There are controversies on whether the effect and the side effects of drugs that were observed in clinical trials can also be maintained in real-life daily medical practice. Observational studies (researchers only watch what happens to patients who take or do not take drugs without interference) using electronic health records of patients can be used to examine this issue. The aim of this study is to assess whether the results of a type of clinical trials called "non-inferiority trials" can be reproduced in an observational study using a group of patients with atrial fibrillation (AF) who resemble, in terms of medical background and other health factors, those who were included in the original trial. We will also test whether the results were also maintained in a broader group that includes AF patients who did not exactly resemble those in the trial. These AF patients are taking either rivaroxaban or warfarin (both drugs are prescribed to prevent blood clots in the brain or the body. In the original trial, rivaroxaban was found to be not unacceptably worse than warfarin, but the use of rivaroxaban was a lot easier compared to warfarin.
Objectives: To assess the generalizability of the results of clinical trials in real-life settings with further investigation of the application of non-inferiority analysis in observational studies.
Methods: the design of the ROCKET AF non-inferiority trial (rivaroxaban vs. warfarin) will be replicated, apart from randomization, in a retrospective cohort study using the CPRD. Patients with atrial fibrillation (AF) who started rivaroxaban or warfarin will be included in the period from 2008 to 2017. The primary aim is to assess whether non-inferiority can be demonstrated for rivaroxaban vs. warfarin in terms of the composite endpoint of stroke (ischaemic or haemorrhagic) or systemic embolism, and whether these results extend to those AF patients who did not meet the eligibility criteria of the trial).
Data analysis: We will perform an analysis that replicates the primary analysis of the ROCKET-AF trial: a Cox proportional hazards regression analysis will be used to estimate the effect of rivaroxaban vs. warfarin, while adjusting for many possible confounders (including time-varying confounders). Non-inferiority will be tested using the same margin as used in the ROCKET-AF trial (HR 1.46).
Health Outcomes to be Measured:
HES Admitted;ONS;Practice IMD (Standard)