Can routine primary care data be used to assess long-term trial outcomes for physical activity interventions? – findings from the PACE-UP and PACE-Lift trials

Conference:

SAPC ASM 2018 - London

Talk Code:

P2.64

Presenter:

Teresa J. Harris

Co-authors:

Liz Limb, Fay Hosking, Iain Carey, Steve DeWilde, Cheryl Watson, Charlotte Wahlich, Shaleen Ahmad, Sally Kerry, Peter Whincup, Michael Ussher, Steve Iliffe, Ulf Ekelund, Judith Ibison, Derek Cook

Author institutions:

St George's University of London, Queen Mary's University of London, Brunel, University of London, University College London, Norwegian School of Sports Sciences.

Problem

The PACE-UP and PACE-Lift trials, two similar primary care pedometer-based walking interventions in adults and older adults, found sustained objectively measured physical activity (PA) increases at 3 and 4 years respectively. To roll such PA interventions out widely through NHS primary care, long-term effectiveness evaluation cannot rely on expensive, objective accelerometery assessment or inaccurate self-report PA measures. However, the availability of routinely collected primary care data provides a potentially valuable alternative approach. In both trials we collected participants’ primary care data on 4-year outcomes including new cardiovascular events, new cases of type 2 diabetes mellitus and depression cases; and falls and fractures and evaluated whether intervention groups had reduced events compared to controls.

Approach

We downloaded primary care record data (medical history, consultations, prescribing) for trial participants who gave written consent, for the 4-year periods following randomisation from 7 PACE-UP and 3 PACE-Lift practices. The following events were counted blind to intervention status: new cardiovascular events (myocardial infarction, new onset angina /ischaemic heart disease, coronary artery bypass graft, angioplasty, stroke / transient ischaemic attack / peripheral vascular disease); new diabetes cases; new depression episodes; and falls or fractures. We modelled the effect of the interventions on outcomes using logistic regression, adjusting for age, sex and practice.

Findings

We downloaded data on 1184/1321 (90%) of original trial participants. Event rates were low (<20 per group) for all outcomes apart from falls /fractures. Logistic regression odds ratios (95% C.I) for interventions vs controls were: cardiovascular events 0.54 (0.25-1.16); new diabetes cases 0.79 (0.42-1.47); new depression episodes 0.87 (0.40-1.89); falls /fractures 0.96 (0.71-1.31).

Consequences

We demonstrated the feasibility of obtaining trial participants’ consent and downloading data from their primary care records. We demonstrated non-significant decreases in new cardiovascular events, diabetes cases and depression episodes in intervention compared to control groups, with no difference in falls and fractures between groups. Our trials were underpowered to find a difference in routinely recorded outcome events, but we demonstrated that an appropriately powered implementation evaluation could use routinely recorded primary care data to evaluate a PA intervention, providing a robust, but simplified, outcome assessment.

Submitted by:

Teresa J. Harris

Funding acknowledgement:

PACE-UP trial was funded by the Health Technology Assessment Programme, NIHR and the CLAHRC South London, NIHR. PACE-Lift was funded by the Research for Patient Benefit Programme, NIHR.