A population-based cohort study was conducted using record-linkage analysis of three databases: drug prescriptions database, civil registry and hospital discharge records (HDR) (data on 6.5 million subjects across 22 local Italian Health Authorities).
All Italian citizens have access to health care services; medical and pharmaceutical services are provided for free or at a minimum charge as part of the National Health Service (NHS).
The Italian national drug database includes the prescriptions reimbursed by the Italian NHS; drugs are coded according to the Anatomical Therapeutic Chemical (ATC) Classification [15] and qualified with respect to dosage, and the date of the first and subsequent prescriptions from which data on adherence can be derived.
This cohort was linked with HDR which includes information on primary diagnoses and up to 5 coexisting conditions, performed procedures, dates of hospital admission and discharge. Diagnoses are classified according to the International Classification of Diseases-Ninth Revision, Clinical Modification (ICD9-CM) [16].
The Italian civil registry provides demographic information.
This study’s methodology has been widely used to produce reliable epidemiological surveys [12,17, 18]. The analysis was carried out in strict compliance with the national Italian regulations for the full protection of the privacy rights of the subjects included in the databases. According to the Italian law, no ethical approval is required to perform this type of analysis and no informed consent from patients is needed. LP, CF and MR had the full access to the prescription database.
As reported in the previous paper [12] the sample population consisted of men ≥40 years who had been prescribed medications for BPH-associated LUTS during the index period (January1st 2004-December31st 2006).
Only ABs and 5ARIs were considered in the analysis (ATC codes: G04CA and G04CB, respectively). During the study period, the first prescription of a drug was considered as “index date” for including a patient. Patient adherence to therapy was estimated only for patients receiving treatment for a minimum of 6 months during the index period. Two different levels of exposure to drugs were set: at ≥ 6 months and at ≥12 months. Patients on treatment for more than 12 months during the index period were followed-up for 4 years (median time). Patients who: a) stopped one of the three regimens (AB monotherapy, 5ARI monotherapy or CT) for at least 2 consecutive months during the first year of treatment and at least 4 months/yearly during the follow-up period, or b) switched regimen were considered as “treatment discontinuation”.
Patients were followed until hospitalization or surgery for BPH occurred or until their last follow-up. Patients were excluded when they were diagnosed with urethral stricture, prostate cancer in the 12 months preceding the index day.
Hospital admissions were recorded for patients receiving ≥1 year of pharmacological therapy and they were considered “BPH-related” when hospital records included a primary diagnosis and/or a surgical procedure related to BPH.
The presence of the ICD9-CM 600.xx code as primary diagnosis without surgical procedures was considered as a “BPH-related hospitalization”. In the absence of clear and universally agreed upon indications for BPH-related hospitalization we included in the analyses all the hospitalizations for haematuria, urinary tract infection, urinary retention, bladder stones, and renal failure due to urinary tract obstruction caused by BPH.
The presence of ICD9-CM codes 57.0,57.91,57.92,60.21, and 60.29,60.3,60.4 as primary or secondary surgical procedures with any primary diagnoses was considered hospitalization for “BPH-related surgery”.
Statistical analysis
For patients with at least 12 months of treatment, the characteristics were reported using descriptive statistics. Differences between patient treatment subgroups were assessed using a standardized difference (SD). Crude incidence rates (IRs) per 1000 men/year and incidence rate ratios (IRRs) with 95 % confidence intervals (CIs) were calculated with the Poisson regression model.
A multivariable Cox proportional hazards regression model was used to account for differences in follow-up and in baseline characteristics among groups. In all Cox models, the associations between groups and all outcomes were adjusted for co-variates known to be of prognostic importance to the outcomes: age and previous hospitalization for BPH, history of BPH-related surgery, and previous pharmacological treatment. Results were expressed as hazard ratios (HRs) and 95 % CIs. Adjusted event-free survival curves were calculated using the corrected group prognosis method. Discontinuation rate according to treatment group was compare using Pearson chi square test.
All reported that p-values are two tailed, and a p-value less than 0.05 was considered statistically different.
Analyses were conducted using SAS Statistical Package Release9.3 (SAS Institute, Cary, NC, USA).