http://orcid.org/0000-0002-5636-4853
Figures
Abstract
Objective
To investigate a possible association between retinal vein occlusion (RVO) and an increased risk of developing acute myocardial infarction (AMI).
Design
A population-based retrospective cohort study using the entire population of the Taiwan National Health Insurance Research Database (NHIRD) from 1st January, 2001 to 31st December, 2013.
Methods
A total of 37921 subjects with RVO were enrolled in the RVO group, and 113763 subjects without RVO were enrolled in the comparison group. The comparison group consisted of randomly selected individuals who were propensity score (PS)-matched with the RVO group at a ratio of 1:3, based on age, gender, obesity, stroke, hyperviscosity syndrome, glaucoma, and the use of antithrombotic drugs. A log-rank test was used to compare the cumulative hazard of AMI between the two groups. A multivariate Cox regression analysis was used to estimate the adjusted hazard ratios (HRs) of AMI, adjusted for PS, diabetes, hypertension, hyperlipidemia, congestive heart failure, and chronic renal failure.
Results
The mean age of the cohort was 62.4±13.2 years. RVO patients had significantly higher proportions of diabetes, hypertension, hyperlipidemia, congestive heart failure, and chronic renal failure than the comparisons. The mean follow-up period was 5.52 years in the RVO group and 5.55 years in the comparison group (p = 0.16). A log-rank test comparing the cumulative hazard curves of AMI for the two groups revealed a significant difference (p<0.0001). In the multivariate Cox regression after adjustment for PS and confounders, the RVO group had a significantly higher risk of AMI (adjusted HR = 1.21; 95% CI: 1.13 to 1.30). When the RVO group was divided into central retinal vein occlusion (CRVO) and branch retinal vein occlusion (BRVO) and analyzed separately, both groups had significantly higher adjusted HRs for developing AMI than the comparison group. Moreover, the CRVO group had a significantly higher risk of AMI than the BRVO group.
Citation: Chen Y-Y, Sheu S-J, Hu H-Y, Chu D, Chou P (2017) Association between retinal vein occlusion and an increased risk of acute myocardial infarction: A nationwide population-based follow-up study. PLoS ONE 12(9): e0184016. https://doi.org/10.1371/journal.pone.0184016
Editor: Stephane Germain, Centre National de la Recherche Scientifique, FRANCE
Received: February 4, 2017; Accepted: August 16, 2017; Published: September 12, 2017
Copyright: © 2017 Chen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data are available from the National Health Insurance Research Database (NHIRD) published by Taiwan National Health Insurance (NHI) Bureau. The data utilized 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 from 2012. Requests for data can be sent as a formal proposal to the NHIRD (http://nhird.nhri.org.tw) or by email to wt.gro.irhn@drihn.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Retinal vein occlusion (RVO) results from thrombosis in the retinal vein, due to compression by a nearby atherosclerotic artery or increased blood viscosity. [1–4] Risk factors of RVO include hypertension, diabetes, hyperlipidemia, and older age. [5–10] All of these factors are also risk factors of acute myocardial infarction (AMI). Because RVO and AMI share a common pathogenesis of atherosclerosis and similar risk factors, these conditions might be associated.
Previous hospital-based studies investigating the relationships between RVO and the risk of developing AMI revealed inconsistent results. [5,10–14] Recently, follow-up studies based on healthcare claims databases were conducted to solve the problem but were also inconclusive. Hu et al utilized the Taiwan National Health Insurance Database (NHIRD) and found that RVO did not independently increase the risk of developing AMI during the 3-year follow-up period. [15] Werther et al analyzed claims data in the US (including Medicare and Medicaid) and showed that RVO was not significantly associated with AMI development during the 4-year follow-up period. [16] However, in South Korea, Rim et al used a database of a nationwide sample cohort and showed that RVO increased the risk of developing AMI [hazard ratio (HR) = 1.25; 95% confidence interval (CI): 1.02 to 1.52] during the 11-year follow-up period. [17] The inconsistent results of previous studies may be due to the insufficient statistical power, non-equivalent study designs, as well as different diagnostic classification methods.
The objective of our study is to investigate whether patients with RVO have a higher risk of developing AMI than controls using the “whole population” NHIRD in Taiwan, not “sample data” as reported in previous studies. Thus, we will have sufficient statistical power. Besides, during the study period, NHIRD adopted the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes, which are generally acknowledged worldwide. Therefore, the results of our study were compared with the results of previous studies. Furthermore, given the completeness of the NHIRD, we can have a long-term study period and can derived the risk ratio of developing AMI after controlling and adjusting for the impact of confounders.
Materials and methods
Study setting
Taiwan’s National Health Insurance (NHI) program, which was launched in 1995, currently covers the health care services of greater than 99% of Taiwan’s 23 million residents. [18] The National Health Insurance Research Database (NHIRD), which is maintained by the National Health Research Institutes of Taiwan, is a collection of all registration file data and claims data for all ambulatory and in-hospital patients in Taiwan. The identification of all patients in the database is encrypted before the data are released for research purpose. [19] Therefore, according to the rules of the Instituitional Review Board, writthen informed consent was waived. [19–20] Based on the healthcare claims of the entire population, we sought to compare the incidence rate of AMI in subjects with and without RVO during the 13-year period. This study was approved by the ethical committee of Yang-Ming University Hospital (2015A018).
Inclusion and exclusion criteria
Using the Taiwan NHIRD from 1996 to 2013, we performed a retrospective cohort study. We first selected patients with RVO during January 1, 2001 to December 31, 2013. The diagnostic codes of RVO included central retinal vein occlusion (CRVO; ICD-9-CM codes 362.35) and branch retinal vein occlusion (BRVO; ICD-9-CM codes 362.36). Patients with RVO diagnoses from January 1, 1996 to December 31, 2000 were excluded to eliminate patients with previous, chronic RVO. The date of the first RVO claim was defined as the index date. We also randomly selected individuals who had never received a diagnosis of RVO as a comparison group. To eliminate the bias caused by the group differences, the RVO group and the comparison group were 1:3 matched using propensity score (PS) matching method, [21] in terms of age, gender, index year (the year of index date), use of antithrombotic drugs, obesity, stroke, hyperviscosity syndrome, and glaucoma. Both the RVO and comparison groups were tracked during the follow-up period to identify the occurrence of AMI (ICD-9-CM codes 410). Patients who received a diagnosis of AMI before the index date were excluded to ensure that patients with only newly diagnosed AMI were included.
Statistical analysis
The demographic/clinical characteristics of the RVO and comparison groups were compared by chi-square test for categorical variables and two-sample t-test for continuous variables. A log-rank test was performed to describe and compare the cumulative hazard curves of AMI between the two groups. A multivariate Cox proportional hazard model was used to estimate the adjusted HRs for the occurrence of AMI. Covariates adjusted in the regression analysis were PS and relevant comorbidities, including diabetes, hypertension, hyperlipidemia, congestive heart failure, and chronic renal failure. [22] To eliminate the effect of the immortal time bias, these comorbidities were regarded as time-dependent variables. [23]
Furthermore, we divided the RVO group into BRVO and CRVO groups and compared the demographic/clinical characteristics between them. Finally we separately compared the HRs for AMI among the BRVO, CRVO and comparison groups. All statistical operations were performed using SAS statistical package, version 9.3 (SAS Institute, Cary, NC, USA).
Results
The study population
This study included 37921 patients with RVO and 113763 PS-matched comparisons. Among the patients with RVO, 11855 (31.3%) were diagnosed with CRVO, and 26066 (68.7%) were diagnosed with BRVO. Table 1 shows a summary of the characteristics of the RVO and the comparison groups. Because the two groups were well-matched on PS derived from age, gender, use of antithrombotic drugs, obesity, stroke, hyperviscosity syndrome, and glaucoma, no differences in theses variables and PS were found between the two groups. The mean age of the overall cohort was 62.4 years, with a standard deviation of 13.2 years. Greater than 60 percent of the subjects were over 60 years old. Males constituted a slightly higher proportion than females (51.1% vs. 48.9%). The mean follow-up period was 5.52 years in the RVO group and 5.55 years in the comparison group, without significant differences (p = 0.16). Comparing the prevalence of comorbidities in the RVO group with the comparison group, we identified significant differences in the prevalence of diabetes, hypertension, hyperlipidemia, congestive heart failure, and chronic renal failure (all p <0.0001). During the 13-year study period, AMI occurred in 1240 (3.22%) patients in the RVO group and in 2616 (2.3%) in the comparison group (p<0.0001).
Cumulative hazard curves and log-rank test
Fig 1 illustrates the cumulative incidence curves of AMI in the RVO group and the comparison group. These two curves were moving apart from each other from the very beginning until the end of the study period. According to the log rank test, the RVO group had a significantly higher cumulative hazard of AMI than the comparison group (p-value<0.0001)
The black line represents the RVO group, and the gray line represents the comparison group. RVO: retinal vein occlusion.
HRs for AMI analyzed using Cox models
Table 2 displays the adjusted HRs for AMI during the 13-year study period calculated with multivariate Cox regression models. After adjusting for PS and confounders, the RVO group had a significantly higher hazard for AMI (adjusted HR = 1.21, 95% CI: 1.13 to 1.30). People with comorbidities, such as diabetes, hypertension, hyperlipidemia, congestive heart failure, and chronic renal failure, had a significantly increased risk of AMI (all p<0.0001). For those with both congestive heart failure and coronary heart disease (CHD, not AMI), the hazard of AMI was significantly increased 4.19-fold compared with those without congestive heart failure. For those with congestive heart failure but without CHD, the hazard for AMI was still statistically higher than those without congestive heart failure (adjusted HR = 1.96 with 95% CI: 1.80–2.14).
Comparison of characteristics between the CRVO and BRVO groups
In Table 3, we divided the RVO group into the CRVO group (with 11855 patients) and BRVO group (with 26066 patients), and compared their demographic/clinical characteristics. The age distribution in the CRVO group was younger than that in the BRVO group. The CRVO group had a significantly higher proportions of hyperviscosity syndrome, glaucoma, and diabetes than the BRVO group. However, the two groups were similar in the proportions of obesity, stroke, and congestive heart failure. The two groups also had similar mean follow-up time (5.51 years in the CRVO group and 5.53 years in the BRVO group). The occurrence rate of AMI is 3.60% in the CRVO group, significantly higher than that of 3.12% in the BRVO group (p<0.0001).
Comparison of the risk of AMI between the CRVO and BRVO groups
Fig 2 compares the adjusted HR for AMI among the comparison, CRVO, and BRVO groups. Compared with the comparison group, the adjusted hazard ratio for AMI was 1.35 (95% CI: 1.22–1.49) in the CRVO group and 1.15 (95% CI: 1.06–1.25) in the BRVO group. When patients with BRVO and CRVO were compared, the CRVO group still had a significantly higher risk of developing AMI than the BRVO group (p-value = 0.01).
AMI: acute myocardial infarction; BRVO: branch retinal vein occlusion; CRVO: central retinal vein occlusion.
Discussion
In this nationwide follow-up study using the complete population data from the Taiwan NHIRD, the risk of developing AMI were compared between 37921 RVO patients and 113763 PS-matched comparisons. After adjusting for PS and possible confounders, the RVO group had a significantly higher risk of developing AMI than the comparison group (adjusted HR = 1.21, 95% CI: 1.13–1.30). When RVO group was divided into CRVO and BRVO groups, we found both of them had significantly higher adjusted HR for AMI than the comparison group. And, the CRVO group had a significantly higher risk of developing AMI than the BRVO group (p = 0.01).
RVO, including CRVO and BRVO, is one of the most common causes of visual impairment, particularly in older people. [24] The pathogenesis of RVO has not been completely described, but occlusion resulting from thrombosis is a well-accepted theory. RVO mainly occurs at arteriovenous crossings where the retinal arteriole and venule share a common adventitia sheath. If the arterial wall is sclerotic, it may compress the underlying vein, and the turbulent blood flow in the vein will facilitate the formation of venous thrombosis and venous occlusion. [25] Similarly, the underlying mechanism of AMI involves a sclerotic artery (coronary artery), which causes thrombus formation and venous obstruction. [26] We postulate that RVO and AMI may be associated because the changes in retinal vessels may somehow imply that changes have also occurred in coronary vessels.
Another explanation for the association between RVO and AMI is that they have common risk factors. As noted in Table 1 in our study, the RVO group exhibits a greater proportion of systemic comorbidities, such as diabetes, hypertension, hyperlipidemia, congestive heart failure, and chronic renal failure, than the comparison group. The result is consistent with the results from previous studies. [9,10,12,16,25,27–31] Notably, these systemic comorbidities are also risk factors for AMI, thus potentially explaining why patients with RVO are predisposed to AMI.
As shown in Fig 1, the cumulative hazard of AMI in the RVO group was significantly higher than the comparison group. The result is compatible with the results of the population-based studies in the US and South Korea. [16,17] Hu’s study in Taiwan, with a follow-up time of 3 years, also revealed a higher cumulative incidence of AMI in the RVO group. [15] With a much larger sample size and a much longer follow-up period, our study further confirms the phenomenon.
Shih et al analyzed claims data from 10,081 patients with RVO aged 70 years in Taiwan and reported that patients with RVO had higher risks of subsequent myocardial infarction (HR = 1.29, 95% CI: 1.15–1.44). [9] Our study, analyzing data of the whole population, not only the elderly, also revealed a significantly higher adjusted HR for the patients with RVO than the individuals without RVO. Bertelsen et al reviewed the registries of 439 photographically verified patients with CRVO from four referral centers in Denmark and found that patients with RVO had a 1.9-fold increased risk of developing ischemic heart disease compared with the control cohort (HR = 1.9, 95% CI: 1.32–2.75). [14] However, in another study by Bertelsen et al utilizing the same data source, BRVO was not significantly associated with subsequent myocardial infarction (MI). [10] In our study, as shown in Fig 2, both the BRVO and CRVO groups have a significantly higher risk of developing AMI than the comparison group. In addition, the risk in patients with CRVO is significantly higher than the risk in BRVO patients. Thus, BRVO and CRVO may not have the same risk factors. [30] To the best of our knowledge, our study is the first to report a significantly higher risk of AMI in the CRVO group compared with the BRVO group.
One strength of our study is the use of the complete population database, precluding possible selection bias, and providing stronger statistical power. Another strength of our study is the accuracy of the diagnoses. In our healthcare system, patients must receive a standard protocol of examinations to confirm the diagnoses or the National Health Insurance Administration will not pay the fees for treatment. Previous studies also validated the accuracy of the diagnoses in the database. [32–34] Therefore, the diagnoses in our study, including RVO, AMI, and comorbidities, were thoroughly verified.
Another strength is the controlling and adjustment for possible confounders using PS matching method and PS analysis. In brief, PS is computed through mathematical algorithms from confounders that have impact on RVO, which means PS represents the overall impact of these confounders. Therefore, if the RVO and comparison groups are well-matched on PS (PS matching method), these confounders would be evenly distributed in the RVO and comparison groups, thus the PS matching method reduces the selection bias. PS analysis means that PS as well as other confounders are regarded as covariates in the regression adjustment, thus we can derive a more accurate relationship between RVO and AMI.
The limitation of our study is that some patients with asymptomatic RVO may not seek medication. Thus, subjects with undetected RVO were classified as the comparison group and not the RVO group. Such a misclassification bias would weaken the effect of RVO on the development of AMI. Even so, RVO patients in our study had a higher hazard of AMI. Therefore, our observation of an increased AMI risk in the RVO group is a real phenomenon.
It is a limitation that our database does not include tobacco use, which is a risk factor of RVO and AMI. However, we have included a wide range of variables to make our RVO and comparison groups balanced. Kolar P, [31] in a review article, has demonstrated that age, gender, obesity, stroke, hyperviscosity syndrome, and glaucoma are risk factors of RVO. In our study through PS matching method, these variables, as well as the use of antithrombotic drugs, have been well-matched and controlled between the RVO and comparison groups. Besides, risk factors that may confound the relationship between RVO and AMI, such as diabetes, hypertension, hyperlipidemia, congestive heart failure, and chronic renal failure, [31] have been adjusted under the multivariate regression adjustment.
Our study has clinical implications and public health implications. Clinically, our study reminds ophthalmologists to pay more attention to the risk of AMI when they are treating patients with RVO. In particular, patients with RVO who have risk factors for AMI, such as diabetes and hypertension, if are not under proper treatment, should be referred to physicians for early diagnosis and treatment to prevent the occurrence of AMI. From the public health perspective, policy makers are encouraged to enforce the surveillance for AMI risks in patients with RVO.
References
- 1. Sodi A, Giambene B, Marcucci R, Sofi F, Fedi S, Abbate R, et al. (2011) Atherosclerotic and thrombophilic risk factors in patients with ischemic central retinal vein occlusion. Retina 31: 724–729. pmid:21178660
- 2. Kesler A, Shalev V, Rogowski O, Shimron O, Shainberg B, Sela BA, et al. (2008) Comparative analysis of homocysteine concentrations in patients with retinal vein occlusion versus thrombotic and atherosclerotic disorders. Blood Coagul Fibrinolysis 19: 259–262. pmid:18469545
- 3. McGimpsey SJ, Woodside JV, Cardwell C, Cahill M, Chakravarthy U (2009) Homocysteine, methylenetetrahydrofolate reductase C677T polymorphism, and risk of retinal vein occlusion: a meta-analysis. Ophthalmology 116: 1778–1787.e1771. pmid:19729099
- 4. Sofi F, Marcucci R, Bolli P, Giambene B, Sodi A, Fedi S, et al. (2008) Low vitamin B6 and folic acid levels are associated with retinal vein occlusion independently of homocysteine levels. Atherosclerosis 198: 223–227. pmid:17945240
- 5. Di Capua M, Coppola A, Albisinni R, Tufano A, Guida A, Di Minno MN, et al. (2010) Cardiovascular risk factors and outcome in patients with retinal vein occlusion. J Thromb Thrombolysis 30: 16–22. pmid:19705255
- 6. Newman-Casey PA, Stem M, Talwar N, Musch DC, Besirli CG, Stein JD (2014) Risk factors associated with developing branch retinal vein occlusion among enrollees in a United States managed care plan. Ophthalmology 121: 1939–1948. pmid:24953793
- 7. Ponto KA, Elbaz H, Peto T, Laubert-Reh D, Binder H, Wild PS, et al. (2015) Prevalence and risk factors of retinal vein occlusion: the Gutenberg Health Study. J Thromb Haemost 13: 1254–1263. pmid:25894549
- 8. Zhou JQ, Xu L, Wang S, Wang YX, You QS, Tu Y, et al. (2013) The 10-year incidence and risk factors of retinal vein occlusion: the Beijing eye study. Ophthalmology 120: 803–808. pmid:23352194
- 9. Shih CH, Ou SY, Shih CJ, Chen YT, Ou SM, Lee YJ (2015) Bidirectional association between the risk of comorbidities and the diagnosis of retinal vein occlusion in an elderly population: a nationwide population-based study. Int J Cardiol 178: 256–261. pmid:25464265
- 10. Bertelsen M, Linneberg A, Rosenberg T, Christoffersen N, Vorum H, Gade E, et al. (2012) Comorbidity in patients with branch retinal vein occlusion: case-control study. Bmj 345: e7885. pmid:23204001
- 11. Mansour AM, Walsh JB, Henkind P (1992) Mortality and morbidity in patients with central retinal vein occlusion. Ophthalmologica 204: 199–203. pmid:1513551
- 12. Tsaloumas MD, Kirwan J, Vinall H, O'Leary MB, Prior P, Kritzinger EE, et al. (2000) Nine year follow-up study of morbidity and mortality in retinal vein occlusion. Eye (Lond) 14: 821–827.
- 13. Capua MD, Minno MN, Guida A, Loffredo M, Cuccaro C, Coppola A, et al. (2012) Coronary artery disease, cerebral non-fatal ischemic stroke in retinal vein occlusion: an 8-yr follow-up. Nutr Metab Cardiovasc Dis 22: 23–27. pmid:20674314
- 14. Bertelsen M, Linneberg A, Christoffersen N, Vorum H, Gade E, Larsen M (2014) Mortality in patients with central retinal vein occlusion. Ophthalmology 121: 637–642. pmid:24053999
- 15. Hu CC, Ho JD, Lin HC (2009) Retinal vein occlusion and the risk of acute myocardial infarction (correction of infraction): a 3-year follow-up study. Br J Ophthalmol 93: 717–720. pmid:19208680
- 16. Werther W, Chu L, Holekamp N, Do DV, Rubio RG (2011) Myocardial infarction and cerebrovascular accident in patients with retinal vein occlusion. Arch Ophthalmol 129: 326–331. pmid:21402990
- 17. Rim TH, Han JS, Oh J, Kim DW, Kang SM, Chung EJ (2016) Retinal vein occlusion and the risk of acute myocardial infarction development: a 12-year nationwide cohort study. Sci Rep 6: 22351. pmid:26924150
- 18.
The Taiwan National Health Research Institutes. National Health Insurance Research Database [Background]. Available at: http://nhird.org.tw/en/index.html. Accessed 30 June,2017.
- 19.
The Taiwan National Health Research Institutes. National Health Insurance Research Database [Data Subsets]. Available at: http://nhird.org.tw/en/index.html. Accessed 20 June, 2017.
- 20.
The Executive Yuan Gazette Online. Available at: http://gazette.nat.gov.tw/EG_FileManager/eguploadpub/eg018127/ch08/type1/gov70/num36/Eg.htm. Accessed 30 June, 2017.
- 21.
Parsons LS. Perfroming a 1:N Case-Control Match on Propensity Score. Available at: http://www2.sas.com/proceedings/sugi29/165-29.pdf. Accessed 30 June, 2017.
- 22.
Bonow RO, Mann DL, Zipes DP, Peter L. Braunwald's Heart Disease: A Textbook of Cardiovascular Medicine. Philadelphia: Saunders/Elsevier, 2012.
- 23. Jones M, Fowler R (2016) Immortal time bias in observational studies of time-to-event outcomes. J Crit Care 36: 195–199. pmid:27546771
- 24. Rogers S, McIntosh RL, Cheung N, Lim L, Wang JJ, Mitchell P, et al. (2010) The prevalence of retinal vein occlusion: pooled data from population studies from the United States, Europe, Asia, and Australia. Ophthalmology 117: 313–319.e311. pmid:20022117
- 25. Cheung N, Klein R, Wang JJ, Cotch MF, Islam AF, Klein BE, et al. (2008) Traditional and novel cardiovascular risk factors for retinal vein occlusion: the multiethnic study of atherosclerosis. Invest Ophthalmol Vis Sci 49: 4297–4302. pmid:18539932
- 26. Libby P (2001) Current concepts of the pathogenesis of the acute coronary syndromes. Circulation 104: 365–372. pmid:11457759
- 27. Klein R, Klein BE, Moss SE, Meuer SM (2000) The epidemiology of retinal vein occlusion: the Beaver Dam Eye Study. Trans Am Ophthalmol Soc 98: 133–141; discussion 141–133. pmid:11190017
- 28. Lim LL, Cheung N, Wang JJ, Islam FM, Mitchell P, Saw SM, et al. (2008) Prevalence and risk factors of retinal vein occlusion in an Asian population. Br J Ophthalmol 92: 1316–1319. pmid:18684751
- 29. Yasuda M, Kiyohara Y, Arakawa S, Hata Y, Yonemoto K, Doi Y, et al. (2010) Prevalence and systemic risk factors for retinal vein occlusion in a general Japanese population: the Hisayama study. Invest Ophthalmol Vis Sci 51: 3205–3209. pmid:20071683
- 30. Appiah AP, Trempe CL (1989) Risk factors associated with branch vs. central retinal vein occlusion. Ann Ophthalmol 21: 153–155, 157. pmid:2729819
- 31. Kolar P (2014) Risk factors for central and branch retinal vein occlusion: a meta-analysis of published clinical data. J Ophthalmol 2014: 724780. pmid:25009743
- 32. Cheng CL, Chien HC, Lee CH, Lin SJ, Yang YH (2015) Validity of in-hospital mortality data among patients with acute myocardial infarction or stroke in National Health Insurance Research Database in Taiwan. Int J Cardiol 201: 96–101. pmid:26292275
- 33. Cheng CL, Kao YH, Lin SJ, Lee CH, Lai ML (2011) Validation of the National Health Insurance Research Database with ischemic stroke cases in Taiwan. Pharmacoepidemiol Drug Saf 20: 236–242. pmid:21351304
- 34. Cheng CL, Lee CH, Chen PS, Li YH, Lin SJ, Yang YH (2014) Validation of acute myocardial infarction cases in the national health insurance research database in taiwan. J Epidemiol 24: 500–507. pmid:25174915