Statin Use and the Risk of Graves’ Orbitopathy: A Nationwide Population-Based Cohort Study

National health insurance research database

Diagnostic information related to GD and GO was obtained from Taiwan’s National Health Insurance Research Database (NHIRD) from 2010 to 2020. The NHIRD has been described in detail in a previous study. ¹⁷ The NHIRD originates from the National Health Insurance (NHI) program, a compulsory health insurance program in Taiwan. Initiated in 1995, the NHI covered more than 99% of the residents in Taiwan. ¹⁷ This healthcare program comprehensively covers an individual’s medical expenses including expenses associated with inpatient and outpatient care, prescribed medications, and dentistry services from contracted medical care institutions. Each NHI beneficiary’s personal medical data, including records of medical visits (inpatient, outpatient, and emergency visits), surgical procedures, prescriptions, and death records are all incorporated into the NHIRD. ¹⁷

Study population

The study cohort included NHI beneficiaries aged ≥40 years with newly diagnosed GD between January 1, 2010, and December 31, 2019. The first date of GD diagnosis was defined as the index day. We excluded subjects with the following conditions: (1) diagnosis of GO before the index day, (2) incomplete data about sex or birthdate, (3) concomitant or consecutive usage of both statin and nonstatin lipid-lowering agents within six months of the index day, and (iv) follow-up period of less than 365 days. The study selection process is shown in Figure 1. The study protocol was approved by the institutional review board of National Cheng Kung University Hospital (IRB number: B-ER-111-214).

OPEN IN VIEWER

FIG. 1.

Flow diagram showing the exclusion process for selecting eligible participants.

Diagnosis of GD and GO

The diagnosis of GD and GO was based on the International Classification of Diseases (ICD) codes, Anatomical Therapeutic Chemical (ATC) codes, and the surgery/procedure codes derived from the NHID. An individual was diagnosed with GD if she/he fulfilled the following two conditions: (1) having ICD-9-CM codes 242.0 and 242.9 or ICD-10 codes E05.0 and E05.9 at least twice; (2) had a prescription of antithyroid drugs at least twice. GO was defined by specific ICD codes with or without oral/intravenous glucocorticoids or specific surgical codes for the treatment of thyroid eye diseases. The detailed diagnostic criteria for GD and GO and the associated ATC codes of antithyroid drugs and glucocorticoids are shown in Supplementary Table S1.

Statin use

Statin use was evaluated based on the patients’ medical records from the NHIRD. The detailed ATC codes for statins are displayed in Supplementary Table S2. Those with a recorded prescription at least twice within six months before the index date were defined as statin users. The patients enrolled in this study were then categorized as statin users or nonusers. All patients were followed from the date of their first GD diagnosis recorded in the NHIRD until 31 December 2020 or until the occurrence of the following conditions during the follow-up period: (1) first visit for GO, (2) switching to a different statin, (3) discontinuation of statins for more than three months, and (4) receipt of a statin or nonstatin lipid-lowering drug among patients initially categorized as nonusers. In addition, since the prescription of statins was mainly restricted to those with specific risk factors of cardiovascular disease or stroke, ¹⁸ we also collected diagnostic information related to diabetes, hypertension, coronary artery disease, and stroke to minimize the possible confounding by indication (Supplementary Table S4). In order to evaluate the potential impact of statin’s potency on the protective effect of GO, we also calculated the treatment dosage of statins for each subject and categorized the statin users into (1) low-to-moderate intensity and (2) high-intensity statin users. ¹⁹ The detailed definition of statin intensity is shown in Supplementary Table S3.

Other covariates

We obtained information for the subject’s demographic characteristics including age, sex, urbanization of residence, and personal income (monthly) to control for potential confounding factors. We also retrieved information regarding risk factors of GO, including cigarette smoking, obesity, and RAI treatment history through the records in the NHIRD. The diagnosis and procedure codes used to identify the covariates are shown in Supplementary Table S4.

Statistical analysis

Propensity score (PS) analysis was used to minimize selection bias and potential confounding by indication by creating more homogenous groups with a balanced distribution of covariates for comparisons. A PS was calculated for each subject based on all the covariates listed in Table 1, which included age, sex, urbanization of residence, personal income, smoking, RAI treatment, diabetes, hypertension, cardiovascular disease, and stroke. We used the 1:4 propensity score matching (PSM) and inverse probability of treatment weighting (IPTW) method ²⁰ to generate risk comparisons for statin users and nonusers. The incidence rate of GO in GD patients was calculated as the number of events divided by the follow-up duration in 1000 person-years for overall population and after matching for propensity scores. Kaplan–Meier survival analysis and multivariable Cox proportional hazards analysis were also performed to compare the risk of GO among statin users and nonusers after examining the proportionality assumption. We use all variables described in Table 1 in the adjusted hazard ratio (HR) analysis. As for HR analysis regarding PSM and IPTW, confounding variables were chosen from factors with a standardized mean difference >0.1 in the population after PSM and IPTW adjustment in Table 1, respectively. Then we further compared the risk of GO among users of rosuvastatin, pitavastatin, and other statins to users of atorvastatin since atorvastatin is the only single statin that showed risk reduction of GO in a large-scale study previously. ⁹ We also evaluated the risk of GO between users of low-to-moderate intensity and high-intensity statins. Then we performed Cox regression analysis for patients who had changed the status of their lipid-lowering drugs (defined as switching to other lipid-lowering medication, stopping statin use, or initiating statin use during the follow-up period) during the follow-up period (intention-to-treat analysis) as an external validation of the protective effect of statins against GO. We used SAS V.9.4 (SAS Institute) for all data analyses.

Principal findings

In this study, we found that the statin users had a significantly lower risk of GO in patients with GD. This is the first large-scale study that investigated the relationship between the use of statins and GO risk in Asian GD patients after adjustment for potential confounders including age, sex, socioeconomic status, smoking, RAI treatment, obesity, cardiovascular disease, and its associated risk factors. One randomized clinical trial conducted by Lazolla et al. showed that adding oral atorvastatin to intravenous glucocorticoid treatment improved outcomes of GO. ⁶ Epidemiological studies have also shown a potential protective effect of statins against GO. ^{9 –12} However, these studies had not adequately considered confounding by indication. Since statins are commonly prescribed for the primary and secondary prevention of cardiovascular and cerebrovascular diseases, ²¹ there were quite a few statin users in this study who were classified as high-risk patients for ophthalmic surgery due to previous stroke or myocardial infarction. Since GO patients are less likely to undergo elective ophthalmological surgery due to unfavorable cardiovascular conditions, the relation between statin use and GO incidence will be confounded, as the diagnosis of GO was based on diagnostic codes and surgery codes in previous studies.

In this study, the use of statin was associated with a 20–35% reduction in risk of GO based on analysis with PSM and IPTW methods. This finding is comparable with the results of previous studies mainly focusing on Western populations by Stein et al. ¹⁰ and Nilsson et al. ⁹ Although racial differences exist in terms of the low-density lipoprotein cholesterol-lowering abilities of statins, ¹⁶ we show that the effect of statins in preventing GO is similar between the Western and Asian populations. Although one recent Taiwanese cohort study showed a substantially lower risk of GO among statin users (more than 80% risk reduction), time bias and confounding by indication were not considered, and therefore, the protective effect of statins against GO could have been overestimated. ¹¹ Furthermore, we found that the GO risk was not statistically different regardless of the kind of statins (i.e., atorvastatin, rosuvastatin, pitavastatin, and other statins) or the intensity of statins. Our finding demonstrated that statin use in GD patients may be effective for preventing GO. Additionally, considering the effectiveness of statins in both primary and secondary prevention of cardiovascular diseases, ^18,21 treatment with statins may be potentially beneficial to the prevention of both GO and cardiovascular diseases for patients with GD and hyperlipidemia.

Possible mechanisms

The association between statin use and the decreased incidence of GO may be explained by the following mechanisms: (1) the lowering of LDL-C levels and (2) the pleiotropic effect of statin on the regulation of the growth, differentiation, apoptosis, and autophagy of the cells in the orbital tissue. ²² Previous studies have shown that elevated LDL-C level is a risk factor for GO. ^23,24 Since hypercholesterolemia results in systemic inflammation that may further trigger the pathogenesis of GO, ²⁴ it seems plausible that lowering the levels of lipids, especially LDL-C, through statin use will reduce the risk of GO. Besides, statins also protect against GO through their pleiotropic anti-inflammatory, immunomodulatory, and antifibrotic effects in the orbital tissue. ²⁵ For example, regarding the IGF-1R pathway, one factor that potentially influences the pathogenesis of GO, ^26,27 atorvastatin lowers the IGF-1 levels ²⁸ while simvastatin and pravastatin also inhibit the signaling of IGF-1R. ^29,30 The suppression of the IGF-1 signaling may help to ameliorate the subsequent hyaluronan production and orbital fibroblast proliferation that potentially results in GO. Besides, simvastatin downregulates adipogenesis and myofibroblast differentiation of cultured orbital fibroblasts ^31,32 and suppresses the expression of the proinflammatory factor cysteine-rich protein 6, which is also implicated in the pathogenesis of GO. ^25,33 Furthermore, statin treatment also blocks the mevalonate pathway, ⁹ which is essential to the autophagic and apoptotic process in developing GO. ³⁴

Age and sex disparities regarding statins in prevention of GO

In this study, the protective effect of statins against GO is more prominent in females than males, which may be the result of the sex-related differences in response to statin treatment. Previous study demonstrated that the disparities in pharmacokinetics features resulted in a higher serum statins level in females than males. ³⁵ It is also reported that females are more responsive to statin’s anti-inflammatory effects than males. ³⁶ In addition, the protective effect of GO seems to be attenuated with elderly age. However, the prevalence of GD and GO is significantly lower in the elderly. ^{37 –40} In one Asian study, the prevalence of thyroid eye disease is about three times higher among middle-aged adults (age 40–50 years) than elderly adults (age >60 years). ³⁸ Furthermore, elderly adults generally tend to choose conservative treatment rather than surgical intervention, which may further underestimate the risk of GO in this age group. Therefore, the relatively low incidence of both GD and GO with potential underdiagnosis of GO subjects may limit the power of our study in evaluating the relationship between statin users and risk of GO in the elderly population.

Limitations

Although the study had many strengths such as a large sample size, long follow-up period, reliable diagnosis, inclusion of different statins, and thorough control of confounding factors, there were still some limitations. First, the diagnosis of GO was based on NHIRD data, glucocorticoid treatment, and surgical codes, which were mainly confined to patients receiving oral/injectable steroid treatment or surgical intervention for GO. Therefore, the incidence of GO might be underestimated since those with minor symptoms of GO were not included for analysis in this study. Besides, information regarding the severity of GO was not available due to the limitation of NHIRD. Second, information related to statin users, including the type of statins and the intensity of different statins, was obtained from prescription records in the NHIRD, and detailed drug compliance data for each patient were not available. Third, several risk factors of GO, including smoking and obesity, were obtained from diagnostic codes, which may underestimate the prevalence and their potential impacts. In addition, the level of anti-TSH-receptor antibody, which is known to be associated with risk of GO, was also unavailable in this study. Fourth, due to the payment restriction of NHI, statins were mainly prescribed to middle-aged and elderly population with risk factors of cardiovascular diseases and were seldomly prescribed in young adults. Hence, only adults aged >40 were included in this study, and further studies may be necessary to clarify whether the statins exhibit similar protective effects against GO in younger population. However, the effect of these limitations may be minimized because of the nationwide population-based cohort study design, which has the advantage of a large sample size with time-dependent analyses. Fifth, those who were diagnosed with GO prior to GD diagnosis were excluded from this study. Whether statin is still effective in preventing early onset GO may still need further investigation.

In conclusion, statin users were significantly associated with a lower risk of GO in patients with GD. Additionally, the risk of GO was not significantly different among users of commonly prescribed statins (i.e., atorvastatin, rosuvastatin, and pitavastatin) or users of low-to-moderate intensity and high-intensity statins. These results indicate that the use of statins may be considered in the prevention of GO for GD patients.