Papers

36 Stamler J. Epidemiology, established major risk factors and the primary 42 Adler AI, Matthews D, Holman RR, Turner RC. Type 2 diabetes and
prevention of coronary heart disease. In: Parmley WW, Chatterjee K, eds. death: causes, estimated life expectancy and mortality rates—the UK pro-
Cardiology. Philadelphia: JB Lippincott, 1987:1-41. spective diabetes study. Diabetes 1998;47(suppl 1):A71.
37 UKPDS Group. UK prospective diabetes study 24: relative efficacy of sul- 43 Palumbo PJ, Elveback LR, Chu CP, Connolly DC, Kurland LT. Diabetes
fonylurea, insulin and metformin therapy in newly diagnosed mellitus: incidence, prevalence, survivorship and causes of death in
non-insulin dependent diabetes with primary diet failure followed for six Rochester, Minnesota, 1945-1970. Diabetes 1976;25:566-73.
years. Ann Intern Med 1998;128:165-75. 44 Panzram G. Mortality and survival in type 2 (non-insulin-dependent)
38 American Diabetes Association. Report of the expert committee on the
diabetes mellitus. Diabetelogia 1987;30:123-31.
diagnosis and classification of diabetes mellitus. Diabetes Care
45 Goodkin G. Mortality in diabetes. A 20 year mortality study. J Occup Med
1998;21(suppl 1):5-19.
39 Franklin GF, Shetterly SM, Cohen JA, Baxter J, Hamman RF. Risk factors 1975;17:716-21.
for distal symmetric neuropathy in NIDDM. The San Luis Valley diabetes 46 Wetterhall SF, Olson DR, DeStefano F, Stevenson JM, Ford ES, German
study. Diabetes Care 1994:1172-7. RR, et al. Trends in diabetes and diabetic complications, 1980-1987.
40 Harris M, Eastman R, Cowie C. Symptoms of sensory neuropathy in Diabetes Care 1992;15:960-7.
adults with NIDDM in the US population. Diabetes Care 1993;16:1446-52. 47 Stamler J, Vaccaro O, Neaton JD, Wentworth D. Diabetes, other risk
41 Adler AI, Boyko EJ, Ahroni JH, Smith DG. Lower extremity amputation factors, and 12 year cardiovascular mortality for men screened in the
in diabetes mellitus: the independent effects of peripheral vascular multiple risk factor intervention trial. Diabetes Care 1993;16:434-44.
disease, sensory neuropathy and foot ulcers. Diabetes Care 1999;22:
1029-35. (Accepted 20 March 2000)

Association of systolic blood pressure with macrovascular

and microvascular complications of type 2 diabetes
(UKPDS 36): prospective observational study
Amanda I Adler, Irene M Stratton, H Andrew W Neil, John S Yudkin, David R Matthews,
Carole A Cull, Alex D Wright, Robert C Turner, Rury R Holman on behalf of the UK Prospective
Diabetes Study Group

Editorial by Abstract Conclusions In patients with type 2 diabetes the risk

Tuomilehto
of diabetic complications was strongly associated with
Diabetes Trial Unit, Objective To determine the relation between systolic raised blood pressure. Any reduction in blood
Oxford Centre for blood pressure over time and the risk of
Diabetes, pressure is likely to reduce the risk of complications,
Endocrinology and macrovascular or microvascular complications in with the lowest risk being in those with systolic blood
Metabolism, patients with type 2 diabetes. pressure less than 120 mm Hg.
University of
Oxford, Radcliffe Design Prospective observational study.
Infirmary, Oxford Setting 23 hospital based clinics in England, Scotland,
OX2 6HE and Northern Ireland.
Amanda I Adler
Introduction
epidemiologist
Participants 4801 white, Asian Indian, and
Afro-Caribbean UKPDS patients, whether The UK prospective diabetes study (UKPDS) has
Irene M Stratton
senior statistician randomised or not to treatment, were included in shown that a policy of tight control of blood pressure,
Carole A Cull which achieved a median blood pressure of 144/82
analyses of incidence; of these, 3642 were included in
senior statistician mm Hg compared with 154/87 mm Hg over median
analyses of relative risk.
Rury R Holman 8.4 years of follow up, substantially reduced the risk of
director Outcome measures Primary predefined aggregate
microvascular disease, stroke, and deaths related to
Division of Public clinical outcomes: any complications or deaths related
Health and Primary
diabetes,1 but not myocardial infarction. Complemen-
to diabetes and all cause mortality. Secondary
Care, Institute of tary information for estimates of the risk of complica-
Health Sciences, aggregate outcomes: myocardial infarction, stroke,
tions including myocardial infarction at different levels
University of lower extremity amputation (including death from
Oxford, OX3 7LF of blood pressure can be obtained from observational
peripheral vascular disease), and microvascular
H Andrew W Neil analysis of the UKPDS data. This information can help
university lecturer in
disease (predominantly retinal photocoagulation). to estimate the expected reduction in the risk of
clinical epidemiology Single end points: non-fatal heart failure and cataract diabetic complications from a given change in blood
continued over extraction. Risk reduction associated with a 10 mm pressure. It can also help to assess whether or not
Hg decrease in updated mean systolic blood pressure thresholds in blood pressure exist below which the risk
BMJ 2000;321:412–9 adjusted for specific confounders of complications is substantially reduced. Such thresh-
Results The incidence of clinical complications was olds would have substantial influence on the establish-
significantly associated with systolic blood pressure, ment of guidelines on clinical care.
except for cataract extraction. Each 10 mm Hg People with type 2 diabetes have a greater
decrease in updated mean systolic blood pressure was incidence of cardiovascular disease, cerebrovascular
associated with reductions in risk of 12% for any disease, and renal disease than the general population.
complication related to diabetes (95% confidence Epidemiological studies suggest that relative hyper-
Details of
interval 10% to 14%, P < 0.0001), 15% for deaths glycaemia accounts for part but not all of the increased
participating related to diabetes (12% to 18%, P < 0.0001), 11% for risk.2–7 Raised blood pressure is more common in
centres, staff, and myocardial infarction (7% to 14%, P < 0.0001), and people with type 2 diabetes than in the general
committees and
additional funding
13% for microvascular complications (10% to 16%, population,8–12 and in people without diabetes it is a
agencies are on the P < 0.0001). No threshold of risk was observed for any major risk factor for myocardial infarction and
BMJ’s website end point. stroke.13 14 Epidemiological studies of the role of blood

003174634
412 1.0 BMJ VOLUME 321 12 AUGUST 2000 bmj.com
 Papers

pressure on the development of cardiovascular disease Blood pressure measurement University College
London Medical
have categorised people as either hypertensive or nor- Blood pressure was measured with the person in a School, Whittington
motensive or have measured blood pressure on a seated position after a five minute rest with a Copal Hospital, London
single occasion,5 6 15-17 whereas repeated measurements UA-251 or a Takeda UA-751 electronic, auscultatory N19 3UA
of blood pressure over several years should be more blood pressure reading machine (Andrew Stephens, John S Yudkin
consultant physician
informative. Brighouse, West Yorkshire). The first reading was
Oxford Centre for
In these analyses, we evaluated the relation between discarded, and the mean of the next three consecutive Diabetes,
systolic blood pressure over time and the development readings with a coefficient of variation below 15% was Endocrinology and
of macrovascular and microvascular complications used. In participants with atrial fibrillation, examiners Metabolism,
University of
using data from the UKPDS and looked for possible used a Hawksley random zero sphygmomanometer. Oxford
thresholds. We compared these results to those of the David R Matthews
UKPDS trial of a policy of tight control of blood pres- Blood pressure exposure consultant
diabetologist
sure.1 When the achieved reduction in risk notably Blood pressure was measured firstly at baseline (mean
exceeded that expected from observational data, analy- Selly Oak Hospital,
of measures taken at two and nine months after Birmingham
ses were performed to evaluate the presence of a treat- diagnosis) and secondly as an updated mean of annual B29 6JD
ment effect beyond that of blood pressure alone.1 measurement of systolic blood pressure, calculated for Alex D Wright
each participant from baseline to each year of follow consultant physician

Methods up. For example, at one year the updated mean is the Diabetes Research
Laboratories,
average of the baseline and one year values and at Oxford Centre for
Participants recruited to the UKPDS three years it is the average of baseline, one year, two Diabetes,
We enrolled 5102 of 7416 patients with newly year, and three year values. Endocrinology and
Metabolism,
diagnosed type 2 diabetes (defined as fasting plasma University of
glucose concentrations over 6.0 mmol/l on two Biochemical methods Oxford
separate mornings) who were referred to the UKPDS The biochemical methods used have been reported Robert C Turner
and were aged 25 to 65 years. Recruitment occurred director
previously.21 Biochemical variables are quoted for
between 1977 and 1991 at 23 clinical centres in measurements after the initial dietary run-in period. Correspondence to:
A Adler
England, Scotland, and Northern Ireland. Exclusion amanda.adler@dtu.
criteria are presented elsewhere18; the main reasons ox.ac.uk
Clinical complications
were severe vascular disease, myocardial infarction or
The clinical end points studied18 and their definitions19 Professor Turner
stroke within the year before recruitment, or major died unexpectedly
were separated into aggregate and single end points after completing
systemic illness.
(see box). work on this paper

Participants in epidemiological analyses

Statistical analysis
We studied the incidence of complications of diabetes
in the 4801 white, Asian Indian, and Afro-Caribbean Incidence rates by category of systolic blood pressure
patients who had blood pressure measured at two and The unadjusted incidence rates were calculated by
nine months after the diagnosis of diabetes. Of these, dividing the number of people with a given complica-
3642 with complete data for potential confounders tion by the person years of follow up for the given
were evaluated in proportional hazards models. Their complication within each category of updated mean
characteristics are presented in table 1. systolic blood pressure and reported as events per
1000 person years of follow up. The categories were
Participants in UKPDS blood pressure control study defined (median values in parentheses) as: < 120 (114),
The UKPDS clinical trials of blood glucose and blood 120-129 (125), 130-139 (135), 140-149 (144), 150-159
pressure control are described elsewhere.19 20 In (154), and >160 (168) mm Hg over the range of
summary, 1148 patients with hypertension, defined as updated mean systolic blood pressures 85-230 mm Hg.
previously receiving antihypertensive treatment and
with a blood pressure >150/85 mm Hg or not
previously receiving antihypertensive treatment and a Table 1 Characteristics of patients included in proportional hazards models measured
blood pressure >160/90 mm Hg, were randomised to after three month dietary run-in after diagnosis of diabetes and those included in
a policy of tight control of blood pressure with a â UKPDS blood pressure control study.1 Figures are means (SD) unless stated otherwise
blocker or an angiotensin converting enzyme inhibitor Proportional hazards model Clinical trial of tight v less tight
of observational analyses blood pressure control policy
or to a policy of less tight control. At entry, the mean (n=3642) (n=1148)
duration of known diabetes was 2.6 years, and the Age (years) 53 (8) 56 (8)
patients were older and heavier than in the whole Proportion of men (%) 60 55
cohort (table 1). The aim in the group allocated to tight Ethnicity (% white/Asian 82/10/8/0 87/8/5/1
control was to achieve blood pressure values Indian/Afro-Caribbean/other)
< 150/ < 85 mm Hg. If this target was not met with Body mass index (kg/m2) 27.7 (5.3) 29.6 (19.2)
maximal doses of a â blocker or angiotensin Fasting plasma glucose (mmol/l)* 7.9 (6.6-10.0) 7.4 (6.2-9.5)

converting enzyme inhibitor, additional agents were Haemoglobin A1c (%) 7.1 (1.8) 7.0 (1.7)
Systolic blood pressure (mm Hg) 135 (19) 159 (19)
prescribed, including a loop diuretic, a calcium channel
Low density lipoprotein cholesterol (mmol/l) 3.5 (1.0) 3.6 (1.1)
blocker, and a vasodilator. The aim in the group
High density lipoprotein cholesterol (mmol/l) 1.06 (0.24) 1.10 (0.27)
allocated to less tight control was to achieve blood
Triglyceride (mmol/l)† 1.5 (0.9-2.5) 1.6 (0.9-2.7)
pressure values < 180/ < 105 mm Hg without the use Albuminuria (%)‡ 13.3 17.7
of a â blocker or an angiotensin converting enzyme
*Median (interquartile range).
inhibitor but using the same stepwise addition of other †Geometric mean (1 SD range).
treatments. ‡>50 mg/l in single morning sample.

003174634
BMJ VOLUME 3211.012 AUGUST 2000 bmj.com 413
Papers

combined—that is, < 130 mm Hg (median 120 mm Hg)

Aggregate end points —to increase the reliability of the results as there were
• Complications related to diabetes (myocardial infarction, sudden death, few of these end points in this range. A separate model
angina, stroke, renal failure, lower extremity amputation or death from with updated mean systolic blood pressure as a continu-
peripheral vascular disease, death from hyperglycaemia or hypoglycaemia, ous variable was used to determine risk reduction asso-
heart failure, vitreous haemorrhage, retinal photocoagulation, and cataract ciated with a 10 mm Hg reduction in blood pressure.
extraction) (This is reflected in the regression line in figures 3 and 4.)
• Death related to diabetes (myocardial infarction, sudden death, stroke, The 95% confidence intervals were calculated on the
lower extremity amputation or fatal peripheral vascular disease, renal
basis of absolute floating risk.22 Log linear relations are
disease, hyperglycaemia or hypoglycaemia)
reported by convention.1 18 The risk reduction associated
• All cause mortality
with a reduction of 10 mm Hg in updated mean systolic
• Myocardial infarction (fatal myocardial infarction, non-fatal myocardial
blood pressure was calculated as 100% minus the recip-
infarction, and sudden death)
rocal of the hazard ratio expressed as a percentage.
• Stroke (fatal and non-fatal stroke)
The regression lines were fitted with updated mean
• Lower extremity amputation or death from peripheral vascular disease
systolic blood pressure as a continuous variable
• Microvascular complications (retinopathy requiring photocoagulation,
centred on the mean of the risk estimates for the
vitreous haemorrhage, and fatal or non-fatal renal failure)
categories 130-139 mm Hg and 140-149 mm Hg. The
Single end points P value reported is that associated with systolic blood
• Heart failure (non-fatal, without a precipitating myocardial infarction) pressure as a continuous variable. Evaluation of a
• Cataract extraction threshold level of systolic blood pressure for each
complication was assessed by visual inspection. The
risk reduction from the continuous variable model
Time of follow up was calculated from the end of the associated with a 10 mm Hg reduction in observed
initial period of dietary treatment to the first systolic blood pressure was compared with the risk
occurrence of that complication or loss to follow up, reduction seen in the UKPDS intervention trial of a
death from another cause, or the end of the study on tight versus a less tight policy of blood pressure
30 September 1997 for those who did not have that control, for which no adjustment for potential
complication. The median follow up time for all cause confounders was required as they were balanced by the
mortality was 10.5 years. For myocardial infarction and randomisation.1 The main exposure of interest for the
stroke, for participants who had a non-fatal event observational analyses was updated mean systolic
before a fatal event, the time to the first event was used. blood pressure regardless of the control policy or the
We calculated adjusted incidence rates for each cat- antihypertensive treatments used.
egory of updated mean systolic blood pressure using a
Poisson regression model adjusted for male sex, white
120
Adjusted incidence per 1000 person years (%)

ethnic group, age at diagnosis 50-54 years, and

duration of diabetes 7.5-12.5 years, and expressed as
events per 1000 person years of follow up. These
parameters were chosen to reflect the cohort’s median
age and duration of diabetes and modal ethnic group 100

and sex.

Hazard ratios and risk reduction

To assess potential associations between updated mean 80
systolic blood pressure and complications we used pro-
portional hazards (Cox) models. The hazard ratio was
used to estimate the relative risk. Potential confounding
risk factors included in all Cox models were sex, ethnic 60
group, age, and smoking (current/ever/never) at
diagnosis of diabetes and baseline concentrations of
high density lipoprotein cholesterol, low density
lipoprotein cholesterol, triglyceride, albuminuria ( > 50 40
mg/l measured on a single morning urine sample), and
haemoglobin A1c. At each event time, the updated mean
systolic blood pressure for a person with an event was
compared with the updated mean systolic blood 20
pressure of those who had not had an event by that time.
The updated mean systolic blood pressure was included
as a time dependent covariate to evaluate systolic blood
pressure during follow up. It was included as a categori- 0
110 120 130 140 150 160 170
cal variable in the categories of blood pressure listed
above, with the lowest category ( < 120 mm Hg) as the Updated mean systolic blood pressure (mm Hg)

reference category assigned a hazard ratio of 1.0. (This is Fig 1 Incidence rate (95% confidence interval) of any aggregate end
reflected in the point estimates shown below in figures 3 point related to diabetes by category of updated mean systolic blood
pressure, adjusted for age, sex, and ethnic group, expressed for
and 4.) In the analyses for stroke, heart failure, and lower white men aged 50-54 years at diagnosis and mean duration of
extremity amputation or deaths from peripheral diabetes of 10 years
vascular disease, the two lowest categories were

003174634
414 1.0 BMJ VOLUME 321 12 AUGUST 2000 bmj.com
 Papers

tion increased from 18 per 1000 patient years in the

Adjusted incidence per 1000 person years (%) 50 group with the lowest systolic blood pressure to 33 per
Myocardial infarction 1000 patient years in the group with blood pressure
Microvascular end points > 160 mm Hg, with the comparable data for microvas-
cular disease being 7 to 21 per 1000 patient years.
40 The estimated hazard ratios associated with each
category of updated mean systolic blood pressure, rela-
tive to the lowest category, are shown as log linear plots
in figures 3 and 4. Mortality related to diabetes and all
30 cause mortality were both strongly associated with
blood pressure (P < 0.001). The risk of each of the
complications evaluated, except cataract extraction,
rose with increasing updated mean systolic blood pres-
sure with and without adjustment for baseline variables
20
including age, sex, ethnic group, lipid concentrations,
HbA1c, smoking, and albuminuria. The decrease in risk
for each 10 mm Hg reduction of updated mean systo-
lic blood pressure was between 12% and 19% for both
10
macrovascular and microvascular complications (table
3 and figures 3 and 4).
There was no indication of a threshold for any of
the complications examined below which risk no
0
110 120 130 140 150 160 170 longer decreased nor a level above which risk no
longer increased. The updated mean systolic blood
Updated mean systolic blood pressure (mm Hg)
pressure showed similar risk relations to baseline
Fig 2 Incidence rates (95% confidence interval) of myocardial systolic blood pressure (table 3).
infarction and microvascular end points by category of updated
mean systolic blood pressure, adjusted for age, sex, and ethnic
group expressed for white men aged 50-54 years at diagnosis and
mean duration of diabetes of 10 years Any end point related to diabetes
4
Hazard ratio

P<0.0001

To assess whether treatment with drugs to lower

blood pressure reduced the complications independ-
ently of the reduction in blood pressure, a proportional
1
hazards model was fitted that included allocation to
tight versus less tight blood pressure policies, updated 12% decrease per 10 mm Hg
reduction in systolic blood pressure
mean systolic blood pressure, age, sex, ethnic group, 0.5
smoking, and concentrations of high and low density Death related to diabetes
Hazard ratio

lipoprotein cholesterol, triglyceride, and albuminuria. P<0.0001

4
An interaction term between blood pressure treatment
and mean updated blood pressure was included.
Statistical analyses were performed with SAS
version 6.12.23
1
Results 17% decrease per 10 mm Hg
reduction in systolic blood pressure
The risk of each of the macrovascular and microvascular 0.5
complications of type 2 diabetes evaluated was strongly All cause mortality
4
Hazard ratio

associated with blood pressure, as measured by updated P<0.0001

mean systolic blood pressure. The incidence of cataract
extraction was not associated with blood pressure.
Figure 1 shows the incidence rated by category of
updated mean systolic blood pressure for any end point 1
related to diabetes adjusted for age, sex, ethnic group,
12% decrease per 10 mm Hg
and duration of diabetes. The increase in risk was reduction in systolic blood pressure
monotonic, showing no evidence of a threshold, and 0.5
110 120 130 140 150 160 170
showed a twofold increase over the range of systolic
Updated mean systolic blood pressure (mm Hg)
blood pressure from < 120 mm Hg (median 114 mm
Fig 3 Hazard rates (95% confidence intervals as floating absolute
Hg) to >160 mm Hg (168 mm Hg). The unadjusted and
risks) as estimate of association between category of updated mean
adjusted rates are shown in table 2. Figure 2 shows the systolic blood pressure and any end point related to diabetes, death
adjusted incidence rates for myocardial infarction and related to diabetes, and all cause mortality with log linear scales.
microvascular end points, both being strongly associated Reference category (hazard ratio 1.0) is systolic blood pressure <120
to a similar degree with increasing blood pressure. Myo- mm Hg; P value reflects contribution of systolic blood pressure to
multivariate model. Data adjusted for age at diagnosis, ethnic group,
cardial infarction, however, occurred about twice as smoking status, presence of microalbuminuria, haemoglobin A1c, high
frequently as microvascular end points at each level of and low density lipoprotein cholesterol, and triglyceride
blood pressure. Thus the incidence of myocardial infarc-

003174634
BMJ VOLUME 3211.012 AUGUST 2000 bmj.com 415
Papers

Table 2 Incidence of complications in patients with type 2 diabetes by category of updated mean systolic blood pressure. Rates per
1000 person years’ follow up adjusted in Poisson regression model to white men aged 50 to 54 years at diagnosis and followed up
for 7.5 to <12.5 years, termed “10 years” (n=4801)
<120 mm Hg 120-129 mm Hg 130-139 mm Hg 140-149 mm Hg 150-159 mm Hg >160 mm Hg
Aggregate end points
Complications related to diabetes:
Events/person years 198/8662 366/9327 375/9946 365/8044 272/5107 262/4493
Unadjusted rate 22.9 28.5 37.7 45.4 53.3 58.3
Adjusted rate (95% CI) 36.0 (29.7 to 43.6) 40.4 (34.0 to 47.9) 51.3 (43.9 to 60.0) 58.3 (49.7 to 68.3) 67.1 (56.4 to 79.8) 76.2 (63.6 to 91.3)
Deaths related to diabetes:
Events/person years 40/9353 59/10 406 115/11 540 104/9589 96/6207 110/5537
Unadjusted rate 4.3 5.7 10.0 10.8 15.5 19.9
Adjusted rate (95% CI) 9.6 (6.5 to 14.1) 10.0 (7.2 to 13.9) 15.7 (11.9 to 20.7) 15.4 (11.6 to 20.5) 21.3 (15.9 to 28.7) 28.7 (21.3 to 38.7)
All cause mortality:
Events/person years 65/9353 133/10 406 183/11 540 184/9589 152/6207 163/5537
Unadjusted rate 6.9 12.8 15.9 19.2 24.5 29.4
Adjusted rate (95% CI) 14.4 (10.7 to 19.5) 20.3 (16.0 to 25.8) 22.2 (17.8 to 27.7) 23.8 (19.1 to 29.7) 29.1 (23.0 to 36.9) 36.2 (28.4 to 46.1)
Fatal or non-fatal myocardial infarction:
Events/person years 85/9083 104/10 064 152/11 048 149/9198 114/5912 115/5326
Unadjusted rate 9.4 10.3 13.8 16.2 19.3 21.6
Adjusted rate (95% CI) 18.4 (13.8 to 24.6) 17.6 (13.5 to 22.9) 21.9 (17.3 to 27.8) 24.2 (19.1 to 30.7) 28.7 (22.2 to 37.3) 33.1 (25.3 to 43.3)
Stroke:
Events/person years* 36/19 523 41/11 334 56/9359 50/6008 58/5271
Unadjusted rate 1.8 3.6 6.0 8.3 11.0
Adjusted rate (95% CI) 2.8 (1.7 to 4.4) 4.4 (2.8 to 6.9) 6.5 (4.3 to 9.9) 8.7 (5.6 to 13.5) 11.9 (7.6 to 18.6)
Lower extremity amputation or death from peripheral vascular disease:
Events/person years* 11/19 573 17/11 393 13/9432 11/6081 15/5409
Unadjusted rate 0.6 1.5 1.4 1.8 2.8
Adjusted rate (95% CI) 1.0 (0.4 to 2.4) 2.4 (1.1 to 5.2) 2.1 (0.9 to 4.8) 2.8 (1.2 to 6.6) 4.8 (2.1 to 11.1)
Fatal or non-fatal microvascular disease:
Events/person years 52/9150 72/10 059 102/10 971 100/9026 67/5769 74/5084
Unadjusted rate 5.7 7.2 9.3 11.1 11.6 14.6
Adjusted rate (95% CI) 7.3 (5.0 to 10.6) 8.9 (6.4 to 12.4) 11.9 (8.8 to 16.1) 14.2 (10.5 to 19.3) 15.2 (10.8 to 21.4) 20.8 (14.8 to 29.2)
Single end point
Heart failure:
Events/person years 26/19 534 34/11 326 33/9356 31/6036 34/5331
Unadjusted rate* 1.3 3.0 3.5 5.1 6.4
Adjusted rate (95% CI) 2.4 (1.4 to 4.2) 4.2 (2.5 to 6.9) 4.1 (2.5 to 6.9) 5.6 (3.3 to 9.6) 7.0 (4.0 to 12.0)
Cataract extraction:
Events/person years 37/9173 53/10 136 55/11 113 62/9190 37/5993 35/5296
Unadjusted rate 4.0 5.2 4.9 6.7 6.2 6.6
Adjusted rate (95% CI) 4.7 (3.0 to 7.5) 5.1 (3.4 to 7.7) 4.4 (2.9 to 6.6) 5.2 (3.5 to 7.8) 4.4 (2.8 to 7.0) 4.4 (2.7 to 7.2)
Person years, events, and unadjusted rates are for all patients.
*For these three end points lowest category was <130 mm Hg.

Table 3 also shows the risk reductions associated There was no interaction between treatment and
with a 10 mm Hg reduction in blood pressure in the updated mean blood pressure.
observational analysis compared with the risk reduc-
tion associated with a 10 mm Hg median difference in
blood pressure from the clinical trial of blood pressure
Discussion
control.1 After exclusion of cataract extractions, the This observational analysis shows an important associ-
relation with blood pressure tended to be stronger in ation between the occurrence of each of the diabetic
the clinical trial than in the observational analysis, complications evaluated (except cataract extraction),
although in the clinical trial the confidence intervals including all cause mortality, and systolic blood
were wider (table 3), reflecting the smaller number of pressure exposure across the range observed in
people at risk and of events.1 With adjustment for patients with type 2 diabetes. This association persisted
updated mean systolic blood pressure in the trial, allo- after adjustment for other characteristics that are asso-
cation to tight blood pressure control had a greater ciated with risk of complications (age, sex, ethnic
effect on reducing the risk of heart failure (P = 0.0054; group, glycaemia, lipid concentrations, smoking, and
17 events in 749 patients allocated to tight control and albuminuria). On average, each 10 mm Hg reduction
20 events in 384 patients allocated to less tight control) in systolic blood pressure was associated with a 12%
than expected from the blood pressure reduction per decrease in the risk of any end point related to diabetes
se. This was also true for stroke (P = 0.027), with 29 and a 15% reduction in the risk of death related to
events in 752 patients allocated to tight control and 26 diabetes. Myocardial infarction occurred more com-
events in 386 patients allocated to less tight control, monly than microvascular complications, but the rela-
and for all deaths related to diabetes (P = 0.038), with tive risk reduction for a 10 mm Hg reduction in systolic
63 events in 752 patients allocated to tight control and blood pressure was similar at 11% and 13%,
49 events in 386 patients allocated to less tight control. respectively.

003174634
416 1.0 BMJ VOLUME 321 12 AUGUST 2000 bmj.com
 Papers

This observational analysis provides an estimate of

the reduction in risk that might be achieved by the Fatal and non-fatal myocardial infarction Fatal and non-fatal stroke
10

Hazard ratio
therapeutic lowering of blood pressure. While it is P<0.0001 P<0.0001
important to realise that epidemiological associations
cannot necessarily be transferred to clinical practice,
the results are consistent with those achieved by the
policy of tight control of blood pressure in the clinical
trial.1 Whereas tight control did not significantly reduce
1
the risk of myocardial infarction in the clinical trial, the 12% decrease per 10 mm Hg 19% decrease per 10 mm Hg
reduction in systolic blood pressure reduction in systolic blood pressure
effect size was commensurate with the observational
0.5
analysis. The risk reductions in the clinical trial of tight Microvascular end points Cataract extraction
10

Hazard ratio
control seemed to be greater than those anticipated
from the epidemiological analyses for any complica- P<0.0001 P=0.41

tions or deaths related to diabetes, stroke, microvascu-

lar disease, and heart failure.1 After allowance for
differences in blood pressure between the tight and less
tight policies in the clinical trial, this apparent
treatment effect, per se, was significant only for deaths 1
13% decrease per 10 mm Hg
related to diabetes, stroke, and heart failure. The reduction in systolic blood pressure
absence of significant interaction suggests that 0.5
treatment effect does not differ by level of blood pres- Amputation or death from
sure. The heart outcomes prevention evaluation peripheral vascular disease Heart failure
10
Hazard ratio
studies (HOPE and MICRO-HOPE) that used ramipril
P<0.0001 P=0.0028 12% decrease per 10 mm Hg
can also be interpreted to have effects beyond those reduction in systolic blood pressure
anticipated by changes in blood pressure alone.24 25
This suggests the possibility that treatment with angio-
tensin converting enzyme inhibitors26 27 and â block-
ers28 29 may have cardioprotective effects separate from
blood pressure reduction. For example, both are 1
16% decrease per 10 mm Hg
beneficial in heart failure.1 28 30 31 The diminished risk of reduction in systolic blood pressure
0.5
heart failure may have reduced the risk of embolic 110 130 140 150 160 170 110 120 130 140 150 160 170
120
stroke, but no direct data are available. Effects greater Updated mean systolic blood pressure (mm Hg)
than anticipated have also been shown in studies in the
Fig 4 Hazard rates (95% confidence intervals as floating absolute risks) as estimate of
general population, where the risk reduction in odds of association between category of updated mean systolic blood pressure and myocardial
stroke from pooled trials of antihypertensive drug infarction, stroke, microvascular end points, cataract extraction, lower extremity amputation,
treatment exceeded the 35-40% expected from epide- or fatal peripheral vascular disease and heart failure, with log linear scales. Reference
miological studies.32 It is possible that the association category (hazard ratio 1.0) is systolic blood pressure <120 mm Hg for myocardial infarction,
microvascular disease, and cataract extraction and <130 mm Hg for stroke, lower extremity
between blood pressure and cardiovascular disease dif-
amputation or fatal peripheral vascular disease, and heart failure; P value reflects contribution
fers in magnitude in diabetic and non-diabetic popula- of systolic blood pressure to multivariate model. Data adjusted for age at diagnosis of
tions, which could not be tested in this study. In diabetes, ethnic group, smoking status, presence of albuminuria, haemoglobin A1c, high and
support of this possibility, the multiple risk factor inter- low density lipoprotein cholesterol, and triglyceride
vention trial (MRFIT) observed that the association of

Table 3 Observational data analysis of relation between systolic blood pressure exposure and complications of diabetes as estimated by risk reduction for
10 mm Hg reduction in systolic blood pressure, measured at baseline and as updated mean, controlled for age at diagnosis of diabetes, sex, ethnic group,
smoking, microalbuminuria, haemoglogin A1c concentration, high and low density lipoprotein cholesterol, and triglycerides (n=3642) compared with results of
clinical trial of tight v less tight blood pressure control policy (n=1148)1
Observational analyses Clinical trial (less tight v tight policy)1
Baseline SBP Updated mean SBP
Decrease in risk Decrease in risk
No of (%)/10 mm Hg reduction (%)/10 mm Hg reduction No of Decrease in risk (%)/10 mm Hg
events (95% CI) P value (95% CI) P value events difference in SBP (95% CI) P value
Aggregate end points
Any end point related to diabetes 1255 9 (7 to 12) <0.0001 12 (9 to 14) <0.0001 429 24 (8 to 38) 0.0046
Deaths related to diabetes 346 19 (15 to 23) <0.0001 17 (13 to 21) <0.0001 144 32 (6 to 51) 0.0019
All cause mortality 597 13 (10 to 17) <0.0001 12 (9 to 16) <0.0001 217 18 (−8 to 37) >0.05
Myocardial infarction 496 13 (9 to 16) <0.0001 12 (7 to 16) <0.0001 176 21 (−7 to 41) >0.05
Stroke 162 13 (7 to 19) 0.0002 19 (14 to 24) <0.0001 72 44 (11 to 65) 0.013
Peripheral vascular disease* 41 30 (20 to 39) <0.0001 16 (9 to 23) <0.0001 16 49 (−37 to 81) >0.05
Microvascular disease 323 10 (4 to 15) 0.0007 13 (9 to 26) <0.0001 122 37 (11 to 56) 0.0092
Single end points
Heart failure 104 14 (5 to 21) 0.0016 15 (4 to 19) <0.0001 45 56 (6 to 80) 0.0043
Cataract extraction 195 7 (0 to 14) 0.043 −3 (−11 to 5) 0.41 42 −34 (40 to −202) >0.05
SBP=systolic blood pressure.
*Lower extremity amputation or fatal peripheral vascular disease.

003174634
BMJ VOLUME 3211.012 AUGUST 2000 bmj.com 417
Papers

systolic blood pressure and death from cardiovascular

disease was of a lower magnitude in diabetic compared
What is already known on this topic
with non-diabetic men.15 Stroke and heart failure were People with diabetes who also have hypertension
the complications least strongly associated with glycae- are more likely to develop complications
mia,2 suggesting that for these complications, by
comparison, raised blood pressure is of greater patho- Treatment of blood pressure in these individuals
genetic importance. reduces the risk of complications

Lack of thresholds
What this study adds
We observed no thresholds of systolic blood pressure for There is a direct relation between the risk of
any complication of diabetes. This suggests that there is complications of diabetes and systolic blood
no specific target blood pressure to aim for but that the pressure over time
nearer to normal systolic blood pressure the lower the
risk of complications, in accord with recommendations No threshold of systolic blood pressure was
to reduce systolic blood pressure to less than 130 mm observed for a substantive change in risk for any
Hg33–35 or less than 125 mm Hg in the presence of of the clinical outcomes examined
microalbuminuria.36 Whether these target values can
realistically be achieved depends on an individual’s The lower the systolic blood pressure the lower
initial blood pressure and willingness to modify life style the risk of complications
or to take several drugs that may have side effects.
There may be additional risk reduction with
Neither our study nor the hypertension optimal
angiotensin converting enzyme inhibitors and
treatment (HOT) study37 found a J or U shaped
â blockers over and above that associated with
association between systolic blood pressure and compli-
lowering of blood pressure
cations, which is now thought to reflect coexisting mor-
bidity with low blood pressures rather than the effect of
treatment.38 Unlike the HOT study the UKPDS did not
with type 2 diabetes is difficult because of the progressive
observe a flattening of the relation at low levels of blood
hyperglycaemia,53 it is easier to maintain improved
pressure. There was no indication of a level above which
blood pressure control, although with time additional
systolic blood pressure was no longer associated with an
blood pressure lowering agents are required.1 As the risk
increased risk of complications. Thus any reduction of
of diabetic complications rises across the range of blood
raised blood pressure is likely to have benefit.
pressures studied, lifestyle interventions that also reduce
As in this analysis, a large study of diabetic
blood pressure in people without hypertension54 or not
applicants for life insurance in the United States
at high risk for diabetes55 should also be beneficial in
showed that hypertension was strongly associated with
diabetic patients. Targeting and treating patients with the
all cause mortality.39 Our finding of a 15% risk
highest blood pressure will reduce individual risk the
reduction for deaths related to diabetes associated with
most, but targeting and treating people with moderately
10 mm Hg systolic change is similar to the 15% risk
raised blood pressure will reduce the risk in greater
reduction observed for cardiovascular death in a global
numbers of people.56 The UKPDS provides an evidence
study40 and to that calculated from the data in the mul-
base for the management of raised blood pressure and
tiple risk factor intervention trial (MRFIT)15 in people
hyperglycaemia to reduce the complications of type 2
with diabetes. Other studies have observed an
diabetes.
association between blood pressure and cardiovascular
mortality.14 39 41–42 The increasing risk of stroke with The cooperation of the patients and many NHS and non-NHS
raised systolic blood pressure has also been shown.43–45 staff at the centres is much appreciated. Details of participating
These observational analyses show that people at centres can be found on the BMJ’s website.
Contributors: AIA coordinated the writing of the paper and
high risk for diabetic complications can be identified participated in interpretation of results. IMS selected the meth-
on the basis of as few as two blood pressure readings odology, carried out the statistical analyses, and participated in
within the year after diagnosis as these measurements interpretation and revision of the paper. HAWN, JSY, DRM, and
were associated with a risk of complications of similar ADW participated in interpretation and revision of the paper.
CAC participated in the preparation of the data and
magnitude to the updated mean systolic blood
interpretation and revision of the paper. RCT and RRH were
pressure measured over many years. Regression the principal investigators, planned and designed the study, and
dilution bias from the baseline measurement was participated in interpretation and revision of the paper. RCT
probably minimised as two readings taken six months was also responsible for the initial draft of the paper. RRH is
apart, rather than a single measurement, were used in guarantor.
Funding: The major grants for this study were from the UK
the model. The similarity of the risk probably reflects Medical Research Council, British Diabetic Association, the UK
tracking of blood pressure in populations over years.46 Department of Health, The National Eye Institute and The
The choice of systolic rather than diastolic pressure National Institute of Digestive, Diabetes and Kidney Disease in
reflects convention and the knowledge that both are the National Institutes of Health, United States, The British
Heart Foundation, Novo Nordisk, Bayer, Bristol-Myers Squibb,
associated with heart disease.32 40 47
Hoechst, Lilly, Lipha, and Farmitalia Carlo Erba. Details of other
This analysis indicates the importance of early funding companies and agencies, the supervising committees,
assessment of blood pressure in the course of diabetes. and all participating staff can be found on the BMJ’s website.
Improved control of blood pressure in diabetic patients Competing interests: AIA has received fees for speaking from
has been shown to be effective in reducing the risk of Bristol-Myers Squibb, SmithKline Beecham, and Pfizer. IMS has
received support for attending conferences from Zeneca and
cardiovascular complications37 48 49 and nephropathy,50 Hoechst and fees for speaking from Hoechst. CAC has received
as well as providing considerable savings in healthcare support for attending conferences from Bristol-Myers Squibb,
costs.51 52 Whereas treatment of glycaemia in patients Novo Nordisk, and Pfizer and fees for speaking from

003174634
418 1.0 BMJ VOLUME 321 12 AUGUST 2000 bmj.com
 Papers

Bristol-Myers Squibb and Novo Nordisk. JSY has received consul- 27 Ferrari R. Effect of ACE inhibition on myocardial ischaemia. Eur Heart J
tancy fees from SmithKline Beecham. DRM has received fees for 1998;19(suppl J):30-5.
28 Doughty RN, Rodgers A, Sharpe N, MacMahon S. Effects of beta-blocker
speaking from Bristol-Myers Squibb, Novo Nordisk, SmithKline
therapy on mortality in patients with heart failure: a systematic overview
Beecham, and Lilly and research funding from Lilly. RRH has of randomised controlled trails. Eur Heart J 1997;18:560-5.
received fees for consulting from Bayer, Boehringer Mannheim, 29 Freemantle N, Cleland J, Young P, Mason J, Harrison J. â Blockade after
Bristol-Myers Squibb, Hoechst, Lilly, Novo Nordisk, Pfizer, and myocardial infarction: systematic review and meta-regression analysis.
SmithKline Beecham; support for attending conferences from BMJ 1999;318:1730-7.
Bayer, Bristol-Myers Squibb, Hoechst, Lilly, Lipha, Novo Nordisk, 30 Psaty BM, Smith NL, Siscovick DS, Koepsell TD, Weiss NS, Heckbert SR,
et al. Health outcomes associated with antihypertensive therapies used as
and SmithKline Beecham; and research funding from Bayer, first-lines agents. JAMA 1997;277:739-45.
Bristol-Myers Squibb, Lilly, Lipha, and Novo Nordisk. 31 UKPDS Group. Efficacy of atenolol and captopril in reducing risk of
macrovascular and microvascular complications in type 2 diabetes
1 UKPDS Group. Tight blood pressure control and risk of macrovascular (UKPDS 39). BMJ 1998;317:713-20.
and microvascular complications in type 2 diabetes (UKPDS 38). BMJ 32 MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J, et al. Blood
1998;317:703-13. pressure, stroke, and coronary heart disease. Part 1: prolonged
2 Stratton IM, Adler AI, Neil HAW, Matthews DR, Manley SE, Cull CA, et al. differences in blood pressure: prospective observational studies corrected
Epidemiological association of glycaemia with macrovascular and micro- for the regression dilution bias. Lancet 1990;335:765-74.
vascular complications of type 2 diabetes (UKPDS 35). BMJ 2000; 33 Joint National Committee on Prevention, Detection, Evaluation and
321:405-12. Treatment of High Blood Pressure. Sixth report. Arch Intern Med
3 Klein R. Hyperglycemia and microvascular and macrovascular disease in 1997;157:2413-46.
diabetes. Diabetes Care 1995;18:258-68. 34 Meltzer S, Leiter L, Daneman D, Gerstein HC, Lau D, Ludwig S, et al.
4 Standl E, Balletshofer B, Dahl B, Weichenhain B, Stiegler, Hormann A, et 1998 clinical practice guidelines for the management of diabetes in
al. Predictors of 10-year macrovascular and overall mortality in patients Canada. Can Med Assoc J 1998;159(suppl 8):1-29.
with NIDDM: the Munich general practitioner project. Diabetologia 35 American Diabetes Association. Standards of medical care for patients
1996;39:1540-5. with diabetes mellitus. Diabetes Care 1998;21(suppl 1):23-31.
5 Hanefeld M, Fischer S, Julius U, Schulze J, Schwanebeck U, Schmechel H, 36 British Cardiac Society, British Hyperlipidaemia Association, British
et al. Risk factors for myocardial infarction and death in newly detected Hypertension Society, British Diabetic Association. Joint British
NIDDM: the diabetes intervention study, 11-year follow-up. Diabetologia recommendations on prevention of coronary heart disease in clinical
1996;39:1577-83. practice. Heart 1998;80(suppl 2):1-29.
6 Lehto S, Ronnemaa T, Haffner SM, Pyörälä K, Kallio V, Laakso M. Dys- 37 Hansson L, Zanchetti A, Carruthers SG, Dahlöf B, Elmfeldt D, Julius S, et
lipidemia and hyperglycemia predict coronary heart disease events in al. Effect of intensive blood-pressure lowering and low-dose aspirin in
middle-aged patients with NIDDM. Diabetes 1997;46:1354-9. patients with hypertension: principal results of the hypertension optimal
7 Wei M, Gaskill SP, Haffner SM, Stern MP. Effects of diabetes and level of treatment (HOT) randomised trial. Lancet 1998;351:1755-62.
glycaemia on all-cause and cardiovascular mortality. Diabetes Care 38 Vatten LJ, Holmen J, Kruger O, Forsen L, Tverdal A. Low blood pressure
1998;21:1167-72. and mortality in the elderly: a 6-year follow-up of 18,022 Norwegian men
8 Kannel W, McGee D. Diabetes and cardiovascular disease. The Framing- and women age 65 years and older. Epidemiology 1995;6:70-3.
ham study. JAMA 1979;241:2035-8. 39 Goodkin G. Mortality in diabetes. A 20 year mortality study. J Occup Med
9 Cowie CC, Harris MI. Physical and metabolic characteristics of persons 1975;17:716-21.
with diabetes. In: Harris MI, Cowie CC, Stern MP, Boyko EJ, Reiber GE, 40 Van den Hoogen PCW, Feskens EJM, Nagelkerke NJD, Menotti A,
Bennett PH, eds. Diabetes in America. 2nd ed. Washington, DC: National Nissinen A, Kromhout D. The relation between blood pressure and mor-
Institutes of Health, National Institute of Diabetes and Digestive and Kid- tality due to coronary heart disease among men in different parts of the
ney Diseases, 1995. world. N Engl J Med 2000;342:1-8.
10 Manson JAE, Colditz GA, Stampfer MJ, Willett WC, Krolewski AS, 41 Rosengren A, Welin L, Tsipogianni A, Wilhelmsen L. Impact of
Rosner B, et al. A prospective study of maturity-onset diabetes mellitus cardiovascular risk factors on coronary heart disease among middle aged
and risk of coronary heart disease and stroke in women. Arch Intern Med diabetic men: a general population study. BMJ 1989;299:1127-31.
1991;151:1141-7. 42 Bruno G, Merletti F, Boffetta P, Cavallo-Perin P, Bargero G, Gallone G, et
11 Barrett-Connor EL, Cohn BA, Wingard DL, Edelstein SL. Why is al. Impact of glycaemic control, hypertension and insulin treatment on
diabetes mellitus a stronger risk factor for fatal ischemic heart disease in general and cause-specific mortality: an Italian population-based cohort
women than in men? JAMA 1991;265:627-31. of type II (non-insulin-dependent) diabetes mellitus. Diabetologia
12 Hypertension in Diabetes Study Group. HDS I: prevalence of 1999;42:297-301.
hypertension in newly presenting type 2 diabetic patients and the associ- 43 Lehto S, Ronnemaa T, Pyörälä K, Laakso M. Predictors of stroke in
ation with risk factors for cardio-vascular and diabetic complications. middle-aged patients with non-insulin-dependent diabetes. Stroke
J Hypertens 1993;11:309-17. 1996;27:63-8.
13 Keil JE, Sutherland SE, Knapp RG, Lackland DT, Gazes PG, Tyroler HA. 44 Barrett-Connor E, Khaw K-T. Diabetes mellitus: an independent risk fac-
Mortality rates and risk factors for coronary disease in black as compared tor for stroke? Am J Epidemiol 1988;128:116-23.
with white men and women. N Engl J Med 1993;329:73-8.
45 Palumbo PJ, Elveback LR, Whisnant JP. Neurological complications of
14 Collins R, Peto R, MacMahon S, Herbert P, Fiebach N, Eberlein K, et al.
diabetes mellitus: transient ischemic attack, stroke, and peripheral
Blood pressure, stroke, and coronary heart disease. Lancet 1990;335:827-
neuropathy. Adv Neurol 1978;19:593-601.
38.
46 Tate RB, Manfreda J, Krahn AD, Cuddy TE. Tracking of blood pressure
15 Stamler J, Vaccaro O, Neaton JD, Wentworth D. Diabetes, other risk
over a 40-year period in the University of Manitoba follow-up study,
factors, and 12 year cardiovascular mortality for men screened in the
1948-1988. Am J Epidemiol 1995;142:946-54.
multiple risk factor intervention trial. Diabetes Care 1993;16:434-44.
47 Cutler JA. High blood pressure and end-organ damage. J Hypertens Suppl
16 Kuusisto J, Mykkänen L, Pyörälä K, Laakso M. NIDDM and its metabolic
1996;14:3-6.
control predict coronary heart disease in elderly subjects. Diabetes
48 Tuomilehto J, Rastenyte D, Birkenhäger WH, Thijs L, Antikainen R, Bul-
1994;43:960-7.
pitt CJ, et al. Effect of calcium-channel blockade in older patients with
17 Uusitupa MI, Niskanen LK, Siitonen O, Voutilainen E, Pyörälä K.
Ten-year cardiovascular mortality in relation to risk factors and diabetes and systolic hypertension. N Engl J Med 1999;320:677-84.
abnormalities in lipoprotein composition in type 2 (non-insulin- 49 Curb JD, Pressel SL, Cutler JA, Savage P, Applegate WB, Black H, et al.
dependent) diabetic and non-diabetic subjects. Diabetologia Effect of diuretic-based antihypertensive treatment on cardiovascular dis-
1993;36:1175-84. ease risk in older diabetic patients with isolated systolic hypertension.
18 UKPDS Group. Intensive blood glucose control with sulphonylureas or Systolic hypertension in the elderly program cooperative research group.
insulin compared with conventional treatment and risk of complications JAMA 1996;276:1886-92.
in patients with type 2 diabetes (UKPDS 33). Lancet 1998;352:837-53. 50 Lewis EJ, Hunsicker LG, Bain RP, Rohde RD. The effect of
19 UKPDS Group. UK prospective diabetes study VIII: study design, angiotensin-converting-enzyme inhibition on diabetic nephropathy. N
progress and performance. Diabetologia 1991;34:877-90. Engl J Med 1993;329:1456-62.
20 Hypertension in Diabetes Study Group. HDS III: prospective study of 51 Rodby RA, Firth LM, Lewis EJ. An economic analysis of captopril in the
therapy in type 2 diabetic patients—efficacy of ACE inhibitor and â treatment of diabetic nephropathy. Diabetes Care 1996;19:1051-61.
blocker. Diabetic Med 1994;11:773-82. 52 UKPDS Group. Cost effectiveness analysis of improved blood pressure
21 UKPDS Group. UK prospective diabetes study XI: biochemical risk control in hypertensive patients with type 2 diabetes (UKPDS 40). BMJ
factors in type 2 diabetic patients at diagnosis compared with 1998;317:720-6.
age-matched normal subjects. Diabetic Med 1994;11:534-44. 53 UKPDS Group. UK prospective diabetes study 16: overview of six years’
22 Easton DF, Peto J, Babiker AG. Floating absolute risk: an alternative to therapy of type 2 diabetes—a progressive disease. Diabetes 1995;44:1249-
relative risk in survival and case-control analysis avoiding an arbitrary 58.
reference group. Stat Med 1991;10:1025-35. 54 Svetkey LP, Simons-Morton D, Vollmer WM, Appel LJ, Conlin PR, Ryan
23 SAS. Version 6. Cary, NC: SAS Institute, 1990. DH, et al. Effects of dietary patterns on blood pressure: subgroup analysis
24 Heart Outcome Prevention Evaluation Study Investigators. Effect of an of the dietary approaches to stop hypertension (DASH) randomized
angiotensin-converting enzyme inhibitor, ramipril, on death from clinical trial. Arch Intern Med 1999;159:285-93.
cardiovascular causes, myocardial infarction and stroke in high-risk 55 Eriksson J, Lindström J, Valle T, Aunola S, Hämäläinen H, Ilanne-Parikka
patients. N Engl J Med 2000;342:145-60. P, et al. Prevention of type II diabetes in subjects with impaired glucose
25 Heart Outcomes Prevention Evaluation Study Investigators. Effects of tolerance: the diabetes prevention study in Finland. Diabetologia
ramipril on cardiovascular and microvascular outcomes in people with 1999;42:793-801.
diabetes mellitus: results of the HOPE study and MICRO-HOPE 56 Rose G. Strategy of prevention: lessons from cardiovascular disease. BMJ
substudy. Lancet 2000;355:253-9. 1981;282:1847-51.
26 Borghi C, Ambrosioni E. Evidence-based medicine and ACE inhibition.
J Cardiovasc Pharmacol 1998;32(suppl 2):24-35. (Accepted 20 March 2000)

003174634
BMJ VOLUME 3211.012 AUGUST 2000 bmj.com 419