Adherence to Dabigatran and Longitudinal Outcomes
Adherence to Dabigatran and Longitudinal Outcomes
This is a retrospective cohort study of patients receiving health care in the VA between October 1, 2010 and September 30, 2012. We included all patients who filled a dabigatran prescription of at least 30 days duration at a VA pharmacy and had at least 30 days of follow-up. During this time, the criteria for dabigatran use were based on national, standardized VA Pharmacy Benefits Management criteria which included patients with non-valvular atrial fibrillation and a CHADS2 or CHA2DS2-VASc score ≥1, consistent with the inclusion criteria of the RE-LY trial.
Patients are excluded from receiving dabigatran within the VA if they have any of the following contra-indications: (a) stroke in the preceding 14 days (b) history of valvular heart disease (c) active infective endocarditis (d) active liver disease (e) concurrent indication for anticoagulant therapy such as deep venous thrombosis/pulmonary embolism or prosthetic heart valve (f) pregnancy (g) known hypersensitivity to dabigatran (h) active pathological bleeding (i) concurrent therapy with P-glycoprotein inducers (j) severe renal impairment (creatinine clearance <30 mL/min) or (k) moderate renal impairment (creatinine clearance 30–50mL/min) and concomitant dronedarone or ketoconazole.
Data for this study were obtained from the VA Corporate Data Warehouse (CDW), which is a national data repository from several VA clinical and administrative systems. Data sources for the CDW have been previously validated and described. It has also been previously used to measure anticoagulation usage in atrial fibrillation. The CDW data includes patient demographics, vital status, the date, time and location of service, diagnoses and procedures of all inpatient and outpatient visits provided at the VA or performed at non-VA facilities that were paid for by the VA. In addition, we obtained data on medications dispensed at VA pharmacies. We extracted data on dabigatran utilization including fill dates for prescriptions, prescription cancel dates and amount dispensed. Since the VA has a closed pharmacy system with a fixed, non-tiered copayment, patients have a strong financial incentive to fill prescriptions within the system, particularly for newer therapies that have higher copayments in the private sector.
Exposure Variable. The primary exposure variable was dabigatran adherence calculated in the first year of therapy. We measured adherence using the proportion of days covered (PDC), defined as the total number of non-hospitalized days in which dabigatran was supplied divided by the observation time interval. The number of outpatient days supplied for dabigatran was determined from prescription fill dates and number of pills dispensed. If outpatient dabigatran supply was interrupted secondary to hospitalization for any cause, duration of inpatient stay was excluded from the denominator but resumed after discharge. We also accounted for any provider prescription cancellation orders by excluding all days after the cancel date from our PDC calculation (Figure 1). Pre-determined end-points for calculation of PDC included death, transition to warfarin or end of study period. Consistent with prior literature, patients with PDC ≥80% were classified as adherent.
(Enlarge Image)
Figure 1.
Pictorial representation of Proportion of Days covered calculation. Time periods labeled a, b, d, e and g represent time during which patient had dabigatran supply. Time period c represents duration of hospitalization and time period f represents duration following physician ordered prescription cancellation. Time period h represents total follow-up duration.
Further, among non-adherent patients (PDC<80%) gaps in dabigatran therapy were identified as any time duration for which a patient did not have any dabigatran supply available and the patient had not refilled the medication. The total number of gaps and their duration was assessed for every non-adherent patient during follow-up.
Outcome Measures. We assessed a composite of all-cause mortality and stroke as our primary outcome for this analysis. The VA vital status file used to assess mortality outcome has a 98.3% sensitivity and 97.6% exact agreement with dates when compared with the National Death Index. All-cause mortality was included to ensure complete capture of all events given the high sensitivity and specificity of VA vital status index file. Stroke was ascertained using previously validated primary or secondary International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis codes (433–437). We also looked at non-fatal bleeding events and myocardial infarction (MI) as our other outcome measures. Non-fatal bleeding events were identified with previously validated ICD-9-CM codes for intra-cranial hemorrhage (430–432), gastrointestinal hemorrhage (456–580), hemarthrosis (719), hemopericardium (423), hematuria (599), vaginal bleeding (626–627), hemoptysis (786), epistaxis (784), and hemorrhage not otherwise specified (459). The assessment of MI as an outcome following initiation of dabigatran was based on primary discharge ICD-9-CM diagnosis code of 410.xx for any hospitalization within the VA.
Other Covariates. We included other covariates based on prior literature and/or clinical rationale. Demographic covariates included age, sex and race (white vs. other). Clinical covariates included history of hypertension, diabetes mellitus, congestive heart failure, prior MI, prior stroke, chronic kidney disease, chronic liver disease, peripheral arterial disease, bleeding requiring hospitalization in the year prior to dabigatran initiation, depression, alcohol abuse and drug abuse. Treatment covariates included warfarin use in the 100 days preceding dabigatran initiation and concomitant clopidogrel use.
Comparisons of patient characteristics were made between adherent (PDC ≥80%) and non-adherent (PDC<80%) patients using χ test for categorical variables, independent samples t tests for normally distributed continuous variables and Wilcoxon rank-sum test for non-normally distributed continuous variables. Next, Cox proportional hazards regression models were used to assess the association between dabigatran adherence and each of our four outcomes adjusted for the aforementioned covariates. Adherence based on PDC was recalculated at every unique event time and treated as a continuous, time-varying covariate to account for survival bias. For this analysis, pre-determined censoring events included death, transition to warfarin or end of study period (September 30, 2012). Hazard ratios with 95% confidence intervals were obtained per 10% increase in PDC. The proportional hazards assumption was evaluated and found to be met.
Sensitivity Analyses. First, as adherence to dabigatran may vary by time since therapy initiation, we conducted stratified analyses for adherence based on duration of follow-up after initiation (stratified into ≥30 days, ≥60 days, ≥90 days and ≥180 days). Second, because the effect of adherence on outcomes may vary by time since therapy initiation, we conducted separate Cox regression analyses assessing the adjusted association between PDC and effectiveness outcomes for stratified time periods (less than 90 days, 90–180 days, ≥180 days). Lastly, as adherence to dabigatran might differ among patients transitioned from warfarin to dabigatran, we assessed if prior warfarin use was a predictor of adherence. Accordingly, we constructed a logistic regression model with prior warfarin use as the independent variable and dabigatran adherence as the dependent variable, adjusted for the covariates described above. All analyses were performed using SAS software version 9.3 (SAS Institute, Cary, NC). The Colorado Multiple Institutional Review Board approved this study, and waiver of informed consent was granted.
No extramural funding was used to support this work. The authors are solely responsible for the design and conduct of this study all study analyses, the drafting and editing of the paper and its final contents.
Methods
Study Design, Setting and Population
This is a retrospective cohort study of patients receiving health care in the VA between October 1, 2010 and September 30, 2012. We included all patients who filled a dabigatran prescription of at least 30 days duration at a VA pharmacy and had at least 30 days of follow-up. During this time, the criteria for dabigatran use were based on national, standardized VA Pharmacy Benefits Management criteria which included patients with non-valvular atrial fibrillation and a CHADS2 or CHA2DS2-VASc score ≥1, consistent with the inclusion criteria of the RE-LY trial.
Patients are excluded from receiving dabigatran within the VA if they have any of the following contra-indications: (a) stroke in the preceding 14 days (b) history of valvular heart disease (c) active infective endocarditis (d) active liver disease (e) concurrent indication for anticoagulant therapy such as deep venous thrombosis/pulmonary embolism or prosthetic heart valve (f) pregnancy (g) known hypersensitivity to dabigatran (h) active pathological bleeding (i) concurrent therapy with P-glycoprotein inducers (j) severe renal impairment (creatinine clearance <30 mL/min) or (k) moderate renal impairment (creatinine clearance 30–50mL/min) and concomitant dronedarone or ketoconazole.
Data Sources
Data for this study were obtained from the VA Corporate Data Warehouse (CDW), which is a national data repository from several VA clinical and administrative systems. Data sources for the CDW have been previously validated and described. It has also been previously used to measure anticoagulation usage in atrial fibrillation. The CDW data includes patient demographics, vital status, the date, time and location of service, diagnoses and procedures of all inpatient and outpatient visits provided at the VA or performed at non-VA facilities that were paid for by the VA. In addition, we obtained data on medications dispensed at VA pharmacies. We extracted data on dabigatran utilization including fill dates for prescriptions, prescription cancel dates and amount dispensed. Since the VA has a closed pharmacy system with a fixed, non-tiered copayment, patients have a strong financial incentive to fill prescriptions within the system, particularly for newer therapies that have higher copayments in the private sector.
Variables
Exposure Variable. The primary exposure variable was dabigatran adherence calculated in the first year of therapy. We measured adherence using the proportion of days covered (PDC), defined as the total number of non-hospitalized days in which dabigatran was supplied divided by the observation time interval. The number of outpatient days supplied for dabigatran was determined from prescription fill dates and number of pills dispensed. If outpatient dabigatran supply was interrupted secondary to hospitalization for any cause, duration of inpatient stay was excluded from the denominator but resumed after discharge. We also accounted for any provider prescription cancellation orders by excluding all days after the cancel date from our PDC calculation (Figure 1). Pre-determined end-points for calculation of PDC included death, transition to warfarin or end of study period. Consistent with prior literature, patients with PDC ≥80% were classified as adherent.
(Enlarge Image)
Figure 1.
Pictorial representation of Proportion of Days covered calculation. Time periods labeled a, b, d, e and g represent time during which patient had dabigatran supply. Time period c represents duration of hospitalization and time period f represents duration following physician ordered prescription cancellation. Time period h represents total follow-up duration.
Further, among non-adherent patients (PDC<80%) gaps in dabigatran therapy were identified as any time duration for which a patient did not have any dabigatran supply available and the patient had not refilled the medication. The total number of gaps and their duration was assessed for every non-adherent patient during follow-up.
Outcome Measures. We assessed a composite of all-cause mortality and stroke as our primary outcome for this analysis. The VA vital status file used to assess mortality outcome has a 98.3% sensitivity and 97.6% exact agreement with dates when compared with the National Death Index. All-cause mortality was included to ensure complete capture of all events given the high sensitivity and specificity of VA vital status index file. Stroke was ascertained using previously validated primary or secondary International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis codes (433–437). We also looked at non-fatal bleeding events and myocardial infarction (MI) as our other outcome measures. Non-fatal bleeding events were identified with previously validated ICD-9-CM codes for intra-cranial hemorrhage (430–432), gastrointestinal hemorrhage (456–580), hemarthrosis (719), hemopericardium (423), hematuria (599), vaginal bleeding (626–627), hemoptysis (786), epistaxis (784), and hemorrhage not otherwise specified (459). The assessment of MI as an outcome following initiation of dabigatran was based on primary discharge ICD-9-CM diagnosis code of 410.xx for any hospitalization within the VA.
Other Covariates. We included other covariates based on prior literature and/or clinical rationale. Demographic covariates included age, sex and race (white vs. other). Clinical covariates included history of hypertension, diabetes mellitus, congestive heart failure, prior MI, prior stroke, chronic kidney disease, chronic liver disease, peripheral arterial disease, bleeding requiring hospitalization in the year prior to dabigatran initiation, depression, alcohol abuse and drug abuse. Treatment covariates included warfarin use in the 100 days preceding dabigatran initiation and concomitant clopidogrel use.
Analysis Plan
Comparisons of patient characteristics were made between adherent (PDC ≥80%) and non-adherent (PDC<80%) patients using χ test for categorical variables, independent samples t tests for normally distributed continuous variables and Wilcoxon rank-sum test for non-normally distributed continuous variables. Next, Cox proportional hazards regression models were used to assess the association between dabigatran adherence and each of our four outcomes adjusted for the aforementioned covariates. Adherence based on PDC was recalculated at every unique event time and treated as a continuous, time-varying covariate to account for survival bias. For this analysis, pre-determined censoring events included death, transition to warfarin or end of study period (September 30, 2012). Hazard ratios with 95% confidence intervals were obtained per 10% increase in PDC. The proportional hazards assumption was evaluated and found to be met.
Sensitivity Analyses. First, as adherence to dabigatran may vary by time since therapy initiation, we conducted stratified analyses for adherence based on duration of follow-up after initiation (stratified into ≥30 days, ≥60 days, ≥90 days and ≥180 days). Second, because the effect of adherence on outcomes may vary by time since therapy initiation, we conducted separate Cox regression analyses assessing the adjusted association between PDC and effectiveness outcomes for stratified time periods (less than 90 days, 90–180 days, ≥180 days). Lastly, as adherence to dabigatran might differ among patients transitioned from warfarin to dabigatran, we assessed if prior warfarin use was a predictor of adherence. Accordingly, we constructed a logistic regression model with prior warfarin use as the independent variable and dabigatran adherence as the dependent variable, adjusted for the covariates described above. All analyses were performed using SAS software version 9.3 (SAS Institute, Cary, NC). The Colorado Multiple Institutional Review Board approved this study, and waiver of informed consent was granted.
No extramural funding was used to support this work. The authors are solely responsible for the design and conduct of this study all study analyses, the drafting and editing of the paper and its final contents.