizes channeling bias driven by cardiometabolic risk profiles. Importantly, in patients with IBD, bile acid sequestrants are frequently prescribed for bile acid malabsorption–associated diarrhea rather than dyslipidemia, which likely contributes to the relatively high prevalence of non-statin use observed prior to matching. We cannot exclude the possibility of residual comparator-related bias if these agents exert independent effects; therefore, the observed association should be interpreted cautiously as hypothesis-generating. Finally, although we incorporated claims-based proxies for IBD phenotype/treatment and colonoscopy utilization, NHIRD cannot capture endoscopic extent, inflammatory activity/biomarkers, or colonoscopy findings; thus, residual confounding and healthy-adherer/selection bias may remain. Using time-updated modeling, registry-confirmed CRC outcomes, and spline-based dose–response analysis (Table 3; Supplemental Figs. 1–6), we observed a nonlinear inverse association that plateaued at higher exposure levels. Notably, the magnitude of the observed associations in the highest exposure stratum (aHR ≈ 0.08) is larger than typically seen in observational pharmacoepidemiology. While this pronounced risk reduction confirms a strong dose–response signal (P for trend < 0.0001), the numerical estimate likely reflects an amplification of the biological effect by cumulative selection processes and survival bias among long-term adherers. Therefore, these estimates should be viewed as a proof-of-concept for chemopreventive potential rather than a precise prediction of clinical efficacy. Accordingly, these results should be interpreted as hypothesis-generating and require confirmation in future studies, ideally randomized trials.

Although prior studies have assessed statins and CRC risk in the general population, findings in IBD patients have been inconsistent [11–15]. While Ananthakrishnan et al. reported a protective effect, population-based studies in Canada and the U.S. found no association after adjustment [34] The Manitoba cohort showed a modest, non-significant reduction, and some analyses suggested increased risk [15]. These discrepancies likely reflect limitations in exposure assessment and absence of time-updated modeling [15, 34]. A Spanish case–control study reported a dose-dependent benefit but lacked spline modeling and was not IBD-specific [35]. In contrast, our study used a nationwide IBD cohort, registry-confirmed outcomes, and time-dependent modeling with quarterly statin updates. We conducted the first spline-based dose–response analysis in this population, showing a consistent inverse association with a plateau at higher exposures, supporting statins as a potential chemopreventive strategy.

The association between statin use and reduced CRC risk in IBD is biologically plausible. Chronic intestinal inflammation promotes colitis-associated carcinogenesis [36]. Statins exert systemic and local anti-inflammatory effects, modulating cytokines and immune responses that may disrupt the inflammation–dysplasia–carcinoma sequence [37]. They also promote apoptosis, inhibit proliferation, and block oncogenic signaling [38]. Simvastatin has reduced tumor burden in colitis-associated cancer models. Statins may influence the BMP pathway, frequently dysregulated in CRC, and potentially restore gut microbial balance [39–41]. IBD-associated CRC involves early p53 mutations and chromosomal instability, which statins may counteract [42, 43]. The dose–response patterns observed across subgroups in our study are consistent with a potential biological effect, although residual confounding cannot be excluded.

Our findings suggest differential CRC risk reduction among statins in IBD. Rosuvastatin and simvastatin were significantly protective, while pravastatin and fluvastatin showed weaker or non-significant effects. This heterogeneity may reflect differences in chemical structure, pharmacokinetics, and tissue distribution; however, it may also be attributable to chance given the relatively small and unstable sample sizes in certain statin-specific subgroups. Although lipophilic statins like simvastatin may better penetrate colonic tissue, rosuvastatin’s strong associations may suggest lipophilicity alone is not explanatory [35]. Prior studies and pharmacologic data highlight the importance of drug potency, systemic exposure, and pleiotropic effects, including BMP pathway modulation [35, 44, 45]. Preclinical evidence suggests rosuvastatin may activate tumor-suppressive mechanisms beyond cholesterol lowering [35]. These findings support the need to evaluate individual statins in IBD for optimized chemoprevention.

The observed dose–response relationship—linking higher cumulative exposure and greater daily statin intensity to lower CRC risk is consistent with a potential chemopreventive association; however, the unusually large effect estimates—particularly in the highest exposure stratum—raise concern for residual confounding or selection effects and do not establish causality. Because exposure intensity and cumulative dose were derived from dispensing/claims data (i.e., refill-based measures such as MPR), they may not fully reflect actual medication intake; therefore, the dose–response patterns should be interpreted cautiously. Similar trends were noted in general population studies by Lee et al. and others [46, 47]. Rodriguez-Miguel et al. reported reduced CRC risk with moderate-to-high intensity statin use, particularly simvastatin and rosuvastatin [35]. In IBD-specific studies, findings have been mixed: Shah et al. found no association [48], while Ananthakrishnan et al. reported a protective effect without dose–response analysis [11–15, 34]. Our time-updated spline-based models demonstrate a graded inverse association, with a plateau at higher doses suggesting biological saturation or selection bias [35, 40]. However, as Lochhead and Chan emphasized, prospective IBD-specific trials are needed to define optimal regimens [20].

This study provides novel evidence that sustained statin use—particularly at higher cumulative doses and daily intensity—is associated with a significantly lower CRC risk in IBD patients. These findings may complement current prevention strategies that rely heavily on colonoscopic surveillance, which remains essential, especially in high-risk subgroups [34]. Given the established cardiovascular benefits and safety of statins, their potential dual role in IBD is noteworthy—particularly for patients with overlapping cardiometabolic risk [34]. Rosuvastatin and simvastatin demonstrated the strongest protective effects, highlighting the importance of agent selection and dosing. While statins are not yet recommended solely for CRC prevention in IBD, our results should not guide statin initiation for cancer prevention, but may inform hypothesis generation and risk–benefit considerations in patients already meeting cardiovascular indications [34]. Future trials should evaluate whether specific statins or higher-intensity regimens confer additional benefit without harm. Until then, statins should be prescribed for cardiovascular indications, with CRC risk reduction as a potential secondary benefit. Despite the observational nature of this study, its large sample size, time-updated exposure modeling, and rigorous confounder control support the need for randomized trials in this setting.

This study has several limitations. First, as an observational design, causality cannot be established. The unusually large effect sizes observed in some strata (e.g., the highest cumulative exposure group) further increase concern for residual confounding and selection bias, despite extensive adjustment and sensitivity analyses. While we applied time-dependent Cox models, PSM, and adjusted for major confounders, residual bias from unmeasured or incompletely captured factors (e.g., BMI (quantitative BMI is unavailable and obesity codes are an imperfect proxy), dietary factors, true medication adherence, first-degree CRC family history, endoscopic extent/severity of IBD, inflammatory biomarkers, and surveillance patterns beyond what can be inferred from claims procedure codes) may persist. Notably, the very low recorded prevalence of family history of gastrointestinal malignancy in claims data suggests under-ascertainment, and NHIRD cannot reliably capture a first-degree CRC family history, which may contribute to residual confounding. Although we stratified analyses by colonoscopy surveillance frequency using procedure claims, this approach cannot fully capture specific indications (symptomatic vs. screening), procedural quality (e.g., bowel preparation), or findings (e.g., polyp detection rate); therefore, we cannot exclude the possibility that residual surveillance bias may remain. Nevertheless, the strong, consistent, and dose-dependent association observed supports the rationale for conducting future randomized trials. Second, misclassification of exposure or outcome may occur in claims data. We attempted to address this by requiring MPR > 80%, using cDDD thresholds, and confirming CRC diagnoses through the Taiwan Cancer Registry. However, OTC drug use and true adherence remain unmeasured. While MPR reflects prescription refill behavior rather than confirmed ingestion, sustained refilling over years—particularly in the highest exposure strata—strongly implies actual medication intake, as it is unlikely that patients would maintain high refill adherence without consumption. Furthermore, regarding the potential for survivor bias (where higher cumulative doses imply longer survival), we employed time-dependent modeling to prevent immortal time bias; however, residual healthy-survivor selection bias (where long-term adherers represent an inherently healthier subset) may still contribute to the magnitude of the observed benefit. Third, residual confounding, including healthy user/healthy adherer bias, is a potential concern if statin users are inherently healthier, and this may persist even with an active-comparator design; the consistent effects observed in patients with low comorbidity (CCI = 0) help mitigate this concern but do not eliminate it. In addition, NHIRD lacks lipid values, formal cardiovascular risk scores, and explicit prescribing indications; therefore, we could not distinguish primary versus secondary cardiovascular prevention, which may contribute to selection bias. Effect estimates could differ under an untreated comparator; we did not use untreated IBD patients by design because such comparisons are typically more vulnerable to healthy-user bias and confounding by healthcare utilization. Because NHIRD does not provide lipid measurements, we could not directly adjust for baseline lipid levels; however, lipids are not expected to lie on the causal pathway for CRC. Finally, generalizability may be limited to systems with similar access to statins and surveillance programs. Nonetheless, the biological and pharmacological consistency of statins supports the broader applicability of these findings.

Conclusions

In this nationwide IBD cohort, sustained statin use was linked to a significantly lower risk of colorectal cancer, with consistent effects across subgroups and a clear dose–response relationship that plateaued at higher exposure. While causality cannot be confirmed, the alignment of statistical strength and biological plausibility offers a compelling rationale for further investigation. Statins may serve not only as cardiovascular agents, but as silent modulators of cancer risk in chronic inflammatory disease. Among statin types, rosuvastatin and simvastatin demonstrated the strongest protective associations and may warrant prioritization in future chemoprevention trials. What begins as an observational signal here may one day inform intentional prevention—if prospective studies confirm what this study now suggests.

Abbreviations

CRC

Colorectal cancer

IBD

Inflammatory bowel disease

UC

Ulcerative colitis

CD

Crohn’s disease

NHIRD

National Health Insurance Research Database

PSM

Propensity score matching

MPR

Medication possession ratio

cDDD

Cumulative defined daily dose

DDD

Defined daily dose

ICD-9-CM

International Classification of Diseases, Ninth Revision, Clinical Modification

ICD-10-CM

International Classification of Diseases, Tenth Revision, Clinical Modification

TCRD

Taiwan Cancer Registry Database

aHR

adjusted Hazard Ratio

CI

Confidence interval

Q1-Q4

Quartiles 1–4

COPD

Chronic obstructive pulmonary disease

NSAIDs

Nonsteroidal anti-inflammatory drugs

WHO

World Health Organization

IRR

Incidence rate ratio

CCI

Charlson Comorbidity Index

Authors’ contributions

Conception and Design: Yan-Jiun Huang; Jui-An Lin; Wan-Ming Chen; Ben-Chang Shia; Szu-Yuan WuFinancial Support: Lo-Hsu Medical Foundation, LotungPoh-Ai Hospital, supports Szu-Yuan Wu’s work (Funding Number: 10908, 10909, 11001, 11002, 11003, 11006, and 11013).Collection and Assembly of Data: Szu-Yuan Wu; Yan-Jiun Huang; Jui-An LinData Analysis and Interpretation: Yan-Jiun Huang; Jui-An Lin; Wang-Ming ChenAdministrative Support: Szu-Yuan Wu*Manuscript Writing: Yan-Jiun Huang; Jui-An Lin; Szu-Yuan WuFinal Approval of Manuscript: All authors.

Funding

Lo-Hsu Medical Foundation, LotungPoh-Ai Hospital, supports Szu-Yuan Wu’s work (Funding Number: 11403, 11404).

Data availability

The data sets supporting the study conclusions are included in the manuscript. We used data from the National Health Insurance Research Database and Taiwan Cancer Registry database. The authors confirm that, for approved reasons, some access restrictions apply to the data underlying the findings. The data used in this study cannot be made available in the manuscript, the supplemental files, or in a public repository due to the Personal Information Protection Act executed by Taiwan’s government, starting in 2012. Requests for data can be sent as a formal proposal to obtain approval from the ethics review committee of the appropriate governmental department in Taiwan. Specifically, links regarding contact info for which data requests may be sent to are as follows: [http://nhird.nhri.org.tw/en/Data_Subsets.html#S3] (http://nhis.nhri.org.tw/point.html) and [http://nhis.nhri.org.tw/point.html] (http://nhis.nhri.org.tw/point.html) . Informed consent was waived because the data sets are covered under the Personal Information Protection Act. Szu-Yuan Wu, MD, PhD had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. The data has not been previously presented orally or by poster at scientific meetings.

Declarations

Ethics approval approval and consent to participate

The study protocols were reviewed and approved by the Institutional Review Board of Tzu-Chi Medical Foundation (IRB109-015-B), and the study was conducted in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.