Introduction of the Taiwan NHIRD
The Taiwan National Health Insurance program provides health care to 99% of the population of 24 million and is linked to 97% of hospitals and clinics in Taiwan (http://nhird.nhri.org.tw/en/) [13]. The NHIRD is a considerable source of information on medical facilities, inpatient and outpatient order details, dental services, drug prescriptions, patient care, and other paramedical registration files (eg, payment, regions, and catastrophic illness), but it does not include laboratory data. Diagnoses are entered based on the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). The NHIRD longitudinal data collection started in 2000, and 1 000 000 beneficiaries were randomly sampled. The database contains the entire original claims data from 1997 to 2013. The study design and protocol were approved by the Institutional Review Board of Chang Gung Memorial Hospital (No. 201800664B1).
Study population
We selected female patients with endometriosis (ICD-9-CM code 617) from 1 million individuals in the NHIRD between January 01, 1997 and December 31, 2013 to conduct a retrospective nationwide population-based cohort study. Women aged between 18 and 50 with a record of at least 3 medical visits or 1 admission for endometriosis were included in the study. To confirm the diagnosis of endometriosis, we linked the relevant clinical information with the corresponding vaginal ultrasound, surgical findings, and medications. Those patients with clinically suspicious diagnosis of endometriosis without image- or procedure-proven evidence were not included. We excluded patients aged <18 or >50 years, male patients, patients experiencing menopause (627), patients with missing or incomplete baseline characteristics data, and patients with a follow-up duration of less than 1 year. For the first purpose of the study, that is, the incidental MACCE in women with endometriosis, patients with diagnoses of acute myocardial infarction (AMI; 410-411), heart failure (HF; 428), or cerebrovascular accident (CVA; 430-436) in the beginning or prior to endometriosis were excluded. Thus, we identified a total of 17 543 women with endometriosis (EM group) at the beginning of the study.
A comparison cohort of 70 172 non-endometriosis (non-EM) women was selected from January 01, 1997 and December 31, 2013 after matching the study cohort at a 4:1 ratio with age and socioeconomic background, including income and urbanization level (Fig. 1). Urbanization was categorized into 4 levels, from level 1 indicating the least urbanized level (country) to level 4 indicating the most urbanized level (city). The insurance taxable income level was stratified into 4 categories based on the monthly insurance payment of each insured participant (ie, level 1: none; level 2: 1-15 840; level 3: 15 841-25 000; and level 4: > 25 000 New Taiwan dollars per month).
Definition
The MACCE were divided into 2 disease entities, namely CVD and CVA. Major CVD involved AMI or HF. CVA encompassed acute ischemic (433-436) or hemorrhagic stroke (430-432). Moreover, information on medications was acquired based on the World Health Organization Anatomical Therapeutic Chemical classification (WHO ATC codes) system. Medical treatment for endometriosis comprised danazol, gestrinone, oral contraceptives, and GnRH agonists. Patients with prescriptions exceeding a month were considered medication users. By contrast, we considered patients with a regimen duration of less than a month as nonusers, because drugs might have been prescribed temporarily for symptomatic relief, diagnostic test, or purposes other than standard treatment for endometriosis. Surgical treatment for endometriosis mainly consisted of therapeutic laparotomy and laparoscopy (ICD-9-CM procedure codes 541 and 542). The time frame of medical and surgical treatment was from the diagnosis of endometriosis to the end of follow-up.
Identification of symptoms, comorbidities, location, and outcomes
We conducted a retrospective longitudinal study. The date of initial diagnosis of endometriosis was defined as the index date. To clarify the association between MACCE and endometriosis, potential risk factors for MACCE and possible confounders were retrieved for further analysis. We assessed gynecological presentations and relevant comorbidities for each patient during the follow-up period (from the diagnosis of endometriosis to the end of study), including dysmenorrhea (ICD-9-CM code 625.3), amenorrhea (ICD-9-CM code 626.0), infertility (ICD-9-CM code 628), ovarian cancer (ICD-9-CM code 183), hypertension (ICD-9-CM codes 401-405), diabetes (ICD-9-CM code 250), dyslipidemia (ICD-9-CM code 272), gout (ICD-9-CM code 274), chronic ischemic heart disease (ICD-9-CM codes 412-414 and 429.2), peripheral vascular disease (ICD-9-CM codes 440, 443.9, 444.0, 444.2, 444.8, 444.9, 447.8, 447.9, 445.0, and 445.02), atrial fibrillation (ICD-9-CM code 427.31), and chronic kidney disease (ICD-9-CM code 585). Endometriosis was further divided into ovarian and extraovarian groups according to its occurrence location.
With respect to study outcomes, the day of the first event occurrence was defined as the event date. The diagnosis of MACCE, including AMI, HF, and CVA, was confirmed by 3 consecutive records of outpatient visits or a one-time diagnosis on admission with the corresponding standard treatment during the whole study period. Endpoints between the 2 groups were censored, whereas clinical events were identified by ICD-9-CM codes and the event date. We investigated the frequency and incidence of MACCE in both EM and non-EM Asian groups. Additionally, we analyzed whether the risk of MACCE was higher in the EM group than in the normal population in Asian women. Furthermore, we studied the differences in the risk of MACCE among diseased population according to none, surgical alone, medical alone, or combination treatment for endometriosis.
Statistical analysis
The ratios of demographic data and comorbidities between the study cohort (EM group) and the matched control cohort (non-EM group) were compared using the independent t test and chi-square test, as appropriate. The incidence rates and 95% CIs of MACCE were calculated for the entire follow-up period. We also examined the outcomes of MACCE and stratified them by subgroups according to age, amenorrhea diagnosis, and comorbidities. Additionally, the Kaplan–Meier method was used to estimate the cumulative incidences, and a log-rank test was performed to examine differences between disease and non-disease groups. Using Cox proportional hazards regression models, we analyzed hazard ratios (HRs) and corresponding 95% CIs after adjusting for age, medications, surgery types, and associated comorbidities. To further understand the effect of surgical or medical treatment for endometriosis, a multivariate Cox regression model was used to examine the adjusted HRs of endometriosis for the occurrence of MACCE in the subgroups. A two-tailed P value of <.05 was considered statistically significant. All analyses were conducted using the SAS statistical software program (Version 9.4; SAS Institute, Cary, NC, USA).