The n e w e ng l a n d j o u r na l of m e dic i n e

Review Article

The Changing Face of Clinical Trials

Jeffrey M. Drazen, M.D., David P. Harrington, Ph.D., John J.V. McMurray, M.D., James H. Ware, Ph.D., and
Janet Woodcock, M.D., Editors

Lessons in Uncertainty and Humility —

Clinical Trials Involving Hypertension
Marc A. Pfeffer, M.D., Ph.D., and John J.V. McMurray, M.D.​​

T
From the Cardiovascular Division, Brigham he central tenet of clinical research in humans is that par-
and Women’s Hospital, Harvard Medical ticipation is informed and voluntary. Participation is considered to be in-
School, Boston (M.A.P.); and the British
Heart Failure Cardiovascular Research formed only when there are properly scrutinized protocols and when par-
Centre, University of Glasgow, Glasgow, ticipants are fully informed of the known risks, discomforts, and inconveniences
United Kingdom (J.J.V.M.). Address reprint related to their involvement.1 Across the many types of clinical investigation, the
requests to Dr. Pfeffer at the Cardiovascu-
lar Division, Brigham and Women’s Hos- randomized, controlled trial is the key scientific tool used to measure the efficacy
pital, 75 Francis St., Boston, MA 02115, and adverse effects of a diagnostic or therapeutic approach.2 Randomization, the
or at ­mpfeffer@​­rics​.­bwh​.­harvard​.­edu. critical element used to reduce bias, not uncommonly creates ethical quandaries,
N Engl J Med 2016;375:1756-66. since the treating physician and the patient, by agreeing to participate in the trial,
DOI: 10.1056/NEJMra1510067 forgo their choice of the intervention being tested, leaving this to chance.3 For
Copyright © 2016 Massachusetts Medical Society.
those designing the trial, making the decision as to whether the comparator treat-
ment should be an inert substance — that is, a placebo — or an active treatment
can be difficult. For those making the decision as to whether to participate, it is
important that they fully understand the medical question being explored and the
role they will play in helping to obtain an answer.
Clinical trials do not occur in a vacuum. Before a major clinical-outcome trial
is started, there is often substantial information available about the intervention
from smaller, hypothesis-generating trials that are usually intended to examine
the effects of the intervention on less definitive outcomes, such as a biomarker or
a surrogate end point that is believed to correlate with changes in clinical events.
In some cases observational data may be available, and in other instances previous
clinical-outcome trials have established the effectiveness of an agent but have done
so in a different patient population. Sometimes this existing information engen-
ders divergent and even polarized opinions. In that case, a randomized trial of
high quality should be welcomed by all parties as the best means to resolve the
controversy. Even so, strong preferences may undermine support for the random-
ization process, affecting the veracity of the trial by injecting bias into the selec-
tion of patients. In other instances, a solid consensus coalesces around what are
perceived to be the best existing data. When the intervention in question is already
generally part of clinical practice, many health professionals find comfort in its
use and view a randomized trial as unnecessary or even unethical, since some
patients would not receive the therapy that they would have received in the absence
of the trial. In this article, we use examples from cardiovascular medicine to dis-
cuss the effects of the existing and evolving information on trial design with re-
gard to the choice of placebo or active comparator agents and to the definition of
the population of patients in the trial.
Regardless of whether the prestudy perception is one of clinical equipoise or
involves a range of uncertainty, if a randomized trial is to be initiated and con-

1756 n engl j med 375;18 nejm.org  November 3, 2016

The New England Journal of Medicine

Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.
 Clinical Trials Series

ducted, the fundamental question it addresses emphasis was placed on detecting patients with
must be worthy of the major collective efforts asymptomatic ventricular arrhythmias in order
required of patients, investigators, and spon- to initiate antiarrhythmic agents that would pro-
sors.4 It is important to note that regardless of vide the assumed benefits. The totally unexpected
the level of confidence in the prestudy informa- result — that treating these abnormal rhythms
tion, a randomized, controlled trial remains a with antiarrhythmic agents was associated with
scientific experiment. Properly worded consent increased mortality — painfully disclosed the
forms should not lead a participant to expect any false sense of security of what was then a widely
unproven benefits. adopted clinical practice. The prompt translation
Notwithstanding the careful planning of the of these findings into reductions in the use of
experts who design the trial, the scrutiny of these unsafe therapies improved public health
regulatory agencies, and the commitment of re- and created additional uncertainties. These sur-
sources from sponsors to demonstrate proposed prising findings also raised questions about other
theoretical benefits, many researchers conclude cardiovascular therapies whose use had became
that the intervention they had hoped would offer widespread after regulatory evaluations were
a diagnostic or therapeutic advance is no better satisfied, predominantly on the basis of altera-
than the existing options.5,6 The effectiveness of tions of surrogate measures rather than clinical
an intervention is best evaluated in randomized events.
trials in which well-defined clinical events are
the primary outcome. E a r ly Data on Bl o od Pr e ssur e
In the absence of clinical-outcome data that a nd Ou t c ome s
provide definitive direction, surrogate measures
are often used to direct therapeutic decisions. Our story is about the treatment of hypertension
Consistent linkage from reliable observational and the finding in multiple trials that the normal-
data between a commonly used biologic measure- ization of blood pressure has been consistently
ment and prognosis can understandably lead to accompanied by beneficial clinical outcomes
a perceived association that changes in that bio- (Fig. 1). Elevations in systemic arterial pressure
marker can be reasonably assumed to cause direc- were identified in early pioneering epidemio-
tionally similar alterations in clinical outcomes. logic studies (e.g., the Framingham Heart Study
For many of these biologic measures, such as and the Multiple Risk Factor Intervention Trial
levels of high-density lipoprotein cholesterol or [MRFIT]) as being tightly coupled to a height-
hemoglobin, the density of ventricular arrhyth- ened risk of illness and death from cardiovascu-
mias, and left ventricular ejection fraction, multi- lar disease.9,10 A historical review of some of the
ple observational studies are available that show predominantly government-sponsored random-
such quantitative associations with cardiovascular ized trials that were designed to ascertain wheth-
risk.7 The finding of a quantitative relationship er and when long-term antihypertensive therapy
between the degree of deviation of the measured lowers these risks illustrates several of the ethi-
marker from normality and adverse outcomes cal issues that influence the comparators and
has on occasion fostered the reasonable expecta- the patient populations chosen for study as new
tion for both drug discovery and clinical practice information is generated.
that therapies that restore the abnormal mea- One of the early randomized, placebo-con-
surement toward normal would have a favorable trolled, clinical-outcome trials in cardiovascular
outcome. medicine, commencing in 1963, was conducted
In cardiovascular medicine, this comfort zone by the Veterans Administration (VA) Cooperative
was shattered by the results of the Cardiac Ar- Study Group on Antihypertensive Agents. In its
rhythmia Suppression Trial, which showed that trial involving 143 men whose diastolic blood
despite the clear association between the de- pressure was between 115 and 129 mm Hg, the
gree of ventricular ectopy and risk of death, anti­ group found that the use of a combination of
arrhythmic therapy, which was effective in sup- three antihypertensive agents resulted in fewer
pressing ventricular premature beats, increased cardiovascular events than the use of placebo.11
rather than decreased mortality.8 Before this Since the risks and the potential benefits of us-
placebo-controlled trial was conducted, a major ing pharmacologic agents to reduce blood pres-

n engl j med 375;18 nejm.org  November 3, 2016 1757

The New England Journal of Medicine
Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

160
Systolic ≥160 JNC 4 ≥160 JNC 8*
blood pressure JNC 3 (1988) ESH–ESC (2014)
Blood-Pressure Threshold at Which Antihypertensive

150 (1984) (2003)

≥150
Medication Should Be Initiated (mm Hg)

No systolic ≥60 yr
140 recommendation
≥140 JNC 6 JNC 7 ≥140 ESH–ESC <60 yr
JNC 5 (1997) (2003) ESH–ESC (2013) JNC 8*
130 (1993) (2007) (2014)

120

110 Diastolic ESH–ESC

blood pressure (2003)
≥100
≥105 JNC 2
100
JNC 1 (1980)
(1977)
≥95 JNC 4
90
JNC 3 (1988) ≥90 ≥90
JNC 6 JNC 7 ESH–ESC JNC 8*
(1984) ESH–ESC
JNC 5 (1997) (2003) (2013) (2014)
80 (1993) (2007)

0
Trial 1966 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
VA 1 VA 2 HDFP SHEP Syst-Eur ALLHAT HYVET SPRINT
MRC

Australian Trial STOP HOT ALLHAT ACCORD HOPE-3

Figure 1. Trials Influencing Blood-Pressure Thresholds at Which Antihypertensive Medications Should Be Used.
ACCORD denotes Action to Control Cardiovascular Risk in Diabetics, ALLHAT Antihypertensive and Lipid-Lowering Treatment to Prevent
Heart Attack Trial, Australian Trial the Australian Therapeutic Trial, ESH–ESC the European Society of Hypertension–European Society
of Cardiology Guidelines, HDFP Hypertension Detection and Follow-up Program, HOPE Heart Outcomes Prevention Evaluation, HOT
Hypertension Optimal Treatment, HYVET Hypertension in the Very Elderly Trial, JNC Joint National Committee on Prevention, Detection,
Evaluation, and Treatment of High Blood Pressure, MRC Medical Research Council, SHEP Systolic Hypertension in the Elderly Program,
SPRINT Systolic Blood Pressure Intervention Trial, STOP Swedish Trial in Old Patients with Hypertension, Syst-Eur Systolic Hypertension in
Europe, and VA Veterans Administration Cooperative Study Group. The asterisk that follows JNC 8 denotes findings that appear as they
were submitted by the appointed panel members, without official endorsements.

sure were unknown, placebo was unquestion- ventive medicine. However, at the time, the clini-
ably the correct comparator. The same group cal importance of treating an asymptomatic per-
concurrently conducted another trial of 380 men son with the available therapies for lowering
whose diastolic pressures were between 90 and blood pressure was not readily accepted by the
114 mm Hg.12 The lower diastolic pressure of the medical community.
participants in this trial made placebo the appro-
priate control for this previously unstudied popu- R ec o gni t ion of the Da nger s
lation. The combination of the participants’ lower of H y per tension
diastolic pressure and the fact that this popula-
tion had not been studied previously also made The National High Blood Pressure Education Pro-
placebo the appropriate control in this trial. The gram was established in 1972 by philanthropist
findings of each of these placebo-controlled Mary Lasker and then-Secretary of Health, Edu-
trials, which showed that the adverse cardiovas- cation, and Welfare Elliot Richardson with the
cular consequences of elevated blood pressure objective of alerting physicians and the public to
could be reduced by the administration of anti- the risks of hypertension.13 This government-
hypertensive medications, are now considered to funded educational program used phrases such
be foundational evidence for the concept of pre- as “the silent killer” to heighten awareness of

1758 n engl j med 375;18 nejm.org  November 3, 2016

The New England Journal of Medicine

Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.
 Clinical Trials Series

the dangers of elevated blood pressure and to ing a diastolic pressure of less than 90 mm Hg
encourage the use of antihypertensive therapy was achieved in 56.7% of the patients in the SC
in patients with hypertension. The reluctance of group and in only 33.7% of the patients in the
the medical community to accept the findings RC group. Even this difference in the intensity of
of the two VA trials was reflected in the design of blood-pressure treatment was associated with a
subsequent major trials focused on hypertension. statistically and clinically significant difference in
The investigators conducting the Hyperten- mortality, which was 17% lower in the SC group
sion Detection and Follow-up Program (HDFP) than in the RC group.14
that was sponsored by National Institutes of The response of the international medical
Health designed their trial to determine whether community to the results of the two relatively
improved blood-pressure control could prolong small initial VA studies was even more restrained
survival. The trial included nearly 11,000 people, than that in the United States. Near the time at
and enrollment began in 1974. Given that study which the HDFP was being conducted in the
entry required a diastolic blood pressure of at United States, three trials related to hypertension
least 90 mm Hg (with more than 1000 partici- control were being performed outside the United
pants having a diastolic pressure of more than States. In each of these trials, placebo was select­
115 mm Hg at study entry), the HDFP was in ed as the comparator (Table 1). Entry into the
some respects an effort to confirm the results of European Working Party Trial16 and the Medi-
the VA trials while also expanding the population cal Research Council Trial (conducted in the Unit­
covered to include women and younger people. ed Kingdom)17 required a diastolic pressure of
Despite these population differences, the design- 90 mm Hg or higher, and entry into the Austra-
ers of the HDFP “agreed that the results of the lian Therapeutic Trial15 required a diastolic pres-
Veterans trials made it inappropriate to use pla- sure of 95 mm Hg or higher. The use of placebo
cebo controls.”14 This clear statement and early in all three trials, which commenced in the
shift from placebo to an active comparator indi- 1970s, was a clear indication that the research-
cated that the investigators believed that it was ers’ level of uncertainty with regard to the need
unethical to withhold the opportunity for anti- to treat people with this level of blood pressure
hypertensive therapy from participants during a was higher than it was for the U.S. investigators
long-term clinical trial. However, the tension be- in the HDFP. Nonetheless, hypertension knows
tween clinical investigators and community prac- no borders, and all three trials showed that fewer
titioners with regard to the acceptance of the cardiovascular events occurred among patients
existing data was reflected in the trial design with better blood-pressure control. With these
and the differential use of antihypertensive drugs findings, the importance of lowering blood pres-
between groups. sure was solidified (Table 1).
In this 5-year HDFP, a total of 10,940 volun- The design of the Hypertension Optimal Treat-
teers (46% of whom were women) were randomly ment (HOT) trial, conducted in the mid-1990s,
assigned to receive either stepped care (SC), in showed that the international community of re-
which antihypertensive therapy was provided at searchers who were investigating hypertension
specialized sites and doses increased to achieve had accepted the proposition that in patients
and maintain specific blood-pressure goals, or with elevated diastolic blood pressure, the re-
referred care (RC), in which participants were duction of that pressure with antihypertensive
referred to their usual sources of care. The de- drugs improved cardiovascular outcomes.21 In the
sign was based on an inherent assumption that HOT trial, the diastolic pressure of participants
despite the absence of placebo, the ambivalence was between 100 and 115 mm Hg, and all re-
of the medical community toward antihyperten- ceived active antihypertensive therapy. Participants
sive drugs would result in fewer prescriptions in were randomly assigned to one of three groups
the RC group than in the SC group. Although a in which the targeted diastolic blood pressures
considerable number of patients in each group were less than 90, less than or equal to 85, or less
remained untreated, the percentage was more than or equal to 80 mm Hg. Although no signifi-
than twice as high in the RC group at year 2, cant differences in the rates of major cardiovas-
with 18.5% of the SC group and 41.7% of the RC cular events were observed across the three levels
group remaining untreated. The goal of attain- of treatment intensity, this trial, which included

n engl j med 375;18 nejm.org  November 3, 2016 1759

The New England Journal of Medicine
Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.
 Table 1. Major Outcome of Trials Assessing the Risks and Benefits of Treating Elevated Blood Pressure.*

1760
Blood Pressure Placebo Issues Regarding
Trial Eligibility Comparators at Inclusion Outcome Used Use of Placebo

mm Hg
11
VA Group 1 (1967) Men with diastolic pressure Antihypertensive treat- Treatment: 186/121 Decrease in composite of Yes Unstudied population
Duration: 1963–1967 115–129 mm Hg ment vs. placebo Placebo: 187/121 cardiovascular events
Participants: 143 and deaths
VA Group 2 (1970)12 Men with diastolic pressure Antihypertensive treat- Treatment: 162/104 Decrease of 70% in com- Yes Diastolic pressure different at
Duration: 1963–1969 90–114 mm Hg ment vs. placebo Placebo: 165/105 posite of cardiovascular entry; unstudied population
Participants: 380 events and deaths
HDFP (1979)14 Diastolic pressure Stepped care vs. referred Stepped care: 159/101 Decrease of 17% in all-cause No All patients could be treated
Duration: 1974–1979 90 to >115 mm Hg care Referred care: 159/101 mortality (P<0.01)
Participants: 10,940
Australian Therapeutic Diastolic pressure 95– Antihypertensive treat- Treatment: 158/101 Decrease of 20% in com- Yes Previous data considered insuf­
Trial (1980)15 109 mm Hg, systolic ment vs. placebo Placebo: 157/100 posite of morbid cardio- ficient for general recom-
The

Duration: 1973–1979 pressure <200 mm Hg vascular events and mendations

Participants: 3427 deaths (P<0.05)
European Working Party Diastolic pressure Antihypertensive treat- Treatment: 183/101 Decrease of 27% in cardio- Yes Previous data considered insuf­
Trial (1985)16 90–119 mm Hg, systolic ment vs. placebo Placebo: 182/101 vascular death (P = 0.04) ficient for general recom-
Duration: 1972–1984 pressure 160–239 mm Hg mendations
Participants: 840
MRC (1985)17 Diastolic pressure Thiazide or propanolol Men Decrease of 46% in stroke Yes Previous data considered insuffi-
Duration: 1977–1985 90–109 mm Hg, systolic vs. placebo Thiazide: 158/98 (P<0.01) and 19% in cient for general recommen-
Participants: 17,354 pressure <200 mm Hg Propanolol: 158/98 ­cardiovascular events dations; active drug initially
Placebo: 158/98 (P<0.05) recommended for patients
Women with systolic pressure ≥210
Thiazide: 165/98 mm Hg or diastolic pressure
Propanolol: 165/98 ≥115 mm Hg; reduced during
n e w e ng l a n d j o u r na l

The New England Journal of Medicine

Placebo: 165/98 trial to ≥200 mm Hg and
of

≥110 mm Hg, respectively

SHEP (1991)18 Systolic pressure Antihypertensive treat- Treatment: 171/77 Decrease of 36% in fatal Yes Active therapy recommended for
Duration: 1985–1991 >160 mm Hg, diastolic ment vs. placebo Placebo: 170/76 and nonfatal stroke patients with systolic pres-

n engl j med 375;18 nejm.org  November 3, 2016

Participants: 4736 pressure <90 mm Hg; (P<0.001) sure >240 mm Hg or diastol-
age ≥60 yr ic pressure >115 mm Hg at

Copyright © 2016 Massachusetts Medical Society. All rights reserved.

m e dic i n e

single visit and patients with

sustained systolic pressure
>220 mm Hg or sustained dia-
stolic pressure >90 mm Hg
STOP (1991)19 Diastolic pressure 105–120 Antihypertensive treat- Treatment: 195/102 Decrease of 38% in cardio- Yes Blood pressures required for eli-
Duration: 1985–1991 mm Hg or diastolic ment vs. placebo Placebo: 195/102 vascular death, myocar- gibility overlapped with those

Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Participants: 1627 pressure ≥90 mm Hg dial infarction, or stroke of some previous trials; open-
with systolic pressure (P = 0.003) label antihypertensive treat-
180–230 mm Hg; age ment for those with blood
70–84 yr pressure >230/120 mm Hg
 Blood Pressure Placebo Issues Regarding
Trial Eligibility Comparators at Inclusion Outcome Used Use of Placebo

mm Hg
Syst-Eur (1997)20 Diastolic pressure Antihypertensive treat- Treatment: 174/86 Decrease of 42% in stroke Yes Eligibility requirements similar
Duration: 1989–1997 <95 mm Hg with ment vs. placebo Placebo: 174/86 (P = 0.003) to SHEP and STOP; active
Participants: 4695 ­systolic pressure treatment for systolic pres-
160–219 mm Hg; sure >219 mm Hg or diastolic
age ≥60 yr pressure >99 mm Hg; during
trial, ethics committee low-
ered eligible systolic pressure
to 200 mm Hg
HOT trial (1998)21 Diastolic pressure Different intensities of Participants in each of No significant difference in No Therapy started with calcium-
Duration: 1992–1997 100–115 mm Hg antihypertensive three treatment groups: rate of cardiovascular channel blocker, with ACE
Participants: 18,790 ­treatment intended 170/105 death, myocardial infarc- ­inhibitor, beta-blocker, di-
to target diastolic tion, or stroke across all uretic, or any combination
blood pressures three groups; however, thereof added; three levels
of <90, ≤85, or fewer clinical events in of reduction of diastolic
≤80 mm Hg patients with diabetes in ­pressure targeted
subgroup targeting low-
er diastolic pressure
ALLHAT (2000)22 Systolic pressure Chlorthalidone vs. doxa- Chlorthalidone: 145/83 No significant difference in No Intended to control blood pres-
Duration: 1994–1999 ≥140 mm Hg or zosin Doxazosin: 145/84 primary outcome of sure and address potential
Participants: 24,335 ­diastolic pressure death from coronary differences between agents
≥90 mm Hg and one heart disease or nonfatal
or more risk factors for myocardial infarction;
coronary heart disease; trial terminated prema-
age ≥55 yr turely when rate of heart
failure with doxazosin
­increased 104% vs.
Clinical Trials Series

chlorthalidone (P<0.001)
ALLHAT (2002)23 Diastolic pressure Chlorthalidone vs. lisino- Chlorthalidone: 146/84 As compared with chlorthal- No Intended to control blood pres-
Duration: 1994–2002 ≥90 mm Hg or systolic pril vs. amlodipine Lisinopril: 146/84 idone, no significant dif- sure and address potential

The New England Journal of Medicine

n engl j med 375;18 nejm.org  November 3, 2016
Participants: 33,357 pressure ≥140 mm Hg Amlodipine: 146/84 ference with lisinopril or differences between agents
and one or more risk amlodipine for primary
factors for coronary outcome of death from
heart disease; age coronary heart disease
≥55 yr or nonfatal myocardial
infarction

Copyright © 2016 Massachusetts Medical Society. All rights reserved.

HYVET (2008)24 Systolic pressure ≥160 Antihypertensive treat- Treatment: 173/91 No significant difference in Yes Intended to address problem of
Duration: 2000–2007 mm Hg; age ≥80 yr ment vs. placebo Placebo: 173/91 rate of stroke (P<0.06), limited data or risks and ben-
Participants: 3845 but 21% decrease in efits of treatment availability
mortality with treatment among octogenarians; active
(P<0.02) treatment provided for pa-
tients with systolic blood

Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
pressure ≥220 mm Hg or dia-
stolic pressure ≥110 mm Hg
ACCORD (2010)25 Type 2 diabetes, risk factors Intensive control (target Intensive control: 139/76 No significant difference No Addressed effects of treatment
Duration: 2001–2009 for cardiovascular dis- <120 mm Hg) or Standard control: 139/76 in primary outcomes of intensity rather than treat-
Participants: 4733 ease (can be taking standard control death from cardiovascu- ment initiation
­antihypertensive ­(target <140 mm Hg) lar disease or nonfatal
­medication) myocardial infarction
or stroke (P = 0.20)

1761
 The n e w e ng l a n d j o u r na l of m e dic i n e

18,790 participants, made a clear statement that

Attack Trial, HDFP Hypertension Detection and Follow-up Program, HOPE Heart Outcomes Prevention Evaluation, HOT Hypertension Optimal Treatment, HYVET Hypertension in the
­effects of treatment intensity
rather than treatment initia-

pressure spectrum and risk

tion but in different patient

Addressed lower end of blood-

elevated diastolic pressure should be lowered to

* ACCORD denotes Action to Control Cardiovascular Risk in Diabetics, ACE angiotensin-converting–enzyme, ALLHAT Antihypertensive and Lipid-Lowering Treatment to Prevent Heart
Like ACCORD, trial addressed

Very Elderly Trial, MRC Medical Research Council, SHEP Systolic Hypertension in the Elderly Program, SPRINT Systolic Blood Pressure Intervention Trial, STOP Swedish Trial in Old
at least 90 mm Hg.
Issues Regarding
Use of Placebo

Shif t t o S ys t ol ic Pr e ssur e

population

spectrum
The Systolic Hypertension in the Elderly Pro-
gram (SHEP), which was conducted from 1985
through 1991, involved 4736 patients, was one
of the first trials to address the use of systolic
blood pressure as a guide for initiating anti­
Placebo
Used

Yes
No

hypertensive treatment. In order to participate,

a patient had to have a systolic blood pressure
between 160 and 219 mm Hg and a diastolic
Decrease of 25% for primary

Patients with Hypertension, Syst-Eur Systolic Hypertension in Europe, and VA Group Veterans Administration Cooperative Study Group.
dial infarction, or stroke
death, nonfatal myocar-
No significant difference in
cardiovascular disease,
outcome of death from

heart failure (P<0.001);

pressure of less than 90 mm Hg (average blood

rates of cardiovascular
drome, myocardial in-
farction, or stroke, or

­mortality (P = 0.003)
acute coronary syn-

pressure, 170/77 mm Hg).18 Although these levels

decrease of 27% in

of systolic pressure are very high by today’s stan-

Outcome

dards, the benefits and risks of antihypertensive

therapy in this generally older population were
unknown at the time the trial was conducted.
Although placebo was used as the control, the
trial included an “escape” clause that allowed for
Intensive control (target Intensive control: 140/78
Standard control: 140/78

the initiation of open-label therapy for patients

Blood Pressure

whose systolic blood pressure was higher than

Treatment: 138/82
at Inclusion

mm Hg

220 mm Hg. At year 2, a total of 6.7% of the

Placebo: 138/82

active-treatment group and 70.9% of the placebo

group were not being treated, and the between-
group difference in systolic pressure was approxi-
mately 13 mm Hg. The active-treatment group
had a rate of stroke that was 36% lower than
standard control (tar-

Candesartan plus hydro-

­placebo, with 2-by-2

that in the placebo group and a composite rate

factorial design of
get <140 mm Hg)

chlorothiazide vs.

statin vs. placebo

<120 mm Hg) or
Comparators

of nonfatal myocardial infarction, death from

coronary heart disease, or heart failure that was
33% lower.
The European Working Party on High Blood
Pressure in the Elderly investigated isolated sys-
tolic hypertension in the Systolic Hypertension in
treatment determined on
initiation of hypertensive
risk factor for cardiovas-
disease, but at least one
stroke excluded; can be
taking antihypertensive

basis of local standards

ACCORD or history of

cular disease; need for

risk factors for cardio-

Europe (Syst-Eur) trial.20 The entry criterion for

vascular disease (per-

No known cardiovascular
sons participating in

this randomized, placebo-controlled trial was a

130–180 mm Hg,
Eligibility

systolic blood pressure of 160 to 219 mm Hg,

Systolic pressure

medication)

with a diastolic pressure of less than 95 mm Hg.

The Syst-Eur trial, which was conducted from
1989 through 1997, was continued after the re-
sults of SHEP were known “because of remain-
ing uncertainties about the treatment of isolated
Duration: 2010–2015

Duration: 2007–2015
Participants: 12,705

systolic hypertension in the elderly.”20 The trial

Participants: 9361
Table 1. (Continued.)

was stopped after an interim analysis showed that

HOPE–3 (2016)27
26

the incidence of the primary end point (stroke)

SPRINT (2015)

was 42% lower in the active-treatment group

than in the placebo group (P = 0.003). The inci-
dence of heart failure and myocardial infarction
Trial

was 25% lower in the active-treatment group.

1762 n engl j med 375;18 nejm.org  November 3, 2016

The New England Journal of Medicine

Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.
 Clinical Trials Series

The Swedish Trial in Old Patients with Hyper- sive-therapy group confirmed the importance of
tension (STOP-Hypertension) was also conducted lowering blood pressure regardless of age and
during this period (1985–1991) and also used closed the chapter on the use of placebo in con-
placebo rather than active treatment in the trolled trials addressing this level of hyperten-
comparator group.19 On entry into the trial, par- sion (Table 1).
ticipants had to have a diastolic blood pressure In the Heart Outcomes Prevention Evalua-
between 105 and 120 mm Hg (irrespective of tion (HOPE)–3 trial,27 the results of which were
systolic pressure) or both a diastolic pressure reported in spring 2016, the investigators used
of at least 90 mm Hg and a systolic pressure be- placebo as the comparator with active combi-
tween 180 and 230 mm Hg. The diastolic criteria nation antihypertensive therapy to evaluate the
overlapped with those for participation in previous risks and benefits of blood-pressure control in
trials, as the investigators acknowledged in their persons whom investigators believed did not
stated objective: “to confirm the value of antihy- have a clear current indication for the active
pertensive treatment in people age 70–74 years therapies being evaluated. Participants could
and to expand the database up to the age of 84 enter the trial after blood-pressure control was
years.” The blood-pressure levels at year 2 of the attained with lifestyle interventions or with drugs
study were 188/97 mm Hg in the placebo group other than an angiotensin-receptor–blocker, an
and 166/87 mm Hg in the treatment group. The angiotensin-converting-enzyme inhibitor, or a
rate of cardiovascular death, myocardial infarc- thiazide diuretic. In effect, this design, which
tion, or stroke was 38% lower in the active- permitted some participants in the control group
treatment group (Table 1). to receive antihypertensive therapy, was a hybrid
These trials from the 1980s and early 1990s between no treatment and add-on therapy, since
generated the data needed to develop a systolic at baseline 22% of patients were already receiv-
blood-pressure target for the initiation of anti- ing treatment intended to lower blood pressure.
hypertensive therapy. The cumulative evidence With the mean blood pressure at baseline at
provided the basis for the recommendations 138/82 mm Hg, some participants had a blood-
initially made in the 1993 report of the Joint pressure level that was above the existing treat-
National Committee on Prevention, Detection, ment targets and the targets being introduced at
Evaluation, and Treatment of High Blood Pres- the time the trial started (Fig. 1).
sure (Fig. 1) to start long-term antihypertensive The trial results were neutral, with no signifi-
therapy at a systolic pressure of 140 mm Hg or cant difference reported between placebo and
higher or a diastolic pressure of 90 mm Hg or active therapy for the composite rate of cardio-
higher. vascular events. However, in a subgroup analysis
of the participants with the highest systolic
pressure at trial entry (>143.5 mm Hg), those
Emergence of E thic a l Issue s
a s K now l ed ge Ac cumul ate s who received combination antihypertensive ther-
apy had fewer cardiovascular events than those
The ethical boundaries of finding sufficient un- receiving placebo. What made this modern pla-
certainty for the use of placebo in patients with cebo-controlled trial distinctive was the fact that
hypertension were probed more recently in the it explored the lower limits at which antihyper-
Hypertension in the Very Elderly Trial (HYVET), tensive therapy should be provided by offering
which was conducted from 2000 through 2007. treatment to participants who were not consid-
In this major outcome trial, active antihyperten- ered to need either initiation of such treatment
sive therapy was compared with placebo in pa- or any additional treatment — an understudied
tients 80 years of age or older who had a systolic population with an annual risk of major cardio-
pressure of 160 mm Hg or higher.24 The ratio- vascular events of approximately 1%. The trial
nale was that minimal data on the effectiveness also had a 2-by-2 factorial design in which a
or safety of antihypertensive therapy were avail- statin was provided for the purpose of preventing
able in this population. Although the rate of the cardiovascular events without the use of lipid or
primary end point of stroke was not significantly blood-pressure targets or monitoring.
lower than the rate with placebo, the 21% lower As the data on the benefits of antihyperten-
rate of death (P<0.02) in the active antihyperten- sive therapy have increased, ethical standards in

n engl j med 375;18 nejm.org  November 3, 2016 1763

The New England Journal of Medicine
Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

this field have continued to evolve.28 Generally, in the trial who had diabetes, there was a grada-
institutional review boards will permit only short- tion in the observed rate of cardiovascular events
term withholding (approximately 2 to 4 weeks) across the three groups that had met the target
of antihypertensive medications (within a non- diastolic pressure.21 This finding led to the rec-
severe blood-pressure range) to establish baseline ommendation of an even lower target blood
blood pressure. Most current trials assessing pressure in these patients.28 However, it was ac-
the effects of lowering blood pressure compare knowledged that this target was based on sub-
the effects of a new therapy with those of an group data.
established comparator rather than placebo.23,28-31 There was sufficient uncertainty regarding the
The comparisons between specific antihyperten- recommendation that one aspect of the NHLBI-
sive agents are often made at doses that achieve sponsored Action to Control Cardiovascular Risk
similar degrees of blood-pressure control. Al- in Diabetics (ACCORD) trial was specifically de-
though some large clinical-outcome trials that signed to address this question.25 In ACCORD, an
compare specific antihypertensive agents have open-label trial that commenced in 2001, patients
been conducted to identify possible benefits of were randomly assigned to receive either inten-
treatment other than that of lowering blood pres- sive antihypertensive therapy targeting systolic
sure,23,29,31 for the most part reductions in the blood pressures of less than 120 mm Hg or
occurrence of major cardiovascular events are standard therapy targeting systolic blood pres-
more closely related to the extent to which blood sures of less than 140 mm Hg. The group receiv-
pressure is lowered than to the class of agent ing standard treatment was to have at least the
used.32 One of the largest trials in which active currently recommended level of blood-pressure
agents were used as comparators was the Anti- control, and the focus was on between-group
hypertensive and Lipid-Lowering Treatment to differences in blood-pressure levels, not on the
Prevent Heart Attack Trial (ALLHAT) sponsored specific antihypertensive agents used. The group
by the National Heart, Lung, and Blood Institute receiving intensive treatment was in effect pro-
(NHLBI), in which a diuretic was compared with viding the unknown relative efficacy and safety
three other classes of antihypertensive drug.23 In data related to the use of a greater number of
this trial, which was conducted from 1994 through antihypertensive medications intended to lower
2002, the finding that the risk of the develop- blood pressure to levels not previously studied in
ment of heart failure was twice as high with major clinical-outcome trials.
doxazosin as it was with chlorthalidone provid- Despite achieving an impressive average dif-
ed the research community with an important ference of 14 mm Hg in systolic pressure, the
lesson in humility.22 primary composite outcome of time to cardio-
vascular death, myocardial infarction, or stroke
was not lower in the group receiving intensive
Be yond Pl acebo
therapy than in the group receiving standard
The clinical-outcome trials involving hyperten- therapy. The trial also showed that there was a
sion that were conducted in the past decade ad- downside to the strategy of striving for greater
dressed the level at which antihypertensive treat- and greater reductions in blood pressure, since
ment should be initiated and at what intensity; a higher number of serious adverse events (i.e.,
these trials did not use a placebo. Since the epi- events that were life-threatening, caused perma-
demiologic data relating blood pressure and car- nent disability, or necessitated hospitalization)
diovascular risk do not show a clear threshold that were attributed to antihypertensive medica-
(i.e., a level below which risk does not appear to tions were reported in the group receiving inten-
diminish),9,10 there was a widely held belief that sive treatment. The results of the ACCORD trial
the use of antihypertensive drugs to bring blood did not align with expert consensus,34 which
pressure below the operational definition of provided another lesson in humility from a well-
hypertension would further improve prognosis. conducted outcome trial that improved practice
For patients with type 2 diabetes, this belief was and defied expert opinion.
supported by a subgroup analysis from the HOT The even more recent NHLBI-sponsored Sys-
trial that showed that among the 1501 patients tolic Blood Pressure Intervention Trial (SPRINT),26

1764 n engl j med 375;18 nejm.org  November 3, 2016

The New England Journal of Medicine

Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.
 Clinical Trials Series

the results of which were published in 2015, L e ssons L e a r ned

addressed the same question in a population with
a high level of cardiovascular risk. However, in The privilege of conducting human research
light of the neutral findings related to target requires investigators to minimize the risks to
blood pressures from the ACCORD trial and the participants and to inform participants of these
Secondary Prevention of Small Subcortical Strokes risks by providing them with the most accurate
trial,33 the SPRINT investigators excluded patients current information available. Randomized trials
with diabetes or a history of stroke. Again, the should be conducted within an ethical frame-
design was a randomized, controlled trial in work that allows all participants access to the
which patients used open-label antihypertensive best possible care (as determined regionally) and
agents to a target systolic pressure of either less to an experimental comparator about which the
than 120 or less than 140 mm Hg. Although in preliminary information is sufficient to suggest
the ACCORD trial the group receiving the more a degree of efficacy that justifies the conduct of a
intensive approach to lowering systolic blood clinical trial (given the inherent uncertainty in-
pressure did not have a benefit–risk profile that volved). As new information is generated and the
justified this approach in persons with diabetes, degree of uncertainty diminished, these develop-
the SPRINT investigators felt justified in testing ments should be reflected in the selection of
the same hypothesis in patients with elevated proper comparators. The history of randomized,
blood pressure who were not known to have controlled trials involving hypertension shows
diabetes. On the heels of the ACCORD trial, the that the level of uncertainty or, conversely, the
early termination of the SPRINT trial for reasons interpretation of the robustness of the currently
of efficacy came as a surprise. The data showed available data can vary among physicians within
that more intensive lowering of blood pressure, a country and across geographic regions. A re-
below the currently recommended level, signifi- view of this history underscores the sanctity of
cantly reduced rates of cardiovascular end points the central tenet of the consent process: prop-
and the risk of death.26 Once scrutinized, it seems erly informing the voluntary participants of rel-
likely that this new information will once again evant information at the time of enrollment so
change current authoritative recommendations for that they can make a truly informed decision.
standard practice in an effort to improve public Disclosure forms provided by the authors are available with
health (Fig. 1).34 the full text of this article at NEJM.org.

References
1. Grady C. Enduring and emerging chal- 7. Fleming TR, DeMets DL. Surrogate Study Group on Antihypertensive Agents.
lenges of informed consent. N Engl J Med end points in clinical trials: are we being Effects of treatment on morbidity in hyper-
2015;​372:​855-62. misled? Ann Intern Med 1996;​125:​605-13. tension. II. Results in patients with dia-
2. Califf RM, DeMets DL. Principles from 8. Echt DS, Liebson PR, Mitchell LB, et al. stolic blood pressure averaging 90 through
clinical trials relevant to clinical practice. Mortality and morbidity in patients re- 114 mm Hg. JAMA 1970;​213:​1143-52.
Circulation 2002;​106:​1015-21. ceiving encainide, flecainide, or placebo: 13. Jones DW, Hall JE. The National High
3. Chalmers TC, Celano P, Sacks HS, the Cardiac Arrhythmia Suppression Trial. Blood Pressure Education Program: thirty
Smith H Jr. Bias in treatment assignment N Engl J Med 1991;​324:​781-8. years and counting. Hypertension 2002;​
in controlled clinical trials. N Engl J Med 9. Kannel WB, Dawber TR, Kagan A, 39:​941-2.
1983;​309:​1358-61. Revotskie N, Stokes J III. Factors of risk in 14. Hypertension Detection and Follow-
4. Eisenstein EL, Lemons PW II, Tardiff the development of coronary heart dis- up Program Cooperative Group. Five-year
BE, Schulman KA, Jolly MK, Califf RM. ease — six year follow-up experience: the findings of the hypertension detection and
Reducing the costs of phase III cardiovas- Framingham Study. Ann Intern Med 1961;​ follow-up program. I. Reduction in mor-
cular clinical trials. Am Heart J 2005;​149:​ 55:​33-50. tality of persons with high blood pres-
482-8. 10. Multiple Risk Factor Intervention Trial sure, including mild hypertension. JAMA
5. Ridker PM, Torres J. Reported out- Research Group. Multiple Risk Factor In- 1979;​242:​2562-71.
comes in major cardiovascular clinical tervention Trial: risk factor changes and 15. The Australian Therapeutic Trial in
trials funded by for-profit and not-for- mortality results. JAMA 1982;​248:​1465-77. mild hypertension: report by the Manage-
profit organizations: 2000-2005. JAMA 11. Veterans Administration Cooperative ment Committee. Lancet 1980;​1:​1261-7.
2006;​295:​2270-4. Study Group on Antihypertensive Agents. 16. Amery A, Birkenhäger W, Brixko P, et
6. Gordon D, Taddei-Peters W, Mascette Effects of treatment on morbidity in hyper- al. Mortality and morbidity results from
A, Antman M, Kaufmann PG, Lauer MS. tension: results in patients with diastolic the European Working Party on High
Publication of trials funded by the Nation- blood pressures averaging 115 through Blood Pressure in the Elderly trial. Lancet
al Heart, Lung, and Blood Institute. N Engl 129 mm Hg. JAMA 1967;​202:​1028-34. 1985;​1:​1349-54.
J Med 2013;​369:​1926-34. 12. Veterans Administration Cooperative 17. Medical Research Council Working

n engl j med 375;18 nejm.org  November 3, 2016 1765

The New England Journal of Medicine
Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.
 Clinical Trials Series

Party. MRC trial of treatment of mild hyper- 23. The ALLHAT Officers and Coordina- tality in the Losartan Intervention For
tension: principal results. Br Med J (Clin tors for the ALLHAT Collaborative Re- Endpoint reduction in hypertension study
Res Ed) 1985;​291:​97-104. search Group. Major outcomes in high- (LIFE): a randomised trial against ateno-
18. SHEP Cooperative Research Group. risk hypertensive patients randomized to lol. Lancet 2002;​359:​995-1003.
Prevention of stroke by antihypertensive angiotensin-converting enzyme inhibitor 30. Julius S, Kjeldsen SE, Weber M, et al.
drug treatment in older persons with iso- or calcium channel blocker vs diuretic: Outcomes in hypertensive patients at high
lated systolic hypertension: final results the Antihypertensive and Lipid-Lowering cardiovascular risk treated with regimens
of the Systolic Hypertension in the Elderly Treatment to Prevent Heart Attack Trial based on valsartan or amlodipine: the
Program (SHEP). JAMA 1991;​265:​3255-64. (ALLHAT). JAMA 2002;​288:​2981-97. VALUE randomised trial. Lancet 2004;​
19. Dahlöf B, Lindholm LH, Hansson L, 24. Beckett NS, Peters R, Fletcher AE, et al. 363:​2022-31.
Scherstén B, Ekbom T, Wester PO. Mor- Treatment of hypertension in patients 80 31. Jamerson K, Weber MA, Bakris GL, et
bidity and mortality in the Swedish Trial years of age or older. N Engl J Med 2008;​ al. Benazepril plus amlodipine or hydro-
in Old Patients with Hypertension (STOP- 358:​1887-98. chlorothiazide for hypertension in high-
Hypertension). Lancet 1991;​338:​1281-5. 25. The ACCORD Study Group. Effects of risk patients. N Engl J Med 2008;​359:​2417-
20. Staessen JA, Fagard R, Thijs L, et al. intensive blood-pressure control in type 2 28.
Randomised double-blind comparison of diabetes mellitus. N Engl J Med 2010;​362:​ 32. Blood Pressure Lowering Treatment
placebo and active treatment for older pa- 1575-85. Trialists’ Collaboration. Effects of differ-
tients with isolated systolic hypertension. 26. The SPRINT Research Group. A ran- ent blood-pressure-lowering regimens on
Lancet 1997;​350:​757-64. domized trial of intensive versus standard major cardiovascular events: results of
21. Hansson L, Zanchetti A, Carruthers blood-pressure control. N Engl J Med 2015;​ prospectively-designed overviews of ran-
SG, et al. Effects of intensive blood-pres- 373:​2103-16. domised trials. Lancet 2003;​362:​1527-35.
sure lowering and low-dose aspirin in pa- 27. Lonn EM, Bosch J, López-Jaramillo P, 33. The SPS3 Study Group. Blood-pressure
tients with hypertension: principal results et al. Blood-pressure lowering in interme- targets in patients with recent lacunar
of the Hypertension Optimal Treatment diate-risk persons without cardiovascular stroke: the SPS3 randomised trial. Lancet
(HOT) randomised trial. Lancet 1998;​351:​ disease. N Engl J Med 2016;​374:​2009-20. 2013;​382:​507-15.
1755-62. 28. Chobanian AV, Bakris GL, Black HR, 34. James PA, Oparil S, Carter BL, et al.
22. The ALLHAT Collaborative Research et al. The Seventh Report of the Joint Na- 2014 Evidence-based guideline for the
Group. Major cardiovascular events in tional Committee on Prevention, Detec- management of high blood pressure in
hypertensive patients randomized to doxa- tion, Evaluation, and Treatment of High adults: report from the panel members
zosin vs chlorthalidone: the antihyperten- Blood Pressure: the JNC 7 report. JAMA appointed to the Eighth Joint National
sive and lipid-lowering treatment to pre- 2003;​289:​2560-72. Committee (JNC 8). JAMA 2014;​311:​507-
vent heart attack trial (ALLHAT). JAMA 29. Dahlöf B, Devereux RB, Kjeldsen SE, 20.
2000;​283:​1967-75. et al. Cardiovascular morbidity and mor- Copyright © 2016 Massachusetts Medical Society.

specialties and topics at nejm.org

Specialty pages at the Journal’s website (NEJM.org) feature articles in cardiology,
endocrinology, genetics, infectious disease, nephrology, pediatrics, and many other
medical specialties. These pages, along with collections of articles on clinical and
nonclinical topics, offer links to interactive and multimedia content and feature
recently published articles as well as material from the NEJM archive (1812–1989).

1766 n engl j med 375;18 nejm.org  November 3, 2016

The New England Journal of Medicine

Downloaded from nejm.org on September 9, 2018. For personal use only. No other uses without permission.
Copyright © 2016 Massachusetts Medical Society. All rights reserved.