Article Text

Download PDFPDF
19 Can cluster randomisation of prescribing policy be used to efficiently generate drug safety and effectiveness data within the NHS? pilot data from the evidence study
  1. Amy Rogers,
  2. Angela Flynn,
  3. Alex Doney,
  4. Robert Flynn,
  5. Tom McDonald
  1. University of Dundee, Dundee, UK


Objectives The Evaluating Diuretics in Normal Care (EVIDENCE) study aims to demonstrate whether a pragmatic cluster randomisation methodology, using existing NHS prescribing policies and mechanisms combined with routinely collected data, can be used to answer an important clinical question in the absence of head-to-head randomised trials.

Method Changes in prescribing policy are rarely formally evaluated. In 2011, NICE hypertension guidelines included a recommendation that indapamide or chlorthalidone (thiazide-like diuretics) should be used in preference to bendroflumethiazide (thiazide diuretic) in the management of hypertension.1 This guidance has not been fully implemented. 70% of prescriptions for thiazide or thiazide-like diuretics dispensed in England in November 2017 were for bendroflumethiazide.2 Reasons for non-implementation may include inertia, drug pricing and availability. However, it is notable that many physicians felt that the guidance was based upon insufficient evidence. EVIDENCE uses a cluster randomised, parallel group design to evaluate the NICE recommendation. Randomisation is at GP practice level with allocation to either bendroflumethiazide or indapamide as preferred diuretic for use in hypertension. Routine prescriptions are switched, where necessary, to comply with the policy, using existing mechanisms. Routinely collected NHS data will be used to monitor prescribing and to identify cardiovascular endpoints.

Results The EVIDENCE study protocol has been approved by a Research Ethics Committee and the initial pilot phase has commenced in Tayside, Scotland. Between May 2017 and October 2017, nearly 2.9 million doses of thiazide or thiazide like diuretics were dispensed in NHS Tayside. 84% of these were bendroflumethiazide and only 16% were indapamide. We will present data on a pilot cohort of at least ten randomised practices within NHS Tayside. These data will include the numbers of patients prescribed study medications before and after the policy implementation along with measures of adherence to applied switching and reasons for non-adherence. It is likely that cardiovascular events in this pilot phase will be few but results will demonstrate the utility of routinely collected prescribing and hospitalisation data.

Conclusions EVIDENCE will test a novel methodology for conducting comparative effectiveness research efficiently within the NHS. It is anticipated that this methodology will be applicable to the assessment of many diverse medications and interventions in current routine use where there is insufficient evidence to guide clinical practice.


  1. (2011) NICE Update of clinical guidelines 18 and 34 Hypertension The clinical management of primary hypertension in adults Clinical Guideline 127 Methods, evidence, and recommendations August 2011. NICE.

  2. Powell-Smith A, Goldacre B. (2015) In:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.