To assess the clinical relevance of regression to the mean for clinical trials and clinical practice.
MEDLINE was searched until February 2018 for randomized trials of BP lowering with over 1000 patient-years follow-up per group. We estimated baseline mean BP, follow-up mean (usual) BP amongst patients grouped by 10 mmHg strata of baseline BP, and assessed effects of BP lowering on coronary heart disease (CHD) and stroke according to these BP levels.
Eighty-six trials (349 488 participants), with mean follow-up of 3.7 years, were included. Most mean BP change was because of regression to the mean rather than treatment. At high baseline BP levels, even after rigorous hypertension diagnosis, downwards regression to the mean caused much of the fall in BP. At low baseline BP levels, upwards regression to the mean increased BP levels, even in treatment groups. Overall, a BP reduction of 6/3 mmHg lowered CHD by 14% (95% CI 11–17%) and stroke by 18% (15–22%), and these treatment effects occurred at follow-up BP levels much closer to the mean than baseline BP levels. In particular, more evidence was available in the SBP 130–139 mmHg range than any other range. Benefits were apparent in numerous high-risk patient groups with baseline mean SBP less than 140 mmHg.
Clinical practice should focus less on pretreatment BP levels, which rarely predict future untreated BP levels or rule out capacity to benefit from BP lowering in high cardiovascular risk patients. Instead, focus should be on prompt, empirical treatment to maintain lower BP for those with high BP and/or high risk.
aThe George Institute for Global Health, University of New South Wales, Hyderabad, India
bThe George Institute for Global Health, University of New South Wales, Sydney, Australia
cUppsala University, Uppsala, Sweden
dThe George Institute for Global Health, University of Oxford, Oxford, UK
eUniversity of Sydney, Australia
fOslo University Hospital, Oslo
gPopulation Health Research Institute, McMaster University, Norway
Correspondence to Kazem Rahimi, DM, The George Institute for Global Health, University of Oxford, 1st Floor, Hayes House, George Street, Oxford OX1 2BQ, UK. Tel: +44 1865617200; fax: +44 1865617202; e-mail: email@example.com
Received 6 June, 2018
Accepted 25 September, 2018
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Website (http://www.jhypertension.com).