Real-World Patient Data Can Emulate Clinical Trial Results

A study led by the London School of Hygiene and Tropical Medicine (LSHTM) has shown that real-world patient data can emulate results of a traditional randomised clinical trial (RCT). The researchers said the method could be used to explore effects of treatments in groups normally underrepresented in or excluded from clinical trials.The team used primary and secondary care data from the UK clinical practice research datalink Aurum to investigate whether they could emulate the pivotal (ARISTOTLE) trial. That trial compared direct oral anticoagulants apixaban and warfarin for prevention of stroke and other thromboembolic events in atrial fibrillation (AF). 
The researchers explained that it was uncertain if the treatment recommendations based on results from this RCT could be applied to populations of patients who were excluded or underrepresented from the trial, such as older people, those with multiple comorbidities, or patients with a higher risk of bleeding. 
Trial Emulation With Real-World Data Could Improve Outcomes 
The team wanted to emulate an RCT for stroke prevention in patients with AF to understand how transferable RCT results are to real-world practices, and clarify whether this methodological approach could help improve treatment options and outcomes for patient groups currently underrepresented in clinical trials.
For their study, published in PLOS Medicine, they attempted to emulate the patient eligibility, selection, and analysis approaches from ARISTOTLE in extracted data for patients prescribed apixaban or warfarin between January 1, 2013 and July 31, 2019. The dataset comprised 8734 apixaban users and propensity score-matched 8734 warfarin users. 
Their results confirmed those of the formal RCT in demonstrating apixaban non-inferiority for stroke, systemic embolism, or all-cause mortality. However, unlike ARISTOTLE, they did not indicate apixaban superiority. Patients prescribed apixaban had similar outcomes to those prescribed warfarin. The team suggested this may have been due to higher quality of warfarin control, suboptimal dosing of apixaban, differences in ethnicity of patients, or use of concomitant medications compared with the clinical trial population.
In a statement, the authors concluded: “Reference-trial-informed design provides a framework for the study of treatment effects in patient groups excluded from or under-represented in trials.”
Medscape News UK asked the lead author, Dr Emma Maud Powell from LSHTM, about the implications of the study.
Does your study have wider relevance? 
This is a challenge that is not specific to ARISTOTLE but rather of relevance to many therapeutic areas and treatments, given that clinical trials are not always able to include the full range of patients who actually end up being treated in primary care. Use of routinely collected data may help provide evidence in these patient groups.
Are your results mainly applicable to specific subgroups or is there a more global implication?
Well-designed and conducted RCTs provide high-quality evidence, and our study suggests that for some treatments the differences in standard of monitoring, ethnicity, and typical concomitant medication use may lead to slightly different treatment effect estimates in primary care compared with what was seen in the RCT. 
Is there any evidence that trial results fail to replicate in real populations, or that significant additional findings [e.g. unanticipated adverse events] come to light when a drug enters more general use? 
Numerous recent observational studies based on existing RCTs have been able to generate results consistent with the reference RCT when analysing populations selected to be as similar as possible to the reference RCT. When the observational analysis is subsequently extended to include groups excluded or underrepresented in trials, whether results are consistent or divergent from the reference RCT in these understudied groups will depend very much on the therapeutic area and research question. 
Your paper commented that your research “provides a framework that can be adapted to investigate treatment effects for other conditions”. Assuming your findings do have more general application, are there other areas particularly worthy of investigation?
There remain challenges in accounting for the impact of switching treatments and in looking at therapeutic areas that typically use repeated measures as the primary outcome in the RCT.
Are you planning further research along similar lines?
We are working on extending the analysis in this paper to look at groups that were excluded or underrepresented in ARISTOTLE, including the elderly, patients with increased bleeding risk, and people with high BMI. 
Another extension of the framework could be to perform head-to-head comparisons of drugs that are unlikely to be compared in an RCT due to being developed by different pharmaceutical companies.
What is the take-home message for Medscape readers?
The routinely collected data generated by GPs is an incredible resource that has the potential to help understand the risks and benefits of treatments in understudied patient groups. 
The study was funded by a grant from the Medical Research Council. 
 
Any views expressed above are the author’s own and do not necessarily reflect the views of WebMD or Medscape.