Beta-blockers and Cardiac Outcomes in Dialysis Patients
Beta-blockers and Cardiac Outcomes in Dialysis Patients
We conducted a retrospective cohort study employing linked healthcare databases in Ontario, Canada. Ontario's population is ~13 million, of which 14% are >65 years of age, 51% are female and 77% are Caucasian. Emigration from the province is <1% annually. Ontario residents have universal access to physician services and hospital care and those older than 65 years have universal prescription drug coverage. We conducted this study according to a prespecified protocol approved by the Research Ethics Board of Sunnybrook Health Sciences Centre (Toronto, Ontario, Canada). We report this study as per guidelines in the Strengthening the Reporting of Observational Studies in Epidemiology statement (as detailed in supplementary Appendix A1).
Prescription medication use was ascertained using the Ontario Drug Benefit Program (ODB) database, which accurately records all outpatient drug prescriptions for residents aged ≥65 years. Chronic dialysis use was determined using the Ontario Health Insurance Plan (OHIP) claims database, which records all inpatient and outpatient physician service claims. Baseline characteristics and outcomes were ascertained using the Canadian Institute for Health Information Discharge Abstract Database (CIHI-DAD), which records detailed admission, diagnostic and procedural information and the Ontario Registered Persons Database which records demographic and vital statistics information on all Ontarians who have been issued a health insurance card. These four databases are effectively complete for all variables examined and have been routinely used to study clinical outcomes.
We studied consecutive chronic dialysis patients, aged ≥66 years, who initiated dialysis between 1 July 1991 and 31 July 2007. Incident chronic dialysis was defined by first evidence of dialysis treatments that persisted for 90 days. We followed patients with chronic dialysis forward in time and separated them into three exposure groups when two prescriptions were filled for a drug in one of the following classes: beta-adrenergic blockers; calcium channel blockers or statins, respectively. For the patient to be considered in these analyses, the medication specified by the two prescriptions had to be dispensed within 60–120 days of each other. Chronic dialysis patients who did not meet these criteria were excluded. Our cohort consisted of chronic dialysis patients at the time of their first use of these drugs following dialysis initiation. Patients with previous evidence of any prescriptions for beta-blockers or calcium channel blockers in the 180 days prior to the first of the two identifying prescriptions were not entered into the study cohort. The index date was defined as the date of the second prescription for an exposure drug and represented the first day of follow-up. supplementary Appendix A2 illustrates the timelines for patient accrual as well as the initiation and termination of follow-up.
Patients could receive different types of beta-blockers and calcium channel blockers during follow-up. Patients within the beta-blocker and calcium channel blocker groups could also be concurrently taking a statin. Once patients were accrued into a group, they remained within that group regardless of initiation or discontinuation of the study exposure drugs. Our reason for categorizing patients according to these three drug classes (beta-blockers, calcium channel blockers and statin-only) is that they are all taken chronically for presumed cardioprotective benefit and this would minimize confounding by indication. We used two comparator groups to more convincingly show a hypothesized benefit of beta-blockers. Prior to November 2010 (release of SHARP trial), there was no evidence to support statin efficacy for our outcomes of interest in patients receiving dialysis. Similarly, with the exception of a single study by Tepel et al., there were no randomized trials demonstrating improved outcomes associated with calcium channel blocker use in this population.
We included patients both with and without known cardiovascular disease in all three groups. We excluded the following patients from the analysis: patients on both beta-blockers and calcium channel blockers and patients with evidence of myocardial infarction in the 3 months prior to the index date. We excluded this latter group, as clinical practice guidelines recommend that patients be prescribed a beta-blocker after a myocardial infarction. As these patients are at high risk of death in the subsequent 90 days, we wanted to avoid accruing such patients preferentially into the beta-blocker group.
We examined demographic characteristics, comorbidities and measures of healthcare access and screening in the 3 years prior to the index date. Diagnostic and procedural codes with established validity were used where possible [20–25] (see supplementary Appendices A3 and A4). We also assessed concomitant medication exposure in the year prior to the index date as another measure of comorbidity.
The primary outcome was a composite of death, hospitalization for myocardial infarction, stroke or coronary revascularization. These outcomes are accurately coded in our data sources with sensitivity and specificity of 81–96% and 69–100%, respectively (detailed in supplementary Appendix A5). Secondary outcomes included the individual components of the primary outcome. Patients were censored after emigration from the province (evidenced by 1 year without healthcare contact in the absence of death), 5 years after the index date or the end of the follow-up period, 31 March, 2009.
We compared baseline characteristics in the three groups using standardized differences. This metric compares differences between group means relative to pooled standard deviations and unlike conventional significance testing. It is unconfounded by sample size. Standardized differences were deemed significant if >10%. We used a Cox proportional hazards model to estimate hazard ratios and 95% confidence intervals (CIs), using the statin group as the referent. We performed a multivariable analysis forcing the following variables into the model: age (66–70, 71–75, 76–80, 81–85, ≥86), gender, diabetes, hypertension, coronary artery disease (including previous myocardial infarction, angina and percutaneous coronary intervention), heart failure, duration of dialysis until the time of index date and the Deyo Revised Charlson Comorbidity score. We assessed the following additional covariates for inclusion in the model by performing univariate regression with a P-value <0.2 resulting in model inclusion: socioeconomic status (by quintile), year of cohort entry (before or after 2000), cerebrovascular disease, peripheral arterial disease, atrial fibrillation/flutter and the number of distinct drugs prescribed in the year preceding the index date, as an additional measure of comorbidity. We conducted all statistical analyses with SAS software Version 9.2 (SAS Institute Inc., Cary, NC).
We examined the primary outcome within six predefined patient subgroups: male and female, age greater than or less than the median, year of cohort entry before or after year 2000, previous history of coronary artery disease, heart failure and dialysis modality (hemodialysis, peritoneal dialysis or unspecified). We also studied the association between various beta-blocker characteristics and outcome: high and low (or unknown) removal with dialysis (classification described in supplementary Appendix A6), high and low dose (threshold doses for each drug were selected using dose ranges established in randomized controlled trials of beta-blockers in patients with adequate kidney function as described in supplementary Appendix A7), cardioselective/beta-1 selective drugs and non-selective drugs and lipophilic and hydrophilic drugs (classifications described in supplementary Appendices A8 and A9, respectively). Carvedilol, which is the only beta-blocker shown to have efficacy in a randomized trial of dialysis patients was also compared against all other beta-blockers.
Materials and Methods
Design and Setting
We conducted a retrospective cohort study employing linked healthcare databases in Ontario, Canada. Ontario's population is ~13 million, of which 14% are >65 years of age, 51% are female and 77% are Caucasian. Emigration from the province is <1% annually. Ontario residents have universal access to physician services and hospital care and those older than 65 years have universal prescription drug coverage. We conducted this study according to a prespecified protocol approved by the Research Ethics Board of Sunnybrook Health Sciences Centre (Toronto, Ontario, Canada). We report this study as per guidelines in the Strengthening the Reporting of Observational Studies in Epidemiology statement (as detailed in supplementary Appendix A1).
Data Sources
Prescription medication use was ascertained using the Ontario Drug Benefit Program (ODB) database, which accurately records all outpatient drug prescriptions for residents aged ≥65 years. Chronic dialysis use was determined using the Ontario Health Insurance Plan (OHIP) claims database, which records all inpatient and outpatient physician service claims. Baseline characteristics and outcomes were ascertained using the Canadian Institute for Health Information Discharge Abstract Database (CIHI-DAD), which records detailed admission, diagnostic and procedural information and the Ontario Registered Persons Database which records demographic and vital statistics information on all Ontarians who have been issued a health insurance card. These four databases are effectively complete for all variables examined and have been routinely used to study clinical outcomes.
Patients
We studied consecutive chronic dialysis patients, aged ≥66 years, who initiated dialysis between 1 July 1991 and 31 July 2007. Incident chronic dialysis was defined by first evidence of dialysis treatments that persisted for 90 days. We followed patients with chronic dialysis forward in time and separated them into three exposure groups when two prescriptions were filled for a drug in one of the following classes: beta-adrenergic blockers; calcium channel blockers or statins, respectively. For the patient to be considered in these analyses, the medication specified by the two prescriptions had to be dispensed within 60–120 days of each other. Chronic dialysis patients who did not meet these criteria were excluded. Our cohort consisted of chronic dialysis patients at the time of their first use of these drugs following dialysis initiation. Patients with previous evidence of any prescriptions for beta-blockers or calcium channel blockers in the 180 days prior to the first of the two identifying prescriptions were not entered into the study cohort. The index date was defined as the date of the second prescription for an exposure drug and represented the first day of follow-up. supplementary Appendix A2 illustrates the timelines for patient accrual as well as the initiation and termination of follow-up.
Patients could receive different types of beta-blockers and calcium channel blockers during follow-up. Patients within the beta-blocker and calcium channel blocker groups could also be concurrently taking a statin. Once patients were accrued into a group, they remained within that group regardless of initiation or discontinuation of the study exposure drugs. Our reason for categorizing patients according to these three drug classes (beta-blockers, calcium channel blockers and statin-only) is that they are all taken chronically for presumed cardioprotective benefit and this would minimize confounding by indication. We used two comparator groups to more convincingly show a hypothesized benefit of beta-blockers. Prior to November 2010 (release of SHARP trial), there was no evidence to support statin efficacy for our outcomes of interest in patients receiving dialysis. Similarly, with the exception of a single study by Tepel et al., there were no randomized trials demonstrating improved outcomes associated with calcium channel blocker use in this population.
We included patients both with and without known cardiovascular disease in all three groups. We excluded the following patients from the analysis: patients on both beta-blockers and calcium channel blockers and patients with evidence of myocardial infarction in the 3 months prior to the index date. We excluded this latter group, as clinical practice guidelines recommend that patients be prescribed a beta-blocker after a myocardial infarction. As these patients are at high risk of death in the subsequent 90 days, we wanted to avoid accruing such patients preferentially into the beta-blocker group.
Baseline Characteristics
We examined demographic characteristics, comorbidities and measures of healthcare access and screening in the 3 years prior to the index date. Diagnostic and procedural codes with established validity were used where possible [20–25] (see supplementary Appendices A3 and A4). We also assessed concomitant medication exposure in the year prior to the index date as another measure of comorbidity.
Outcomes
The primary outcome was a composite of death, hospitalization for myocardial infarction, stroke or coronary revascularization. These outcomes are accurately coded in our data sources with sensitivity and specificity of 81–96% and 69–100%, respectively (detailed in supplementary Appendix A5). Secondary outcomes included the individual components of the primary outcome. Patients were censored after emigration from the province (evidenced by 1 year without healthcare contact in the absence of death), 5 years after the index date or the end of the follow-up period, 31 March, 2009.
Statistical Analyses
We compared baseline characteristics in the three groups using standardized differences. This metric compares differences between group means relative to pooled standard deviations and unlike conventional significance testing. It is unconfounded by sample size. Standardized differences were deemed significant if >10%. We used a Cox proportional hazards model to estimate hazard ratios and 95% confidence intervals (CIs), using the statin group as the referent. We performed a multivariable analysis forcing the following variables into the model: age (66–70, 71–75, 76–80, 81–85, ≥86), gender, diabetes, hypertension, coronary artery disease (including previous myocardial infarction, angina and percutaneous coronary intervention), heart failure, duration of dialysis until the time of index date and the Deyo Revised Charlson Comorbidity score. We assessed the following additional covariates for inclusion in the model by performing univariate regression with a P-value <0.2 resulting in model inclusion: socioeconomic status (by quintile), year of cohort entry (before or after 2000), cerebrovascular disease, peripheral arterial disease, atrial fibrillation/flutter and the number of distinct drugs prescribed in the year preceding the index date, as an additional measure of comorbidity. We conducted all statistical analyses with SAS software Version 9.2 (SAS Institute Inc., Cary, NC).
Additional Analyses
We examined the primary outcome within six predefined patient subgroups: male and female, age greater than or less than the median, year of cohort entry before or after year 2000, previous history of coronary artery disease, heart failure and dialysis modality (hemodialysis, peritoneal dialysis or unspecified). We also studied the association between various beta-blocker characteristics and outcome: high and low (or unknown) removal with dialysis (classification described in supplementary Appendix A6), high and low dose (threshold doses for each drug were selected using dose ranges established in randomized controlled trials of beta-blockers in patients with adequate kidney function as described in supplementary Appendix A7), cardioselective/beta-1 selective drugs and non-selective drugs and lipophilic and hydrophilic drugs (classifications described in supplementary Appendices A8 and A9, respectively). Carvedilol, which is the only beta-blocker shown to have efficacy in a randomized trial of dialysis patients was also compared against all other beta-blockers.
Source...