Parts of this chapter have been reproduced with permission from Wing et al.1 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.
Background
Chronic obstructive pulmonary disease (COPD) affects 3 million people in the UK.2 The most common cause is smoking, and patients exhibit airflow obstruction that is not fully reversible. The disease is progressive, with declining lung function and a worsening of symptoms. Most troublesome are acute exacerbations manifested as a sudden worsening of symptoms (e.g. severe coughing, shortness of breath and chest congestion) that require urgent treatment and possibly hospitalisation. Although smoking cessation remains the most effective intervention, the rate of exacerbation can be reduced by regular medication, such as combination long-acting beta agonists (LABAs) and inhaled corticosteroids (ICSs) or long-acting muscarinic antagonists (LAMAs).3,4
Chronic obstructive pulmonary disease treatment guidelines are largely informed by randomised controlled trial (RCT) results,5 but it is not clear if these findings apply to the large patient populations who are not studied in these trials. Fluticasone propionate plus salmeterol (FP-SAL) [seretide (GlaxoSmithKline plc)] is a LABA/ICS combination and is one of the most widely used COPD treatments. It was studied in large randomised trials [e.g. the TORCH (TOwards a Revolution in COPD Health) trial],3 but the effects of treatment in important patient groups who were not studied are unknown. Some patients were excluded from trials (e.g. those aged > 80 years, those with concomitant asthma or those with substantial comorbidity), whereas others are under-represented (e.g. people with mild COPD),3,6 meaning that conclusions about these groups are difficult to make.
Although the conduct of non-interventional studies (sometimes also referred to as ‘observational studies’) to investigate possible drug harms is well established, the use of these studies to estimate treatment effectiveness is in its infancy. Issues of treatment channelling and indication bias mean that measuring the intended benefit of a treatment is beset with difficulties. Over the next few years, we believe that we will see more non-interventional studies of drug effectiveness emerging because of recent legislation that requires pharmaceutical companies to study the real-world effects of medications;7,8 however, rigorous, validated methodology is needed to translate these complex data into reliable evidence.
For example, the availability of anonymised individual patient data from RCTs provides the potential for ‘RCT-analogous’ cohorts to be selected from non-interventional data sources (by first applying the trial inclusion and exclusion criteria to a non-interventional data source and then matching patient records from non-interventional data to the RCT patient records on key characteristics). Once a cohort of patients has been selected from non-interventional data with very similar characteristics to the original trial population, analysis can be performed of this cohort, looking at the same outcomes as the trial, but applying statistical methods for analysing non-interventional data. If the results of analysing a RCT-analogous cohort in this way are different from the trial results, this shows that issues with the validity of the analysis remain, even after creating a non-interventional cohort that is highly comparable to the trial. If, however, subsequent analysis of this non-interventional RCT-analogous cohort generates results that are similar to those generated by the reference RCT, one could be confident in the validity of the results and also in the non-interventional methods used to obtain these results in this setting. This would then provide confidence that if one applies similar analysis approaches to cohorts of patients who were excluded from the trial but have been selected in a similar way to the trial (in terms of inclusion and exclusion criteria but not trial matching, as the trial did not include these patients by design so they would not be available for matching), then the results obtained are likely to be valid.
In this study, we used TORCH3 individual trial data to validate non-interventional methods for assessing COPD treatment effectiveness, before going on to apply these methods to the analysis of treatment effectiveness within people excluded from, or under-represented in, the TORCH trial.3 Non-interventional data were obtained from the UK Clinical Practice Research Datalink (CPRD) [linked to the Hospital Episode Statistics (HES) database].9 The results generated could aid patients, prescribers and policy-makers in deciding the most appropriate treatment for COPD for all types of patients. The approach used can also provide a template for treatment effectiveness research using non-interventional data with inbuilt validation against a randomised trial.
Aims and objectives
The aims of our study were as follows:
- to measure the association between treatments for COPD and a number of COPD outcomes, including exacerbation rate, mortality, pneumonia and time to treatment change, among patients not included in randomised clinical trials for COPD treatments
- to develop a methodological framework with inbuilt validation against RCT data for using non-interventional electronic health records (EHRs) to answer questions about drug treatment effects (i.e. both benefits and risks).
Specific objectives were to:
- validate methods for measuring COPD medication effectiveness in EHR data by comparing with trial results
- use EHR data to measure COPD medication effectiveness in patients excluded from trials (most importantly, those aged > 80 years or those with comorbidities)
- determine COPD treatment effectiveness in an understudied disease stage (i.e. mild COPD).
Figure 1 provides a high-level overview of the study approach, detailing each objective and data source used. Figure 1 illustrates how existing RCT data were used in objective 1 to validate methods for analysing COPD in routinely collected electronic data for application to unanswered questions in objectives 2 and 3.
FIGURE 1 Overview of the COPD real-world medicines effects study. A, Work performed by others prior to this study. Of the total population of people with COPD, only a subset are included in RCTs of COPD treatments, based on the RCT inclusion/exclusion criteria. The RCT generates results that inform clinical practice, and the anonymised raw data for the study can be made available to other researchers via the Clinical Study Data Request website. For this study, the specific COPD treatment RCT of interest is the TORCH trial,3 which investigated the effect of FP-SAL on COPD exacerbations. B, Work performed as part of this study. Objective 1: a cohort of TORCH trial-analogous3 patients was selected from the UK CPRD by matching people with COPD within CPRD to the records of people included in the trial. Analyses of the effect of FP-SAL on COPD exacerbations were then performed on this TORCH-analogous3 CPRD cohort. The result obtained were then compared with the TORCH trial3 itself, serving as a validation step, with comparable results indicating that the data from the non-interventional (‘real-world’) CPRD source can reliably be used to study COPD treatment effects. Objective 2: the validated analysis techniques used for objective 1 were used to study people in CPRD who would not have been eligible for inclusion in a RCT because of their age and the presence of other comorbidities, and for whom the effect of FP-SAL is currently unknown. Objective 3: the validated analysis techniques were then used to study people with only mild COPD who have been under-represented in RCTs and for whom the effect of COPD treatments is unclear. Reproduced with permission from Wing et al.1 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original. (Figure not reproduced here; see the original article.)