European Heart Journal (1998) 19, 372377
Clinical Perspectives
Diagnosis and investigation of essential and secondary
hypertension
Introduction
Hypertension is a common and important modifiable
risk factor for cardiovascular disease. It has been
convincingly demonstrated that reducing blood pres-
sure, even modestly, in hypertensive patients, reduces
the subsequent rate of stroke and coronary heart
disease. A long-term reduction of 56 mmHg in di-
astolic blood pressure is associated with approxi-
mately 3540% less stroke and 2025% less coronary
heart disease related deaths[1]. Because of the com-
monness of hypertension, the supposed simplicity of
blood pressure measurement, and the widespread
availability of effective and well-tolerated anti-
hyptensive agents the management of hypertension
has been predominantly the task of primary care
doctors. Despite this, cases exist which are more
complicated, either because of hypertensive compli-
cations, co-incident but unrelated diseases, or because
of difficulties in diagnosis or management, such as
white coat hypertension, resistant hypertension or
suspected secondary hypertension. Detecting second-
ary hypertension, and so identifying potentially cur-
able causes without excessive investigation of low risk
patients or undue use of expensive and potentially
harmful tests, is a problem often faced by cardiac
specialists to whom such patients may frequently be
referred.
Hypertension definition, diagnosis
and treatment
There is no natural dividing line between normo-
tension and hypertension. Essential hypertension is
a numerical description not a distinct disease. In a
meta-analysis of nine prospective observational
studies a continuous direct positive relationship
between diastolic blood pressure and both stroke and
Submitted 13 May 1997, and accepted 9 June 1997.
Correspondence: Dr J. Mayet, Department of Cardiology, Royal
Brompton Hospital, London SW3 6NP, U.K.
0195-668X/98/030372+10 $18.00/0 hj970595
coronary heart disease was observed[2]. Therefore any
definition of essential hypertension based on blood
pressure readings is inevitably arbitrary. In theory
hypertension could be defined as a level of blood
pressure above that at which intervention has been
shown to reduce risk[3]. This, of course, means that
different populations (e.g. young vs old, men vs
women) should have different diagnostic thresholds.
However, there are also the problems of proving
mortality reduction in a randomized controlled trial
for each new threshold level of blood pressure.
Another factor is that in most large prospective
trials blood pressure has been measured in special
trial settings, with trained nursing personnel using
more rigorously maintained sphygmomanometers
than is typical in routine practice. It is uncertain
whether other forms of blood pressure measurement,
such as routine blood pressure estimation by doctors
or newer more sophisticated measurement by ambu-
latory blood pressure monitoring would lead to the
same diagnostic and treatment thresholds. It is
usually recommended that clinicians base clinical
decisions on multiple measurements made over
multiple visits with the patient in the sitting position
after several minutes rest. A cuff of appropriate size
for the patients arm should be used with Korotkoff
phase V (when the sound disappears) taken to be
diastolic blood pressure[4]. At least three measure-
ments should be taken, with a minimum of 3 min
between each, and the values averaged. Blood pres-
sure levels often settle on further assessments as
patients become used to the clinic environment. Un-
less the hypertension is severe or there is associated
target organ damage e.g. left ventricular hypertrophy,
measurements should be repeated over a 36 month
period. This is the ideal as it closely follows the
methods of the randomized control trials, upon which
our evidence-to-treat is based. In practice, however,
this is far from what actually happens. Frequent
errors are made with sloppy technique, poorly main-
tained apparatus and far too few readings prior to a
diagnosis being made and treatment commenced[5,6].
As a result people may be unnecessarily treated.
? 1998 The European Society of Cardiology

Hypertension in practice is commonly defined
as a persistent resting diastolic blood pressure above
90 mmHg. This recommendation is based on the fact
that large studies assessing the treatment of mild
hypertension randomized subjects with a diastolic
blood pressure of greater than either 90[79] or
95 mmHg[10] and demonstrated a reduced cardio-
vascular risk in the actively treated group. However,
it has become increasingly apparent that systolic
blood pressure is also important in assessing cardio-
vascular risk and more recent studies have demon-
strated that treatment of subjects with high systolic,
but normal diastolic pressures can reduce such
risk[11,12]. It is therefore important to define a level of
systolic blood pressure above which a subject can be
labelled as being hypertensive. The range of systolic
blood pressure corresponding to diastolic pressures of
90105 mmHg is 140180 mmHg[13]. The level at
which treatment of isolated systolic hypertension has
been shown to be beneficial is 160 mmHg[11,12] and
although these studies were in subjects over 60 years
of age it would seem reasonable to extrapolate them
to younger subjects.
Based largely on the above information the
World Health Organization/International Society of
Hypertension guidelines subcommittee define mild
hypertension as a systolic blood pressure of 140180
and/or a diastolic blood pressure of 90105 mmHg[14].
They label a subgroup of this as borderline hyper-
tensives (systolic blood pressure of 140160 and/or a
diastolic blood pressure of 9095 mmHg). The U.S.
Joint National Committee on detection, evaluation,
and treatment of high blood pressure, in their fifth
report, proposes that hypertension be divided into
stages[15]. Stage 1, which they suggest is the equivalent
of mild hypertension is defined as a systolic blood
pressure of 140159 and/or a diastolic blood pressure
of 9099 mmHg. Other national guidelines use differ-
ent diagnostic and treatment thresholds, so that there
is little consensus between the experts as to what level
of blood pressure is hypertension. There is also the
problem of what to do about ubiquitously inaccurate
blood pressure measurement or how to incorporate
newer blood pressure measurement techniques, such
as ambulatory blood pressure monitoring or home
monitoring with semi-automatic digital devices. De-
spite the advantages of these (especially ambulatory
blood pressure monitoring) in terms of reliable accu-
racy, greater reproducibility and better correlations
with hypertensive end-organ damage, the published
guidelines usually discount the newer blood pressure
measurement techniques and leave the practising doc-
tor either to stick to conventional blood pressure
measurement with all its flaws, or to use the newer
techniques without the backing of guidelines.
Clinical Perspectives 373
White-coat hypertension and ambulatory
blood pressure monitoring
White-coat hypertension may be defined as increased
blood pressure confined to the setting of the physi-
cians office or clinic[16]. It affects approximately
2030% of a population otherwise thought to be
hypertensive based on their clinic blood pressure
values. The aforementioned large prospective rand-
omized hypertension treatment studies have all in-
cluded such patients since they recruited subjects on
the basis of clinic blood pressure alone. Although not
tested in a randomized control trial, many physicians
suspect that subjects with white-coat hypertension are
not prone to the same degree of cardiovascular risk
as those subjects with sustained hypertension. This
supposition is strongly supported by cross-sectional
studies suggesting that target organ damage e.g. left
ventricular hypertrophy is less likely in white-coat
hypertensives than in sustained hypertensives[17]. The
studies also suggest that the development of compli-
cations in the two largest prospective studies pub-
lished to date were at a lower rate[18,19]. This issue is
important because with the widespread availability of
24 h ambulatory blood pressure monitoring such
subjects can now be readily identified. It should be
realized, however, that ambulatory blood pressure
readings averaged over the day-time hours are gener-
ally several mmHg lower than pressures measured
in the clinic[20]. Until prospective data regarding
ambulatory blood pressure monitoring become
available it would seem reasonable in general, not to
recommend drug therapy for patients who have
white-coat hypertension in the absence of any evi-
dence of hypertension-related target organ damage.
Mild/borderline hypertension
The diagnosis and management of mild/borderline
hypertension can be difficult and patients with this
degree of hypertension may be referred to specialist
units in order to answer the question of whether drug
therapy is indicated. In the assessment of such
patients ambulatory blood pressure monitoring and
echocardiography are often useful. The former test
helps to screen for a significant white-coat effect,
and given its greater reproducibility allows a defini-
tive assessment of usual blood pressure level upon
which diagnostic and treatment decisions can be
made[21]. The question of whether a patient has a
degree of white-coat hypertension is often raised since
if even a small effect is present it may change the
advice that is given to the patient. In this situation it
is particularly important to realize that the equivalent
Eur Heart J, Vol. 19, March 1998
374 Clinical Perspectives
ambulatory blood pressure readings are several
mmHg lower than clinical blood pressures. Echocar-
diography can be useful since if left ventricular hyper-
trophy is detected it places the patient in a higher risk
group[22] and strongly favours the case of antihyper-
tensive therapy. Unfortunately electrocardiography,
although more widely available, is very insensitive for
detecting left ventricular hypertrophy[23] and so is
often unhelpful in stratifying patients. Echocardiog-
raphy is, unfortunately, still not widely available for
routine use in assessing suspected hypertensives.
Once a diagnosis of mild or borderline hyper-
tension is made treatment dilemmas may follow.
Although patients with diastolic blood pressures in
the range of 9099 mmHg have been found to benefit
from anti-hypertensive therapy, the magnitude of such
benefit is small and relates largely to the reduction in
the rate of stroke[7]. Some patients are not convinced
about the need for lifelong anti-hypertensive therapy
for such mildly elevated blood pressure values,
especially if they are aware that their blood pressure is
higher when seen by a doctor than at other times. An
ambulatory blood pressure monitoring convincingly
demonstrating an elevated average blood pressure
level will often be helpful in supporting advice for the
need to treat. It is important to stress, however, that
hypertension should not be assessed in isolation of
other cardiovascular risk factors. The presence of
other risk factors, such as advanced age, male sex,
smoking, hyperlipidaemia, diabetes mellitus or a
family history of cardiovascular disease argues for the
recommendation of anti-hypertensive therapy more
rigorously and at lower thresholds, as does the pres-
ence of end-organ damage, e.g. renal failure or left
ventricular hypertrophy. Many experts argue that the
thresholds for anti-hypertensive treatment should be
a full 5 mmHg lower in diabetics.
If a decision is made not to treat a mild
hypertensive with drugs it is important that the blood
pressure is closely monitored. Additionally, lifestyle
advice is mandatory, as it should be for all hyperten-
sive patients. Reducing salt (and calorie, if over-
weight) intake, avoiding excessive alcohol, cessation
of smoking, avoiding a high saturated fat intake and
taking regular exercise are all recommended[24].
Hypertension in the elderly
Elderly patients are particularly important because
they have a very high prevalence of hypertension and
a high absolute risk of cardiovascular complications.
Earlier treatment studies had been based on middle-
aged subjects. However, more recent studies have
confirmed that blood pressure lowering is also
Eur Heart J, Vol. 19, March 1998
beneficial in the elderly[11,12,25], making it important
to screen for hypertension in this age group. The
SHEP study evaluated patients aged over 60 years,
the MRC studied patients aged 6574 years and the
STOP-Hypertension study patients aged 7084 years.
All demonstrated a reduction in cardiovascular mor-
bidity or mortality in the actively treated group. It
should be appreciated, however, that the patients
enrolled in these studies in general were fit elderly.
Although the STOP-Hypertension study included
patients in their eighties, there remains uncertainty
about whether the very elderly benefit from treat-
ment. The SHEP and MRC study suggested that
treating isolated systolic hypertension (systolic blood
pressure >160 mmHg and diastolic blood pressure
<90 mmHg) was beneficial in this group of patients.
Therefore in patients in their 60s and 70s there is
convincing evidence that treatment of systolic blood
pressure >160 mmHg and/or diastolic blood pressure
>90 mmHg is beneficial. One complication is that in
the elderly the frequency of drug side effects is higher.
In addition, with stiffer arteries there may be an
exaggerated alerting response to the clinic setting, so
the argument for use of ambulatory blood pressure
monitoring to confirm hypertension is even more
cogent in the elderly population.
Secondary hypertension
Although the vast majority of cases of hypertension
are essential, a variety of diseases can cause raised
blood pressure. Most estimates place the prevalence
of secondary hypertension at less than 5% of an
unselected hypertensive population, with renal dis-
ease being by far the most common. Most can be
easily detected, but there remains a difficult 1% or so
in whom special testing may be necessary to detect
the underlying cause and no consensus exists to guide
the clinician as to when or how he or she should look
for an underlying cause. This is an area in which
evidence-based medicine is lacking to guide practice,
and yet we all have to face this problem regularly. At
what age and in what circumstances is special testing
required? History and examination may give clues to
the presence of an underlying disease such as renal
failure, renovascular disease, hyperaldosteronism,
phaeochromocytoma or aortic coarctation. Other
suggestive factors are early age of onset, lack of
family history, unusual course, early complications,
or resistance to therapy. The use of ambulatory blood
pressure monitoring can also give clues to the possi-
bility of secondary hypertension as detailed below.
Causes of secondary hypertension can be grouped
into categories:

v Drug induced: cocaine, cyclosporin, erythro-
poietin, non-steroidals, licorice derivatives,
sympathomimetics, some antidepressants, oestro-
gens and excess sodium chloride can all cause
hypertension[26].
v Renal disorders: chronic renal failure, nephritis,
papillary necrosis, renovascular disease, renal tu-
mours and infections can all cause hypertension,
and renal function testing and imaging are import-
ant tests.
v Endocrine cause: diabetes, thyroid disorder, phaeo-
chromocytoma, Conns syndrome, Cushings
disease, SAME (syndrome of apparent mineralo-
corticoid excess), glucocorticoid resistance, ac-
romegaly, and others can cause hypertension and
an endocrine cause should always be considered
if the course of the hypertension is unusual or if
associated features suggest the possibility.
v Cardiovascular causes: coarctation, Takayasus
and other arterial stenotic disorders are rare causes.
Certain patients are at particularly high risk of
a secondary cause of hypertension and warrant
specific investigation. Such routine screening for
secondary causes usually consists of renal ultrasonog-
raphy, 24 h urinary collection for catecholamines,
electrolyte estimation, plasma renin and aldosterone
measurement and a test for renal artery stenosis.
History and examination findings may point towards
performing one or more of these, or occasionally
other investigations, such as cortisol, thyroid function
or other specific endocrine testing. In addition, an
increase in serum creatinine after commencing
angiotensin-converting enzyme inhibitors may
prompt investigations for renovascular disease. Hy-
pertension that is resistant (poor blood pressure con-
trol while on three or more antihypertensive agents)
more frequently has an underlying cause as does high
blood pressure occurring in the young, particularly
when there is no family history of hypertension.
Another feature more recently associated with an
increased prevalence of secondary hypertension is the
detection of an absent fall in blood pressure at night
during 24 h ambulatory blood pressure monitoring
despite an apparently good nights sleep by diary
entry. This latter finding is not diagnostic but it raises
the suspicion of secondary hypertension and should
trigger further consideration of this possibility.
A working rule is that routine investigation in
all new hypertensives should include:
1. Urea and electrolytes before and after commenc-
ing antihypertensive therapy. This helps detect
endocrine causes, e.g. low potassium in Conns,
renal failure and renovascular hypertension
(increase in creatinine after commencing ACE
inhibitor).
Clinical Perspectives 375
2. Chest X-ray: for coarctation, cardiomegaly.
3. Fundoscopy.
4. ECG.
5. Echocardiography would be very helpful but at
the moment it is not sufficiently available to be
used in all subjects.
6. Ambulatory blood pressure monitoring: this
should be considered in all hypertensives, as it
can confirm hypertension, accurately document
the severity, detect white coat hypertension, and
give a pointer to an increased possibility of
secondary hypertension (absence of fall in blood
pressure at night).
Some have argued that routine hypertensives
should not be offered this test on the grounds of
cost, but given the implications of life-long anti-
hypertensive therapy the cost of a decent initial
assessment is small in comparison to the life-time cost
of an inaccurate initial assessment.
When should further investigation be insti-
tuted? In the presence of abnormalities detected in the
routine procedures above, in the presence of height-
ened clinical suspicion (e.g. extensive peripheral vas-
cular disease makes renovascular hypertension much
more likely) or in patients presenting with definite
hypertension under the age of 30 years. The exact
choice of tests to exclude the individual cause of
secondary hypertension depend on the cause of
suspicion, the availability of local renal imaging
modalities and the specifics of the case.
Renovascular hypertension
Renal artery stenosis is rare in patients with mild
hypertension without any suggestive clinical clues and
in these patients there is little point in further inves-
tigation[27]. In contrast, some patients have a very
high prevalence of renal artery stenosis. This includes
those with accelerated or malignant hypertension,
those with renal asymmetry detected on non-invasive
assessment and those whose renal function deterior-
ates on an angiotensin-converting enzyme inhibi-
tor[28]. In these patients it seems worth proceeding
directly to renal angiography since a negative non-
invasive test would not be sufficient adequately to
exclude an underlying renal artery stenosis. It is in the
in-between group of patients that non-invasive assess-
ment, as a screening method for those that require
further assessment with angiography, is important.
Before fully investigating for renovascular hyperten-
sion the question of whether its detection would alter
patient management should be raised e.g. if blood
pressure is adequately controlled on drug therapy in
an elderly patient is there any reason to attempt renal
Eur Heart J, Vol. 19, March 1998
376 Clinical Perspectives
artery revascularization if a diagnosis of renal artery
stenosis is made? Good reasons to exclude the diag-
nosis include hypertension that is uncontrolled on
drug therapy, deteriorating renal function and inter-
mittent acute episodes of pulmonary oedema. Other
indications might include unacceptable side-effects
with drug therapy and a young age when renal artery
stenosis due to fibromuscular hyperplasia is relatively
more common. This type of pathology responds well
to percutaneous transluminal renal angioplasty and
intervention may prevent life-long drug therapy.
There is no universally agreed age at which screening
for secondary causes should become routine; ages of
under 20 and under 25 have both been suggested[27,28]
although we favour a high index of suspicion ap-
proach under the age of 30 years. The elderly have a
particularly high incidence of renovascular disease,
usually due to atheromatous disease, but again, this
diagnosis is only worth pursuing if it will influence
clinical management.
Where a non-invasive test is required to screen
a patient prior to possible angiography, the test used
will depend largely on local expertise. Intravenous
pyelography (sensitivity 75%, specificity 86%), plasma
renin activity (sensitivity 5080%, specificity 84%),
captopril plasma renin activity (sensitivity 74%,
specificity 89%), captopril renography (sensitivity
93%, specificity 95%), Doppler renal artery ultra-
sonography (sensitivity 90%, specificity 9095%) and
magnetic resonance angiography (sensitivity 9095%,
specificity 95%) have all been used. The figures quoted
are derived from a review by Mann and Pickering[27]
and are a summary of many studies. These studies
however, usually occur in centres with expertise in this
area, and so may be different in the real world.
Mineralocorticoid hypertension
It has been appreciated recently that mineralocorti-
coid associated hypertension is more common than
previously thought. Milder cases of hyperaldos-
teronism can be treated medically and can have well
controlled hypertension for many years with mild
hypokalaemia. There has been a profusion of infor-
mation of the molecular defects in rare forms of
mineralocorticoid associated hypertension and more
subtle derangements of the steroid pathways seem
likely to account for a greater proportion of cases of
hypertension when we investigate this area in greater
detail with newer techniques.
Conclusion
We have highlighted some of the difficult areas
in hypertension management, especially when to
Eur Heart J, Vol. 19, March 1998
diagnose and treat hypertension and when to initiate
investigation for secondary hypertension. It is
important to remember, however, that hypertension
can no longer be treated as a single, isolated risk
factor. It is important to perform a full cardiovascu-
lar risk factor profiling on all patients, including
assessment of smoking status and serum cholesterol.
The aim of management should be to reduce the
overall risk of future cardiovascular events.
J. MAYET
A. J. S. COATS
Imperial College of Science, Technology and Medicine,
at the National Heart and Lung Institute and Royal
Brompton Hospital
London, U.K.
References
[1] Collins R, Peto R, McMahon S et al. Blood pressure, stroke
and coronary heart disease. Part 2, short-term reductions in
blood pressure: overview of randomised drug trials in their
epidemiological context. Lancet 1990; 335: 82738.
[2] McMahon S, Peto R, Cutler J et al. Blood pressure, stroke
and coronary heart disease. Part 1, prolonged differences in
blood pressure: prospective observational studies corrected for
regression dilution bias. Lancet 1990; 335: 76574.
[3] Evans JG, Pose G. Hypertension. Br Med Bull 1971; 27:
3742.
[4] Petrie JC, OBrien ET, Littler WA, de Swiet M. Blood
pressure measurement. London: Br Med J, 1987.
[5] OBrien E, Atkins N, Mee F, Coyle D, Syed S. A new
audiovisual technique for recording blood pressure in re-
search? The sphygmocorder. J Hypertens 1995; 13: 17347.
[6] Smith TDW, Clayton D. Individual variation between general
practitioners in labelling of hypertension. Br Med J 1990; 300:
745.
[7] Medical Research Council Working Party. MRC trial of
treatment of mild hypertension: principal results. Br Med J
1985; 291: 97104.
[8] Veterans Administration Cooperative Study Group on Anti-
hypertensive Agents. Effects of treatment on morbidity in
hypertension. II. Results in patients with diastolic blood
pressure averaging 90 through 114 mmHg. JAMA 1970; 213:
114352.
[9] Hypertension Detection and Follow-up Program Cooperative
Group. Five-year findings of the hypertension detection and
follow-up program. I. Reduction in mortality of persons with
high blood pressure, including mild hypertension. JAMA
1979; 242: 256271.
[10] Australian National Blood Pressure Management Committee.
The Australian Therapeutic trial in mild hypertension. Lancet
1980; I: 12617.
[11] SHEP Cooperative Research Group. Prevention of stroke by
antihypertensive drug treatment in older persons with isolated
systolic hypertension. JAMA 1991; 265: 25564.
[12] Medical Research Council Working Party. MRC trial of
treatment of hypertension in older adults: principal results. Br
Med J 1992; 304: 40512.
[13] Kannel WB, Dawber TR, McGee DL. Perspectives on systolic
hypertension: the Framingham study. Circulation 1986; 61:
117982.
[14] 1993 guidelines for the management of hypertension: memo-
randum from a World Health Organization/International
Society of Hypertension meeting. J Hypertens 1993; 11: 905
18.

[15] The fifth report of the Joint National Committee on detection,
evaluation, and treatment of high blood pressure (JNC V).
Arch Intern Med 1993; 153: 15483.
[16] Pickering TG. Blood pressure measurement and detection of
hypertension. Lancet 1994; 344: 315.
[17] Gosse P, Promax H, Durandet P, Clementy J. White coat
hypertension: no harm for the heart. Hypertension 1993; 22:
7669.
[18] Perloff D, Sokolow M, Cowan RM, Juster RP. Prognostic
value of ambulatory blood pressure measurements: further
analyses. J Hypertens 1989; 7 (Suppl 3): S3S10.
[19] Verdecchia P, Porcellati C, Schillaci G et al. Ambulatory
blood pressure: an independent predictor of prognosis in
essential hypertension. Hypertension 1994; 24: 793801.
[20] Staessen JA, OBrien ET, Atkins N, Amery AK. Short report:
ambulatory blood pressure in normotensive compared with
hypertensive subjects. J Hypertens 1993; 11: 128997.
[21] Coats AJS, Radaelli A, Clark SJ, Conway J, Sleight P. The
influence of ambulatory blood pressure monitoring on the
design and interpretation of trials in hypertension. J Hyper-
tens 1992; 10: 38591.
[22] Kannel WB. Left ventricular hypertrophy as a risk factor in
arterial hypertension. Eur Heart J 1992; 13 (Suppl D): 828.
European Heart Journal (1998) 19, 377381
Survival analysis and classification of death in patients
under anti-arrhythmic treatment
Introduction
The practice of medicine is becoming increasingly
complex and, paradoxically, despite greater knowl-
edge, even more uncertain[1]. Today, knowledge itself
is defined on the basis of an arbitrary but accepted
statistical test, performed in a randomized clinical
trial[2]. What the physician thinks, suspects, believes,
or has a hunch about, is assigned to the not knowing
category. Technical advances expected to reduce
clinical uncertainty have not only contributed to its
increase, but have even been used to obscure it[1].
Obscurity and uncertainty are also promoted by the
lack of statistical expertise of the average clinician,
and by the sophisticated, sometimes overtly unclear,
presentation of trial results.
As regards modern anti-arrhythmic treatment,
we struggle with the fact that arrhythmic death is an
almost impossible definition[3,4]. Thus, total mortality
seems to be the only reliable end-point. In our
opinion, however, in the study of arrhythmias, sur-
vival analysis cannot be reduced to a superficial
matter of life and death[3]. This contrasts with other
opinions, which consider sudden cardiac death as a
soft surrogate end-point[5,6]. Total mortality as a
Clinical Perspectives 377
[23] Devereux RB, Koren MJ, De Simone G, Okin PM, Kligfield
P. Methods for detection of left ventricular hypertrophy:
application to hypertensive heart disease. Eur Heart J1993; 14
(Suppl D): 815.
[24] Sever PS, Beevers G, Bulpitt C et al. Management guidelines
in essential hypertension: report of the second working part
of the British Hypertension Society. Br Med J 1993; 306:
9837.
[25] Dahlof B, Lindholm LH, Hansson L, Scherston B, Ekbom T,
Wester P-O. Morbidity and mortality in the Swedish Trial in
Old Patients with Hypertension (STOP-Hypertension). Lancet
1991; 338: 12815.
[26] Streeten DH, Anderson GH. Secondary hypertension. An
overview of its causes and management. Drugs 1992; 43:
80519.
[27] Mann SJ, Pickering TG. Detection of renovascular hyperten-
sion. State of the art: 1992. Ann Intern Med 1992; 117:
84553.
[28] National High Blood Pressure Education Program Working
Group. 1995 update of the working group reports on chronic
renal failure and renovascular hypertension. Arch Intern Med
1996; 156: 193847.
primary end-point avoids difficult issues, but it may
reduce the sensitivity of a trial below acceptable
levels. Other causes of mortality are also likely to be
recorded particularly in older patient populations.
An analysis of cause-specific mortality, rather than
intention-to-treat analysis of total mortality, is to be
preferred, and should be pre-specified in the protocol,
in line with the hypothesis under test[7].
More complications and confusion arise from
the sloppy application of the intention to treat
principle[7]. In addition, clinicians and trialists face
controversies in counting and attributing events in
clinical trials[8].
Classification of death
Presently accepted classifications of death do not fully
describe or tabulate all significant aspects of terminal
events, nor do they consider unique aspects of ar-
rhythmia investigations[9]. A. E. Epstein et al. present
a new classification scheme. This uses the following
categories: (i) primary organ cause (cardiac [arrhyth-
mic, non-arrhythmic or unknown], non-cardiac or
unknown); (ii) temporal course (sudden, non-sudden
378 Clinical Perspectives
or unknown); (iii) documentation (witnessed, moni-
tored [yes, no or unknown]); (iv) operative relation
(pre-operative, peri-operative or post-operative); and
(v) system relation (procedure related, pulse gener-
ator related and lead related [yes, no or unknown]). A
NASPE (North American Society of Pacing and
Electrophysiology) Policy Conference recommended
a set of definitions, statistical considerations, and
minimal standards for reporting the outcome of
implantable cardioverter defibrillator therapy[10]. All
deaths should be classified as cardiac or non-cardiac
deaths. Cardiac deaths should further be subclassified
as sudden or non-sudden cardiac deaths. In addition,
operative mortality, waiting period deaths and deaths
related to hardware problems should be identified.
In CAST, death was judged to be due to
arrhythmia if it was characterized in any of the
following ways: (i) witnessed and instantaneous, with-
out new or accelerating symptoms; (ii) witnessed and
preceded or accompanied by symptoms attributable
to myocardial ischaemia in the absence of shock or
Class IV congestive heart failure as categorized by the
New York Heart Association; (iii) witnessed and
preceded by symptoms attributable to cardiac ar-
rhythmia, e.g. syncope or near-syncope; or (iv) un-
witnessed but without evidence of another cause. In
the presence of severe congestive heart failure, death
was judged to be not to arrhythmia if death from
heart failure appeared probable within 4 months of
the fatal episode[11].
In the CAPS study, it was suggested that,
considering all the possible effects of anti-arrhythmic
drugs as well as the difficulty in classifying events, it
may be more practical simply to evaluate total car-
diac mortality in the conduct of any future rand-
omized intervention trial[12]. The authors stress that
such a classification could increase the sample size
necessary to detect a difference between treatment
and placebo, or could even obscure an improvement
in arrhythmic mortality, if the active drug caused
death by some other mechanism, such as congestive
heart failure. Unless more precise measurements of
the mechanism of death are developed, the combi-
nation of a time-based and aetiology-based assess-
ment of cause of death will be necessary, since
sudden death is not equivalent to arrhythmic
death.
A classification based on the condition of the
circulation immediately before death appears to be
the most relevant to studies of sudden death. In 58%
of the 142 deaths[13], the subject collapsed abruptly
and his pulse ceased without prior circulatory col-
lapse (arrhythmic death); in 42%, the pulse ceased
only after the peripheral circulation had collapsed
(deaths in circulatory failure). Ninety-three percent of
Eur Heart J, Vol. 19, March 1998
S1 = instantaneous SD
S2 = SD in < 60 min
S: 14
S3 = unwitnessed SD
S4 = other SD
D1 = SD in NYHA Class I
D2 = SD in NYHA Class II
D: 14
D3 = SD in NYHA Class III
D4 = SD in NYHA Class IV
Figure 1 Classification of death in cardiac patients. The
S14, D14 code is applicable for drug-treated and device-
treated patients. SD=sudden death.
final illnesses that lasted less than 1 h ended in
arrhythmic deaths; 74% that lasted more than 1 day
ended in deaths in circulatory failure. Eighty-eight
percent of deaths that occurred outside of the hospi-
tal were arrhythmic; 71% of deaths that occurred in
the hospital were deaths in circulatory failure. Ninety
percent of deaths in which the primary illness was
heart disease were arrhythmic; 86% of deaths in
which the primary cause was other than heart disease
were deaths in circulatory failure. Ninety-one percent
of deaths precipitated by an acute cardiac event were
arrhythmic; 98% precipitated by acute respiratory
obstruction, haemorrhage, infection, stroke or other
non-cardiac events were deaths in circulatory failure.
A classification of deaths as arrhythmic or circulatory
failure cannot be entirely accurate unless the ECG is
observed or recorded at the time of death.
In this study, an estimate, based upon the
proportion of cases in which the observational data
were least complete or somewhat ambiguous, sug-
gests that the number of misclassifications is not
greater than 5%, and if present, primarily occurred
among the unwitnessed deaths.
In clinical reports, clarity competes with
details. An appropriate code, together with total
mortality and total cardiac mortality, is necessary for
adequate reports on the natural history of life-
threatening arrhythmias and on the results of modern
treatment[14,15]. There are examples of coding systems
in cardiology. NYHA functional class IIV is a
parameter used in nearly every clinical report[16]; the
Lown grading of ventricular arrhythmias has gained
wide acceptance in the world of arrhythmias[17,18].
Thus, a similar concise, and descriptive code for
reporting sudden death would be extremely con-
venient. Figure 1 describes the code S14, D14 which
has been explained previously[14,15]. The code is

equally applicable to drug-treated and device-treated
patients. We have offered suggestions for a more
complicated description of events in implantable
cardioverter defibrillator patients[15], but complexity
creates a barrier, as in the case of other quoted
classification systems. If the code S14, D14 would
limit the number of misclassifications to 5%[13], it
would be a handy tool in all trials and studies on
anti-arrhythmic treatment.
Survival analysis and other statistics in
clinical medicine
Today, for unproved treatments, it is a standard
requirement that a properly designed randomized
clinical trial be performed[19]. However, in view of the
limitations of randomized trials, valuable obser-
vational studies remain necessary. We have to pay
strict attention to the events and observations that
occur in the ordinary circumstances of clinical prac-
tice. Randomized trials are unfeasible for studying
multiple therapeutic candidates, instabilities due to
rapid technological improvements, and long-term ad-
verse effects[20]. In comparison to the normal, ex-
pected life expectancy of individuals included in
major trials, the mean follow-up duration is very
short: CAST[11]: 10 months; MADIT[21]: 27 months;
CAMIAT[22]: 179 years; EMIAT[23]: 21 months
(median). Thus, it is obvious that, when the trial ends
because of excess mortality in one treatment arm, the
long-term complications and side effects remain
unknown.
The most notorious techniques for survival
analysis are the KaplanMeier method, life-tables,
log-rank statistics, Coxs proportional hazards
model, multivariate logistic regression, odds ratios,
relative risk calculations and, more recently, a trian-
gular sequential design modified for two-sided alter-
natives. Books listed as references[2427], have,
in particular, been crucial for our understanding of
survival analysis.
For the clinician it is necessary to differentiate
between statistical significance and clinical relevance.
Many studies refer to risk reduction, without clearly
accentuating the absolute value of risks. Confidence
limits and attention to the power of a study are of
major interest. The statistical representation of results
is too often not as simple and clear as it should and
could be.
Age, gender and duration of follow-up deserve
full attention. These demographic data allow the
calculation of expected survival and mortality. An
easy and attractive tool is the standardized mortality
ratio[27]: the number of observed deaths divided by
Clinical Perspectives 379
the number of expected deaths. Life-tables and
KaplanMeier curves only reflect the duration of
follow-up, the occurrence of death and time of cen-
soring. There is no correction for the age at the time
of inclusion in the study. The standardized mortality
ratio adjusts the mortality rate to eliminate the effects
of age, sex, and follow-up duration[28]. Thus, it con-
tributes to a better comparison of outcomes in rand-
omized and observational studies. It is a parameter to
be added to the standard patient characteristics, as
means and standard deviations of age and follow-up.
It is a pity that standard statistical packages do not
provide easy processing of the procedure.
Intention-to-treat analysis and analysis by
actual treatment
The intention-to-treat analysis (analysis as rand-
omized) is a method of handling survival data for
randomized studies to minimize biases resulting from
exclusion of selected patients[29]. It includes all rand-
omized patients in the group to which they were
randomly assigned, regardless of their compliance
with the entry criteria, regardless of the treatment
they actually received, and regardless of subsequent
withdrawal from treatment or deviation from the
protocol[30]. It has been argued that intention-to-treat
encourages sloppiness: Whatever we do, give the
treatment or not, it is OK since the analysis is by
intention-to-treat![7]. It is very likely that the % of
nonsense conclusions equals the % discontinuation of
allocated treatment for reasons other than outcome
events. Maximal adherence to the original treatment
allocation is essential. An analysis by actual treat-
ment, including only those patients who satisfy the
entry criteria and who adhere to the protocol subse-
quently, is mandatory, but requires a strict follow-up.
If both analyses lead to the same conclusion, the
strength of that conclusion is considerably increased.
When they lead to different conclusions, troubles
arise[7,10,29]. A secondary analysis by actual treat-
ment, in which data are censored at a specific time,
such as in the case of implantable cardioverter de-
fibrillator explant (without replacement) or cardiac
transplantation, may provide useful information[10].
The intention-to-treat principle has different impli-
cations as regards the function of the test hypothesis:
comparison of an active agent with placebo, assess-
ment of equivalence between two treatments, compet-
ing risk situation, safety analysis[7].
Modern trials and some fallacies
For the clinician, it is important to be concerned
about the classical statistical errors[31]. Type 1 errors
Eur Heart J, Vol. 19, March 1998
380 Clinical Perspectives
are false positive errors that assign statistical benefit
to a given modality, when such benefit does not exist.
Type 2 errors are false negative errors: the observed
distinction is regarded as not significant (no benefit),
whereas in reality a benefit exists. Type 2 errors can
arise in circumstances in which the numbers of obser-
vations are too small: insufficient power of the study.
Another report advocates type 3 and type 4 errors[31].
Type 3 errors are errors in which the risk of a given
medical or public health approach is under-estimated,
undetected or not specifically sought, leading to an
under-estimate of the risk-benefit balance. Type 4
errors arise because the risks of a given medical
intervention are over-estimated, leading to under-use
or abandonment of a useful intervention. It should be
emphasized that type 3 errors appear to be much
more frequent than type 4 errors. This stems, in part,
from the failure of medicine to establish effective
mechanisms for the rapid recognition and correction
of errors. Most recorded medical history deals with
triumphs of medicine. The disasters are usually in-
terred along with the victimpatients. The introduc-
tion into the field of anti-arrhythmic treatment, of
notions as type 3 and type 4 errors, could offer
interesting perspectives.
Besides these classical errors, other misunder-
standings can arise from insufficient data analysis
and/or presentation. Examples of these are: omission
of simple absolute risks[23], omission of simple
mortality data[11], conventional medical treatment
without preset minimum requirements[21,32], one-
sided significance tests[22], inappropriate drug selec-
tion[11,33], and cumulation of results for different
drugs in one group[11,21,34]. Discontinuation of anti-
arrhythmic drug treatment, especially when com-
pared to implantable cardioverter defibrillator
therapy, can be a fatal error in high risk patient
groups. In the MADIT study[21,32], only 45% of the
patients with drug treatment were taking amiodarone
at the time of the last follow-up visit. Study medi-
cation withdrawal in the placebo/amiodarone
arm was 214/385% in EMIAT[23], 255/364% in
CAMIAT[22].
Conclusions
The most provocative challenge in cardiology for the
next decade will be the attempt to delay sudden
death as long as possible. To replace this final event
by another terminal endpoint such as death from
heart failure, carcinoma, cerebrovascular accident, is
not at all attractive. The implantable cardiac de-
fibrillators cut deaths[34], by reverting sudden ar-
rhythmic death. Amiodarone can prevent arrhythmic
Eur Heart J, Vol. 19, March 1998
death[22,23]. Length and quality of life interfere with
cost/benefit considerations. In the MADIT patient
population, the cost of implantable cardioverter de-
fibrillator treatment per life-year saved was U.S.
$27 000, and in the AVID study U.S. $114 000[35].
Can survival and financial calculations ever
offer a definite solution for individual patients in
different socio-economic and philosophical back-
grounds? We believe not. We concur with Voltaire
that common sense is not so common, but the
clinician hardly needs unbiased information, based
on adequate and objective data analysis. The study of
money is one in which complexity is used to disguise
truth or to evade it, but not to reveal it[36]. Let us
hope that clinical trials and practice will not suffer
from this syndrome, described by Galbraith[36].
What should the practitioner do when a new
treatment is described? The most concise answer
comes from Kassirer[37]. We cannot uncritically adopt
the new strategy, but are obliged to use our inferential
skills to answer several questions. Is my patient
similar to the subjects studied? Is the magnitude of
the difference between the new treatment and the
existing one sufficient to justify a switch? Can the
results of the study be generalized to patients in my
practice and to the type and quality of care available
locally? Affirmative answers to these questions will
encourage a switch to the new therapy.
H. ECTOR
H. HEIDBU } CHEL
F. VAN DE WERF
University Hospital Gathuisberg
Leuven, Belgium
References
[1] Logan RL, Scott PJ. Uncertainty in clinical practice: impli-
cations for quality and costs of health care. Lancet 1996; 347:
5958.
[2] Hellman S, Hellman DS. Problems of the randomized clinical
trial. N Engl J Med 1991; 324: 15859.
[3] Ector H. Endpoints and trials: a matter of life and death.
PACE 1994; 17: 107981.
[4] Ector H, Rogers R, Rubens A, De Geest H. Classification of
death in patients under antiarrhythmic treatment. PACE
1993; 16: 22504.
[5] Epstein AE. AVID necessity. PACE 1993; 16: 17735.
[6] Gottlieb SS. Dead is dead artificial definitions are no
substitute. Lancet 1997; 349: 6623.
[7] Lewis JA, Machin D. Intention to treat who should use
ITT? Br J Cancer 1993; 68: 64750.
[8] Sackett DL, Gent M. Controversy in counting and attrib-
uting events in clinical trials. N Engl J Med 1979; 301:
141012.
[9] Epstein AE, Carlson MD, Fogoros RN, Higgins SL, Venditti
FJ. Classification of death in antiarrhythmic trials. J Am Coll
Cardiol 1996; 27: 43342.
[10] Kim SG, Fogoros RN, Furman S, Connolly SJ, Kuck KH,
Moss AJ. Standardized reporting of ICD patient outcome: the

report of a North American Society of Pacing and Electro-
physiology Policy conference, February 910, 1993. PACE
1993; 16: 135862.
[11] Echt DS, Liebson PR, Mitchell LB et al. Mortality and
morbidity in patients receiving encainide, flecainide, or pla-
cebo. The cardiac arrhythmia suppression trial. N Engl J Med
1991; 324: 7818.
[12] Greene HL, Richardson DW, Barker AH et al.Classification
of deaths after myocardial infarction as arrhythmic or non-
arrhythmic (the cardiac arrhythmia pilot study). Am J Cardiol
1989; 63: 16.
[13] Hinkle Jr LE, Thaler HT. Clinical classification of cardiac
deaths. Circulation 1982; 65: 45764.
[14] Rogers R, Ector H, Rubens A, Timmermans C, Heidbuchel
H, De Geest H. Classification of death in patients
under antiarrhythmic treatment. In: Aubert AE, Ector H,
Stroobandt R, eds. Cardiac Pacing and Electrophysiology.
Dodrecht, The Netherlands: Kluwer Academic Publishers,
1994: 4148 .
[15] Ector H. Endpoints and trials: a matter of life and death.
Classification of death in patients under antiarrhythmic treat-
ment. In: Oto AM, ed. Practice and Progress in Cardiac
Pacing and Electrophysiology. Dordrecht, The Netherlands:
Kluwer Academic Publishers, 1996: 15 .
[16] The Criteria Committee of the New York Heart Association.
Nomenclature and Criteria for Diagnosis of Diseases of the
Heart and Great Vessels. Boston: Little, Brown and Com-
pany, 1973: 286 .
[17] Lown B, Wolf M. Approaches to sudden death from coronary
heart disease. Circulation 1971; 44: 13042.
[18] Lown B. Sudden cardiac death 1978. Circulation 1979; 60:
15939.
[19] Passamani E. Clinical trials, are they ethical? N Engl J Med
1991; 324: 158992.
[20] Feinstein AR. The limitations of randomized trials. Ann
Intern Med 1983; 99: 54450.
[21] Moss AJ, Hall WJ, Cannom DS et al. Improved survival with
an implanted debrillator in patients with coronary disease at
high risk for ventricular arrhythmia. N Engl J Med 1996; 335:
193340.
[22] Cairns JA, Connolly SJ, Roberts R, Gent M. Randomized
trial of outcome after myocardial infarction in patients with
Clinical Perspectives 381
frequent or repetitive ventricular premature depolarizations:
CAMIAT. Lancet 1997; 349: 67582.
[23] Julian DG, Camm AJ, Frangin G. et al. Randomized trial of
effect of amiodarone on mortality in patients with left-
ventricular dysfunction after recent myocardial infarction:
EMIAT. Lancet 1997; 349: 66774.
[24] Lee ET. Statistical methods for survival analysis, 2nd edn.
New York: John Wiley & Sons, Inc., 1992 .
[25] Cox DR, Oakes D. Analysis of survival data. London,
England: Chapman & Hall, 1992.
[26] Whitehead J. The design and analysis of sequential clinical
trials, 2nd edn. Chichester, England: Ellis Horwood Limited,
1992 .
[27] Gardner MJ, Altman DG. Statistics with confidence. London,
England: British Medical Journal, 1989 .
[28] Ector H, Jordaens L, Vanhaecke J. Survival analysis and
clinical medicine. An observational comparison of the im-
plantable cardioverter defibrillator, amiodarone treatment,
and heart transplantation. Eur Heart J 1996; 17: 14447.
[29] Yusuf S, Garg R, Zucker D. Analyses by the intention-to-
treat principle in randomized trials and databases. PACE
1991; 14: 207882.
[30] Fischer LD, Dixon DO, Herson J, Frankowski RK, Hearson
MS, Peace KE. Intention to treat in clinical trials. In: Peace
KE, ed. Statistical issues in drug research and development.
New York: Marcel Dekker, 1990: 33150..
[31] Robin ED, Lewiston NJ. Type 3 and type 4 errors in the
statistical evaluation of clinical trials. Chest 1990; 98: 4635.
[32] Friedman PL, Stevenson WG. Unsustained ventricular
tachycardia to treat or not to treat. N Engl J Med 1996;
335: 19845.
[33] Mason JW. A comparison of seven antiarrhythmic drugs in
patients with ventricular tachyarrhythmias. N Engl J Med
1993; 329: 4528.
[34] McCarthy M. Implantable cardiac defibrillators cut deaths.
Lancet 1997; 349: 1225.
[35] Muschlin AI, Zipes DP. NASPE 18th Annual Scientific Ses-
sions, New Orleans, May 710, 1997.
[36] Galbraith JK. Money. Whence it came, where it went. New
York: Bantam Books, 1975: 6 .
[37] Kassirer JP. Clinical trials and meta-analysis. What do they
do for us. N Engl J Med 1992; 327: 2734.
Eur Heart J, Vol. 19, March 1998