Statin use in the secondary prevention of coronary heart disease in primary care: cohort study and comparison of inclusion and outcome with patients in randomised trials
BMJ 2005; 330 doi: https://doi.org/10.1136/bmj.38398.408032.8F (Published 07 April 2005) Cite this as: BMJ 2005;330:821
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Dear Sir,
Wei et al reported that statins are as effective in the community as
in clinical trials in the secondary prevention of coronary heart disease,
but have not commented on some even more surprising results(1). They
reported benefit for calcium blockers, in contradiction of previous
findings (2-4), and found beta blockers to be much more effective than
previously thought-OR for risk reduction 0.38 compared to 0.81 in a
previous systematic overview(2).
We are concerned that these unusually positive findings across the
board may reflect an undetected bias in the study design, rather than the
true benefit of drug treatment. The marked differences between the
intervention and control groups has already been pointed out (Penston,
Chalk and cheese, BMJ rapid response 12 April 2005), and most of these
differences were controlled for in the analysis. However, the significant
rise in statin use from 3% in 1993 to 63% in 2001 means that the
intervention and control groups are effectively cohorts from different
time periods. This would also affect the results for beta blockers and
calcium blockers, both of which were much more commonly used in the statin
group.
It is not possible to draw meaningful conclusions from this study
until the effect of date of discharge from hospital is allowed for in the
analysis, and we therefore suggest that the authors may wish to reanalyse
the data to take this into account.
Reference List
(1) Wei L, Ebrahim S, Bartlett C, Davey PD, Sullivan FM, MacDonald
TM. Statin use in the secondary prevention of coronary heart disease in
primary care: cohort study and comparison of inclusion and outcome with
patients in randomised trials. BMJ 2005; 330(7495):821.
(2) Teo KK, Yusuf S, Furberg CD. Effects of prophylactic
antiarrhythmic drug therapy in acute myocardial infarction. An overview of
results from randomized controlled trials. JAMA: The Journal of the
American Medical Association 1993; 270(13):1589-1595.
(3) Aronow WS. Management of the Elderly Person After Myocardial
Infarction. J Gerontol A Biol Sci Med Sci 2004; 59(11):1173-1185.
(4) van der Elst ME, Buurma H, Bouvy ML, de Boer A. Drug Therapy for
Prevention of Recurrent Myocardial Infarction. Ann Pharmacother 2003;
37(10):1465-1477.
Competing interests:
None declared
Competing interests: No competing interests
We are pleased our paper has aroused some interest in the role of
observational data in examining treatment effects. Clearly there can be
no substitute for properly randomised controlled trials in evaluating the
effects of drug treatments. But, it is important to realise that
selection criteria used in trials may result in study populations that are
substantially different from the populations who will receive treatment in
routine clinical practice. For example, in the case of antihypertensive
drugs, the MRC mild hypertension trial cardiovascular mortality rates were
similar to those of normotensive men in a population sample.[1] For this
reason, examining the associations between treatment and outcome in
typical patients is potentially useful in understanding generalisability
of treatment effects.
Penston suggests that those who are treated are too dissimilar to
make fair comparisons, highlighting the demographic and co-morbidity
differences. However, he misses the important point that they have all
suffered from a cardiovascular event that makes them eligible for drug
treatment. Multivariable logistic regression is a standard and robust
statistical technique for making adjustments for potential confounding
factors; it is not statistical trickery! Penston asserts that the
differences between all-cause mortality (31%, 95% CI 20%, 41%) and fatal
and non-fatal outcomes (18%, 95% CI 5%, 29%) indicate the analyses are
invalid. Examining point estimates is misleading and is why we presented
the findings with the 95% confidence intervals which overlap,
demonstrating that the difference between these two outcomes may simply
represent the play of chance.
Improvements in treatment and survival prospects of patients have
occurred over the period of the study. However, an analyses of the
relationship between statin use in the first and second halves of our
study showed similar associations – relative risk for all cause mortality
was 0.62 (95%CI 0.49-0.77) for the period of 1993-1997 and 0.73 (0.54-
0.98) for the period 1998-2001, suggesting that temporal trends do not
explain our findings.
Mills is correct; secondary prevention generally means early
detection and management of pre-symptomatic disease and tertiary
prevention applies to the management of those with diagnosed disease. In
the prevention of cardiovascular disease idiosyncratic terminology has
arisen and is widely used and understood. For example, the American Heart
Association defines secondary prevention as “identifying and treating
people with established disease and those at very high risk of developing
cardiovascular disease”. [2] This definition is also used by the World
Health Organisation.[3]
1. Ebrahim S, Davey Smith G. Numbers needed to treat: does it help
clinical decision making? J Human Hypertension 1999;13:721-4
2. http://www.americanheart.org/presenter.jhtml?identifier=4723
(accessed 18 April 2005)
3. World Health Organisation. Prevention of recurrent heart attacks
and strokes in low and middle income countries. WHO, Geneva, 2003
Competing interests:
None declared
Competing interests: No competing interests
I am puzzled by the title of the article, "Statin use in the
secondary prevention of coronary heart disease in primary care."
Secondary prevention refers to early detection of disease with a view to
minimizing morbidity and disability (see Last JM. A Dictionary of
Epidemiology, 4th ed. New York: Oxford University Press, 2001), but the
article describes the use of statins in tertiary prevention.
Competing interests:
None declared
Competing interests: No competing interests
able to find women of sufficient coronary risk to find even a hint of all-cause
mortality benefit from statin. JAMA
(1) said last year that this is now 20,000 women in studies lasting
over 1 year. Women should be told so -before they embark on the costly
path of long-term side effects of statins.
Regarding 'subgroup' analysis: women may be 'subgroups' in studies but
in most Western countries the majority dying of cardiovascular causes,
and can 20,000 women in subgroups be all wrong?
Sir Richard Peto has argued the dangers of subgroup analysis, yet he
included subgroups of diabetics and women in the HPS study and now refuses,
3 years after original publication, to release the Kaplan-Meier all-cause statin
mortality curves for men, women and diabetics. Let's hope he responds
to this posting with those crucial graphs on eBMJ.
Not to make this a battle of studies, let's pick one statin, atorvastatin
(Lipitor). In ASCOT, placebo does marginally better than 10 mg/d at 1.6
years and the mortality curves again virtually touch at mean study end
(3.3 years). Next, the even larger TNT trial compares 10 mg/ with 80 mg/d
in extreme risk patients (effectively: angioplasty and/or bypass) to find
5 more nonvascular deaths than vascular deaths at top dose Lipitor
after 4.9 years. Maybe Dr. Lewis has patients at greater
cardiovascular risk but I doubt it.
For some perspective, ASCOT found NO cardiovascular benefit in women
and diabetics, while part of the primary endpoint benefit was in 'silent
heart attacks' (my preferred kind) while all-cause mortality was relegated
to secondary endpoint. In TNT, there were 114 fewer cardiac endpoints
yet a non significant increase in overall deaths on top dose statin.
When statins are prescribed off-label (i.e. other than for dyslipidemia)
they'd better show all-cause mortality benefit in such fatal disease.
Clearly in almost all people on them this is not the case, such are the
studies. More on the reality of anti-mevalonate (statin) therapy is here
health-heart.org/cholesterol.htm.
1) Walsh JM, Pignone M. Drug treatment of hyperlipidemia in women. JAMA
2004 291(18):2243-52. Medline
15138247
Competing interests:
None declared
Competing interests: No competing interests
Sir
Referring to the results of their recently published study in the
BMJ, Wei et al. stated: “The effects of statins in routine clinical
practice were consistent with, and similar to, those reported in clinical
trials.” [1] Whether or not the data support such a conclusion is,
however, a matter of debate.
The study population comprised a cohort of 4,892 patients from
Tayside defined by a discharge diagnosis of myocardial infarction between
January 1993 and December 2001. In the routine course of their medical
management, some of these patients were treated with statins. The cohort
was divided into two groups – those receiving statins (50.3%) and those
not treated with these drugs (49.7%). The two groups, however, were
markedly different in many important respects. Those not receiving statins
were more likely to be older (mean age 73 v 63 years; proportion >75
years old, 49% v 15%), to be female (42% v 37%) and to have other co-
existent diseases, but less likely to be receiving other cardiovascular
drugs (for example, beta-blockers (33% v 69%), ACE inhibitors (30% v 57%)
and anti-platelet therapy (65% v 89%)) than those given statins. But the
differences do not end here. As described in the paper, in 1993 only 3.1%
of patients received statins compared with 62.9% in 2001. It follows,
therefore, that the majority of patients receiving statins came from the
later period of the study (1998-2001) whereas those not given these drugs
were recruited from the earlier period (1993-1997). This accounts, at
least in part, for the differences in treatment between the two groups in
terms of other cardiovascular medication. But many other changes over this
time are also relevant: demographic changes (for example, a reduction in
smoking), changes in the acute management of myocardial infarction (in
particular, the increasing use of thrombolytic therapy) and, most
importantly, changes in the definition of myocardial infarction. These
factors would tend to reduce the long-term mortality following myocardial
infarction in those patients diagnosed in the latter part of the study
period.
The authors placed their trust in multivariate analysis of the data
to determine whether or not treatment with statins reduced the occurrence
of the end-points. There must, though, be some concern about the validity
of the results. For example, the finding that statins reduced all cause
mortality by 31% compared with a reduction in the composite end-point of
myocardial infarction plus cardiovascular mortality of only 18% is, to say
the least, strange. Surely any effect on the composite end-point should be
greater than that on all cause mortality? Indeed, this finding encourages
the conclusion that the difference in all cause mortality between the two
groups is not so much a treatment effect as the result of the action of
other relevant variables.
The two groups compared in this study were as different as chalk and
cheese. In such circumstances, no amount of statistical trickery would
deliver any worthwhile conclusions.
References
[1] Wei L, Ebrahim S, Bartlett C, Davey PG, Sullivan FM, MacDonald
TM. Statin use in the secondary prevention of coronary heart disease in
primary care: cohort study and comparison of inclusion and outcome with
patients in randomised trials. BMJ 2005;330;821-4.
Competing interests:
None declared
Competing interests: No competing interests
Eddie Vos is correct in saying that women have yet to show a
mortality benefit, in published Statin trials. For example, in the 4S
study of 4444 subjects, the 444 women studied demonstrated a HIGHER crude
mortality when given Simvastatin versus placebo. This may be real or
misleading, for the statistical reasons which caution against making
unintended subset analysis. Women do however benefit significantly in
reduced cardiovascular events.
But Eddie goes on to say that the rationale for giving a Statin is
'high cholesterol'. This is no longer the case. ALL those patient groups
with sufficiently high coronary risk have been shown to benefit from
Statin prescription, in all the trials, in disregard of starting
cholesterol. Patients with low cholesterol benefitted to the same extent
as patients with high cholesterol. Benefit was related to the extent of
cholesterol-lowering, but it remains moot whether this is a marker for
dose, or a causal mechanism. Subsets selection of patients at low
coronary risk can be expected not to have sufficient numbers to reach
statistical significance. Young people, and women are at relatively low
risk in the short term, subject to the presence of other risk factors.
I conclude that ALL patients with SUFFICIENTLY HIGH CORONARY RISK,
REGARDLESS OF CHOLESTEROL, should have a Statin.
Dr Sam Lewis
Newport Surgery
Competing interests:
Cost versus Benefit
Competing interests: No competing interests
higher cholesterol live longer (ex: Schatz
IJ et al and Weverling-Rijnsburger
et al) either because cholesterol protects against infection, brain function decline or cancer, or simply being a marker for a well functioning biochemistry. It is therefore logical to find lower mortality, as do Wei et al, in an older group selected for having higher rather than
lower cholesterol, the prescribing rationale for statin use.
With no trial ever having even found a hint of a mortality benefit in women (2004: BMJ, JAMA, JACC), the mortality difference in the Wei et al study is indeed unlikely to result from lowered mevalonate, the rate limiting cholesterol precursor targeted by statin.
The suggestion by Wei et al that statins have a 'beneficial effect' in their cohort of an older and more female population is simply not provable by this kind of study, and is not supported, regarding all-cause mortality,
by the vast majority of clinical trials [EXCEL, PROSPER, ASCOT, ALLHAT, etc.]. vos@health-heart.org
Competing interests:
None declared
Competing interests: No competing interests
From time and again some stray comments are seen suggesting incidence
of stroke and AMI may go up following sudden withdrawal of statins and
resultant rebound dyslipidemia.Are there any such events from the subjects
comprising this or other statin trials?
Competing interests:
None declared
Competing interests: No competing interests
Still chalk and cheese...
Sir
Ebrahim et al. assert that I missed “the important point that they
[the patients in their study] have all suffered from a cardiovascular
event that makes them eligible for drug treatment”. Quite what they mean
by this is unclear. Presumably, they are attempting to argue that, in some
sense or other, this single characteristic of the study population is
sufficient to dispel any concerns stemming from the heterogeneous nature
of the patients, their disease and its management. Nonetheless, the
differences between the two groups that I summarised in my letter remain.
Furthermore, there is no reason to believe that the list of differences
ends here. By their own admission, the covariates used in the multivariate
analysis were limited by the data available and, of course, it is likely
that many more factors – known or unknown – were omitted.
Ebrahim et al, have misinterpreted my remarks about the outcomes of
their study. I did not claim that the reduction in all cause mortality
with statins was greater than the reduction in the composite end-point. I
merely pointed out that their results were rather strange. For example,
there was no evidence of a greater reduction in the risk of the composite
end-point than that of all cause mortality; this is contrary to what would
be expected both intuitively as well as on the basis of the results of
previously published, large-scale randomised controlled trials of statin
therapy. Similarly, the finding of a 31% reduction in all cause mortality
with statins is inconsistent with the results of other trials. Indeed,
many of the recent trials have failed to demonstrate any significant
reduction in all cause mortality. Hence, I stand by my assertion that the
results of the Tayside study raise questions about its validity.
The allegiance shown by Ebrahim et al. to multivariate analysis is,
of course, entirely in keeping with the current indulgent attitude to
statistics throughout medical research. Surely, though, when the two
groups are as different as chalk and cheese, and when the results are
anomalous, those with a sceptical disposition may, with justification,
draw attention to the air of statistical trickery?
Competing interests:
None declared
Competing interests: No competing interests