Data sources
Danish nationwide administrative registries were used to collect data at individual levels by use of a unique personal identification number which is assigned to all residents in Denmark. For this study, three nationwide registers were linked on an individual level to obtain information on all Danish residents aged 18 years or older and who had claimed prescriptions of bromocriptine between 1 July 1995 to 26th June 2018. The Danish National Patient Register holds information about all admissions to Danish hospitals since 1978, and outpatients visits since 1990, including diagnoses coded according to the International Classification of Diseases, eighth edition (ICD-8) and ICD-10 and all procedures, surgeries included, are coded according to the Nordic Medico-Statistical Committee (NOMESCO) classification [15]. The Danish National Prescription Registry, which holds information on all claimed prescriptions in Denmark since 1995, ensuring complete data on date of dispensing, strength of the tablets, number of pills dispensed and of cause the prescribed drug grouped according to the Anatomical Therapeutic Chemical (ATC) codes [16]. And the civil register contains information about vital status of all Danish residents [17]. Register-based studies in which individuals cannot be identified do not require ethical approval in Denmark.
Study population
We included female patients aged 18 years or older treated with bromocriptine and matched controls identified from the background population. Men treated with bromocriptine were excluded due to different indications for treatment, older age at treatment initiation, and differences in clinical characteristics [18]. Patients entered the study cohort when claiming a second prescription of bromocriptine with this date defining the study inclusion date. Each patient was matched by age and sex to 5 matched controls, and patients and controls were followed from index date until occurrence of the first of following events: hospitalization for heart valve disease or an outpatient contact with a heart valve diagnosis, death, or end of study period (31th June 2018). We excluded patients who prior to study entry, had a history of: heart valve disease, rheumatic heart disease, rheumatic fever, chronic heart failure, congenital heart and valve disease, endocarditis, carcinoid syndrome, Parkinson’s disease, pergolide-, levodopa-, cabergoline-, quinagolide-use, or patients who had been treated with medications that may induce fibrosis (fenfluramine, dexfenfluramine, ergotamine) [19]. To ensure patients were naive bromocriptine users, we excluded patients who claimed a prescription in the first half of 1995 (first year of full data coverage in the registry). All patients were individually risk set matched by age, sex and year of inclusion with 5 controls from the background population. In a sensitivity analysis we restricted the population to patients who claimed five or more prescriptions with the date of the fifth prescription claim set as index date. We also repeated the analysis with a more detailed matching of each patient by age, sex, prior diabetes, deep venous thrombosis, bleeding, chronic renal failure, ischemic heart disease, stroke, chronic obstructive lung disease and ongoing use of beta blockers, loop diuretics and aspirin, with two controls from the background population.
The majority of patients (82%) had no prior in or outpatient hospital diagnosis of hyperprolactinaemia before initiating bromocriptine treatment. We had no access to patient charts from the general practitioner but we assume these patients were treated by their general practitioner as we included other in and out hospital diagnoses which could indicate bromocriptine treatment.
Comorbidities and concomitant pharmacotherapy were identified by assessing all hospital discharge codes prior to index and information on claimed prescriptions one year before index date. ICD, procedure, and ATC codes used in this study are listed in the Additional file 1: Tables 2 and 3.
Outcomes
The primary outcome was development of heart valve disease as defined by an outpatient clinic visit or hospital admission with a diagnosis of heart valve disease or valvular heart surgery. As secondary outcomes we assessed the risk of valvular heart surgery as proxy for the severity of heart valve disease.
Statistical analysis
Baseline characteristics for bromocriptine-treated patients and the control cohort were described by use of numbers and percentages for categorical variables and medians and interquartile ranges for continuous variables. Differences between the bromocriptine group and controls were obtained by use of Chi-square test for categorical variables.
Cumulative incidence curves for heart valve disease, with death as a competing risk, were estimated, and differences between patients and control cohort were compared using Gray’s test. Crude incidence rates were calculated. Cause-specific Cox regression were used to compare risk of heart valve disease between patients and the control cohort. Cox regression analyses were adjusted for age at index, year of inclusion and history of hypertension, ischemic heart disease, acute myocardial infarction and diabetes mellitus. The variables adjusted for were chosen based on clinical relevance and known prognostic importance in heart valve disease. The variable age and comorbidities (hypertension, ischemic heart disease, acute myocardial infarction and diabetes) were tested for interactions with the use of bromocriptine in relation to both outcomes and, unless stated otherwise, found absent. Interactions were considered significant if they yielded a p-value < 0.05. Log (-log(survival)) curves were used to evaluate the proportional hazard assumption. The assumption of linearity of age was tested by including a variable of age squared.
Furthermore, the interaction between patients with and without an ICD-code for hyperprolactinemia in addition to use of bromocriptine were tested for all outcomes.
In a supplementary analysis, we calculated the accumulated dosage of claimed bromocriptine for patients with and without a heart valve disease diagnosis. Differences between the groups were obtained by use of Wilcoxons test.
Results were considered significant if p < 0.05. The SAS statistical software package, version 9.4 (SAS Institute, Cary, North Carolina; USA) and R, version 3.5.0 (R development Core Team) were used for all analyses.