Challenges in Interpreting Real-World Studies Involving the NOACs

Phase III randomized controlled trials (RCTs) are viewed as the gold standard for comparing the efficacy and safety of a new drug with an existing drug. RCTs assess whether a drug works under strict conditions in a carefully selected patient population and using a set protocol. However, results from RCTs may not necessarily be generalizable to all patients treated in routine practice. Real-world evidence (RWE) has a higher generalizability and can play a key role in establishing the safety and effectiveness of new medicines in routine clinical practice. A comparison between different data sources, across prospective and retrospective studies, is summarized in Table 1.

EHR, electronic health record; RCT, randomized controlled trial

Real-world evidence

RWE studies in all their forms should be viewed as complementary to RCTs. RWE studies include patients who are non-randomized and whom cover a more diverse patient population compared with RCTs. A larger and more diverse patient population monitored regularly over a longer time frame may better reflect patients in routine practice and capture long-term effectiveness and safety data (including rare adverse outcomes). RWE studies can be prospective or retrospective, with data gathered from sources such as non-interventional studies or practice registries, electronic health records, chart reviews or administrative claims databases. Studies employing RWE methodologies, are associated with biases, in particular, selection and misclassification bias. However, proper study design, careful and detailed data collection, robust statistical analysis, (using methods including propensity score matching or adjustment), along with proper data interpretation, can help mitigate the impact of these biases on the conclusions derived from RWE (although bias can never be completely eliminated).

The efficacy and safety of non-vitamin K antagonist oral anticoagulants (NOACs) have been demonstrated in phase III trials, and NOACs have been routinely used in clinical practice for several years. A wealth of RWE is available, and studies analysing these data complement, and in some cases extend, the body of evidence supporting the use of NOACs (such as rivaroxaban) in clinical practice. RWE can aid clinicians with treatment decisions, particularly for patients not included in RCTs, and help us better understand anticoagulant prescribing and use in patients (e.g. adherence and persistence to therapy). However, these studies are not without challenges or limitations, as outlined by two examples below.

XANTUS pooled analysis real-world study

The XANTUS (Xarelto for Prevention of Stroke in Patients with Atrial Fibrillation) programme has demonstrated the safety and effectiveness of rivaroxaban in the general population. XANTUS was a prospective, international, observational, non-interventional clinical study. The study was designed to investigate outcomes in patients with non-valvular atrial fibrillation (NVAF) prescribed rivaroxaban in routine treatment conditions to prevent stroke or non-central nervous system systemic embolism. A total of 6785 patients were enrolled from across Canada, Europe and Israel, and the study results were generally consistent with those of the phase III ROCKET AF study. Additional XANTUS studies were performed in the Asia-Pacific region (XANAP) and in Africa, Eastern Europe, Latin America and the Middle East (XANTUS-EL). A pooled XANTUS analysis of these three studies included 11,121 patients and showed low bleeding and stroke rates in rivaroxaban-treated patients with NVAF. The major bleeding rate was 1.7 events per 100 patient-years, and the rate of stroke or systemic embolism was 1.0 events per 100 patient-years. All-cause death was 1.9 events per 100 patient-years, and the 1-year treatment persistence was 77.4%. A major strength of the XANTUS pooled analysis was that the outcome events were adjudicated by a central committee, which may have reduced any reporting bias. However, limitations include possible selection bias driven by patients (e.g. self-selection for inclusion) or study investigators. To find out more, click here.

Real-world study of frail patients with non-valvular AF who received NOACs

In a retrospective study by Martinez et al., the effectiveness and safety of apixaban, dabigatran and rivaroxaban were separately assessed in patients who were frail and with NVAF compared with warfarin. Information gathered from a US claims database, Truven MarketScan, between November 2011 and December 2016 was used, and patients receiving each of the three NOACs were separately propensity score matched 1:1 to warfarin users. MarketScan data includes patient demographics, diagnosis and procedure codes, admission and discharge dates, outpatient data and prescription records. Patients were monitored regularly for up to 2 years or until an event (stroke or systemic embolism or major bleeding), insurance disenrollment or the end of follow-up. In total, 2700, 2784 and 5270 patients were included in the apixaban, dabigatran and rivaroxaban 1:1 matched analyses to warfarin, respectively. The study found rivaroxaban, but not apixaban or dabigatran, was associated with a reduced rate of stroke or systemic embolism versus warfarin in frail patients with NVAF (albeit some analyses may have lacked adequate power to detect differences between cohorts); none of the NOACs were associated with a significant difference in major bleeding versus warfarin. In addition, these results highlighted the relative safety and effectiveness of initiating oral anticoagulation in frail patients, a population that typically does not receive adequate oral anticoagulation. A strength of this study was the ability to identify and evaluate a large sample size of frail patients (conservatively applying a validated algorithm) whom are often under-represented in RCTs. Limitations for this type of claims database study included selection (leading to confounding bias) and misclassification bias. However, the researchers used appropriate statistical methods (propensity score matching and Cox regression) to adjust for confounding bias and utilized validated International Classification of Diseases, Ninth- and Tenth- Revision (ICD-9 and -10) coding schemas to identify frail patients and outcomes of interest, potentially reducing the impact of these biases. Additionally, the Martinez et al. study used US claims data, meaning the conclusions are most readily generalizable to a US population (lower external validity). For example, the 110 mg dose of dabigatran is not approved in the US; therefore, no conclusions can be made regarding the use of this dose versus warfarin in frail patients.

In conclusion, RWE studies can be useful for assessing different NOACs in routine practice when viewed in context to corresponding RCTs. They can also provide data on types of outcomes and measurements (i.e., prescribing patterns and adherence) that cannot be assessed as part of an RCT. It is important to be aware of the different methodologies used in generating RWE, and the strengths and limitations of each, before any conclusions are drawn.