Association Between Statin Use and Non-Melanoma Skin Cancer Risk: A Distributed Network Analysis of 11 Real-World Databases

Abstract

Background

Given the dual potential of statins to act as both carcinogenic and antineoplastic agents, evidence from previous investigations into statin use and the risk of nonmelanoma skin cancer (NMSC) remains inconclusive.

Objective

To demonstrate the impact of statin exposure on the NMSC risk in Korean patients.

Methods

We carried out a multicenter cohort study based on electronic health record data aggregated from 11 Korean hospitals. Patients with hyperlipidemia who had been taking antilipidemic agents, including statins were included in this study. We applied 1:1 propensity score matching to create balanced cohorts and used Cox regression to assess the hazard ratio (HR) for NMSC.

Results

A total of 18,579 statin users and a matched number of non-users were included across 11 databases. Statin use showed no significant association with an increased risk of NMSC (HR, 1.03; 95% confidence interval, 0.75–1.43). Subgroup analyses of drug exposure, age, and sex depicted no significant HR trends.

Conclusion

Statin exposure demonstrated no significant association with NMSC development in Korean patients with hyperlipidemia.

INTRODUCTION

Statins, or 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors, are clinically prescribed antilipidemic agents that have recently been used to manage high cholesterol levels and atherosclerosis. Statin are well-known for their phototoxicity potential, and several biological mechanisms have been proposed to elucidate the apparent association between statin use and an increased risk of nonmelanoma skin cancer (NMSC). For instance, statins are reported to increase the sensitivity of human skin cells to UV radiation mediated by the formation of photoproducts¹. Certain statins have also been shown to enhance ultraviolet A-induced cellular damage in low-cholesterol environments². However, statins appear to decrease the NMSC risk by regulating the Rapidly Accelerated Fibrosarcoma-Mitogen-Activated Protein/Extracellular Signal-Regulated Kinase-Extracellular Signal-Regulated Kinase pathway, which induces the apoptosis of keratinocytes³. Additionally, statins show regulatory effects of Akt blockage, thereby stimulating cell apoptosis in human squamous cell carcinoma lines⁴,⁵. Alternatively, statins may prevent the development of basal cell carcinoma (BCC) by inhibiting the Hedgehog signaling pathway⁶.

Given these conflicting mechanisms, the carcinogenic potential of statins may be hampered by their protective effects. The increase in NMSC risk associated with statin use has been explored in several studies, yielding mixed results. Prior studies have indicated that the relationship between statin use and increased risk of NMSC remains a subject of ongoing debate⁷,⁸,⁹, although several studies have reported a significant association¹⁰,¹¹. The conflicting findings in the literature regarding statin-associated NMSC prompted us to investigate whether statin exposure increases the risk of NMSC in Korean patients with hyperlipidemia, which is what we aimed to achieve via this study.

MATERIALS AND METHODS

Data sources

We conducted a retrospective, multicenter, new user cohort study, based on the real-world clinical data from 11 secondary or tertiary hospitals in the Republic of Korea. The data were analyzed from a total of 14,474,239 patients and transformed into the Observational Medical Outcomes Partnership Common Data Model (OMOP-CDM), version 5.3.1, to enable standardized analysis across institutions. Local codes for diagnoses, medications, and procedures from these hospitals were converted to the standard OMOP vocabulary, thereby facilitating the integration of disparate electronic health records¹².

The electronic health records from the following institutions were converted into CDM data for the distributed research network and included in this study: (1) Kyung Hee University Hospital at Gangdong (KHNMC; 887,370 patients; Jun 2006–Mar 2023), (2) Kyung Hee University Medical Center (KHMC; 1,168,640 patients; Jan 2008–Feb 2022), (3) Ewha Womans University Medical Center (EUMC; 1,667,671 patients; Jan 2001–Dec 2021), (4) Pusan National University Hospital (PNUH; 1,753,001 patients; Feb 2011–Aug 2019), (5) Ajou University Medical Center (AUMC; 2,873,443 patients; Jan 1994–Feb 2022), (6) Kangdong Sacred Heart Hospital (KDH; 1,724,052 patients; Oct 1986–Dec 2019), (7) Kangwon National University Hospital (KWMC; 567,439 patients; Jan 2003–Jan 2022), (8) Daegu Catholic University Medical Center (DCMC; 906,587 patients; Jan 2005–Oct 2021), (9) Soonchunhyang University Seoul Hospital (SCHSU; 1,098,041 patients; May 2003–May 2021), (10) Soonchunhyang University Bucheon Hospital (SCHBC; 940,767 patients; Feb 2001–Sep 2019), and (11) Soonchunhyang University Cheonan Hospital (SCHCA; 887,228 patients; Jun 2006–Sep 2019).

This study was approved by the Institutional Review Board (IRB) of Kyung Hee University Hospital, Gangdong (IRB No. KHNMC 2024-10-001).

Study design

Patients aged ≥40 years who were newly prescribed lipid-lowering agents were included in this study. The study drug included statins approved for use in Korea: atorvastatin (concept identification [ID] #1545958), cilastatin (#1797258), fluvastatin (#1549686), nystatin (#922570), pitavastatin (#40165636), pravastatin (#1551860), rosuvastatin (#1510813), simvastatin (#1539403), somatostatin (#19136066), and ulinastatin (#43009071). To ensure consistency in the definition of “lipid-lowering agents except statins,” we only included acipimox (#19029644), bezafibrate (#19022956), bile acid sequestrants (#21601875), ethyl linoleate (#37489441), evolocumab (#46287466), ezetimibe (#1526475), fenofibrate (#1551803), fibrates (#21601864), nicotinic acid and derivatives (#21601880), and proprotein convertase subtilisin/kexin type 9 inhibitor (#46287792), based on Korea pharmaceutical information center.

The index date was defined as the date of the initial exposure of any lipid-lowering agents. The target cohort included patients diagnosed with hyperlipidemia (concept ID # 432867) who met all of the following criteria: (1) initial prescription of statins, (2) continuous statin use at any dose for ≥90 days, (3) an observation period of ≥30 days prior to the index date, and (4) aged ≥40 years. The comparator cohort included patients diagnosed with hyperlipidemia who met all of the following criteria: (1) initial prescription of lipid-lowering agents except statins, (2) continuous use of lipid-lowering agents except statins at any dose for ≥90 days, (3) an observation period of ≥30 days prior to the index date, and (4) aged ≥40 years. Exclusion criteria for both the target and comparator cohorts included: (1) absence of hyperlipidemia diagnosis or (2) any history of primary skin malignancy before the index date. The time at risk began 90 days following the index date and ended at the date of last follow-up.

The primary endpoints were the incidence rate and hazard ratio (HR) for non-melanoma skin cancer (NMSC) observed during the risk period. NMSC diagnosis was identified using the International Classification of Diseases (ICD)-10 code C44, which includes malignant neoplasm of the skin (concept ID #4155297), squamous cell carcinoma in situ of the skin (#4112745), and Bowen's disease of the skin (#44505397) while excluding malignant melanoma (#4162276).

To assess the robustness of our finding, three subgroup analyses were conducted: 1) duration of drug exposure (prescribed the study drug at any dose 90–364 days and ≥365 days), 2) age (aged 40–64 years and ≥65 years), and 3) sex (male and female).

Statistical analysis

We performed a cohort study using ATLAS version 2.7.6, an analysis platform that provides an open-source cohort method R package developed by Observational Health Data Sciences and Informatics. The analyses were conducted using Federated Electronic and Data Exchange for Research-Network, a Korean nationwide health data platform based on the OMOP-CDM. Large-scale 1:1 propensity score matching (PSM) was performed to balance the two cohorts. Propensity scores were computed through binary logistic regression, aggregating all baseline confounders into a single score to efficiently match patients from each group¹³. Greedy matching was performed to match patients using a caliper width equal to 0.2 of the standard deviation of the logit of the propensity score. Overall, the Cox regression model was employed to estimate the HR of NMSC between statin users and non-users, and the Kaplan–Meier plot was employed to illustrate the cumulative hazard of NMSC in statin users and non-users. The incidence rate of NMSC was calculated as the number of NMSC cases per 1,000 person-years (PY) of follow-up. A p-value of<0.05 was considered statistically significant. All statistical analyses were conducted using R software (version 3.6.1; R Foundation of Statistical Computing, Vienna, Austria).

RESULTS

Study population

A total of 82,474 statin-treated and 20,794 non-statin-treated patients with hyperlipidemia were identified from 11 databases and included in the present study. After PSM, baseline covariates–including age group, sex, comorbidities, and medications—were balanced across 11 databases, with most standardized mean differences falling below 0.1 (Supplementary Fig. 1). The aggregated baseline characteristics of the two groups and their matched controls are shown in Table 1. In total, 18,579 patients treated with statins (111,417 PY) and 18,579 patients treated with lipid-lowering agents other than statins (110,568 PY) were pooled, as depicted in Fig. 1.

Table 1. Baseline demographics of the study population.

Variables		Stain users	Statin non-users
All		18,579	18,579
Age group (yr)
	40–44	1,375 (7.4)	1,451 (7.8)
	45–49	2,116 (11.4)	2,067 (11.1)
	50–54	2,968 (16.0)	3,015 (16.2)
	55–59	3,010 (16.2)	2,954 (15.9)
	60–64	3,140 (16.9)	3,189 (17.1)
	65–69	2,280 (12.3)	2,306 (12.4)
	70–74	1,502 (8.1)	1,510 (8.1)
	75–79	1,208 (6.5)	1,169 (6.3)
	80–84	629 (3.4)	602 (3.2)
	85–89	294 (1.6)	267 (1.4)
	90–94	57 (0.3)	49 (0.3)
	95–99	0	0
Sex
	Female	9,800 (52.7)	9,889 (53.2)
	Male	8,779 (47.3)	8,690 (46.8)
Medical history
	Chronic liver disease	299 (1.6)	142 (0.8)
	Diabetes mellitus	3,686 (19.8)	3,685 (19.8)
	Hypertensive disorder	7,649 (41.2)	7,564 (40.7)
	Obesity	138 (0.7)	164 (0.9)
	Osteoarthritis	322 (1.7)	345 (1.9)
	Renal impairment	530 (2.9)	565 (3.0)
	Rheumatoid arthritis	69 (0.4)	70 (0.4)
	Cerebrovascular disease	1,709 (9.2)	1,757 (9.5)
	Ischemic heart disease	1,837 (9.9)	1,783 (9.6)
	Malignant neoplastic disease	378 (2.0)	434 (2.3)
Medications
	Agents acting on the renin-angiotensin system	6,452 (34.7)	6,342 (34.1)
	Anti-inflammatory and anti-rheumatic agents	2,257 (12.1)	2,549 (13.7)
	Anti-neoplastic agents	627 (3.4)	731 (3.9)
	Beta blocking agents	2,705 (14.6)	2,754 (14.8)
	Calcium channel blockers	4,604 (24.8)	4,701 (25.3)
	Diuretics	2,653 (14.3)	2,594 (14.0)
	Anti-diabetic agents	5,067 (27.3)	4,985 (26.8)
	Immunosuppressants	338 (1.8)	376 (2.0)
	Lipid modifying agents	84 (0.5)	81 (0.4)

Values are presented as number (%).

Fig. 1. Flowchart showing the procedures conducted to select a cohort of 1:1 propensity score-matched-pairs of statin users and never users from the Observational Medical Outcomes Partnership-Common Data Model database.

Primary outcomes

Table 2 shows The HRs for NMSCs following statin exposure across individual databases, none of which demonstrated a statistically significant differences in NMSC risk between statin users and non-users. After aggregating the results from the 11 databases, the incidence rate of NMSC was 0.68 per 1,000 PY among statin users, and 0.66 per 1,000 PY among non-users. The pooled analysis of HRs across the 11 databases revealed no significant relationship between statin use and the increased NMSC risk (HR, 1.03; 95% confidence interval [CI], 0.75–1.43), as reflected in the cumulative hazard curves of NMSC (Fig. 2).

Table 2. Incidence and hazard ratio of non-melanoma skin cancer according to statin exposure across 11 databases.

Variables	Statin users				Statin non-users				Hazard ratio	p-value
	Patients	Cases	Time at risk (PY)	Incidence (per 1,000 PY)	Patients	Cases	Time at risk (PY)	Incidence (per 1,000 PY)	(95% confident interval)
KHNMC	2,372	14	14,724	0.95	2,372	11	14,514	0.76	1.00 (0.39–2.56)	1.00
KHMC	4,643	15	25,324	0.59	4,643	14	24,983	0.56	1.12 (0.43–3.00)	0.81
EUMC	500	3	2,641	1.14	500	4	2,710	1.48	1.00 (0.12–8.33)	1.00
PNUH	726	2	2,219	0.90	726	4	2,235	1.79	0.50 (0.02–5.22)	0.62
AUMC	1,718	4	13,110	0.31	1,718	5	12,878	0.39	1.50 (0.25–11.39)	0.68
KDH	1,016	3	8,109	0.37	1,016	3	8,172	0.37	1.50 (0.25–11.39)	0.68
KWMC	1,048	8	6,913	1.16	1,048	7	6,752	1.04	1.00 (0.31–3.20)	1.00
DCMC	2,207	15	13,490	1.11	2,207	11	13,734	0.80	1.38 (0.56–3.55)	0.50
SCHSU	1,516	6	8,446	0.71	1,516	4	8,377	0.48	3.00 (0.38–60.62)	0.39
SCHBC	1,923	6	12,178	0.49	1,923	8	11,823	0.68	1.00 (0.28–3.60)	1.00
SCHCA	910	0	4,262	0	910	2	4,389	0.46	0.14
Total	18,579	76	111,417	0.68	18,579	73	110,568	0.66	1.03 (0.75–1.43)	0.841

PY: person-year, KHNMC: Kyung Hee University Hospital at Gangdong, KHMC: Kyung Hee University Medical Center, EUMC: Ewha Womans University Medical Center, PNUH: Pusan National University Hospital, AUMC: Ajou University Medical Center, KDH: Kangdong Sacred Heart Hospital, KWMC: Kangwon National University Hospital, DCMC: Daegu Catholic University Medical Center, SCHSU: Soonchunhyang University Seoul Hospital, SCHBC: Soonchunhyang University Bucheon Hospital, SCHCA: Soonchunhyang University Cheonan Hospital.

Fig. 2. Kaplan-Meier plot comparing cumulative hazard of nonmelanoma skin cancer among statin users and non-users.

HR; hazard ratio, CI; confidence interval.

Subgroup outcomes

Table 3 presents the results of the subgroup analyses of NMSC risk according to three factors: (1) duration of drug exposure, (2) age, and (3) sex (Table 3). Stratification by duration of drug exposure and age yielded no statistically significant patterns in HR (p>0.05). The corresponding HRs were as follows; (1) 90–364 days (HR, 1.07; 95% CI, 0.62–1.84), (2) ≥365 days (HR, 0.74; 95% CI, 0.48–1.12), (3) aged 40–64 years (HR, 0.95; 95% CI, 0.59–1.55), and (4) aged ≥65 years (HR, 1.02; 95% CI, 0.66–1.57). Similarly, sex-specific analysis noted the tendency towards a lower HR in the statin users, but no clear association with an increased risk of NSMC was observed (male: HR, 0.95; 95% CI, 0.55–1.64; female: HR, 0.86; 95% CI, 0.57–1.27).

Table 3. Incidence and hazard ratio of non-melanoma skin cancer in statin users and non-users stratified by subgroup.

		Statin users				Statin never users				Hazard ratio
Variables		Patients	Cases	Time at risk (PY)	Incidence (per 1,000 PY)	Patients	Cases	Time at risk (PY)	Incidence (per 1,000 PY)	(95% confident interval)	p-value
All		18,579	76	111,417	0.68	18,579	73	110,568	0.66	1.03 (0.75–1.43)	0.841
	<365	9,302	27	49,016	0.55	9,302	25	48,165	0.52	1.07 (0.62–1.84)	0.811
	≥365	9,286	38	62,009	0.61	9,286	52	62,285	0.84	0.74 (0.48–1.12)	0.151
Age (yr)
	40–64	12,560	32	80,773	0.40	12,560	32	79,370	0.42	0.95 (0.59–1.55)	0.839
	≥65	6,143	40	31,117	1.29	6,143	41	31,888	1.29	1.02 (0.66–1.57)	0.938
Sex
	Male	8,800	25	51,776	0.48	8,800	26	51,130	0.51	0.95 (0.55–1.64)	0.854
	Female	9,913	45	60,804	0.74	9,913	52	60,211	0.86	0.86 (0.57–1.27)	0.441

PY: person-year.

DISCUSSION

Within this multicenter cohort study, no significant link was observed between statin use and the NMSC risk. Moreover, the risk estimates for NMSC did not differ according to the duration of drug exposure, age, or sex.

Several biological pathways have been suggested to explain the potential association between statin exposure and NMSC risk, though none are explicit. However, several studies have reported a statistically significant link between satin treatment and the risk of NMSC. The Women’s Health Initiative study identified a statistically significant increase in NMSC risk among women using statins, (odd ratio [OR], 1.21; 95% CI, 1.07–1.35), with the association being particularly pronounced for lipophilic statins, showing OR of 1.39¹⁴. Additionally, a population-based study from Iceland reported that statin exposure increased the risk of squamous cell carcinoma (SCC), however not BCC, in a low UV environment¹⁵. In a subgroup analysis, the risk of invasive SCC and SCC in situ significantly increased in statin-exposed patients aged >60 years (invasive SCC: adjusted odd ratio [aOR], 1.29; 95% CI, 1.11–1.50; SCC in situ: aOR, 1.43; 95% CI, 1.24–1.64).

In contrast to these findings, a large meta-analysis by Li et al.¹⁶ reported that the use of statin did not significantly increase NMSC risk (relative risk, 1.03; 95% CI, 0.90–1.19). Similarly, a nationwide study in Denmark found no clear increase in NMSC risk associated with statin use, regardless of histological subtype (BCC: aOR, 1.09; 95% CI, 1.06–1.13; SCC: aOR, 1.01; 95% CI, 0.91–1.11)¹⁷. A prospective cohort study also showed no elevated risk of SCC in relation to statin use (SCC: HR, 0.89; 95% CI, 0.66–1.19), whereas statin use materially decreased the risk of BCC (BCC: HR, 0.89; 95% CI, 0.79–0.996)¹⁸. Overall, our findings, which indicate no statistically significant statin-related increase in risk of NMSC, align with most prior studies.

This study has several strengths. Based on our literature review, this is the first and largest population-based cohort study to investigate association between statin exposure and NMSC risk in Korea. Given that the study population predominantly consisted of Korean individuals, the findings are likely to be representative of the Korean population and may also be relevant to Asian individuals with Fitzpatrick skin types III and IV. In addition, we extracted a sufficiently large cohort groups using OMOP-CDM, a large nationwide population-based database, which provided sufficient statistical power for the study. As antihyperlipidemic drugs can only be obtained by prescription in Korea (Ethical Therapeutic Category), this minimized the attrition bias of the findings.

Our study has some potential limitations. First, we were unable to distinguish between SCC and BCC within NMSC. The OMOP-CDM does not contain pathological data, and the ICD-10 system does not distinguish NMSCs by pathological subtype. Second, we lacked individual-level data on UV exposure, which is major risk factor for NSMC. The OMOP-CDM does not contain information on personal sun exposure patterns such as occupational or recreational sun exposure, sunscreen use, or sun protection behaviors. Therefore, residual confounding due to unmeasured individual UV exposure cannot be excluded in the study. Third, this study did not include sub-analyses based on statin subtypes. A recent meta-analysis by Wang et al.¹⁹ revealed that skin cancer risk is affected by subtypes of statin. This meta-analysis indicated an increased risk of SCC with lipophilic statin exposure and a decrease in the risk of BCC under hydrophilic statin exposure. Notably, none of the studies included patients with NMSC from Asian countries, which points to the need for validation of the findings in Asian patients. Further research should be conducted to analyze the association between exposure to statin subtypes and the risk of NMSC in Korea. Fourth, several potential sources of residual confounding may have influenced our findings. One important issue is that we were unable to adjust for actual low-density lipoprotein (LDL) cholesterol levels or the degree of lipid control. Statins typically achieve consistent LDL reductions of 30%–50%, whereas the comparator group included diverse non-statin lipid-lowering agents with variable efficacy²⁰. This heterogeneity in lipid-lowering capacity may have led to differences in LDL levels between groups, which could independently affect NMSC risk. Another consideration is that the participating databases encompassed different calendar periods (ranging from the 1990s to 2023), during which diagnostic awareness and clinical coding of NMSC, as well as prescribing patterns of lipid-lowering agents, have evolved substantially. These factors may introduce residual confounding, and the influence of temporal heterogeneity cannot be entirely ruled out.

This study suggests that Korean patients with hyperlipidemia in Korea may take statins without the potential for high NMSC risk.

SUPPLEMENTARY MATERIAL

Supplementary Fig. 1

Covariate balance plots before and after 1:1 propensity score matching across 11 databases.