Download Are Observational Studies More Informative Than Randomized

Controversies in
Hypertension
Are Observational Studies More Informative
Than Randomized Controlled Trials
in Hypertension?
Downloaded from http://hyper.ahajournals.org/ by guest on May 8, 2017
Con Side of the Argument
Giuseppe Schillaci, Francesca Battista, Giacomo Pucci
T
he very first rigorously designed, randomized controlled
trial (RCT) in the history of modern medicine was published in 1948. The study, designed and carried out by the
Streptomycin in Tuberculosis Trials Committee of the Medical
Research Council, demonstrated the efficacy of streptomycin
in treating pulmonary tuberculosis on top of bed rest, which
was the best available treatment at that time.1 In the field of
hypertension treatment and cardiovascular prevention, the
history of event-based RCTs began in 1967, when the cardiovascular benefits and risks of antihypertensive drug treatment
were evaluated in the setting of the Veterans Administration
Cooperative Study in Hypertension.2 In that trial, 143 patients
with severe hypertension were randomized to receive blood
pressure (BP)–lowering drug treatment or matched placebo for
11 months. A major morbid or fatal event was observed in 26
of 70 patients who received placebo (37%) and in only 1 of 73
patients under active treatment (1%). The amazing results of
this landmark trial established beyond any doubt the striking
cardiovascular benefits of antihypertensive drug treatment of
hypertension at a time when controversy still persisted regarding the supposedly deleterious effects of BP reduction on
organ perfusion.3
Ever since, RCTs have represented a formidable tool for
evaluating risks and benefits of hypertension treatment. Over
the past 50 years, the remarkable progress in treatment and
control of high BP, one of the most outstanding achievements of modern medicine, has been driven by the results
of many large, event-based RCTs. The influential position
of RCTs in clinical and therapeutic research as opposed to
real-world observational studies (surveys, registries, and
retrospective analyses of existing databases) originates from
their high internal validity (ie, the power to address clinical
questions with a low level of internal bias). A list of strengths
and weaknesses of RCTs and observational studies is given
in Table 1. The present review examines values and limitations of RCTs and observational studies, with a focus on the
problems arising when evidence from observational studies
is used to inform decisions in the clinical practice, in the
absence of less biased sources of evidence.
Internal Versus External Validity
In studies of disease risk, dependable results derive from the
minimization of both random errors (imprecision) and systematic errors (biases). Because random errors are inversely
proportional to the number of individuals developing the disease during follow-up, the first key determinant of the internal
validity of a clinical study is represented by the numbers of
observed events. This is of particular concern when effects
of plausibly moderate size are to be detected. This requirement has led to the conception of RCTs with many hundreds
or even thousands of incident cardiovascular events (megatrials). Other studies that also tend to minimize random errors
are meta-analyses, which combine the results of multiple
individual studies using appropriate statistical methods, and
observational studies drawn from large databases based on
routinely collected data.
The second key determinant of the internal validity of a
clinical study is to what extent the results of the research may
be affected by systematic errors or biases. Minimization of
systematic error is necessary to establish cause-and-effect
The opinions expressed in this editorial are not necessarily those of the editors or of the American Heart Association.
From the Dipartimento di Medicina, Terni University Hospital, Università di Perugia, Terni, Italy.
Correspondence to Giuseppe Schillaci, Dipartimento di Medicina, Università di Perugia, Struttura Complessa di Medicina Interna, Azienda OspedalieroUniversitaria di Terni, Viale Tristano di Joannuccio, 1, IT 05100 Terni, Italy. E-mail [email protected]
(Hypertension. 2013;62:470-476.)
© 2013 American Heart Association, Inc.
Hypertension is available at http://hyper.ahajournals.org
DOI: 10.1161/HYPERTENSIONAHA.113.01501
470
Schillaci et al Observational Studies More Informative Than RCTs? 471
Table 1. Strengths and Weaknesses of Randomized
Controlled Trials and Observational Studies
Table 2. Suitability of Various Study Designs to Address
Different Clinical Research Questions
Randomized
Controlled Trials
Observational
Studies
Yes*
Generally not†‡
No/little*
Yes†
Yes*
Less well
established†
Is previous registration required?
Usually yes*
Generally not†
Is selective reporting a concern?
No*
Yes†
Feature of the Study
Are groups well matched?
Is interpretation exposed to the effects
of confounding factors?
Are methodological rules well
established?
Cost
Higher†
Lower*
Time needed for carrying out the study
Longer†
Shorter*
Number of participants
Lower†
Greater*
Does protocol reflect clinical care?
Duration of observation
Usually not†
Usually yes*
Shorter†
May be longer*
Downloaded from http://hyper.ahajournals.org/ by guest on May 8, 2017
*Strength.
†Weakness.
‡Adjustment for propensity to receive a particular treatment (propensity
score matching) may increase the comparability of study groups.
relationships between a given intervention (or independent
variable) and its outcome (dependent variable). The pure
experiment in the form of a randomized controlled longitudinal study, also referred to as an RCT, is often regarded as the
gold standard, and is believed to yield the lowest chance of
bias. Nonrandomized studies, also referred to as quasiexperimental, observational, or correlation studies, are considered
as having a lower internal validity. A graphic representation of
the pyramidal hierarchy of the different types of studies based
on their internal validity is given in the Figure.
The internal validity of a given study has little relation with
its external validity, or generalizability. For instance, an RCT
has a high internal validity, but may be less suited to generalization, which restricts its practical usability. Nonrandomized
longitudinal studies, however, have a lower internal validity,
but can nevertheless be very useful for management practice.
In clinical practice, an answer to the question “How much
does this treatment work?” is sometimes less relevant than
the response to questions like “In what circumstances does
it work or for whom does it work?”. A study design is never
strong or weak in itself: it all depends on the question(s) to be
addressed (Table 2).
Figure. Levels of evidence for various research designs, based on
their internal validity. RCTs indicates randomized controlled trials.
Domain
Effectiveness
Randomized Observational CrossControlled Longitudinal Sectional
Research Question
Trials
Studies
Surveys
Does it work? Does it
work better than other
treatments?
++
+
−
Safety
Will it do more good
than harm?
++
+
−
Context
In what circumstances
does it work, for
whom?
−−
−
+
Will the target
group accept the
intervention?
−−
−
+
Prevalence
How often is this
intervention applied?
−−
−
++
Appropriateness
Is this the right
intervention for this
target group?
−−
+
++
How does it work, why
does it work?
−−
−
+
Acceptability
Explanation
++ indicates very good; +, good; −, fair; and −−, poor.
Cause-and-Effect Relationship: Misinformation
Resulting From Observational Studies
As shown in Table 2, observational studies can be very informative when addressing issues like the acceptability and appropriateness of a given intervention for a certain target group.
However, because of the presence of a high degree of bias,
their findings should be taken with greater caution or can even
be misleading when they are used to establish cause-and-effect
relationships. Although estimates of treatment effects are often
similar in observational studies and in RCTs,4,5 this may not
always be the case. We discuss a number of clinically relevant
conclusions drawn from real-world observational studies in the
field of hypertension treatment and cardiovascular prevention,
which have subsequently shown to be false (Table 3).
Does Antihypertensive Treatment Prevent Ischemic
Heart Disease?
Observational studies have reported increased risks of coronary
heart disease in patients treated with antihypertensive drugs. In
a population-based cohort study of elderly Swedish men, myocardial infarction or death because of chronic ischemic cardiac
disease were more frequent over a 10-year follow-up in those
subjects who were taking antihypertensive drugs than in those
who were not, and this rate was reduced but still raised after
adjustment for potential confounders, such as smoking habits,
BP, dyslipidemia, diabetes mellitus, obesity, and serum creatinine concentration.6 Similarly, in a large Norwegian nationwide
screening of 21 314 men aged 35 to 49 years, BP-lowering treatment was associated with increased mortality from all causes,
coronary heart disease, and noncardiovascular causes. Again,
these associations remained significant after controlling for pretreatment BP, cholesterol, age, smoking, and body mass index.7
These data have been overturned by a large number of
RCTs, which have uniformly and unequivocally demonstrated
472 Hypertension September 2013
Table 3. A List of Conclusions Derived From Observational
Studies, Which Have Been Eventually Confuted by Properly
Designed Randomized Controlled Trials
Conclusion
Antihypertensive drug treatment
increases the risk of ischemic heart
disease
Randomized
Observational Controlled Trials and
Studies
Their Meta-Analyses
Downloaded from http://hyper.ahajournals.org/ by guest on May 8, 2017
6, 7
8, 9
Aspirin use increases the risk of
coronary heart disease
10, 11
12
Calcium-channel blockers increase
the risk of myocardial infarction in
hypertension
15, 16
18–20
Postmenopausal estrogen replacement
reduces cardiovascular morbidity and
mortality
23–25
26, 27
Achieved blood pressure values well
lower than the usual goal of below
140/90 mm Hg carry substantial
benefits (the lower, the better)
28, 29
34–37
that the same treatments produce a marked reduction in stroke
and coronary events, as well as in cardiovascular and all-cause
mortality, both in systolic–diastolic and in isolated systolic
hypertension.8,9
Does Aspirin Increase the Risk of Ischemic
Heart Disease?
In some observational studies, regular aspirin use has been
associated with an increased incidence of coronary heart
disease. In a cohort of 13 987 elderly US subjects without a
reported history of angina, myocardial infarction, or stroke
who were followed up for 6.5 years, the risk of incident ischemic heart disease was increased by 70% to 90% in those who
took aspirin daily compared with nonusers.10 Similarly, in
the Swedish National Diabetes Register, a population-based
cohort study of 18 646 patients with type 2 diabetes mellitus
free from cardiovascular disease followed up for 3.9 years,
there was a >19% risk of incident coronary heart disease in
subjects treated with aspirin. This difference remained significant after adjustment for several baseline risk factors and
characteristics by stratification with a propensity score.11 The
authors’ conclusions10,11 that the daily use of aspirin increases
the risk of ischemic heart disease were refuted later. Indeed,
aspirin has clearly been shown in carefully conducted RCTs
to reduce the risks of myocardial infarction and coronary heart
disease in people without prevalent cardiovascular disease,
without any difference in proportional benefits in diabetic
versus nondiabetic individuals.12 Also, given the absence of
large trials specifically conducted in diabetic subjects, 2 current large placebo-controlled trials are specifically testing the
effects of aspirin in adults with diabetes mellitus but no known
vascular disease.13,14
Do Calcium-Channel Blockers Increase the Risk of
Myocardial Infarction?
In 1995, the publication of 2 studies reporting adverse coronary
events with calcium-channel blockers15,16 generated enduring
concern in the scientific and lay communities.17 In a population-based case-control study of hypertensive patients carried
out among enrollees of the Group Health Cooperative of Puget
Sound, Psaty et al15 reported a 60% increase in the risk of myocardial infarction for calcium-channel blockers compared with
other BP-lowering drug classes. Soon after, Pahor et al16 published the data from the hypertensive subset of a large European
cohort, which documented higher mortality and coronary events
for calcium-channel blockers than for β-blockers.
Over the ensuing years, the circumstantial evidence for the
potentially dangerous side effects of calcium-channel blockers15,16 was contradicted when put to the test of double-blind
RCTs. A number of large, active-controlled RCTs failed to
show any association between the use of calcium-channel
blockers and increased coronary or cardiovascular morbidity.18–20 Subsequent meta-analyses confirmed, with a greater
statistical power, that the benefits of a treatment based on calcium-channel blockers are similar to or only slightly different
from that based on other drug classes, with the possible exception of a somewhat better protection against stroke afforded by
calcium-channel blockers.8,9
How can we explain these discrepant findings between
observational studies and RCTs? Observational studies evaluating drug effects are affected by the problem of confounding by indication.21 In the observational design, treatment
allocation is not under the control of the investigator, but at
the discretion of the prescribing physician. Because the physician’s choice of the drug of interest depends on a number of
patient characteristics, exposed and unexposed subjects will
be fundamentally different. Such differences may affect the
study outcome, and the design of observational studies does
not allow controlling completely for this effect. In the case of
calcium-channel blockers, these drugs might have been more
frequently prescribed to patients with higher comorbidities,
also in the light of the fact that they are approved by the Food
and Drug Administration for the treatment of stable angina
pectoris. The same applies to the above-reported effects of
antihypertensive drug treatment6,7 and aspirin10,11 when evaluated within a nonrandomized database. Techniques to adjust
for these dissimilarities are generally incomplete: the totality
of clinical characteristics driving treatment selection is almost
never fully measured and may, in fact, never be fully measurable.21,22 As a consequence, it is difficult to attribute a given
effect to the drug rather than to the clinical characteristics that
prompted its use.
However, negative results of RCTs may also be influenced
by their design. First, a substantial proportion of patients in
these trials withdrew from their assigned treatment (drop-out
effect). Because the main results of the RCTs are reported
according to the intention-to-treat analysis, this may tend to
dampen any true differences. Second, combination treatment
is more the rule than the exception in hypertension management and in event-based RCTs. In most trials involving
calcium-­channel blockers, although the randomized comparison was between single-drug classes, most patients were
receiving >1 drug during the study (add-on effect). This
means that the comparison is not among given drugs but
among regimens based on given drugs (or drug classes). This
would similarly blunt any differences among drug classes. In
Schillaci et al Observational Studies More Informative Than RCTs? 473
summary, long-term, event-based active-controlled RCTs may
generally tend to overestimate the similarity among different
drug classes, although this does not diminish the reliability of
the observed differences.
Does Postmenopausal Estrogen Replacement
Reduce Cardiovascular Morbidity and Mortality?
Downloaded from http://hyper.ahajournals.org/ by guest on May 8, 2017
Another classic example of the superiority of RCTs over observational studies in establishing cause-and-effect links is represented by the controversies about the cardiovascular benefits
of postmenopausal estrogen replacement. Case-control studies and longitudinal investigations of large population-based
cohorts with several hundreds or even thousands of major incident cardiovascular events uniformly indicated that postmenopausal women who have elected to use hormone-replacement
therapy develop coronary morbidity and mortality at only
about half the rate of other postmenopausal women.23–25
On the contrary, treatment with estrogen plus progestin did
not reduce the overall rate of coronary heart disease events
in postmenopausal women with established coronary disease
in the randomized, blinded, placebo-controlled Heart and
Estrogen/progestin Replacement Study.26 These data were
confirmed and extended by the Women’s Health Initiative,
a randomized, placebo-controlled trial of estrogen plus progestin in >16 000 generally healthy postmenopausal women,
in which estrogen plus progestin actually increased the risk
of coronary heart disease.27 Of course, incidence of myocardial infarction in long-term users of combined hormone therapy may not be well gauged by randomized trials in which
hormone use does not extend beyond several years—the
Women’s Health Initiative was stopped after 5.2 years for risk
excess in the estrogen/progestin group, instead of a planned
duration of 8.5 years. These trials also leave several questions
open on the multiple pharmacological actions of combined
hormone therapy, which bear on the risk of myocardial infarction. Nevertheless, only these RCTs were able to provide the
straightforward, undisputed conclusion that, “estrogen plus
progestin therapy should not be initiated or continued for the
prevention of cardiovascular disease.”27
BP Treatment Goals: The Lower, The Better?
Observational studies clearly show that BP has a direct linear relationship with cardiovascular events as low as 115 to
110 mm Hg systolic and 75 to 70 mm Hg diastolic, without
indication of a J-curve phenomenon within this BP range.28,29
It should be kept in mind that this evidence has profoundly
influenced clinical practice and expert recommendations. It
has indeed been one of the main arguments in favor of the
repeated suggestion, in several authoritative guidelines,
to pursue BP values lower than the usual target values of
<140 mm Hg systolic and <90 mm Hg diastolic, if tolerated. Examples of such guidelines include the ones prepared
by the World Health Organization/International Society of
Hypertension Liaison Committee in 1999,30 and those issued
by the European Society of Hypertension and the European
Society of Cardiology in 200331 and 2007.32 Extrapolation of
the above data to people at higher cardiovascular risk has also
formed the basis for the American Heart Association recommendation to aim at achieving a BP of <130/80 mm Hg in
hypertensive subjects with diabetes mellitus, coronary heart
disease, or without coronary disease but at risk of developing
it (such as those with abnormal carotid ultrasound, peripheral
arterial disease, abdominal aortic aneurysm, and an estimated
10-year Framingham risk score of ≥10%).33
Unfortunately, the expectations that BP reduction to values lower than the goal of <140/90 mm Hg carries substantial additional benefits in hypertensive patients could not be
fully met in specifically designed RCTs. In the Hypertension
Optimal Treatment study, the first randomized study aimed
at comparing the cardiovascular outcomes of different BP
targets, lower treatment goals (<80 mm Hg or <85 mm Hg
diastolic) were not associated with lower rates of major cardiovascular events than the standard goal of <90 mm Hg in the
whole population, as well as in individuals with overt coronary heart disease.34 Subsequent post hoc subgroup analyses
of the Hypertension Optimal Treatment study suggested that
more intensive diastolic BP lowering was actually associated
with increased risk of cardiovascular events among smokers,
whereas significant benefits could not be demonstrated in any
of the other subgroups, with the possible exception of diabetic
patients.35 In the Action to Control CardiOvascular Risk in
Diabetes, 4733 patients with type 2 diabetes mellitus, mostly
treated hypertension, high cardiovascular risk, and without
overt nephropathy were randomized to achieve and maintain
a systolic BP <120 mm Hg, or a systolic BP <140 mm Hg.36
Over a 4.7-year follow-up period, the primary composite end
point of myocardial infarction, stroke, or cardiovascular death
was not different in the arm randomized at achieving intensive
BP control (systolic BP <120 mm Hg; average in-treatment
BP 119/64 mm Hg) versus standard BP control (systolic BP
<140 mm Hg; average in-treatment BP 133/70 mm Hg). Stroke
was the only prespecified end point for which a significant
reduction was observed in the intensive BP control group.36
Subsequent meta-analyses suggested that aiming at lower BP
goals may lead to barely significant reductions in major cardiovascular events and myocardial infarction and a significant
reduction in stroke.37 However, these meta-analyses combine
studies that compared groups in which a variety of different
BP targets were pursued, thus no definite recommendations
can be directly obtained as to the optimal target BP values.37
Thus, at variance with observational studies, evidence from
RCTs supporting the incremental benefits and safety of a
lower BP goal is scanty in most categories of hypertensive
individuals. In the light of the abovementioned limitations of
observational studies and in the absence of a clear-cut demonstration of the benefits and harms of such treatment strategy,
a BP target of <140/90 mm Hg remains an appropriate goal
for the majority of hypertensive subjects, including most of
the individuals at high risk of developing coronary artery disease, as more recently acknowledged in a reappraisal of the
European guidelines on hypertension management.38
Putting Randomized Versus Nonrandomized
Research Into Context
The applicability of an RCT to everyday clinical practice is
necessarily limited because of its very nature and goal. An
RCT is designed with the precise aim of eliminating, where
possible, those effects that are on the contrary very important in
474 Hypertension September 2013
Downloaded from http://hyper.ahajournals.org/ by guest on May 8, 2017
clinical practice, such as the placebo effect,39,40 the physician–
patient relationship,41 and patient preference.42,43 Although placebo control, blinded treatment allocations, and exclusion of
patients with strong treatment preferences increase the internal validity of a trial, they also separate the world of RCTs
and their results from the real world and its expectations. To
be enrolled in an RCT, a patient needs to be selected to meet
the trial protocol’s requirements. Going back to the landmark
Veterans Administration Cooperative Study in Hypertension,2
exclusion criteria comprised “patients who wished to return
to care of their private physicians, those who for geographical or other reasons would be unable to attend clinic regularly,
and patients of dubious reliability such as alcoholics, vagrants
and poorly motivated patients.”2 In the Systolic Hypertension
in the Elderly Program, which demonstrated the benefits of
BP-lowering treatment in elderly subjects with isolated systolic hypertension, >400 000 subjects were screened, but as few
as 4736 were eventually randomized.44 Although this selection process increases internal validity by creating comparable groups, it also limits the generalizability of the otherwise
compelling conclusions of RCTs, which cannot be necessarily
regarded as general principles valid for all subjects.
RCTs are not only a source of data obtained in a rigorously
controlled experimental setting. They can also be used to obtain
a number of additional findings, besides those for which they
have been designed primarily. In this context, subgroup analyses are widely used. Although such analyses may generate new
hypotheses, they can also falsely indicate that treatment is beneficial in a particular subgroup when the trial shows no overall
benefit, or that there is no treatment effect in a particular subgroup when the trial shows benefit overall. Subgroup analyses
must be predefined, carefully justified, limited to a few clinically important questions, and performed and reported following formal rules.45 Another post hoc extrapolation of a trial’s
findings is to compare the occurrence of the study outcomes
among groups based on in-treatment characteristics irrespective of treatment allocations, an approach that is typically
fraught with bias. For example, in the guidelines issued by the
European Society of Hypertension and the European Society
of Cardiology in 2003,31 the recommendation for a lower BP
goal was supported by a post hoc analysis of the Hypertension
Optimal Trial, indicating the lowest event incidence to be at
in-treatment BP values ≈138/83 mm Hg.4 One should be aware
that, irrespective of its statistical significance, any post hoc
analysis of a given trial has to be considered with skepticism.
Of course, it is impractical to conduct separate large RCTs
to study the effect of every treatment or treatment combination
or every dose of a specific treatment, neither can a given intervention be evaluated with high statistical power in patients
in every possible subgroup considered individually. In this
regard, nonrandomized treatment comparisons based on routinely collected data have generally a greater external validity
than RCTs because they include all patients, are performed in
the real world, and include the effects of the doctor–patient
relationship and patient preference. Although adjustment for
between-group differences can be performed with appropriate techniques,46 it is impossible to be confident that the
bias inherent in observational research will be corrected.
Rothwell45 has acutely pointed out that “routinely collected
data are useful where RCTs are impractical, such as in evaluating rare adverse events, but are an adjunct rather than an
alternative.”
Conclusions
We do not believe that dependable clinical evidence only
comes from RCTs. The practice of medicine cannot and
should not be reduced to the hierarchy of evidence-based
medicine (Figure). If RCTs were required for proof of efficacy
of a given treatment, the practice of clinical medicine would
indeed be reduced to a relatively few verified treatments. The
process of clinical thinking involves interpretation and synthesis of relevant evidence from all sources and extrapolation to
the clinical situation. This complex method is at risk of oversimplification in the approach advocated in evidence-based
medicine.47 Having said that, there is no question that RCTs
and their systematic reviews provide the most reliable source
of evidence on the effects of treatment. To our knowledge,
in no case the main conclusions of a large RCT in hypertension have been frankly misleading, or eventually confuted by
observational studies. RCTs (and their meta-analyses) provide
direct, trustworthy answers to simple, straightforward clinical questions, and establish cause-and-effect relationships.
In this field, they provide unique information, which cannot be obtained in a different way with the same degree of
confidence.
However, RCTs are not able to address more complex questions (eg, the role of treatment in a real-world setting in which
various combinations of medications are used in a wide variety
of clinical, social, and economic conditions). Observational
studies and real-world data (surveys, registries, and retrospective analyses of existing databases) can, and should, be used to
explore these more complex domains, although their findings
should be taken with due caution. There is a need for a more
integrated approach to clinical research in the field of hypertension, which includes both RCTs and observational studies.
Despite their limitations, RCTs remain a cornerstone of
sound, dependable clinical research. The evidence pyramid
has been built on a good foundation, and is still standing.
Disclosures
None.
References
1. Medical Research Council. Streptomycin treatment of pulmonary tuberculosis. Br Med J. 1948;2:769.
2.VA Cooperative Study Group. Effects of treatment on morbidity in hypertension: results in patients with diastolic blood pressures averaging 115
through 129 mm Hg. JAMA. 1967;202:1028–1034.
3. Friedberg CK. Diseases of the heart. Philadelphia, PA: WB Saunders Co;
1949.
4. Benson K, Hartz AJ. A comparison of observational studies and randomized, controlled trials. N Engl J Med. 2000;342:1878–1886.
5. Concato J, Shah N, Horwitz RI. Randomized, controlled trials, observational studies, and the hierarchy of research designs. N Engl J Med.
2000;342:1887–1892.
6. Merlo J, Ranstam J, Liedholm H, Hedblad B, Lindberg G, Lindblad U,
Isacsson SO, Melander A, Råstam L. Incidence of myocardial infarction in elderly men being treated with antihypertensive drugs: population
based cohort study. BMJ. 1996;313:457–461.
7.Thürmer HL, Lund-Larsen PG, Tverdal A. Is blood pressure treatment
as effective in a population setting as in controlled trials? Results from a
prospective study. J Hypertens. 1994;12:481–490.
Schillaci et al Observational Studies More Informative Than RCTs? 475
Downloaded from http://hyper.ahajournals.org/ by guest on May 8, 2017
8. Blood Pressure Lowering Treatment Trialists’ Collaboration. Effects of
different blood-pressure-lowering regimens on major cardiovascular
events: results of prospectively-designed overviews of randomised trials.
Lancet. 2003;362:1527–1535.
9. Law MR, Morris JK, Wald NJ. Use of blood pressure lowering drugs
in the prevention of cardiovascular disease: meta-analysis of 147 randomised trials in the context of expectations from prospective epidemiological studies. BMJ. 2009;338:b1665.
10.Paganini-Hill A, Chao A, Ross RK, Henderson BE. Aspirin use and
chronic diseases: a cohort study of the elderly. BMJ. 1989;299:1247–1250.
11.Ekström N, Cederholm J, Zethelius B, Eliasson B, Fhärm E, Rolandsson
O, Miftaraj M, Svensson AM, Gudbjörnsdottir S. Aspirin treatment and
risk of first incident cardiovascular diseases in patients with type 2 diabetes: an observational study from the Swedish National Diabetes Register.
BMJ Open. 2013;3:e002688. doi:10.1136/bmjopen-2013-002688.
12. Antithrombotic Trialists’ Collaboration. Aspirin in the primary and secondary prevention of vascular disease: collaborative meta-analysis of individual participant data from randomised trials. Lancet. 2009;373:1849–1860.
13. De Berardis G, Sacco M, Evangelista V, Filippi A, Giorda CB, Tognoni
G, Valentini U, Nicolucci A; ACCEPT-D Study Group. Aspirin and
Simvastatin Combination for Cardiovascular Events Prevention Trial in
Diabetes (ACCEPT-D): design of a randomized study of the efficacy of
low-dose aspirin in the prevention of cardiovascular events in subjects
with diabetes mellitus treated with statins. Trials. 2007;8:21.
14. ASCEND: A Study of Cardiovascular Events iN Diabetes. http://clinicaltrials.gov/show/NCT00135226. Accessed May 16, 2013.
15.Psaty BM, Heckbert SR, Koepsell TD, Siscovick DS, Raghunathan TE,
Weiss NS, Rosendaal FR, Lemaitre RN, Smith NL, Wahl PW. The risk
of myocardial infarction associated with antihypertensive drug therapies.
JAMA. 1995;274:620–625.
16.Pahor M, Guralnik JM, Corti MC, Foley DJ, Carbonin P, Havlik RJ. Longterm survival and use of antihypertensive medications in older persons. J
Am Geriatr Soc. 1995;43:1191–1197.
17. Ad Hoc Subcommittee of the Liaison Committee of the World Health
Organisation and the International Society of Hypertension. Effects of
calcium antagonists on the risks of coronary heart disease, cancer and
bleeding. J Hypertens. 1997;15:105–115.
18. Hansson L, Lindholm LH, Ekbom T, Dahlöf B, Lanke J, Scherstén B,
Wester PO, Hedner T, de Faire U. Randomised trial of old and new antihypertensive drugs in elderly patients: cardiovascular mortality and morbidity the Swedish Trial in Old Patients with Hypertension-2 study. Lancet.
1999;354:1751–1756.
19.Hansson L, Hedner T, Lund-Johansen P, Kjeldsen SE, Lindholm LH,
Syvertsen JO, Lanke J, de Faire U, Dahlöf B, Karlberg BE. Randomised
trial of effects of calcium antagonists compared with diuretics and betablockers on cardiovascular morbidity and mortality in hypertension: the
Nordic Diltiazem (NORDIL) study. Lancet. 2000;356:359–365.
20. Brown MJ, Palmer CR, Castaigne A, de Leeuw PW, Mancia G, Rosenthal
T, Ruilope LM. Morbidity and mortality in patients randomised to double-blind treatment with a long-acting calcium-channel blocker or diuretic
in the International Nifedipine GITS study: Intervention as a Goal in
Hypertension Treatment (INSIGHT). Lancet. 2000;356:366–372.
21. Johnston SC. Identifying confounding by indication through blinded prospective review. Am J Epidemiol. 2001;154:276–284.
22. Walker AM, Stampfer MJ. Observational studies of drug safety. Lancet.
1996;348:489.
23.Stampfer MJ, Colditz GA. Estrogen replacement therapy and coronary
heart disease: a quantitative assessment of the epidemiologic evidence.
Prev Med. 1991;20:47–63.
24.Ettinger B, Friedman GD, Bush T, Quesenberry CP Jr. Reduced mortality associated with long-term postmenopausal estrogen therapy. Obstet
Gynecol. 1996;87:6–12.
25. Grodstein F, Manson JE, Colditz GA, Willett WC, Speizer FE, Stampfer
MJ. A prospective, observational study of postmenopausal hormone therapy and primary prevention of cardiovascular disease. Ann Intern Med.
2000;133:933–941.
26. Hulley S, Grady D, Bush T, Furberg C, Herrington D, Riggs B, Vittinghoff
E. Randomized trial of estrogen plus progestin for secondary prevention of coronary heart disease in postmenopausal women. Heart and
Estrogen/progestin Replacement Study (HERS) Research Group. JAMA.
1998;280:605–613.
27. Manson JE, Hsia J, Johnson KC, Rossouw JE, Assaf AR, Lasser NL,
Trevisan M, Black HR, Heckbert SR, Detrano R, Strickland OL, Wong
ND, Crouse JR, Stein E, Cushman M; Women’s Health Initiative
Investigators. Estrogen plus progestin and the risk of coronary heart disease. N Engl J Med. 2003;349:523–534.
28. MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J, Abbott
R, Godwin J, Dyer A, Stamler J. Blood pressure, stroke, and coronary
heart disease. Part 1, prolonged differences in blood pressure: prospective
observational studies corrected for the regression dilution bias. Lancet.
1990;335:765–774.
29.Prospective Studies Collaboration. Age-specific relevance of usual blood
pressure to vascular mortality: a meta-analysis of individual data for one
million adults in 61 prospective studies. Lancet. 2002;360:1903–1913.
30.
Guidelines Sub-Committee. 1999 World Health Organization–
International Society of Hypertension guidelines for the management of
hypertension. J Hypertens. 1999;17:151–183.
31.Guidelines Committee. 2003 European Society of Hypertension–
European Society of Cardiology guidelines for the management of arterial
hypertension. J Hypertens. 2003;21:1011–1053.
32. Mancia G, De Backer G, Dominiczak A, et al.; Management of Arterial
Hypertension of the European Society of Hypertension; European
Society of Cardiology. 2007 Guidelines for the Management of
Arterial Hypertension: The Task Force for the Management of Arterial
Hypertension of the European Society of Hypertension (ESH) and of
the European Society of Cardiology (ESC). J Hypertens. 2007;25:
1105–1187.
33.Rosendorff C, Black HR, Cannon CP, Gersh BJ, Gore J, Izzo JL Jr,
Kaplan NM, O’Connor CM, O’Gara PT, Oparil S. Treatment of hypertension in the prevention and management of ischemic heart disease. A
Scientific Statement From the American Heart Association Council for
High Blood Pressure Research and the Councils on Clinical Cardiology
and Epidemiology and Prevention. Circulation. 2007;115:2761–2788.
34. Hansson L, Zanchetti A, Carruthers SG, Dahlöf B, Elmfeldt D, Julius S,
Ménard J, Rahn KH, Wedel H, Westerling S. Effects of intensive bloodpressure lowering and low-dose aspirin in patients with hypertension:
principal results of the Hypertension Optimal Treatment (HOT) randomised trial. HOT Study Group. Lancet. 1998;351:1755–1762.
35.Zanchetti A, Hansson L, Clement D, Elmfeldt D, Julius S, Rosenthal
T, Waeber B, Wedel H; HOT Study Group. Benefits and risks of more
intensive blood pressure lowering in hypertensive patients of the HOT
study with different risk profiles: does a J-shaped curve exist in smokers?
J Hypertens. 2003;21:797–804.
36. Cushman WC, Evans GW, Byington RP, et al.; ACCORD Study Group.
Effects of intensive blood-pressure control in type 2 diabetes mellitus. N
Engl J Med. 2010;362:1575–1585.
37. Lv J, Neal B, Ehteshami P, Ninomiya T, Woodward M, Rodgers A, Wang
H, MacMahon S, Turnbull F, Hillis G, Chalmers J, Perkovic V. Effects of
intensive blood pressure lowering on cardiovascular and renal outcomes:
a systematic review and meta-analysis. PLoS Med. 2012;9:e1001293.
38.Mancia G, Laurent S, Agabiti-Rosei E, et al.; European Society of
Hypertension. Reappraisal of European guidelines on hypertension management: a European Society of Hypertension Task Force document. J
Hypertens. 2009;27:2121–2158.
39. Kleijnen J, de Craen AJ, van Everdingen J, Krol L. Placebo effect in double-blind clinical trials: a review of interactions with medications. Lancet.
1994;344:1347–1349.
40. Kaptchuk TJ. Powerful placebo: the dark side of the randomised controlled trial. Lancet. 1998;351:1722–1725.
41.Di Blasi Z, Harkness E, Ernst E, Georgiou A, Kleijnen J. Influence
of context effects on health outcomes: a systematic review. Lancet.
2001;357:757–762.
42. Benson J, Britten N. Patients decisions about whether or not to take antihypertensive drugs: qualitative study. Br Med J. 2002;325:873–876.
43.Redelmeier DA, Rozin P, Kahneman D. Understanding patients’ decisions. JAMA. 1993;270:72–76.
44.SHEP Cooperative Research Group. Prevention of stroke by antihypertensive drug treatment in older persons with isolated systolic hypertension.
Final results of the Systolic Hypertension in the Elderly Program (SHEP).
JAMA. 1991;265:3255–3264.
45.Rothwell PM. Subgroup analysis in randomised controlled trials: importance, indications, and interpretation. Lancet. 2005;365:176–186.
46. Guo SY, Frazer MW. Propensity score analysis: statistical methods and
applications. London: SAGE; 2009.
47.Swales JD. Evidence-based medicine and hypertension. J Hypertens.
1999;17:1511–1516.
476 Hypertension September 2013
Response to Are Observational Studies More Informative Than Randomized
Controlled Trials in Hypertension? Con Side of the Argument
Giovanni de Simone, Raffaele Izzo, Paolo Verdecchia
Downloaded from http://hyper.ahajournals.org/ by guest on May 8, 2017
We appreciate the defense of randomized controlled trials (RCTs) by Schillaci et al. However, their message that observational studies inevitably carry a high degree of bias is an oversimplification. It may be true when investigators use data
repositories of clinical records and take selected patients to be included in a given registry. However, it is less prominent
in large registries specifically designed to collect information from unselected and largely consecutive patients in a given
context or disease. This is facilitated by the widespread use of electronic databases of clinical records.
We are now aware of the problems listed in Table 1 and will take steps to account for them. As touched on in their review,
in large all-inclusive registries, reliable case-control matching is actually possible using propensity scores. Also the methodology of data collection and management can be substantially improved using advanced informatic tools, including cloud
repositories. In our opinion, registration should be compulsory for data repositories and registries, as recently suggested.1
Schillaci et al also state that it is impossible to compare treatments within observational studies. They report a number
of anecdotal examples. We agree that allocation to a specific treatment in observational studies is determined by the physician’s attitude and the clinical condition of patients. However, the inconsistencies reported by Schillaci et al have often
been caused by overestimation of small effects reported in many observational studies. Although the measures of risk are
in general adjusted for potential confounders, statistical adjustment is not a panacea.2 If the odds ratios or hazard ratios
in observational studies are small to moderate (ie, between 0.6 and 1.6 for a binary predictor), the observed effects could
largely be because of residual confounding.2 This is the case in many of the reported anecdotal examples. The reader needs
to be aware that in observational studies the effects have to be reported as odds ratios or hazard ratios, and need to be twice,
ideally, or ≥60% (or lower) than the control, to be fully reliable. Thus, special attention should be paid to the statistics used
in observational studies.
Every inconsistency between observational studies and RCTs should be carefully evaluated before drawing conclusions
about the unreliability of carefully designed observational studies.1 We still believe that RCTs should be in the middle of
a process of acquiring knowledge in which data repositories and registries can promote the need for particular RCTs to be
conducted. At the same time, data repositories and registries can be used to externally validate RCT results. There are circumstances in which the evidence provided by observational studies is ironclad, and we think that medical action may take
place even in the absence of confirming RCTs. The Parachute paradox3 should constantly be taken into account.
References
1.Tzoulaki I, Siontis KC, Ioannidis JP. Prognostic effect size of cardiovascular biomarkers in datasets from observational studies versus randomised trials:
meta-epidemiology study. BMJ. 2011;343:d6829. doi:10.1136/bmj.d6829.
2.Sainani K. The limitations of statistical adjustment. PMR. 2011;3:868–872.
3.Smith GC, Pell JP. Parachute use to prevent death and major trauma related to gravitational challenge: systematic review of randomised controlled trials.
BMJ. 2003;327:1459–1461.
Are Observational Studies More Informative Than Randomized Controlled Trials in
Hypertension?: Con Side of the Argument
Giuseppe Schillaci, Francesca Battista and Giacomo Pucci
Downloaded from http://hyper.ahajournals.org/ by guest on May 8, 2017
Hypertension. 2013;62:470-476; originally published online July 22, 2013;
doi: 10.1161/HYPERTENSIONAHA.113.01501
Hypertension is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 2013 American Heart Association, Inc. All rights reserved.
Print ISSN: 0194-911X. Online ISSN: 1524-4563
The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://hyper.ahajournals.org/content/62/3/470
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Hypertension can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial
Office. Once the online version of the published article for which permission is being requested is located,
click Request Permissions in the middle column of the Web page under Services. Further information about
this process is available in the Permissions and Rights Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Hypertension is online at:
http://hyper.ahajournals.org//subscriptions/