Fewer Gastrointestinal Bleeds With Ticagrelor and Prasugrel Compared With Clopidogrel in Patients With Acute Coronary Syndrome Following Percutaneous Coronary Intervention

Neena S. Abraham; Eric H. Yang; Peter A. Noseworthy; Jonathan Inselman; Xiaoxi Yao; Jeph Herrin; Lindsey R. Sangaralingham; Che Ngufor; Nilay D. Shah

Disclosures

Aliment Pharmacol Ther. 2020;52(4):646-654. 

In This Article

Methods

Data Source

We used medical, and pharmacy claims data from the OptumLabs Data Warehouse. A national data source that includes physician, hospital and prescription drug claims of >100 million privately insured and Medicare Advantage enrolees across the United States.43 Medical claims include International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM & ICD-10-CM) diagnosis codes, ICD-9 & ICD-10 procedure codes, Current Procedural Terminology, Version 4 (CPT-4) procedure codes, Healthcare Common Procedure Coding System procedure codes, site of service codes and provider specialty codes. This study was exempt from Institutional Review Board approval as it involved analysis of pre-existing, de-identified data.

Study Population

We identified patients 18 years of age or older with an index prescription of clopidogrel, prasugrel or ticagrelor between 01/01/10 and 07/31/18. The date of the first prescription fill was defined as the index date and used to stratify patients to their exposure group. We required at least 12 months of health plan enrolment before the index date and excluded patients with evidence of a dispensed prescription in the 12 months before the index date to ensure a new-user cohort. We excluded patients with a cancer diagnosis at risk of malignancy-associated GIB and missing gender data. All patients were required to have a diagnosis of ACS within 90 days of index date and evidence of a recent PCI, within 14 days of index prescription (Appendix 1).

Patient Characteristics

Baseline demographic characteristics (age, gender) and CHA2DS2-VASC score (hypertension, age, diabetes mellitus, congestive heart failure, stroke/transient ischaemic attack/thromboembolic event). Concomitantly prescribed medications, including prescription acetylsalicylic acid (ASA), nonsteroidal anti-inflammatory drugs (NSAIDs), selective serotonin reuptake inhibitors, anticoagulants and gastroprotective agents, were assessed as potential confounding variables. Administrative codes identified co-morbid conditions in the primary or secondary position on any claim during the baseline period, and overall comorbidity burden determined using the Charlson-Deyo index.

Study Outcomes

The primary outcome of interest was total GIB using administrative codes as previously described and validated in prior publications[7–9] (Appendix 1), with each event identified using inpatient hospital claims for relevant primary and secondary discharge diagnoses. Total GIB included upper (including small intestinal) and lower GIB. Drug exposure was considered continuous from the index prescription until GIB or censoring occurred due to end of enrolment (including mortality), switch to another treatment strategy or treatment termination as defined by the absence of prescription supply for 30 days following the last identified prescription fill date for the index medication. Secondary outcomes included GIB-related transfusions and hospital length of stay.

Statistical Analysis

Propensity score and inverse probability treatment weighting (IPTW) were used to balance the differences in baseline characteristics (Appendix 2) among three treatment groups (clopidogrel, prasugrel and ticagrelor). The propensity score was estimated using generalised boosted modelling, which uses an iterative estimation procedure to find a model with the best balance among treatment groups.43 This method is particularly suited to comparing more than two treatment groups and has been used in numerous previous studies.[11–13] The propensity score model included the baseline characteristics listed in Appendix 2. Weights were calculated independently for the overall, non-ST elevated ACS (NSTE-ACS) and ST-elevated myocardial infarction (STEMI) groups. Baseline characteristics are displayed after the inverse probability treatment weighting. Standardised differences are calculated to assess the balance of covariates, with a difference of <10% considered acceptable;[14] whereas covariates that exceed the 10% threshold treated as independent variables in subsequent survival models.

The outcome of interest was calculated using Weighted Cox Proportional Hazards models with a robust variance estimator, stratified by ACS event type (STEMI or NSTE-ACS). Schoenfeld residuals[15] was used to test the proportional hazards assumption. We calculated the event rates per 100 person-years and hazard ratios (HR) with 95% confidence interval (CI) in the overall cohort, and the STEMI and NSTE-ACS subgroups. The number needed to harm (NNH) is used to express the magnitude of risk reduction for each comparison. In addition, we calculated the event rates of GIB-related inpatient transfusions. We examined three independent outcomes (chronic obstructive pulmonary disease, pneumonia and fracture) as falsification tests to assess for residual confounding. We used a Sidak correction to adjust for multiple comparisons.[16] These falsification endpoints[17] could be associated with patient frailty but are unlikely to be related to the choice of antiplatelet agent. Finally, we conducted a sensitivity analysis excluding patients prescribed ASA, NSAIDS and anticoagulants to assess the influence of these important GIB-related covariates on the estimates. The analytic data set was created and manipulated using SAS 9.3 (SAS Institute Inc) and Stata 15.1 (Stata Corp).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....