Thresholds for taking antihypertensive drugs in different professional and lay groups: questionnaire surveyBMJ 2000; 320 doi: https://doi.org/10.1136/bmj.320.7247.1446 (Published 27 May 2000) Cite this as: BMJ 2000;320:1446
- Nick Steel, health services research fellow ()
- Accepted 13 December 1999
Hypertension is a common risk factor for cardiovascular disease, but patients' compliance with medication is low. Patients may be less likely to take antihypertensive drugs if they have accurate information about their levels of risk.1 Doctors who estimate the risk more accurately are less likely to recommend treatment.2 It is not known whether a patient's professional background affects his or her threshold for complying with treatment.
The study compared the threshold at which consultant physicians, general practitioners, nurses attached to a general practice, and the general population would start taking antihypertensive drugs.
Participants, methods, and results
A postal questionnaire asked whether or not respondents would take drugs if one life would be saved for every 12, 33, 50, 100, or 250 people treated for five years. This gave six thresholds of numbers needed to treat (<12 to 250).
A pilot study showed that a sample size of 28 in each group would have 80% power to detect a probability of 0.284 that the number needed to treat for nurses is lower than for general practitioners, using a Wilcoxon (Mann-Whitney) rank sum test for ordered categories with two sided significance of 0.05. To allow for low response rates, all 39 consultant physicians at the city hospitals were recruited. Altogether 39 practice nurses, 39 general practitioners, and 100 adult members of the public were selected from the lists of the local health authority, with systematic sampling from a random starting point. A lower response rate was expected from the public. The local research ethics committee approved the study.
The response rate was 69% (149/217). The threshold for numbers needed to treat chosen by consultant physicians (100) was twice that chosen by general practitioners (50) and three times that chosen by nurses and the public (33) (table). The range of responses within each group was wide, but the difference between the median levels of benefit chosen by the groups was significant (P=0.003, Kruskal-Wallis test). Logistic regression to control for age and sex, with the outcome chosen to be whether or not the number needed to treat was 50 or greater, also showed a significant overall difference between the groups (P=0.005). The adjusted odds ratio for a nurse rather than a consultant choosing a number needed to treat below 50 was 12.5 (95% confidence interval 2.9 to 50).
Treating hypertension entails combining evidence with judgment about risk. This study shows how people's risk judgments differ. The questionnaire posed a hypothetical question. The thresholds chosen by the groups might depend on whether the question was real or hypothetical, or whether the benefits were expressed as reduction of absolute risk (such as numbers needed to treat) or relative risk. Because all groups faced the same question, however, valid comparisons can be made between them. The moderate response rate from members of the public (58%) is a possible source of bias.
The questionnaire referred to reduction in mortality rather than morbidity. Death as an outcome is easier to understand and less subject to interpretation than non-fatal cardiovascular events. For adults aged 60 years or over treatment reduces the absolute risk of both mortality and morbidity by a similar amount. The number needed to treat to prevent one death is about 50, the number chosen by general practitioners.3 For younger adults there is little evidence for a reduction in mortality, but the number needed to treat to prevent one stroke over five years is about 170.4
The 1999 guidelines of the British Hypertension Society (written largely by professors) recommend starting treatment on the basis of risk rather than blood pressure.5 This will encourage doctors to make explicit judgments of risk. Clinicians should not assume that their patients and professional colleagues are likely to share their opinion whether treatment for hypertension is worth while.
I thank all the people who took part in this study, Brian Tom from the Centre for Applied Medical Statistics at Cambridge University for statistical advice, and Peter Brambleby and Malcolm Adams for their constructive comments.
NS is the sole author and acts as guarantor of the study.
Funding NS was supported by a health services research fellowship from the NHS Executive Anglia and Oxford Research and Development Directorate.
Competing interests None declared.