Association of Long-term Use of Low-Dose Aspirin as Chemoprevention With Risk of Lung Cancer

Key Points

Question  Is long-term use of low-dose aspirin associated with reduced risk of lung cancer, and if so, which populations may derive the greatest benefit?

Findings  This cohort study using data from 12 969 400 participants in the Korean National Health Information Database found that the use of low-dose aspirin for more than 5 years was associated with a modest risk reduction of incident lung cancer, with the strongest association observed among elderly participants and among people without diabetes.

Meaning  Intake of low-dose aspirin for more than 5 years may reduce the risk of incident lung cancer, particularly among the elderly and among people without diabetes.

Importance  Low-dose aspirin use for chemoprevention of lung cancer risk remains controversial.

Objectives  To investigate the association between low-dose aspirin use and lung cancer risk, and to identify specific subgroups that may derive the most benefit from low-dose aspirin use.

Design, Setting, and Participants  This nationwide, retrospective, cohort study used data from the Korean National Health Information Database from 2002 to 2015. Data analyses were performed from October 2016 to December 2018. Eligible participants (n = 12 969 400) were people aged 40 to 84 years who had undergone national health screening between 2009 and 2010 and had no history of lung cancer between 2006 and 2010 and no standard-dose aspirin use for 6 months between 2002 and 2010.

Main Outcomes and Measures  The duration of low-dose aspirin use between January 2002 and December 2010 was calculated for each participant. Lung cancer was defined as the first recorded diagnosis of lung cancer—using International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes and expanding benefit coverage—between January 2011 and December 2015.

Results  A total of 63 040 participants with a mean (SD) age of 66.4 (9.3) years received a diagnosis of lung cancer. Of these, 45 156 (71.6%) were men. The incidence rate of lung cancer was 98.8 per 100 000 person-years. The duration of low-dose aspirin use was none for 10 987 417 participants (84.7%), 1 to 2 years for 750 992 participants (5.8%), 3 to 4 years for 506 945 participants (3.9%), 5 to 6 years for 371 062 participants (2.9%), 7 to 8 years for 240 528 participants (1.9%), and 9 years for 112 456 participants (0.9%). Compared with no aspirin use, 5 to 6 years (adjusted hazard ratio, 0.96 [95% CI, 0.92-0.99]), 7 to 8 years (adjusted hazard ratio, 0.94 [95% CI, 0.90-0.99]), and 9 years (adjusted hazard ratio, 0.89 [95% CI, 0.84-0.94]) of aspirin use were significantly associated with reduced lung cancer risk. After stratified analysis, a significant reduction of lung cancer risk was observed among people aged 65 years or older and among people without diabetes.

Conclusions and Relevance  Although the use of low-dose aspirin for more than 5 years was associated with decreased risk of lung cancer, particularly among elderly participants and among people without diabetes, the observed effect size was quite modest. Future prospective studies are needed to determine whether there is a causal association.

Lung cancer was the most commonly diagnosed cancer (1.8 million cases; 12.9% of all cancer cases) and the most common cause of cancer death (1.6 million deaths; 19.4% of all cancer deaths) in 2012 worldwide.¹ The risk of lung cancer remains elevated after smoking cessation,² and 15% of men and 53% of women with lung cancer are lifelong nonsmokers.³ Given this significant health burden, additional approaches for preventing lung cancer have been suggested, including regular aspirin use. Previous studies of aspirin use in patients with cardiovascular disease have suggested that low-dose aspirin use is beneficial in reducing the risk of cancer in a duration-dependent manner.^4,5

The antineoplastic effect of aspirin may be associated with the inhibition of cyclooxygenase-1 (COX-1)–mediated platelet activation.⁶ Inhibition of arterial thrombosis by aspirin impedes the development and progression of certain cancers.⁷ Cyclooxygenase-2 is known to be involved in the inflammatory process, and inhibition of COX-2–derived prostaglandin E₂ formation by aspirin induces inhibition of cell proliferation and reduction of angiogenesis and immunomodulation through the production of lymphocytes in the peripheral blood.⁸ Cyclooxygenase-2 is expressed at high levels in human lung cancer tissue, especially in adenocarcinoma.⁹

According to the results of 2 meta-analyses, despite significant findings on aspirin use and lung cancer risk among case-control studies, the chemopreventive effects of aspirin on lung cancer risk in cohort studies were controversial.^10,11 A pooled analysis of randomized trials with long-term follow-up showed that daily aspirin use reduced death due to lung cancer.¹² Thus, the association between low-dose aspirin use and the risk reduction of lung cancer, the appropriate duration of aspirin use for chemoprevention, and the specific subgroups expected to benefit more from low-dose aspirin use are still inconclusive.

The present study aimed to examine the association between long-term use of low-dose aspirin and the risk of incident lung cancer through a large-scale, nationwide, population-based, cohort study. This study also aimed to investigate the appropriate duration of aspirin intake for chemoprevention and to identify specific subgroups that may benefit more from low-dose aspirin use as chemoprevention.

This nationwide, retrospective, cohort study used data from the Korean National Health Information Database (KNHID) collected between January 1, 2002, and December 31, 2015. Data analysis was performed from October 2016 to December 2018, and preliminary data analysis using a sample cohort was conducted from July 2015 to September 2016. The KNHID includes information about participants who visited hospitals under the Korean National Health Insurance Service program, which covers all legal residents of the Republic of Korea.¹³ The KNHID consists of 5 databases (eligibility database, national health screening database, health care utilization database, long-term care insurance database, and medical institution database) and includes public data on health care use, health screening, sociodemographic variables, and mortality.^13,14 The types and duration of component databases used in this study are illustrated in Figure 1. The present study was approved by the Institutional Review Board of Ewha Womans University Medical Center and the National Health Insurance Service, which also waived the need for obtaining informed patient consent because the KNHID was constructed after anonymization according to strict confidentiality guidelines. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline¹⁵ for reporting the results of the present study.

The study included 12 998 062 participants, aged 40 to 84 years during 2010, who underwent national health screening between January 1, 2009, and December 31, 2010. Of them, 18 804 participants were excluded because they had received a diagnosis of lung cancer between January 1, 2006, and December 31, 2010. In addition, 9858 participants were excluded because they were prescribed standard doses (≥325 mg)¹⁰ of aspirin for more than 6 months between January 1, 2002, and December 31, 2010 (Figure 2; eTable 1 in the Supplement).

Using the heath care utilization database, participants with lung cancer during follow-up between January 1, 2011, and December 31, 2015, were defined as those with International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes C33.x (malignant neoplasm of trachea) or C34.x (malignant neoplasm of bronchus and lung), in the principal or first additional diagnosis with the code for expanding benefit coverage, which is a Korean policy that commenced in September 2005 (eTable 2 in the Supplement). This policy reduces the medical expenses of patients with catastrophic illnesses, such as cancer or cardiovascular disease or severe burn injury.

For the exposure assessment, we used the health care utilization database collected between January 1, 2002, and December 31, 2010. In the present study, low-dose aspirin was defined as a dose of 100 mg or less. The prescription data from eligible participants showed that 94.7% of the prescriptions instructed taking 1 tablet at a time, and 94.6% of the prescriptions directed taking aspirin once daily (eTable 3 in the Supplement).

Participants who were prescribed low-dose aspirin for at least 104 days were defined as low-dose aspirin users in each year, which means that a person took at least 2 aspirin tablets per week during a year.^16,17 Those who were prescribed aspirin for less than 104 days in a year were defined as nonusers of low-dose aspirin in each year. Nonusers of low-dose aspirin also included those who had never used low-dose aspirin. The duration of low-dose aspirin use was then categorized into 2-year ranges: none, 1 to 2, 3 to 4, 5 to 6, 7 to 8, and 9 years (Table 1).

For information on health behaviors (smoking status, alcohol consumed, and physical activity), physical examination (body mass index [BMI]; calculated as weight in kilograms divided by height in meters squared), and fasting serum glucose levels, we used the most recent national health screening results from the national health screening database collected between January 1, 2009, and December 31, 2010. Data on diagnoses of type 1 and type 2 diabetes were collected using ICD-10 codes (E10.x and E11.x) from the 2010 health care utilization database. Participants reported smoking status as nonsmokers, past smokers, or current smokers. Depending on pack-years of smoking, the past smokers and current smokers were reclassified into nonsmokers (0 pack-years of smoking) or smokers of less or more than 30 pack-years of smoking. Alcohol consumption was categorized as no consumption or as 1 to 7, 8 to 14, and more than 15 units of alcohol consumed in a week. The frequency of moderate or vigorous exercise per week was categorized as no exercise, 1 to 4 times per week, and more than 5 times per week. Participant BMI was categorized as less than 18.5 (underweight), 18.5 to 22.9 (normal weight), 23.0 to 24.9 (overweight), 25.0 to 29.9 (obese), and 30 or greater (morbidly obese). Patients with diabetes were defined by their fasting serum glucose level (≥126 mg/dL; to convert to millimoles per liter, multiply by 0.0555) or by a diagnosis of type 1 or type 2 diabetes.

Information regarding family history of cancer was collected from the national health screening database between January 1, 2002, and December 31, 2008. The Charlson comorbidity index score was calculated using ICD-10 codes obtained from the health care utilization database for 2010.¹⁸ Information on variables of age (years), sex, income (quintile), and residential area (metropolitan, city, and rural areas) was obtained from the eligibility database for 2010.

All participants were followed up from baseline (January 1, 2011) until the date of lung cancer diagnosis (event of interest), death by any reason (competing event), censoring due to loss of health insurance eligibility (eg, loss of nationality or moving overseas, or nonpayment of insurance cost), or the end of the follow-up period (December 31, 2015), whichever occurred first (eTable 4 in the Supplement). The follow-up duration in days was used as the time scale.

Competing risk analysis (Fine and Gray model)¹⁹ was performed with consideration of death as a competing event. This analysis was conducted to reduce the bias from competing events that either hinder or alter the chance of an event of interest occurring.²⁰

For the multivariate analysis, the association between duration of low-dose aspirin use and the risk of incident lung cancer was evaluated using 2 models, an age- and sex-adjusted model and a fully adjusted model. The fully adjusted model used possible confounders of age, sex, income, residential area, pack-years of smoking, quantity of alcohol consumed, frequency of moderate or vigorous exercise, BMI, history of diabetes, family history of cancer, and Charlson comorbidity index score; age was a continuous variable and other variables were categorical variables, as shown in Table 2. Missing values in the covariates were included in the model as a missing indicator category. A trend test was performed with the duration of low-dose aspirin use treated as a numeric variable in the models.

Stratified analyses examined differences in associations between long-term use of low-dose aspirin and lung cancer risk, defined by sex (men or women), age at baseline (<65 years or ≥65 years), BMI (<25 or ≥25), history of diabetes (yes or no), and pack-years of smoking (nonsmoker, <30 pack-years, or ≥30 pack-years).

We conducted 8 sensitivity analyses as follows. (1) A conventional survival analysis (Cox proportional hazards model)²¹ was conducted to compare previous cohort studies.^22,23 (2) We analyzed the association in the 1:1 propensity score-matched cohort between aspirin users and nonusers. (3) Because some participants were taking aspirin during the follow-up period (eTable 5 in the Supplement), we analyzed the association between incident lung cancer and the duration of aspirin use from 2002 to the year in which the event occurred or the follow-up was censored. (4) Because 20% of the family history data were missing, we conducted an analysis in the fully adjusted model that excluded the family history variable. (5) Several operational definitions for the assessment of low-dose aspirin intake between January 2002 and December 2010 were used for the fully adjusted model. First, participants who were prescribed low-dose aspirin for at least 156, 208, or 260 days in each year were defined as a person who took at least 3, 4, or 5 aspirin tablets per week, respectively. Second, the total days of low-dose aspirin prescribed was used. (6) In the fully adjusted model, we used smoking status instead of pack-years of smoking for the adjustment. (7) We analyzed the fully adjusted model after excluding cases that occurred during the first year of follow-up. (8) We conducted a subgroup analysis with a combined 7 to 9 years of low-dose aspirin use group.

All statistical analyses were conducted using SAS, version 9.4 (SAS Institute Inc). The criterion for statistical significance was a 2-sided P < .05.

In total, 12 969 400 participants were enrolled in this study (Figure 2). The characteristics of the excluded and included participants are shown in eTable 1 in the Supplement.

The duration of low-dose aspirin use was none for 10 987 417 participants (84.7%), 1 to 2 years for 750 992 participants (5.8%), 3 to 4 years for 506 945 participants (3.9%), 5 to 6 years for 371 062 participants (2.9%), 7 to 8 years for 240 528 participants (1.9%), and 9 years for 112 456 participants (0.9%). Longer duration of low-dose aspirin use was associated with male sex, older age, high income, high BMI, history of diabetes, and high Charlson comorbidity index score (Table 2).

Total person-years of follow-up were 63 787 432.9 person-years for the lung cancer outcome. Among the 12 969 400 participants, 63 040 (0.5%) incident cases of lung cancer were detected. The incidence rate of the event was 98.8 per 100 000 person-years. The mean (SD) age of the patients with lung cancer was 66.4 (9.3) years, and 45 156 (71.6%) were men and 17 884 (28.4%) were women (eTable 6 in the Supplement).

The results of the fully adjusted model are presented in Figure 3. In all participants, compared with no use, low-dose aspirin intake for more than 5 years was significantly associated with reduced lung cancer risk in a duration-dependent manner (adjusted hazard ratio [aHR]: 5-6 years’ use, 0.96 [95% CI, 0.92-0.99]; 7-8 years’ use, 0.94 [95% CI, 0.90-0.99]; and 9 years’ use, 0.89 [95% CI, 0.84-0.94]; P < .001 for trend).

Long-term use of low-dose aspirin was significantly associated with reduced risk of lung cancer among both men (aHR: 5-6 years’ use, 0.97 [95% CI, 0.93-1.01]; 7-8 years’ use, 0.96 [95% CI, 0.91-1.01]; 9 years’ use, 0.88 [95% CI, 0.82-0.94]; P < .001 for trend) and women (aHR [95% CI, 95% CI]: 5-6 years’ use, 0.96 [95% CI, 0.89-1.03]; 7-8 years’ use, 0.91 [95% CI, 0.83-0.99]; 9 years’ use, 0.92 [95% CI, 0.81-1.04]; P = .002 for trend). Long-term use of low-dose aspirin was also significantly associated with reduced risk of lung cancer among both nonobese (aHR: 5-6 years’ use, 0.92 [95% CI, 0.88-0.97]; 7-8 years’ use, 0.95 [95% CI, 0.90-1.00]; 9 years’ use, 0.83 [95% CI, 0.76-0.90]; P < .001 for trend) and obese participants (aHR: 5-6 years’ use, 0.98 [95% CI, 0.92-1.04]; 7-8 years’ use, 0.90 [95% CI, 0.84-0.96]; 9 years’ use, 0.94 [95% CI, 0.85-1.03]; P = .006 for trend). Long-term use of low-dose aspirin was significantly associated with reduced risk of lung cancer among nonsmokers (aHR: 5-6 years’ use, 0.94 [95% CI, 0.89-1.00]; 7-8 years’ use, 0.95 [95% CI, 0.89-1.02]; 9 years’ use, 0.89 [95% CI, 0.81-0.98]; P = .003 for trend), among smokers with less than 30 pack-years (aHR: 5-6 years’ use, 0.99 [95% CI, 0.92-1.06]; 7-8 years’ use, 0.91 [95% CI, 0.83-0.99]; 9 years’ use, 0.90 [95% CI, 0.80-1.01]; P = .03 for trend), and among smokers with more than 30 pack-years (aHR: 5-6 years’ use, 0.97 [95% CI, 0.91-1.04]; 7-8 years’ use, 0.97 [95% CI, 0.90-1.05]; 9 years’ use, 0.89 [95% CI, 0.80-0.99]; P = .04 for trend). Long-term use of low-dose aspirin was significantly associated with reduced risk of lung cancer among participants aged 65 years or more (aHR: 5-6 years’ use, 0.95 [95% CI, 0.92-1.00]; 7-8 years’ use, 0.93 [95% CI, 0.89-0.98]; 9 years’ use, 0.87 [95% CI, 0.82-0.93]; P < .001 for trend) but not among those aged less than 65 years (aHR: 5-6 years’ use, 0.93 [95% CI, 0.86-1.00]; 7-8 years’ use, 0.99 [95% CI, 0.90-1.08]; 9 years’ use, 0.99 [95% CI, 0.85-1.14]; P = .09 for trend). Long-term use of low-dose aspirin was significantly associated with reduced risk of lung cancer among people without diabetes (aHR: 5-6 years’ use, 0.96 [95% CI, 0.92-1.00]; 7-8 years’ use, 0.94 [95% CI, 0.89-0.99]; 9 years’ use, 0.87 [95% CI, 0.81-0.94]; P < .001 for trend) but not among those with diabetes (aHR: 5-6 years’ use, 0.97 [95% CI, 0.91-1.04]; 7-8 years’ use, 0.98 [95% CI, 0.91-1.07]; 9 years’ use, 0.95 [95% CI, 0.85-1.06]; P = .37 for trend).

The results of the sex- and age-adjusted model are presented in eTable 7 in the Supplement. Sensitivity analyses 1 through 7 for all participants showed trends consistent with the main results (eTables 8 through 17 in the Supplement). Sensitivity analysis 8 showed that 7 to 9 years of low-dose aspirin use among women, obese people, and smokers (<30 pack-years) had a significant association with a lower risk of lung cancer (eTable 18 in the Supplement).

This study showed that low-dose aspirin use for more than 5 years was associated with a significantly lower incidence of lung cancer. Furthermore, except for younger people and participants with diabetes, the risk of incident lung cancer decreased with longer duration of low-dose aspirin use. The effective duration of low-dose aspirin use and proper aspirin dose are clinically important factors to avoid adverse effects of aspirin. A previous study reported that the risk of gastrointestinal bleeding increased with increasing frequency of weekly standard-dose aspirin use, and that the adverse effect was more closely associated with aspirin dose than duration of aspirin use.²⁴ To our knowledge, the association between long-term use of low-dose aspirin and the decreased risk of incident lung cancer is a novel finding with implications for a new strategy in lung cancer prevention. In addition, the present study provided evidence for modifying the association of decreased risk by age and history of diabetes, whereas sex, obesity, and smoking history did not modify the association between low-dose aspirin use and the risk of incident lung cancer.

According to previous epidemiologic studies, the association between low-dose aspirin use and risk of lung cancer remains controversial. A case-control study reported that intake of aspirin regularly (≥3 standard [325 mg] tablets/week) for more than 5 years was inversely associated with the risk of lung cancer (odds ratio, 0.36 [95% CI, 0.22-0.58]).²⁵ A pooled analysis of randomized trials with 20 years’ follow-up showed that daily aspirin use reduced deaths due to lung cancer (HR, 0.71 [95% CI, 0.58-0.89]).¹² The beneficial effect increased with the duration of aspirin use and at ages 65 years or older.¹² By contrast, the benefit was not associated with aspirin dose (75 mg and upward), sex, or smoking.¹² Although the effect size of our study was quite modest compared with pooled analyses, other results of pooled analyses were similar to ours. The Nurses’ Health Study demonstrated a nonsignificant 16% lower risk of lung cancer for regular (1 or 2 tablets per week) aspirin users among women, whereas the risk of lung cancer increased by 55% among aspirin users who took 15 or more tablets per week.²² A nested case-control study also showed 53% increased risk of lung cancer was reported among participants who had been prescribed aspirin of 150 mg or more per day for at least 1 year.²⁶ The Women’s Health Study reported no significant association between long-term alternate-day aspirin use (100 mg) and lung cancer risk among women after a median 10 years of follow-up.²⁷ A cohort study reported that daily use of adult-strength aspirin (≥325 mg) for more than 5 years was not significantly associated with the risk of lung cancer.²³ A 32-year prospective cohort study following up 88 084 women from the Nurses’ Health Study and 47 881 men from the Health Professionals Follow-up Study reported no significant association between regular use of aspirin and reduced risk of lung cancer.²⁸

The present study revealed that the use of low-dose aspirin was not associated with a decreased risk of incident lung cancer among individuals with diabetes and participants younger than 65 years old. People with diabetes have increased platelet turnover and protein glycosylation,²⁹ which lowers COX-1 inhibition and aspirin-mediated acetylation.³⁰ This mechanism has been suggested as the cause of aspirin resistance in patients with diabetes. A previous pooled analysis of randomized trials demonstrated that the association of aspirin with reduced risk of death due to lung cancer was stronger with increasing age.¹² The free fraction of salicylate (aspirin) in plasma increases more with age and also significantly reduces clearance of unbounded salicylate in elderly individuals.³¹ However, the precise biological mechanism for this association with age was not clear.

In the present study, low-dose aspirin use for 7 to 8 years, but not 9 years, among women, obese people, or individuals who smoke less than 30 pack-years was associated with a lower risk of lung cancer. Plausible explanations for this result are the relatively small sample size of the 9-year use group (eTable 18 in the Supplement), or the possibility of aspirin resistance in women, obese people, and smokers caused by the presence of estrogen,^32,33 inflammation,³⁴ and thrombus,^35,36 respectively.

Our study had several strengths. First, this was a large-scale, nationwide, population-based, cohort study, and selection bias was reduced by including the majority of the population, which increased statistical power. In addition, a cohort study tends to have better data quality than a case-control study. Second, this study featured a competing risk analysis that represents the absolute risk in the real-world by considering the competing risk.³⁷ Long-term aspirin users had higher mortality rates than nonusers in this study (eTable 4 in the Supplement); therefore, the event of interest (occurrence of lung cancer) could be masked or altered by a competing event, such as death. Third, we conducted several sensitivity analyses. All results showed consistency, which increased the reliability of our results. Fourth, we used the KNHID, which stores data that reflect clinical practice and real-world settings, and therefore the recall bias was reduced when compared with data collected using a questionnaire. Fifth, we used codes from the expanded insurance benefit coverage to define lung cancer. The use of these codes increased the accuracy of lung cancer diagnosis in our study (eTable 2 in the Supplement) because the diagnosis was confirmed by a physician to avail the benefit of expanded insurance coverage.

This study also had some limitations. First, it was a retrospective, nonrandomized, cohort study. However, the KNHID is a large-scale, nationwide, population-based database that provides avoidance of selection bias. Second, the effect size of this study was quite modest. Third, the KNHID database did not permit the distinguishing of cancer subtypes, and we were unable to obtain data to investigate the association between histologic types of lung cancer and aspirin use. Previous studies have reported that chemoprevention was limited, particularly for adenocarcinoma. Fourth, we could not capture the duration of low-dose aspirin use before enrollment. As shown in eTable 5 in the Supplement, those who took low-dose aspirin between 2002 and 2010 were more likely to take low-dose aspirin between 2011 and 2015. Therefore, the duration of low-dose aspirin use in the present study could have been underestimated. Fifth, study results have shown year to year variations by the type of analysis and type of subgroup. Because this study was an observational study using census data, the size and the characteristics of the participants varied according to the years of low-dose aspirin use. We performed adjustments of several confounders, stratification analysis, and competing risk analysis to reduce biases. However, it may be insufficient for clarifying the appropriate duration of aspirin use and subgroups having greater chemopreventive benefit.

The use of low-dose aspirin for more than 5 years may be associated with a lower risk of lung of cancer among the elderly and among people without diabetes, regardless of sex, obesity, and smoking history. In addition, the longer use of low-dose aspirin and the lower risk of lung cancer were mutually associated, as clarified by several sensitivity analyses. However, given the limitations of the present study, further prospective studies are needed to establish whether a causal association exists between aspirin use and risk of lung cancer.

Accepted for Publication: January 10, 2019.

Published: March 1, 2019. doi:10.1001/jamanetworkopen.2019.0185

Correction: This article was corrected on March 22, 2019, to fix an error in the spelling of the first author’s first name.

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2019 Ye S et al. JAMA Network Open.

Corresponding Authors: Eun Mi Chun, MD, PhD, Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, Ewha Womans University, 1071, Anyangcheon-ro, Yangcheon-gu, Seoul 07985, Republic of Korea (cem@ewha.ac.kr); Eun-Hee Ha, MD, PhD, Department of Occupational and Environmental Medicine, School of Medicine, Ewha Womans University, 1071, Anyangcheon-ro, Yangcheon-gu, Seoul 07985, Republic of Korea (eunheeha@ewha.ac.kr).

Author Contributions: Dr Ye had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Ha and Chun contributed equally to this work.

Concept and design: Ye, M. Lee, Ha, Chun.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Ye, M. Lee, Ha, Chun.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Ye, M. Lee, Ha.

Administrative, technical, or material support: Ye, Chun.

Supervision: Ye, M. Lee, Ha, Chun.

Conflict of Interest Disclosures: None reported.

Additional Information: The data can be obtained by applying to the National Health Insurance Sharing Service of Korea; the computing code can be obtained by contacting the corresponding author. The fee for obtaining data is proportional to amount of data requested.