Brief lifestyle interventions for hypertension

Editor—Cappuccio's suggestions do not explain the results.

Nurses measured blood pressure by using semiautomated monitors (minimising measurement bias) and gave structured advice in all groups (minimising placebo effect¹,²). Any bias is likely to favour the active interventions, and there was no evidence of this.

General practitioners and nurses were asked to refer patients after two to three readings, using appropriate cuff sizes on a single occasion—the group targeted for non-pharmacological advice according to guidelines from the British Hypertension Society. After a few weeks the baseline blood pressure (the mean of three readings on the second occasion) was 153/93 mm Hg, similar to the previous smaller Dutch study (158/91 mm Hg), which indicated that low sodium salt was effective.³ When patients with a baseline diastolic blood pressure above 90 mm Hg (n = 171) were selected the estimate for the low salt group was - 1.16 mm Hg (95% confidence interval - 3.5 to 1.18). When patients older than 60 (n=94—a similar power to the Dutch study) were selected the estimate was 0.001 mm Hg (- 3.17 to 3.17).

Robust standard errors, allowing for clustering, were almost identical (slightly lower)—as expected with individual randomisation and a highly structured approach.

There are two likely reasons for the lack of effect: low sodium salt is not very effective, and in a pragmatic trial the control group are not constrained to a constant diet and know that they aren't getting low salt—and thus may be more motivated to change diet in response to basic advice (as we observed). Hence even among elderly patients and those with more definite hypertension the effect may be less than expected. To balance blinded “efficacy” trials we also need evidence from pragmatic open trials designed to look at the effect of advice in everyday practice—where patients' behaviour is rather more realistic.