Jump to comment:

• REACHing the wrong conclusion: residual confounding by indication
  Jonathan A Batty, Alistair S Hall
  Published on: 4 September 2017

• Published on: 4 September 2017

  REACHing the wrong conclusion: residual confounding by indication

  • Jonathan A Batty, Research Fellow University of Leeds
  • Other Contributors:
    • Alistair S Hall, Professor of Cardiology

  Dear Professor Otto –

  In a recent edition of Heart, Potier et al reported the results of a large, observational analysis of the REACH registry[1]. The authors sought to retrospectively compare clinical outcomes in angiotensin receptor blocker (ARB), and angiotensin converting enzyme inhibitor (ACEi) treated patients, using propensity score matching to reduce confounding by indication. They conclude that treatment with ARB was more effective than with ACEi, across a wide spectrum of cardiovascular diseases and with regard to a number of different clinical outcomes. However, we believe that their methodology falls short of the standards expected from a well-conducted pharmacoepidemiological analysis[2].

  Although the REACH Registry is a well-powered cohort of patients at risk of adverse cardiovascular outcomes, several characteristics make it disadvantageous in the context of comparative drug efficacy analysis. Firstly, exposure to ACEi or ARB was established at baseline; all participants were prevalent users of these agents. Much evidence exists to suggest that bias is introduced by such an approach; the characteristics of prevalent users may be affected by the drug itself[3]. A new-user design would have eliminated such concerns.

  The indication for ACE inhibition and angiotensin receptor blockade differed greatly in this cohort of patients, recruited in 2003 and 2004. During this time, the evidence base for the use of ARBs was limited; most patients wo...

  Show More

  Dear Professor Otto –

  In a recent edition of Heart, Potier et al reported the results of a large, observational analysis of the REACH registry[1]. The authors sought to retrospectively compare clinical outcomes in angiotensin receptor blocker (ARB), and angiotensin converting enzyme inhibitor (ACEi) treated patients, using propensity score matching to reduce confounding by indication. They conclude that treatment with ARB was more effective than with ACEi, across a wide spectrum of cardiovascular diseases and with regard to a number of different clinical outcomes. However, we believe that their methodology falls short of the standards expected from a well-conducted pharmacoepidemiological analysis[2].

  Although the REACH Registry is a well-powered cohort of patients at risk of adverse cardiovascular outcomes, several characteristics make it disadvantageous in the context of comparative drug efficacy analysis. Firstly, exposure to ACEi or ARB was established at baseline; all participants were prevalent users of these agents. Much evidence exists to suggest that bias is introduced by such an approach; the characteristics of prevalent users may be affected by the drug itself[3]. A new-user design would have eliminated such concerns.

  The indication for ACE inhibition and angiotensin receptor blockade differed greatly in this cohort of patients, recruited in 2003 and 2004. During this time, the evidence base for the use of ARBs was limited; most patients would have received an ARB for the treatment of hypertension and heart failure (but not in the post-infarction or high cardiovascular risk settings). The authors match only on propensity score, without restricting or forcing matching by indication. Participants receiving ACEi use had a greater burden of cardiovascular morbidities, such as ischaemic heart disease and heart failure, than those receiving ARBs, who tended to have more hypertension (Supplementary Table 1). Although the burden of these comorbidities was marginally balanced in the propensity-matched ACEi and ARB groups, it is plausible that many patients receiving an ACEi in the early post-infarction setting (at high risk of adverse outcomes) were matched with patients receiving an ARB for hypertension (at relatively lower risk), with similar propensity. Clearly, significant confounding by indication persists.

  Given that the conclusions of this analysis are contrary to data from head-to-head randomised controlled trials, and multiple meta-analyses[4], we advise significant caution when interpreting these findings. The analytic methodology is deeply flawed; the conclusions of this study most probably reflect residual confounding by indication.

  Yours sincerely,

  Jonathan A Batty
  Alistair S Hall

  1. Potier L, Roussel R, Elbez Y, Marre M, Zeymer U, Reid CM, et al. Angiotensin-converting enzyme inhibitors and angiotensin receptor blockers in high vascular risk. Heart. 2017.
  2. Glynn RJ. Use of Propensity Scores To Design Observational Comparative Effectiveness Studies. Journal of the National Cancer Institute. 2017;109(8).
  3. Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. American journal of epidemiology. 2003;158(9):915-20.
  4. Li EC, Heran BS, Wright JM. Angiotensin converting enzyme (ACE) inhibitors versus angiotensin receptor blockers for primary hypertension. The Cochrane database of systematic reviews. 2014(8):Cd009096.

  Show Less

  Conflict of Interest:
  Alistair S Hall has received speaker honoraria from Servier.

  • Back to top