Perry, Jr HM, Davis BR, Price TR, Applegate WB, Fields WS, Guralnik JM, Kuller L, Pressel S, Stamler J, Probstfield JL, for the Systolic Hypertension in the Elderly Program (SHEP) Cooperative
Research Group . Effect of Treating Isolated Systolic Hypertension on the Risk of Developing
Various Types and Subtypes of StrokeThe Systolic Hypertension in the Elderly Program (SHEP). JAMA. 2000;284(4):465-471. doi:10.1001/jama.284.4.465
Author Affiliations: Department of Medicine, Washington University, St Louis, Mo (Dr Perry); School of Public Health, University of Texas Health Science Center, Houston (Drs Davis and Fields and Ms Pressel); Department of Neurology, University of Maryland School of Medicine, Baltimore (Dr Price); Department of Medicine, Wake Forest University, Winston-Salem, NC (Dr Applegate); National Institute on Aging, Bethesda, Md (Dr Guralnik); Department of Epidemiology, University of Pittsburgh, Pittsburgh, Pa (Dr Kuller); Department of Preventive Medicine, Northwestern University Medical School, Chicago, Ill (Dr Stamler); and Department of Medicine, University of Washington, Seattle (Dr Probstfield). A complete list of the SHEP Cooperative Research Group was published previously (JAMA. 1991;265:3255-3264).
Context The Systolic Hypertension in the Elderly Program (SHEP) demonstrated
that treating isolated systolic hypertension in older patients decreased incidence
of total stroke, but whether all types of stroke were reduced was not evaluated.
Objective To investigate antihypertensive drug treatment effects on incidence
of stroke by type and subtype, timing of strokes, case-fatality rates, stroke
residual effects, and relationship of attained systolic blood pressure to
Design The SHEP study, a randomized, double-blind, placebo-controlled trial
began March 1, 1985, and had an average follow-up of 4.5 years.
Setting and Participants A total of 4736 men and women aged 60 years or older with isolated systolic
hypertension at 16 clinical centers in the United States.
Interventions Patients were randomly assigned to receive treatment with 12.5 mg/d
of chlorthalidone (step 1); either 25 mg/d of atenolol or 0.05 mg/d of reserpine
(step 2) could be added (n = 2365); or placebo (n = 2371).
Main Outcome Measures Occurrence, type and subtype, and timing of first strokes and stroke
fatalities; and change in stroke incidence for participants (whether in active
treatment or placebo groups) reaching study-specific systolic blood pressure
goal (decrease of at least 20 mm Hg from baseline to below 160 mm Hg) compared
with participants not reaching goal.
Results A total of 85 and 132 participants in the active treatment and placebo
groups, respectively, had ischemic strokes (adjusted relative risk [RR], 0.63;
95% confidence interval [CI], 0.48-0.82); 9 and 19 had hemorrhagic strokes
(adjusted RR, 0.46; 95% CI, 0.21-1.02); and 9 and 8 had strokes of unknown
type (adjusted RR, 1.05; 95% CI, 0.40-2.73), respectively. Four subtypes of
ischemic stroke were observed in active treatment and placebo group participants,
respectively, as follows: for lacunar, n = 23 and n = 43 (adjusted RR, 0.53;
95% CI, 0.32-0.88); for embolic, n = 9 and n = 16 (adjusted RR, 0.56; 95%
CI, 0.25-1.27); for atherosclerotic, n = 13 and n = 13 (adjusted RR, 0.99;
95% CI, 0.46-2.15); and for unknown subtype, n = 40 and n = 60 (adjusted RR,
0.64; 95% CI, 0.43-0.96). Treatment effect was observed within 1 year for
hemorrhagic strokes but was not seen until the second year for ischemic strokes.
Stroke incidence significantly decreased in participants attaining study-specific
systolic blood pressure goals.
Conclusions In this study, antihypertensive drug treatment reduced the incidence
of both hemorrhagic and ischemic (including lacunar) strokes. Reduction in
stroke incidence occurred when specific systolic blood pressure goals were
The Systolic Hypertension in the Elderly Program (SHEP) was the first
completed trial investigating isolated systolic hypertension. Results showed
that treatment of hypertension reduced all strokes, both fatal and nonfatal,
by 36%; all myocardial infarctions (MIs), both fatal and nonfatal, by 27%;
all coronary heart disease by 27%; and all cardiovascular disease by 32%.
Total mortality was reduced by 13%. Reductions were also demonstrated in the
number of transient ischemic attacks (TIAs), and episodes of congestive heart
This article examines, for the first time to our knowledge, the distribution
of stroke types (ischemic or hemorrhagic) and their various subtypes in active
treatment vs placebo group participants. It also examines the timing of strokes,
the number that were fatal, their relationship to attained systolic blood
pressure (SBP), and their residual effects.
SHEP was a double-blind, randomized, placebo-controlled trial designed
to test whether antihypertensive drug therapy reduces the frequency of new
strokes in a multiethnic cohort of 4736 men and women aged 60 years or older
with isolated systolic hypertension (SBP ≥160 mm Hg and diastolic blood
pressure <90 mm Hg).2
The cohort, recruited at 16 clinical centers, was 43.2% male and 13.9%
black. Average age at randomization was 71.6 years, with 42% of participants
in their 60s, 45% in their 70s, and 14% in their 80s or 90s. Baseline blood
pressure averaged 170/77 mm Hg; 57% of participants had SBP from 160 to 169
mm Hg, 27% from 170 to 179 mm Hg, and 15% had 180 mm Hg or higher. Ten percent
of participants had diabetes mellitus, 4.9% had experienced an MI, and 1.4%
had experienced a previous stroke. Subjects with atrial fibrillation were
excluded from the trial. Participants were randomly assigned to double-blind
active treatment or placebo groups. Average follow-up in the trial was 4.5
The first treatment step was chlorthalidone (12.5 mg/d), which could
be doubled if the participant's SBP was not controlled at goal. The second
step, if needed, was atenolol (25 mg/d) or low-dose reserpine (0.05 mg/d)
for those with a contraindication to β-blockers. The second-step drug
also could be doubled if control had not been acheived. At the end of the
trial, 46% of participants randomized to active treatment were receiving the
step 1 drug only and 23% were receiving step 1 and step 2 drugs; however,
89% to 90% of participants in the active treatment group were receiving some
active drug at all 5 annual visits. The percentage of participants in the
placebo group taking active antihypertensive drug(s) increased progressively
from 13% at year 1 to 44% at year 5.
During the trial, the decrease in blood pressure from pretreatment baseline
averaged 26/9 mm Hg for participants in the active treatment group and 15/4
mm Hg for those in the placebo group; the amount of decrease changed little
during the trial.1 At the 5-year visit, 65%
of participants in the active treatment group and 40% in the placebo group
were at goal SBP (decrease in baseline SBP of ≥20 mm Hg to an SBP of <160
Participants with strokes were hospitalized during the acute phase and
were examined by a trial neurologist who completed standardized forms describing
the event. When examination was not feasible, pertinent records were reviewed
by a trial neurologist and the forms completed. Transient ischemic attacks
were handled similarly. Final decisions regarding TIA and stroke end points
(including type and subtype) were made in a face-to-face unanimous decision
process by 3 physicians on the end point subcommittee. Diagnoses were based
on information from hospital charts, standardized forms giving findings on
history and physical examination, evaluations of brain images (computed tomography
or magnetic resonance imaging) by 2 neuroradiologists blinded to treatment
groups, and the clinical center neurologist's diagnosis. All strokes were
evaluated, but only the first stroke was considered unless otherwise stated.
Transient ischemic attack was defined as the
rapid onset of a focal neurological deficit lasting more than 30 seconds and
less than 24 hours without evidence of an underlying nonvascular cause and
presumed to be caused by cerebral ischemia.
Stroke was defined as the rapid onset of a
new neurological deficit attributed to obstruction or rupture in the cerebral
arterial system. The defined deficit had to persist at least 24 hours unless
death supervened and had to include specific localizing findings confirmed
by neurological examination and without evidence of an underlying nonvascular
The method of determining stroke type (ischemic, hemorrhagic, or unknown)
and subtype was similar to that used in the Stroke Data Bank.4
A patient was diagnosed as having a hemorrhagic stroke
if intracranial bleeding was found on brain imaging, by lumbar puncture, or
at autopsy, and there was no evidence on the brain image of late bleeding
into an ischemic infarction. Ischemic stroke was
diagnosed when a focal neurological deficit was present without blood in the
brain image or lumbar puncture. Stroke, type unknown,
was diagnosed when the definition of stroke was satisfied but there was insufficient
evidence to determine whether it was hemorrhagic or ischemic.
Ischemic strokes were subdivided into lacunar, cardioembolic, atherosclerotic,
or other/unknown on the basis of clinical information and brain imaging.
Lacunar stroke was diagnosed either when a
small (<2 cm in diameter) lesion correlated with symptoms and/or signs
of the stroke was seen in deep structures of the hemisphere or in the pons
by brain imaging or if brain imaging showed no lesion responsible for the
signs or symptoms and the patient had pure motor hemiparesis, pure sensory
stroke, dysarthria-clumsy hand syndrome, or ataxia hemiparesis syndrome.
Embolic stroke was diagnosed when there was
insufficient evidence of a lacune and a known source of embolism was present
(eg, atrial fibrillation or recent MI).
Atherosclerotic infarction was diagnosed when
there was evidence by noninvasive test (eg, carotid duplex) or by angiogram
of at least 70% stenosis of the appropriate artery (eg, left carotid compromise
with left cerebral hemisphere infarction).
Ischemic infarction, other/unknown type was
diagnosed when no specific subtype was identified.
Hemorrhagic strokes were subdivided into subarachnoid hemorrhage if
blood was seen in the subarachnoid space, or intraparenchymal hemorrhage if
blood was seen within the brain substance.
All 18 participants who died within 30 days of their first stroke were
considered stroke fatalities. Six additional participants died within 30 days
of subsequent strokes; they too were considered stroke fatalities. None of
the 24 stroke fatalities had other obvious causes of death. For the 38 participants
with strokes who died more than a month after their strokes, other events
were considered to be primary causes of death.
Satisfactory brain images were obtained during hospitalization following
85% of first strokes. For strokes occurring early in the trial, adequate imaging
was not available for 15 participants in the active treatment group and 25
in the placebo group, thus 222 participants had adequate images for evaluation.
However, for 122 participants, no appropriate lesion was recognized, leaving
100 participants with strokes having identifiable lesions. For the 100 participants
with visualized lesions, stroke volumes were approximated using the method
of Kothari et al.5
Blood pressure levels were measured by trained, certified technicians
using random-zero sphygmomanometers. All visit SBP levels were the average
of 2 measurements made at that visit with the subject seated quietly. Baseline
SBP was the average (between 160 and 219 mm Hg) of measurements at the first
and second baseline visits. In-trial visit SBP levels were the average of
measurements at that visit. The 1-year SBP was the value at the first annual
visit. For participants with strokes, in-trial prestroke SBP was the average
SBP at all visits before the first stroke.
Relationships were sought between in-trial prestroke SBP and the likelihood
of stroke. For this purpose, SBP was considered a dichotomous variable, with
participants having average prestroke SBP above or below the cut point. All
participants with strokes were divided into these 2 groups whether randomized
to placebo group or active treatment group and whether taking antihypertensive
drug or not. The nonstroke SHEP participants were also divided by the same
cut points on the basis of SBP at 1 year (when in-trial SBP had been established),
providing the groups from which participants with strokes came.
Three estimates of stroke residual were made. Admissions to skilled
or intermediate care nursing homes were tabulated for all participants. Activities
of daily living (ADL) data were obtained at baseline and annually for each
SHEP participant, using a standardized questionnaire derived from the Barthel
Index.6,7 Seven activities were
assessed: walking across a small room, bathing, grooming oneself, dressing,
eating, getting from bed to chair, and using the toilet. To evaluate stroke-induced
changes in ADL, the last prestroke assessment and the first assessment more
than 6 months after the stroke were compared. With the double-blind intact,
participants reported the number of disability days of 2 types: days of reduced
ordinary activity and days in bed (a subset of disability days) for the 2
weeks before each annual visit.
All analyses were performed on an intention-to-treat basis, and unless
specific exception is indicated, only first strokes were considered. Cumulative
rates were calculated using life-table methods. Comparisons of active treatment
and placebo groups with respect to stroke and its subtypes, and reductions
in risk with control for baseline variables were determined using proportional
hazards analyses.8 The proportional hazards
assumption was tested using Schoenfeld's approach.9
This assumption was not violated. Mean ADL scores were compared between treatment
groups using t tests.10
Days of reduced activity and days in bed were compared between treatment groups
using rank-order tests.11 A priori, there was
relatively low power to detect differences between treatment groups for poststroke
events. The numbers, percentages, and P values are
a more descriptive guide.
The relationship between in-trial SBP during treatment and stroke incidence
was examined using Cox proportional hazards regression models accounting for
both time-constant and time-dependent covariates. Time-constant baseline variables
included in the model were age, race, sex, antihypertensive medication status
at initial contact, history of prior MI and/or stroke, history of diabetes,
current smoking status, presence of specified electrocardiogram abnormalities,
body mass index (BMI), BMI squared, heart rate, total serum cholesterol and
high-density lipoprotein cholesterol levels, and years of education. The time-dependent
variable was SBP, modeled as either a dichotomous or a continuous variable.
Treatment reduced the incidence of all strokes and of both stroke types
(Table 1). There were significantly
fewer ischemic strokes among participants in the active treatment group than
among those in the placebo group (85 vs 132), fewer lacunar and other/unknown
subtypes of stroke, and equal numbers of atherosclerotic strokes. Although
there were fewer hemorrhagic strokes, both subarachnoid and intraparenchymal
numbers were small and differences between participants in the active treatment
and placebo groups were not significant. After the indicated adjustments, Table 1 provides risk ratios (RRs) and
95% confidence intervals (CIs) for stroke incidences.
During the first postrandomization year, there were 2 hemorrhagic strokes
for participants in the active treatment group and 7 in the placebo group,
suggesting a prompt treatment effect. The numbers of hemorrhagic strokes during
the trial totaled 9 and 19, respectively. The numbers of ischemic strokes
during the first year were similar in both groups (25 participants in the
active treatment group and 26 in the placebo group). Thereafter, the difference
between the 2 groups increased to 85 and 132, respectively (Figure 1).
Twenty-four (9.2%) of the 262 participants with strokes died of stroke,
18 within 30 days of their first stroke and 6 within 30 days of a subsequent
stroke (4 second strokes, 1 third stroke, and 1 fourth stroke). Stroke mortality
rates were similar for participants in the active treatment group and in the
placebo group. Thirty-eight other participants (14 in the active treatment
group and 24 in the placebo group) had nonfatal first strokes followed by
fatal nonstroke events an average of 414 days later (Table 2). Of the 38 fatal nonstroke events, 14 were cardiovascular
(5 participants in the active treatment group and 9 in placebo group).
A total of 29 participants (11% of those with strokes) had a second
stroke; 12 were in the active treatment group and 17 were in the placebo group
(P = .34). The median interval between the first
and second stroke was 319 days (375 for the active treatment group and 284
for the placebo group), with a range of 18 to 1561 days.
Sixty-two participants in the active treatment group and 82 in the placebo
group had TIAs during the trial (adjusted RR, 0.74; 95% CI, 0.54-1.04). Of
these 144 participants, 22 had a stroke, but only 15 (5 in the active treatment
group and 10 in the placebo group) had TIAs prior to their first stroke.
Forty-four participants in the active treatment group and 55 in the
placebo group experienced atrial fibrillation during the trial (P = .26). Fourteen of these 99 had strokes, but for only 6 (2 in the
active treatment group and 4 in the placebo group) was atrial fibrillation
recognized before the first stroke.
A total of 63 participants in the active treatment group and 98 in the
placebo group developed symptomatic MI during the trial (P = .005). Fourteen of these 161 had strokes, and for 9 of them (4
participants in the active treatment group and 5 in the placebo group) the
MI followed the first stroke.
Two hundred twenty-two participants with strokes had brain images adequate
for evaluation. Of these, 122 (55%) had no visible lesions related to their
strokes, leaving 100 (42 in the active treatment group and 58 in the placebo
group) with appropriately localized lesions ascribed to their first stroke.
Hemorrhagic strokes were visualized more often (82%; 23/28) than ischemic
strokes (40%; 77/194).
For the 100 visualized strokes, there were no obvious differences between
participants in the active treatment group and those in the placebo group
in their frequency or approximate stroke volumes. Lacunar strokes in both
treatment groups were markedly smaller than other stroke subtypes. Combining
both treatment groups, the median stroke volumes were lacunar, 52 mm3; embolic, 1500 mm3; atherosclerotic, 1470 mm3;
subarachnoid, 1754 mm3; and intraparenchymal, 1960 mm3.
Participants who attained specific in-trial, prestroke SBP goals had
reduced likelihood of developing strokes (Table 3). One year after randomization, an SBP lower than 160 mm
Hg had been achieved by two thirds (3162) of all SHEP participants; 1906 (60.3%)
were in the active treatment group and 1256 (39.7%) were in the placebo group.
Although more participants in the active treatment group than in the placebo
group reached each SBP level of Table 3, reduction in stroke incidence for those who had reached a goal
was similar for both treatment groups. At 1 year, some participants in the
placebo group were taking active drug(s): 15% of those with SBP of lower than
160 mm Hg and 18% of those with SBP of lower than 150 mm Hg.
Participants whose in-trial prestroke SBP was lower than 160 mm Hg experienced
a significant reduction (33%) (adjusted RR, 0.67; 95% CI, 0.51-0.89) in total
stroke incidence compared with participants having higher in-trial prestroke
SBP (Table 3). Participants whose
SBP was lower than 150 mm Hg experienced a 38% reduction in total stroke incidence
compared with those having an SBP of 150 mm Hg or higher. For the smaller
number of participants (1356) with an SBP lower than 140 mm Hg, the 22% reduction
in stroke incidence was not statistically significant compared with those
having an SBP of 140 mm Hg or higher. Data for all strokes and for ischemic
and lacunar strokes are cited in Table 3.
Approximately half (2317) of SHEP participants reached the SHEP goal
(decrease in SBP of ≥20 mm Hg from baseline to <160 mm Hg); they had
a significant decrease (33%) in stroke incidence compared with participants
who had not reached that goal (Table 3).
In a model using in-trial prestroke SBP as a continuous variable, for
every millimeter of mercury decrease in SBP, stroke incidence decreased by
1%. At 1 year, two thirds of all SHEP participants had decreases in SBP of
10 mm Hg or more. Those who had achieved that goal had a significant decrease
(10%) in total stroke incidence compared with those who had not achieved it
Three estimates of stroke residuals were made. Within 6 months of their
first stroke, 4 participants (4%) in the active treatment group and 8 (5%)
in the placebo group were admitted to skilled or intermediate care nursing
homes. By the end of the trial, 6 participants in the active treatment group
and 15 in the placebo group were admitted to such nursing homes; the average
interval between stroke and admission approximated 250 days for both treatment
groups. For comparison, 82 (1.8%) of the 4474 nonstroke SHEP participants
had such admissions during the trial follow-up at 4.5 years (average).
Of the 262 participants with stroke, 87 participants in the active treatment
group and 117 in the placebo group had both prestroke and poststroke ADL assessments.
For them, the mean decrease in ADL score (indicating increased disability)
was 1.3 for participants in the active treatment group and 1.5 for those in
the placebo group. The ADL scores decreased in 1 or more of 7 tested domains
for 31 participants (31%) in the active treatment group and 56 (35%) in the
placebo group. Scores did not change for 51 participants (50%) in the active
treatment group and 66 (42%) in the placebo group. Differences between treatment
groups were not significant.
Self-reported days of reduced activity were consistently, but not significantly,
fewer for participants in the active treatment group than for those in the
placebo group at 3 poststroke intervals (Table 4). Days in bed were also consistently fewer for participants
in the active treatment group than in the placebo group; moreover, at the
last trial report, the difference in the average of 0.25 days in bed for 61
participants in the active treatment group and the average of 1.18 days for
92 participants in the placebo group was significant (P = .03) (Table 4).
One in 10 of the 262 first strokes observed was hemorrhagic, slightly
more than 8 in 10 were ischemic, and the remainder were of unidentified type.
The frequency of stroke type and subtype diagnoses is similar to that observed
in the Stroke Data Bank.4 In both SHEP and
the Stroke Data Bank, no subtype diagnosis could be made for about 45% of
ischemic strokes. Most ischemic strokes in SHEP (60%) were not visualized
and the Stroke Data Bank had comparable findings, with as many as 50% of first
computed tomographic scans for ischemic strokes being normal.4
Hypertension is characterized by microaneurysm, lipohyalinosis, and
fibrinoid necrosis, particularly in penetrating arteries that supply basal
ganglia, cerebral deep white matter, and pons.12- 14
These are sites of lacunar stroke and intraparenchymal hemorrhage, the 2 stroke
subtypes most strongly associated with hypertension, both of which may have
been decreased by treatment in SHEP. Lacunar strokes have been reported to
decline in a community referral hospital after treatment,15
and a decrease in primary intraparenchymal hemorrhage has been attributed
to antihypertensive treatment.16 The apparent
lack of treatment effect on the incidence of atherosclerotic strokes, which
comprised 12% and 9% of ischemic strokes in SHEP and the Stroke Data Bank,4 respectively, was unexpected, since hypertension is
associated with increased atherosclerosis.17- 19
Although there were 40% more fatal strokes among participants in the
placebo group than in the active treatment group, fatal strokes were few and
the percentage of strokes that were fatal were similar for the 2 treatment
groups. The Swedish Trial in Old Patients with Hypertension (STOP-Hypertension),
involving a cohort of high-risk elderly hypertensive patients, showed decreases
in stroke incidence and in all-cause and cause-specific mortality. In the
active treatment group of that trial, 3 (10%) of 29 strokes were fatal, whereas
in the placebo group, 12 (23%) of 53 were fatal.20
In the Medical Research Council Trial of Treatment of Hypertension in Older
Adults, treatment decreased stroke incidence; however, like SHEP, mortality
rates for strokes that did occur were similar for both treatment groups. In
that trial, 37 (37%) of 101 participants in the active treatment group died
vs 42 (31%) of 134 participants in the placebo group.21
The Medical Research Council Trial involved participants seemingly at lower
risk of stroke than SHEP, but it experienced difficulties with adherence.
Finally, SHEP and the Stroke Data Bank4 used
similar methods to identify stroke types. Although SHEP participants were
older, case fatality rates were similar, with the 30-day rate for ischemic
strokes being 5% for SHEP and 8% for the Stroke Data Bank, and the rates for
hemorrhagic stroke being 39% and 30%, respectively.
Usually considered precursors of stroke, TIA and atrial fibrillation
relatively rarely preceded first stroke in SHEP. Only 15 of 262 participants
experienced TIA prior to their first strokes, and atrial fibrillation was
recognized in only 6 prior to their first strokes.
The fact that achieving specific SBP goals significantly decreased the
incidence of stroke should encourage both the physician and patient to strive
for such goals. Thus, decreasing the SBP to less than 160 mm Hg lowered the
stroke rate by one third and decreasing SBP to less than 150 mm Hg lowered
the rate even more. Decreasing the SBP to less than 140 mm Hg did not have
a significant effect, perhaps because fewer participants achieved that goal.
For those reaching a goal SBP, reduction was similar regardless of the participant's
randomization group, although more participants in the active treatment group
than in the placebo group reached each goal. This strongly suggests that the
level of attained SBP rather than treatment or a particular antihypertensive
drug was the paramount factor in reducing stroke incidence.
Three measures of stroke residual were examined in seeking differences
between the 2 treatment groups. First, nursing home admissions were similar
for both treatment groups. Second, ADL scores worsened slightly, but not significantly
more, for participants in the placebo group. Third, the consistently fewer
days of reduced activity, including days in bed, suggest that participants
in the active treatment group were less disabled by strokes than those in
the placebo group. Although these data must be considered "soft," they were
self-reported with the double-blind intact.
Treatment induced a significant reduction in the incidence of all strokes
in elderly patients with isolated systolic hypertension. There were also reductions
in ischemic and hemorrhagic stroke types and in the lacunar subtype. Similar
percentages of strokes in participants in the active treatment group and in
the placebo group were fatal. The treatment effect may have appeared earlier
for hemorrhagic than for ischemic strokes. Significant reduction in stroke
incidence was observed for any participants reaching certain SBP levels: SHEP
goal (20 mm Hg decrease in SBP to <160 mm Hg), SBP below 160 mm Hg, and
SBP decrease of 10 mm Hg. Certain stroke residuals may be less among participants
in the active treatment group than in the placebo group. Elderly patients
and their physicians should be diligent in treating isolated systolic hypertension
to achieve a goal SBP.