Blood pressure and mortality in healthy old people: the r shaped curve

BMJ 1996; 313 doi: (Published 16 November 1996) Cite this as: BMJ 1996;313:1243
  1. John M Starr, senior lecturer in geriatric medicinea,
  2. Susan Inch, research nursea,
  3. Susan Cross, research nursea,
  4. Ian J Deary, professor of psychologya,
  5. William J MacLennan, professor of geriatric medicinea
  1. a Departments of Geriatric Medicine and Psychology, University of Edinburgh, Edinburgh EH4 2DN
  1. Correspondence to: Dr J M Starr, Royal Victoria Hospital, Edinburgh EH4 2DN.
  • Accepted 11 July 1996

In middle age mortality rises exponentially with increased blood pressure. In advanced age, however, people with the highest systolic pressures are less likely to die than those with lower pressures.1 This paradox is usually attributed to the association of diseases, such as cancer and cardiac failure, with low blood pressure. We sought to reduce the confounding effect of disease by prospectively examining blood pressure and mortality in healthy old people.

Subjects, methods, and results

We screened general practice casenotes of 10 000 patients aged over 69 years to identify healthy subjects for a study of the effects of hypertension on cognition.2 We visited 1467 subjects at home and asked about health problems and medications. Six hundred and three subjects (237 men, 366 women), mean age 75.7 years (range 70–88 years), reported no health problems and were taking no regular medication. We asked about educational attainment and occupation and at the end of the 20 minute interview measured blood pressure in a sitting position with a standard sphygmomanometer. Mean systolic blood pressure was 160 mm Hg (range 100–220 mm Hg) and mean diastolic blood pressure 86 mm Hg (range 50–120 mm Hg).

Four years later we determined the outcome in all 603 subjects by revisiting general practitioners' surgeries and tracing those cases not found via the local primary care division. Sixty nine subjects (40 men, 29 women) were identified as having died, and we found death certificates for 67. Logistic regression revealed male gender (odds ratio 1.61, 95% confidence interval 1.23 to 2.12) and older age (odds ratio 1.17, 1.10 to 1.25, per year) as the only factors associated with increased mortality. Systolic blood pressure as a continuous variable improved the model only marginally (P = 0.051, positive relation). Diastolic pressure, education, and occupation had no effects. To investigate the possibility of a non-linear relation between blood pressure and mortality, we stratified systolic pressure into three groups close to the 33rd and 67th centiles: low (<150 mm Hg, n = 192), medium (150–169 mm Hg, n = 182), and high (>/=170 mm Hg, n = 229). Entering these blood pressure groups significantly improved the age-sex mortality model (χ2 = 6.68 with 2 df, P = 0.04), with low systolic blood pressure associated with a lower risk (odds ratio 0.59, 0.38 to 0.92) and medium systolic blood pressure a higher risk (1.47, 1.01-2.15) than high systolic blood pressure. Table 1 shows the crude mortality for each group. The excess deaths in the medium and high groups were mostly due to cardiovascular and cerebrovascular disease.

Table 1

Crude four year mortality rates not adjusted for age or sex by systolic blood pressure group for 603 healthy old people. Expected deaths per group adjusted for age and sex but not blood pressure

View this table:


We found that healthy old people with casual systolic readings below 150 mm Hg have a better survival than those with higher pressures. Only 17 subjects (2.8%) had systolic readings below 120 mm Hg, suggesting that subjects with serious disease had been excluded from the group. Mortality was 5% in subjects with readings below 140 mm Hg, 8% in those with pressures of 140–149 mm Hg, rising to 16% in those with pressures of 150–159 mm Hg and 14% in those with pressures of 160–169 mm Hg. Although regression to the mean will affect high and low casual readings, the medium readings are probably close to prevalent pressure. The group with high blood pressure may fare better than the medium group because they are more likely to be treated.3 We identified 30 subjects started on antihypertensive medication where blood pressure at initiation had been noted by the general practitioner. Median starting pressure was 200 mm Hg: only one person with a systolic pressure below 170 mm Hg was treated. Currently treatment is recommended at 160 mm Hg4; our findings suggest that in otherwise healthy old people we should consider the New Zealand criteria and treat systolic blood pressures above 150 mm Hg.5

We thank all the general practitioners who helped with this study.


  • Funding Chief Scientist's Office, Edinburgh.

  • Conflict of interest None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract