Can we predict who should be tested for postural hypotension?

Conference: 
Talk Code: 
3C.1
Presenter: 
Christopher Clark
Co-authors: 
Christopher Clark, Daniel Thomas, Natasha Mejzner, Fiona Warren, David Llewellyn, Luigi Ferrucci, John Campbell
Author institutions: 
University of Exeter Medical School England, National Institute on Ageing Baltimore Maryland USA

Problem

Over three million people aged over 65 fall each year in the UK. Postural or orthostatic hypotension (PH) is a major risk factor for falls, and is associated with excess mortality. It is also associated with hypertension and the use of antihypertensive drugs. Most elderly patients with postural hypotension are asymptomatic but they may have more subtle adverse effects on wellbeing and cognition. Guidelines differ in recommendations for detection of postural hypotension: NICE (2011) advises testing in the presence of falls or symptoms whilst ESH/ESC (2013) advises checking in the elderly and diabetics. We have found that PH is not actually checked for in UK clinical practice unless patients report postural symptoms. We studied the InCHIANTI dataset to derive and validate a simple prediction tool designed to facilitate identification of subjects to be checked for PH.

Approach

InCHIANTI is a prospective cohort study of ageing. It recruited adult subjects from two small towns in Tuscany, Italy in 1998 and they are followed up triennially. Blood pressure (BP) at recruitment was measured after resting supine, and one and three minutes after standing, using a mercury sphygmomanometer. Systolic PH was defined as a ≥20mmHg fall in supine BP on standing. Subjects were randomised to derivation or validation cohorts; allocation was undertaken blinded to PH status and medical history. Candidate predictor variables identified from literature searches were tested for univariable cross sectional associations with PH using χ2 tests. Those with significant associations were entered into multivariable linear regression models, and used to derive simple and weighted prediction scores (DROP scores). DROP scores were tested in the validation cohort for prediction of PH, future falls, cognitive decline and mortality rates with χ2 or ANOVA as appropriate to the data.

Findings

PH was present in 56/726 (7.7%) of the derivation cohort and 45/727 (6.2%) of the validation cohort (p=0.25). There were no significant differences between the cohorts. On multivariable analysis PH was associated with age ≥65, falls in the preceding year, diabetes, previous stroke, hypertension and Parkinson’s disease. A simple score summing numbers of these variables performed as well as a weighted score (AUROC 0.67 (0.59 to 0.74); p<0.001). Rising DROP scores were incrementally predictive of future falls (χ2 for trend p<0.01), increasing rates of decline in mini mental state examination (ANOVA p<0.001) and mortality (HR 1.8 (1.6 to 2.0) per unit increment in DROP score; p <0.001).

Consequences

The DROP score can predict presence of PH, future falls, and increased rates of mortality and cognitive decline. Application of the score could facilitate detection and management of PH. External validation of the DROP score is underway and future work to implement medicines optimisation on the basis of PH detection is planned.

Submitted by: 
Christopher Clark
Funding acknowledgement: 
CEC is supported by a National Institute for Health Research (NIHR) Clinical Lectureship award. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.