There are a variety of data sources that can be used to collect RWD and these sources can be split into primary research data (e.g. clinical trials, patient-reported data, chart review) and secondary use of data (e.g. from electronic medical records, claims/insurance database, disease/product registry, social media platforms and mobile devices).4 Although not primarily used for research, these secondary data sources can contain a wealth of RWD to be used for RWE generation.4 Advantages of using RWE in treatment effectiveness research include improving generalisability of study findings to patients in the usual practice settings, facilitating the identification of populations with enhanced benefit and risk, assessing heterogenous treatment effects according to patient characteristics and evaluating personalised treatment strategies.6 My colleague, Jian Yi Choy, has touched on how adequate Asian representation in global clinical trials may not always be achievable and how RWE is accepted as a source of evidence for HTA decision-making in Singapore in a separate commentary here (‘How Can We Incorporate Locally Relevant and Culturally Appropriate Evidence for HTA Decision-Making in Asia?’).
However, there are several challenges in the use of RWD for RWE generation and these may vary across sources. For example, medical claims data may not capture all the data required to answer a question of interest. While these sources are likely to contain major events such as hospitalisation or mortality data, they may not have more nuanced records such as quality of life data or changes in symptom severity over time.1 Another limitation could be that these data were recorded in an unstructured manner (e.g. as free-text or lack of specific coding of diagnoses or treatment).1 Chun-Ting Yang (MSc, National Cheng Kung University, Taiwan) emphasised that the first step to develop reliable and robust RWE is to select a high quality RWD source by understanding the research question and being aware of the potential limitations around each source.6
The second consideration would be to develop robust study designs to limit confounding effects or possible selection bias present in RWD. Target Trial Emulation is a study design methodology shared by Felicitas Kuehne (Senior Scientist, Health Services Research and HTA, UMIT TIROL) to limit bias and generate quality RWE in situations when a randomised controlled trial (RCT) is not practical (e.g. too expensive, time-sensitive or due to ethical limitations) or not necessary.7 This approach involves designing a hypothetical randomised trial (i.e. the target trial) and then emulating the target trial using the available RWD. Although this concept could be implicit in many analyses, the target trial itself is rarely defined.8 A target trial should explicitly define the eligibility criteria, treatment strategies, assignment procedures, follow-up period, outcome, casual comparisons of interest (i.e. intention-to-treat effect, per-protocol effect) and the statistical analysis plan (Figure 2).8 Any data that do not meet the criteria should be regarded as lost to follow-up (i.e. censored).8 If the emulation is successful, the RWE should yield similar effect estimates as the target trial if the trial had been conducted.8
Figure 2: Target trial emulation methodology
Today, HTA bodies are increasingly tasked with issuing reimbursement or pricing recommendations on novel technologies in the absence of RCTs or limited RCT data,9 likely in a bid to improve patient access. When decisions must be made in the absence of randomised trials, it is essential to adopt a robust approach to the design and analysis of RWE. With increasing use of RWE in the APAC region, it would be crucial to build local expertise within each market to critically appraise RWD and generate quality RWE for use in HTA and treatment evaluation and these would likely benefit patient access in the long term.