Methods
We performed a registry study using data from the Registration Network Groningen (RNG), one of several registration networks in the Netherlands. These registration networks carry out research on data derived from the electronic registration of daily patient care in their participating general practices. The Registration Network Groningen was established in 1989, and has three practices in the north of the Netherlands, with an annual population of approximately 30,000 patients.
In the Netherlands, the GP is the gatekeeper in the Dutch health-care system controlling access to specialized medical care. Virtually all non-institutionalized Dutch citizens are registered with a GP so the total practice population represents the general population.
GPs working in the RNG-practices use a structured medical record, in which all patient contacts are registered. This includes reason for encounter, medical diagnosis (according to the International Classification of Primary Care (ICPC)), applied treatment (among which prescriptions, using the Anatomical Therapeutic Chemical (ATC) codes, and referrals), but also cause of death. The database also includes population dynamics, such as date of entry and departure from the database.
From the RNG registration, we selected all data from men aged 50 years and older (age at any time during the study period: 1January 1998–31 December 2008).
Patients with a history of Prostate Cancer, as well as men with a history of CVD were excluded for longitudinal analysis. Access to the patient's medical history was a prerequisite. We collected the following data from the registration: date of birth, date of entry in the study, date of and reason for leaving the registration, GP code, patient contacts (ICPC codes), prescriptions (ATC codes) and ICPC codes attached to these medications, and hospital referrals. From this information, we calculated age from date of birth and number of person years in study. All of the data were anonymised. We received ethical permission to access the Registration Network Groningen from the Medical Ethical Research Board, University Medical Center Groningen (M13.132482).
Definitions
Cardiovascular event is defined as a documented acute myocardial infarction, chronic ischaemic heart disease, transient ischemic attack or stroke in the medical records of the GPs.Table 1 provides the ICPC codes as registered by the RNG-GPs that were used for this study and recode this according to the ICPC. For each participant, we defined CVD-status at baseline (01-01-1998), or at date of study entry, as being present (previous cardiovascular event before inclusion in the study) or absent.
Lower urinary tract symptoms (LUTS) include the sensation of not urinating completely, withholding urinating, and difficulty voiding. This may include having a stop-and-go urinary flow and getting up frequently at night to urinate. There is no specific ICPC code for LUTS, therefore we defined LUTS by all relevant ICPC codes or use of LUTS medication (Table 1).
Erectile dysfunction is defined as the persistent inability to achieve or maintain an erection suficient for satisfactory sexual performance, in this study defined by the ICPC code Y07: symptoms sexual potential.
Statistical Analysis
Descriptive statistics were used to compare the baseline characteristics. Continuous variables are presented in means and confidence intervals, and nominal variables reported in modi. Prevalences (and incidence numbers for LUTS and CVD) for LUTS, CVD, hypertension, Erectile dysfunction, and Diabetes mellitus were calculated by means of prevalence/1000 personyears for the first and last studyyear (1998 and 2008).
Cox proportional hazard regression analysis was used to determine the association between the proportions of CVD (outcome) and LUTS in our population. We used age as time factor in a Cox proportional hazard regression analysis to determine the association between CVD (outcome) and LUTS in our population. In this open cohort, time to event (CVD) was calculated from the patient's date of birth to event (i.e. diagnosis of LUTS) and data of subjects were censored in case leaving the cohort, death, or end of study.
Initially, an unadjusted analysis was performed. This association regression model was subsequently corrected for confounders. A covariate was considered a confounder in the event the beta coefficient of LUTS changed by 10 or more percent.
In this analysis, the following potential confounders were tested: hypertension, diabetes mellitus, obesity, dyslipidaemia, depression, and antihypertensive (ACE-inhibitors, all antagonists, beta blockers, calcium antagonists, and diuretics), and statins.
In all analyses, multiple dummy variables were created for the analyses of categorical data. In addition, model assumptions were tested for compliance. All analyses were performed using SPSS version 16 based on a two-sided 0.05 significance level.