Association between statin use and cataract formation in a retrospective cohort study using Japanese health screening and claims data

Abstract

In this retrospective cohort study of data (recorded between 1 January 2005 and 31 December 2017) from a commercially available health screening and insurance claims database of the working-age Japanese population, we examined the association of statin use with cataract formation. Using the health screening data, we identified 1,178,560 patients who met the dyslipidaemia criteria; among them, 724,200 patients were enrolled. Based on person-years, the cohort was categorised by statin non-use and new use. Unadjusted, age–sex-adjusted, and multivariate-adjusted hazard ratios (hazard ratios [HRs]; with their 95% confidence intervals [CIs]) were estimated, and intergroup comparisons were undertaken using Cox proportional hazards regression. An increased risk of cataract incidence was associated with statin use (adjusted HR [95% CI]) compared with statin non-use (1.56 [1.43–1.70]). The adjusted HRs [95% CI] for cataract incidence for low- and high-potency statins were 1.48 [1.30–1.70] and 1.61 [1.44–1.79], respectively, whereas those for lipophilic and hydrophilic statins were 1.56 [1.39–1.75] and 1.56 [1.38–1.75], respectively. The adjusted HR for statin use with incidence of cataract was 1.35–1.73, except for fluvastatin and simvastatin. In the middle-aged Japanese working population, statin use was associated with a 1.5- to 1.6-fold higher risk of cataracts.

Introduction

Dyslipidaemia is a well-known risk factor for atherosclerotic cardiovascular disease^1,2, which has a high prevalence among individuals aged ≥ 21 years in the United States (37%)³and among the middle-aged population in the United Kingdom (50%)⁴. Despite its relatively lower prevalence in Japan, the prevalence of dyslipidaemia in adults aged ≥ 20 years has notably increased from 2016 to 2019 (males, 10–15%; females, 15–21%)⁵; in 2020, approximately 4.01 million people had dyslipidaemia⁶. Therapeutic management and monitoring of lipid levels (which are biomarkers of dyslipidaemia) may be essential to prevent atherosclerotic cardiovascular disease.

In several guidelines, meta-analyses, and reviews^7–13, statin treatment and lifestyle changes are recommended as primary and secondary preventive strategies to reduce the risk of cardiovascular disease, myocardial infarction, stroke, and mortality. Despite its benefits, statin therapy has been associated with several adverse effects, including muscle injury, elevated hepatic transaminase levels, diabetes mellitus, and the risk of kidney disease^14–17. Worldwide, cataracts are among the main causes of blindness and vision impairment¹⁸, and it is estimated that 95 million people are affected by cataracts in 2014¹⁹, which were the leading cause of blindness in those aged ≥ 50 years worldwide in 2020²⁰. A recent study highlighted that statin use may increase the risk of cataracts²¹. A systematic review and meta-analysis of observational studies, including cohort and case–control studies on statin use and cataract development, suggested that statin use was associated with an increased risk of cataracts (odds ratio, 1.11; 95% confidence interval [CI], [1.02–1.21])²². However, the pooled risk ratios (95% CI) of cohort studies, case–control studies, and randomised controlled trials in another systematic review were 1.13 (1.01–1.25), 1.10 (0.99–1.23), and 0.89 (0.72–1.10), respectively²³. The inconsistent findings of these studies^22,23are attributable to their different research designs. In particular, subgroup analyses in a systematic review for those aged < 65 years or those with a follow-up period of < 5 years did not show an increased risk of cataract²². In contrast, subgroup analyses of cohort studies showed that the risk of cataracts in those aged < 60 years and those with < 5 years of follow-up significantly increased by 28%, whereas in a systematic review, subgroup analyses of randomised controlled trials with follow-up period (48 weeks to 5.4 years) and age (18–85 years) revealed no association of these factors with cataract development²³. Furthermore, the statin-specific risk of cataract differed, as atorvastatin and lovastatin significantly increased the risk of cataract by 17–22%, though other statins were not associated with cataract²³.

In some studies, statin users, whose average age was approximately 50 years, had an increased risk of cataracts compared with that in non-users^24–26. Though higher age is a risk factor for cataracts²⁰, it remains unclear whether statin use is associated with the incidence of cataracts in the middle-aged population. Furthermore, in one study, the mean low-density lipoprotein (LDL) cholesterol level was inversely related to the risk of cataracts²⁴. Non-statin users were defined based on a claims database in these studies^24–26. Currently, only few studies have exclusively used populations with dyslipidaemia as a study cohort, though some studies^24–26 have been conducted in the general population, regardless of the lipid profile. However, the clarification of this aspect may be essential to enhance comparability regarding the baseline risk of cataracts.

The primary objective of this study was to investigate the association between statin use and cataract formation. The secondary objective was to clarify whether the risk of cataract differed based on the potency of the statins that were used²⁷, the properties of statins²⁸, and individual statins. Therefore, we conducted a retrospective cohort study using health screening and claims data to compare the incidence of cataract between new users and non-users of statins in a population that met the diagnostic criteria for dyslipidaemia. The detection of an association between statin use and ocular disease may facilitate the safe and effective use of statins in patients with dyslipidaemia in real-world settings.

Results

Using health screening data from January 1, 2005, to December 31, 2017, we identified 1,178,560 patients who met at least one dyslipidaemia diagnostic criterion; however, 69,383 (3.1%) individuals in this cohort (n = 2,233,475) had no data regarding any lipid values. Among them, 724,200 patients who met the diagnostic criteria were included in this study cohort (Fig. 1). Among these included patients, 666,745 had a period of statin non-use, whereas 140 patients had a period of statin use, as they had already started statin therapy on the index date. Furthermore, 57,315 patients had a period of statin non-use before the initiation of statin therapy. Among them, 57,294 patients started statin therapy, whereas the remaining 21 initiated two or more statins simultaneously after a period of statin non-use.

Fig. 1.

Flowchart depicting the screening and selection of the study population, based on the dyslipidaemia criteria using health screening data recorded between 1 January 2005 and 31 December 2017.

Table 1 presents the baseline characteristics of the study population. For statin users and non-users, the mean age was 51.7 and 45.9 years, respectively, whereas the proportion of men was 64.9% and 68.0%, respectively. The mean follow-up periods were 3.2 years for non-users and 1.3 years for statin users. The absolute values of the standardised difference between statin users and non-users for comorbidities and co-medications were > 0.1. Additionally, the mean defined daily dose (DDD) of statin was 0.4 (standard deviation, 0.2) at baseline.

Table 1.

Baseline characteristics of patients with dyslipidaemia with statin non-use or use.

	Statin non-use	Statin use	Standardised differencea
Number of patients	724,060	57,434	-
Age, years, mean (SD)	46 (10.4)	52 (8.7)	0.603
Male, n (%)	492,134 (68.0)	37,282 (64.9)	−0.065
Mean follow-up duration, years	3.2	1.3	−0.892
Comorbidities, n (%)
Hypertension	64,933 (9.0)	18,062 (31.4)	0.583
Diabetes mellitus	6,775 (0.9)	3,276 (5.7)	0.269
Chronic heart failure	8,750 (1.2)	3,715 (6.5)	0.276
Cerebrovascular disease	14,859 (2.1)	5,357 (9.3)	0.318
Cancer	35,803 (4.9)	6,591 (11.5)	0.240
Chronic pulmonary disease	39,103 (5.4)	4,359 (7.6)	0.089
Liver diseases	22,012 (3.0)	6,427 (11.2)	0.321
Myocardial infarction	2,413 (0.3)	1,701 (3.0)	0.208
Peptic ulcer disease	29,641 (4.1)	4,683 (8.2)	0.170
Peripheral vascular disease	9,874 (1.4)	4,061 (7.1)	0.287
Renal disease	7,058 (1.0)	1,954 (3.4)	0.167
Charlson Comorbidity Index score, (mean ± SD)	0.4 ± 1.1	1.1 ± 1.8	0.499
Co-medications, n (%)
Benzodiazepines	40,938 (5.7)	6,213 (10.8)	0.189
Antihypertensives	56,730 (7.8)	16,891 (29.4)	0.577
Anticoagulants	2,006 (0.3)	567 (1.0)	0.090
Antiplatelet drugs	1,399 (0.2)	1,752 (3.1)	0.228
Antipsychotic drugs	11,127 (1.5)	1,479 (2.5)	0.073
Aspirin	3,542 (0.5)	2,896 (5.0)	0.280
Antidepressants	14,945 (2.1)	2,024 (3.5)	0.089
Antidiabetic drugs	10,372 (1.4)	4,476 (7.8)	0.307
Antihyperuricaemic drugs	16,742 (2.3)	5,340 (9.3)	0.302
Ezetimibe	23 (0.0)	677 (1.2)	0.154
Corticosteroids, except for topical formulations	62,345 (8.6)	6,435 (11.2)	0.087

SD, standard deviation.

^aStandardised difference values > 0.1 were considered meaningful⁵¹.

Statin use was associated with an increased risk of cataract formation (hazard ratio [HR], 1.56; 95% CI, 1.43–1.70) compared with statin non-use (Fig. 2). When the grace period for prescription intervals was 60 or 90 days, the results were consistent with those of the sensitivity analysis.

Fig. 2.

Hazard ratios and their 95% confidence intervals for cataract incidence in patients who used any statins and those who used low- and high-potency statins compared with the incidence in the statin non-use group. Statins were classified into low-potency and high-potency statins, based on potency. Hazard ratios and their 95% CI were estimated using Cox proportional hazards regression models. A grace period of 30–90 days was included before the next dispensing date.

The crude incidence rate of new-onset cataracts was 2.4 per 1,000 person-years during the period of statin non-use, whereas the rates during statin use were 8.8 and 8.7 per 1,000 person-years for low- and high-potency statins, respectively (Table 2). The use of high- and low-potency statins was associated with an increased risk of cataract formation compared with statin non-use (Fig. 2; Table 2). The HR [95% CI] for high-potency statins (1.61 [1.44–1.79]) was slightly higher than that for low-potency statins (1.48 [1.30–1.70]). Furthermore, the HRs (95% CI) based on statin properties were 1.56 (1.38–1.75) and 1.56 (1.39–1.75) for hydrophilic and lipophilic agents, respectively. Compared with the cataract incidence in the statin non-use group, these values remained similar regardless of the property and potency of statins.

Table 2.

Association of cataract incidence with Statin use, stratified by the potency, property, and the individual statin.

Statin	Person-yearsa	Patients diagnosed with cataract and treated medically	Incidence rate (per 1,000 person-years)	Rate ratio (95% CI)	Hazard ratio (95% CI)b
					Unadjusted	Age–sex adjusted	Multivariatec adjusted
None,n = 724,060	2,290,188	5,479	2.39	1.0	1.0	1.0	1.0
Statin use, by potency
High-potency statins (atorvastatin and rosuvastatin), n = 35,999	46,058	400	8.68	3.63 (3.28–4.02)	3.24 (2.94–3.58)	1.70 (1.53–1.88)	1.61 (1.44–1.79)
Low-potency statins (pitavastatin, fluvastatin, simvastatin, and pravastatin), n = 21,435	25,389	223	8.78	3.67 (3.21–4.20)	3.24 (2.84–3.69)	1.60 (1.40–1.83)	1.48 (1.30–1.70)
Statin use, by property
Lipophilic statins (atorvastatin, pitavastatin, fluvastatin, and simvastatin), n = 28,577	35,860	312	8.70	3.64 (3.24–4.08)	3.24 (2.90–3.63)	1.66 (1.48–1.86)	1.56 (1.39–1.75)
Hydrophilic statins (rosuvastatin and pravastatin), n = 28,857	35,587	311	8.74	3.65 (3.26–4.09)	3.24 (2.89–3.62)	1.66 (1.48–1.86)	1.56 (1.38–1.75)
Individual statin
Atorvastatin, n = 14,756	19,401	179	9.23	3.86 (3.32–4.48)	3.46 (2.99–4.01)	1.81 (1.56–2.10)	1.73 (1.48–2.03)
Rosuvastatin, n = 21,243	26,657	221	8.29	3.47 (3.03–3.96)	3.09 (2.70–3.53)	1.62 (1.41–1.85)	1.52 (1.32–1.74)
Pitavastatin, n = 11,596	13,950	109	7.81	3.27 (2.70–3.95)	2.88 (2.38–3.48)	1.48 (1.22–1.78)	1.35 (1.12–1.64)
Fluvastatin, n = 1008	1,206	12	9.95	4.16 (2.36–7.33)	3.70 (2.10–6.54)	1.74 (0.97–3.10)	1.64 (0.92–2.93)
Simvastatin, n = 1,217	1,303	12	9.21	3.85 (2.19–6.79)	3.36 (1.90–5.93)	1.49 (0.83–2.66)	1.38 (0.77–2.48)
Pravastatin, n = 7,614	8,930	90	10.08	4.21 (3.42–5.19)	3.72 (3.02–4.57)	1.77 (1.44–2.19)	1.67 (1.36–2.06)

CI, confidence interval.

^aPerson-years are the total number of years from the start date of observation to the date of censoring.

^bHazard ratios and their 95% confidence interval (CI) were estimated using Cox proportional hazards regression models.

^cAdjusted for age, sex, co-medications (benzodiazepines, antihypertensives, anticoagulants, antiplatelet drugs, antipsychotic drugs, aspirin, antidepressants, antidiabetics, ezetimibe, and corticosteroids [except topical formulations]), comorbidities (hypertension, diabetes mellitus, chronic heart failure, cerebrovascular disease, cancer, chronic pulmonary diseases, liver disease, myocardial infarction, peptic ulcer disease, peripheral vascular disease, and renal disease), and the Charlson Comorbidity Index score.

Compared with the risk in the statin non-use group, the risk of cataract formation increased with the use of all statins (atorvastatin, rosuvastatin, pitavastatin, and pravastatin; Table 2; Fig. 3), except for fluvastatin and simvastatin. The HRs for individual statins ranged from 1.4 to 1.7.

Fig. 3.

Hazard ratios and their 95% confidence intervals for cataract incidence in patients by each individual statins as compared with the cataract incidence in the statin non-use group were estimated using Cox proportional hazards regression models.

The sensitivity analysis that was conducted after excluding patients with diabetes mellitus showed that the HRs (95% CI) for the incidence of cataract formation with statin use were 3.20 (2.94–3.47), 1.61 (1.48–1.76), and 1.55 (1.42–1.70) in the unadjusted, age-sex-adjusted, and multivariate-adjusted analyses, and these values were similar to those of the main analysis. Furthermore, the results of analyses stratified by DDDs and age groups at baseline were identical to those of the main analysis (Table 3).

Table 3.

Stratification of cataract incidence by the DDDs of statins and age groups at baseline.

Groups	Hazard ratio (95% CI)a
	Unadjusted	Age–sex adjusted	Multivariateb adjusted
None	1.0	1.0	1.0
DDDs
< 0.5	3.55 (3.20–3.93)	1.75 (1.57–1.94)	1.62 (1.46–1.81)
≥ 0.5	2.93 (2.59–3.32)	1.57 (1.39–1.78)	1.46 (1.28–1.67)
Age, years
< 50	3.07 (2.32–4.06)	2.01 (1.52–2.65)	1.55 (1.14–2.10)
≥ 50	1.86 (1.71–2.02)	1.63 (1.49–1.77)	1.54 (1.40–1.68)

DDD, defined daily dose.

^aHazard ratios with 95% confidence interval (CI) were estimated using Cox proportional hazards regression models.

^bAdjusted for age, sex, co-medications (benzodiazepines, antihypertensives, anticoagulants, antiplatelet drugs, antipsychotic drugs, aspirin, antidepressants, antidiabetics, ezetimibe, and corticosteroids [except topical formulations]), comorbidities (hypertension, diabetes mellitus, chronic heart failure, cerebrovascular disease, cancer, chronic pulmonary diseases, liver disease, myocardial infarction, peptic ulcer disease, peripheral vascular disease, and renal disease), and the Charlson Comorbidity Index score.

Discussion

In this retrospective cohort study, the risk of cataract formation in the statin use group was 1.5–1.6 times higher than that in the statin non-use group, regardless of the different grace periods (60 or 90 days). Additionally, analyses stratified by DDDs and age group (< 50 or ≥ 50 years) at baseline showed similar results as those of the main analysis. This finding was consistent for both high-potency and low-potency statins, as well as for hydrophilic and lipophilic statins. Furthermore, the use of individual statins (except for fluvastatin and simvastatin) increased the risk of cataract incidence. Cataracts constitute a risk factor for blindness and visual impairment^20,21. Middle-aged patients using statins may need regular ophthalmological consultations to manage their statin-use-associated risk of cataract formation.

The findings of systematic reviews regarding the association between statin use and cataract formation^22,23are inconsistent. In particular, subgroup analysis of observational studies in the age group of < 65 years revealed no association between statin use and cataract formation²². However, a subgroup analysis of cohort studies in the age group of < 60 years showed that statin use increased the risk of cataract formation (relative risk, 1.28; 95% CI, 1.19–1.38)²³. Similar to the findings of this previous study²³, the findings from our study population (mean age, approximately 50 years) showed that, compared with statin non-use, new statin use was associated with an increased risk of cataract incidence. Moreover, in studies that revealed an association between statin use and cataract formation compared with that in statin non-users^24–26,29, the mean age of the population in some studies was approximately 50 years^26,29. Furthermore, the mean LDL-C levels might have an inverse effect on cataract incidence²⁴. However, this effect might have been adjusted for in this study, as our study population met the criteria for dyslipidaemia in both the statin use group and statin non-use group. Further studies with larger cohorts (> 60,000) are required to verify these findings, particularly owing to the small cohort sizes for some statins (fluvastatin and simvastatin).

In this study, there was no significant difference in cataract risk from statin use based on hydrophilicity or lipophilicity, and the HRs for both categories were approximately 1.6 (Table 2). In a population-based case–control study in Taiwan, compared with those for statin non-use, the adjusted odds ratios for cataract formation were 3.24 (1.98–5.32) for hydrophilic statins (pravastatin and rosuvastatin) and 3.49 (2.61–4.66) for lipophilic statins (simvastatin, fluvastatin, atorvastatin, and lovastatin)²⁶. In contrast, the HRs (95% CI) for cataracts in a study of kidney transplant recipients in the United States did not show significant differences: 1.10 (0.88–1.39) for hydrophilic statins (rosuvastatin and pravastatin) and 1.14 (0.96–1.35) for lipophilic statins (simvastatin, pitavastatin, and lovastatin), compared with that for other lipophilic statins (atorvastatin and fluvastatin)³⁰. Thus, the result of the present study regarding the non-significant difference in the risk of cataract based on the properties of statins is possibly consistent with previous findings. However, as our study defined the reference group based on the non-use of statins^26,30, we did not directly compare the risk of cataracts between hydrophilic and lipophilic statins. Therefore, the risk of cataract formation among patients who are on statin therapy is independent of the properties of statins.

The detailed mechanism underlying statin-induced cataract formation, based on the properties of statins, remains unknown. Cholesterol is an essential component necessary for membrane formation by proliferative lens epithelial cells³¹. The inhibitory effects of simvastatin (a lipophilic statin) on cholesterol synthesis in the human lens were more prominent than those of pravastatin (a hydrophilic statin)²⁹. Although clinical studies have reported similar risk of cataract with statins (simvastatin and pravastatin)^25,32, the cataract risks associated with these statins differed from those in this study.

The strength of this study includes the identification of the study population based on definite diagnostic criteria for dyslipidaemia^7,33,34 from the health screening database, which overcomes the usual challenge in identifying populations with statin non-use. This approach facilitated the establishment of highly comparable populations for statin use and non-use. Therefore, our cohort study enabled an investigation of the association between the risk of cataract formation based on statin use and non-use.

Nevertheless, this study has certain limitations. First, the diagnostic codes for cataract in the claims database in Japan were not validated and did not include the recording of ophthalmic evaluations. Therefore, for the outcome, we used information pertaining to the diagnosis, medication, and surgery for cataract. In particular, we considered cataract surgery as one of the outcomes for cataracts, as it is likely that only a proportion of those who had cataract were diagnosed. Furthermore, in the case of surgical records, the degree of misclassification may be low. Second, older people aged ≥ 75 years were excluded from our study population. Therefore, our findings may not be highly generalisable for the older population. Third, the 95% CIs for fluvastatin and simvastatin were wide, and these estimates may be unstable because the number of patients who used these statins was smaller than that who used the other statins. Simvastatin reduces the risk of cataracts³⁵; however, further studies are required to confirm this finding. Fourth, in addition to the diagnosis of diabetes and hypertension, adjustments for confounding factors were made for medications used to treat these conditions. Conversely, baseline LDL-C levels³⁶and lifestyle-related factors, such as alcohol consumption³⁷and smoking³⁸, might constitute risk factors for cataracts; however, these factors were not adjusted for in the database. Though we excluded patients with diabetes mellitus³⁹, residual confounding factors may exist if the proportions of individuals with habits of alcohol consumption and smoking differ between the subpopulations who did and did not use statins. Fifth, the mean follow-up time for statin users in this study was 1.3 years, which may not be sufficient to detect long-term outcomes, such as cataract formation. A shorter follow-up period might underestimate the long-term effects of statin use on cataract formation. However, previous studies have shown that long-term statin use is not associated with an increased risk of cataracts⁴⁰and that shorter-term (< 5 years) use of statins increases the risk of cataract surgery⁴¹. Further research is needed to explore the relationship between the duration of statin use and cataract formation. Sixth, we could not fully examine the association of cataract risk with statin dose, although the DDDs at baseline could be considered a relevant parameter. Seventh, to estimate the effect of statins on cataract formation, we excluded patients with dyslipidemia who, before cohort entry, had a diagnosis of cataract, prescriptions that indicated cataract, or a history of cataract surgery. To obtain data on cohort members who were identified through the health screening data, we linked the health screening data to claims data. As we did not obtain medical records, we were unable to confirm whether cohort members were at high risk of cataracts, regardless of whether they were taking statins. Additionally, to identify the cohort with dyslipidemia, it may be beneficial to exclude individuals without lipid values from health screening data, as this approach could obviate the inclusion of relatively healthy individuals and thus reduce potential selection bias.

In conclusion, this study indicates that, compared with statin non-use, statin use is associated with a higher risk of cataract formation in a middle-aged Japanese working population. However, fluvastatin and simvastatin use was not associated with this risk. Moreover, statin users are encouraged to undergo regular ophthalmology consultations to manage the risk of cataract formation. Further studies are required to clarify whether period of statin use and old age are associated with the risk of cataract formation.

Methods

Data sources

On 18 September 2020, we obtained health screening and claims data that were maintained by JMDC Inc.; the database comprises data from approximately 6 million individuals^42,43. The database represented approximately 4.3% of the Japanese population as of 2017, and the data were anonymised. Information regarding diagnosis, procedure, and drug use during hospitalisation was sourced from inpatient claims data, as inpatient claims have the same format as outpatient claims in Japan⁴⁴. The claims database did not contain data of individuals aged ≥ 75 years, as their medical expenses are covered by public health insurance (late-stage medical care system for older individuals), and not by corporate health insurance⁴⁵. The insurance claims data included information on diagnosis, procedure, and prescriptions. The health screening data included values and measurement dates for total cholesterol (TC), LDL cholesterol (LDL-C), high-density lipoprotein (HDL) cholesterol (HDL-C), and triglyceride (TG). The enrolment data included dates (year and month) of enrolment and disenrollment of members. These three types of data were linked using patient IDs for this study. Claims data included sex, year and month of birth, diagnosis codes (The International Classification of Diseases, 10 th revision [ICD-10]), date of diagnosis, and procedures (domestic codes), including operation, and dates of procedures and prescriptions (generic name, daily dose, dispensing date, prescription date, and number of prescription days).

Study cohort

We identified patients with TC level ≥ 220 mg/dL, LDL-C level ≥ 140 mg/dL, HDL-C level < 40 mg/dL, or TG level ≥ 150 mg/dL using annual health screening data recorded between 1 January 2005 and 31 December 2017^7,33,34. We defined the index date as the first date when one of these criteria was met, and baseline periods of at least 6 months prior to the index date after the date of enrolment were available. Patients who had data for a baseline period of < 6 months, had used any statins, had a prescription of anti-cataract medication (pirenoxine and glutathione with the indication for cataracts in Japan), had a diagnosis of cataract (ICD-10 codes: H25 to H28), or had a record of cataract surgery (K282, K282-02, or K283 of domestic procedure code) within the baseline period were excluded. If patients had started any statins on the index date, the date coincided with the day on which both statin start date and dyslipidaemia criteria were met. Patients who started statin therapy on the index date or after a period of statin non-use were categorised as new users of statins⁴⁶; those who were simultaneously prescribed two or more statins on the index date were excluded from the study. In addition, a sensitivity analysis of the study cohort without a diagnosis of diabetes or without the use of antidiabetic drugs (including insulin) at baseline was conducted.

Exposures and outcomes

We categorised person-years of the cohort based on non-use and new use of statins (atorvastatin, rosuvastatin, pitavastatin, fluvastatin, simvastatin, and pravastatin). Furthermore, statins were classified into low-potency (pitavastatin, fluvastatin, simvastatin, and pravastatin) and high-potency (atorvastatin and rosuvastatin) statins based on their potency²⁷. Additionally, based on their structure and pharmacokinetic properties²⁸, statins were divided into hydrophilic (pravastatin and rosuvastatin) and lipophilic (atorvastatin, pitavastatin, fluvastatin, and simvastatin) statins. In the main analysis, continuous use of statins was defined using the dispensing date plus the days of supply, with a grace period of up to 30 days before the next dispensing date⁴⁷.

To evaluate the effects of varying statin dosages in the sensitivity analysis, we standardised the analysis using DDD, based on the Anatomical Therapeutic Chemical classification system⁴⁸. To determine DDDs at baseline, we defined the daily initial dose of individual statin as the dose most commonly prescribed in one or more statin prescriptions prior to censoring a statin user⁴⁹.

Cataract was defined as a diagnosis of cataract (H25 and H26 according to the ICD 10 code), cataract surgery (K282 and H283 according to domestic procedure code), or use of drugs with an indication for cataracts (pirenoxine and glutathione) after the index date. We defined cataract diagnosis and prescription of anti-cataract medications as outcomes for the main analysis because validation studies on cataract diagnosis have not been conducted in Japan.

Participant follow-up

The observation period for assessing cataracts was defined as the period from the index date to the incidence date of the cataract, date of disenrollment, date of discontinuation or switching of a statin, or 31 December 2017 (end of the study period), whichever occurred first.

Covariates

Using the claims database, we obtained the following covariates in the baseline period and considered them for confounding adjustment: demographic characteristics (age and sex), co-medications (benzodiazepines, antihypertensives, anticoagulants, antiplatelets, antipsychotics, aspirin, antidepressants, diabetes medications, gout medications, ezetimibe, and corticosteroids, excluding topical medications) using the generic name of the drug, and comorbidities (hypertension [I10]), diabetes mellitus (E10, E11, E13, and E14), chronic heart failure (I50), cerebrovascular disease (I61 to I63, I67, and I69), cancer (C00 to C97), chronic lung disease (J40, J42 to J45, J84, and J96), liver disease (B16, B18, B19, I85, K70 to K74, and K76), myocardial infarction (I21), peptic ulcer (K22, K25 to K27), peripheral vascular disease (I70, I71, I73, and I74), and renal disease (N03 to N05, N10, N17 to N19, and Q61) using ICD-10 codes. Furthermore, the Charlson Comorbidity Index scores were included as covariates⁵⁰.

Statistical analysis

The study population was divided by the statin use period or statin non-use, and data are presented as summary statistics of baseline characteristics (age, sex, comorbidities, and concomitant medications). The standardised difference was calculated for differences between the groups⁵¹; this difference is important if the absolute values are greater than 0.1. We calculated the incidence rate (per 1,000 person-years) for both the statin use and statin non-use groups, as well as the crude rate ratio and its 95% CI for cataracts, as the outcome. Furthermore, we calculated the unadjusted, age-sex-adjusted, and multivariate-adjusted HRs and their 95% CIs using Cox proportional hazards regression models based on the incidence of cataracts during the non-use of statins. The proportional hazards assumption was checked using Log (-log (St)) plots. Despite adjusting for confounding factors through forward, backward, and stepwise selection of covariates in the model, the same covariates were selected regardless of the method used. Furthermore, regardless of the selected covariates, age, sex, and potential risk factors of cataracts (e.g., diabetes mellitus, antidiabetics, hypertension, antihypertensives, systemic corticosteroids, and the Charlson Comorbidity Index score)^23,24 were included in the adjusted model. Robust variance was used when estimating 95% CIs.

Sensitivity analyses were performed for the incidence of cataracts related to statin use or non-use and for the grace period of 60 or 90 days for prescription intervals, to assess the consistency of our findings. Moreover, we conducted stratified analyses by DDDs (< 0.5 and ≥ 0.5) and age groups (individuals < 50 and ≥ 50 years) at baseline. Furthermore, we estimated the risk associated with the potency of statins compared with that of non-use because the incidence rates of cataracts may differ by the potency of statins. Similarly, we analysed the risk of cataract formation by statin properties, such as hydrophilic and lipophilic properties³⁰. Furthermore, the same analysis was performed for individual statins. Diabetes mellitus is a risk factor for cataract formation³¹; therefore, we examined statin use and the risk of cataract formation by excluding patients with diabetes or those who used glucose-lowering drugs, including insulin, during the baseline period from the study cohort.

Sample size

Assuming that the incidence of cataracts in the statin non-use group would be 0.5% (α = 0.05, β = 0.2, and relative risk = 1.4), the total cohort size was estimated to be approximately 470,000 individuals. All statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary NC, USA), and statistical significance was set at p < 0.05.

Ethical considerations

This retrospective cohort study was conducted in compliance with the principles of the Declaration of Helsinki (as revised in Brazil 2013). The study protocol was approved by the Ethics Committee of the Nihon University School of Pharmacy (20 − 001). The study data were completely anonymised; hence, the need for informed consent from patients was waived by the Ethics Committee of the Nihon University School of Pharmacy.

Author contributions

All authors contributed equally to the study, and they have read and approved the submission of this manuscript. Conceptualization, K.K., N.O.; methodology, K.K., N.O.; analysis, K.K. K.I., N.F., Mo.T., Mi.T., N.O.; writing—original draft preparation, K.K., H.K., Y.S.; writing—review and editing, K.K., N.O.; supervision, N.O.

Data availability

The data that support the findings of this study are available from JMDC Inc., but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are, however, available from the corresponding authors upon reasonable request and with the permission of JMDC Inc.

Declarations

Competing interests

The authors declare no competing interests.