Antibacterial prescribing and antibacterial resistance in English general practice: cross sectional studyCommentary: antibiotic resistance is a dynamic process
BMJ 2001; 323 doi: https://doi.org/10.1136/bmj.323.7320.1037 (Published 03 November 2001) Cite this as: BMJ 2001;323:1037
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.
Editor-The BMJ has now published two cross-sectional ecological
studies showing a modest but significant correlation between practice
levels of prescribing and antimicrobial resistance to amoxy/ampicillin in
urinary coliforms. (1,2) However, the conclusions drawn differ. Magee et
al seem to emphasise the seriousness of the problem while Priest et al
venture that the problem may be over-emphasised. What are GPs and policy-
makers to make of this? There are at least two reasons for thinking that
cross-sectional ecological studies such as these may underestimate the
true effect.
Firstly, the interpretation of ecological data is not
straightforward. Sometimes a weak relationship at an ecological level of
data obscures a stronger relationship at a lower level. (3) This may be
the case if, hypothetically, resistant UTIs are predominantly generated in
an epidemiologically distinct segment of the population (such as young
women living with young families). If much of the variation in measured
exposure (prescribing) between the units studied (practices) is because of
exposure in another group (such as the elderly) and practices differ in
their population make-up then a strong relationship may be obscured when
data is aggregated.
Secondly, after introduction of a resistant organism it takes time
for equilibrium (a stable relationship between prescribing and resistance
levels) to develop. We do not know how long this process takes so it is
difficult to guess what stage of the process we may be at. As Priest et al
acknowledge, their argument that a 20% decrease in prescribing might
result in a decrease in resistance of only 1% holds only if the resistant
population has reached equilibrium and the communities from which the
estimates are derived are independent. The closer we are to approaching
equilibrium the more time will have elapsed for mixing of people and
microbes and the less the independence assumption can be justified. This
may account for the larger correlations between prescribing and resistance
obtained using a larger unit of analysis (primary care groups) as the
assumption of independence may be more justified.
Of course it may be the case that both papers have got the right
answer, especially for antibiotics like amoxycillin which have been
heavily prescribed for many years and where resistance levels are already
high, but we cannot assume that the observed association between
prescribing of amoxy/ampicillin and resistance is generalisable to other
antibiotics. For newer antibiotics (such as the quinolones) we are at a
much earlier stage of the development of resistance so control of
prescribing now is likely to prevent or defer escalation of the problem in
the future.
In order to address these questions in a more valid way we need more
detailed longitudinal data where there is less biased sampling and better
characterisation of the individuals and groups within practices who
receive antibiotics and who develop resistance. We also need to supplement
our insights from other sources of data - such as European data - where
variation in prescribing (and resistance) is greater. (4) (5)
Stephen Conaty, Lecturer in Infectious Disease Epidemiology
Andrew Hayward, Senior Lecturer in Infectious Disease Epidemiology
Richard Morris, Senior Lecturer in Medical Statistics
Department of Primary Care and Population Science, Royal Free and
University College Medical School, Royal Free Campus, London NW3 2PF
Ben Cooper, Research Fellow
Department of Clinical Microbiology, Royal Free Hospital, London, NW3 2QG
References
1. Priest P, Yudkin P, McNulty C, Mant D. Antibacterial prescribing
and antibacterial resistance in English General Practice: cross sectional
study. BMJ 2001; 323: 1037-41.
2. Magee JT, Pritchard EL, Fitzgerald KA, Dunstan FDJ, Howard J.
Antibiotic prescribing and antibiotic resistance in community practice:
retrospective study, 1996-8. BMJ 1999; 319: 1239-40.
3. Morris RW, Whincup PH, Lampe FC, Walker M, Wannamethee, SG, Shaper AG.
Geographic variation in incidence of coronary heart disease in Britain:
the contribution of established risk factors. Heart 2001; 86: 277-283.
4. Kahlmeter G. The ECO.SENS Project: a prospective, multinational,
multicentre epidemiological survey of the prevalence and antimicrobial
sensitivity of urinary tract pathogens - interim report. J Antimicrob
Chemother 2000; 46(S1): 15-22.
5. Cars O, Molstad S, Melander A. Variation in antibiotic use in the
European Union. Lancet 2001; 357: 1851-3.
Competing interests: No competing interests
Editor - While enjoying the study reported by Priest et al (1) on
the relationships between antibiotic prescribing and bacterial resistance
in English general practice, I disagree with two of their three
conclusions.
Surveillance of routine isolates may not be the most sensitive of
techniques for establishing the subtleties of relationships between
antibiotic consumption and resistance but it does give a useful snapshot
of the worst case scenario which can be used in guiding empiric treatment
in appropriate patient populations and detecting problems of emerging
resistance. Given the general ease of availability of such data, I firmly
believe its analysis is worthwhile, while accepting that targeted
surveillance data also needs to be collected and improved laboratory
susceptibility testing methods used in our battle to control antibiotic
resistance (2).
The authors final conclusion, that trying to reduce the overall level
of antibiotic prescribing in UK general practice may not be the most
effective strategy for reducing resistance in the community is also not
warranted. There are several examples in the literature where just such a
strategy has been successful (3). The disappointingly small differences
in resistance rates between patient isolates in low and high prescribing
practices may well be more a reflection of individual patient selection
and widespread environmental pollution by antibiotics, resistant bacteria
and their resistant determinants. It is only by reducing community
prescribing that significant inroads will be made into the antibiotic
resistance problem for the majority of community resistant problems. The
efforts of general practitioners to reduce antibiotic prescribing are to
be applauded and should continue (4). Comparison with prescribing data
from Holland suggests there is further room for improvement (5). At the
same time high quality diagnosis of serious bacterial infection must
remain a priority so that there is no increase in untreated, invasive,
community acquired sepsis (4).
1. Priest, P., Yudkin, P., McNulty, C. & Mant, D. (2001)
Antibacterial prescribing and antibacterial resistance in English general
practice: cross sectional study. British Medical Journal 323 1037-
1041.
2. Gould, I.M. (2000) Towards a common susceptibility testing method?
Journal of Antimicrobial Chemotherapy 45 757-762.
3. Kristinsson, K.G. (2001) Mathematical models as tools for evaluating
the effectiveness of interventions: A comment on Levin. Clinical
Infectious Diseases 33 S174-S179.
4. Gould, I.M., Clarke, R., Hutchinson, S. & Davey, P (2001)
Variation in European antibiotic use. The Lancet 358 1273.
5. Cars, O., Mölstad, S., Melander, A. (2001) Variation in antibiotic
use in the European Union. The Lancet 357 1851-53.
Competing interests: No competing interests
We read with interest the article by Priest et al describing the link
between antibiotic prescribing and antibacterial resistance in English
general practice(1). We agree wholeheartedly for the call to use more
accurate data to estimate community levels of antibiotic resistance. The
results of a recent study in Christchurch New Zealand support this (2).
Using samples collected by GPs in the Christchurch Sentinel Network we
found a significant discrepancy between reported rates of resistance to
trimethoprim based on routinely collected data (19.0%) and our
representative sample ( 11.5%). Clinical behaviour is changing; the move
toward empiric therapy results in a greater proportion of specimens sent
being for complicated infections and treatment failures, thus creating
potential for a bias towards higher reported rates of antibiotic
resistance(3).
We also agree with the point made in the discussion by Priest et al
about potential sources of bias in cross sectional studies on prescribing
and antibiotic resistance. The link between prescribing and resistance is
two way - prescribing changes in response to reported local resistance
patterns - an apparent increase in resistance to an antibiotic may lead to
a decrease in prescription of that antibiotic. These kind of trends cannot
be captured in cross sectional studies, highlighting further the need for
systematic collection of accurate data on antibiotic resistance patterns
in the community and their relationship to prescribing.
Our experience indicates that it is essential that when apparent
increases in resistance are reported from routine data efforts are made to
eliminate sources of bias before recommending changes in prescribing
practice.
Derelie Richards, lecturer in general practice
Les Toop, professor of general practice
Department of Public Health and General Practice,
Christchurch School of Medicine and Health Sciences,
University of Otago,
PO Box 4345,
Christchurch,
New Zealand
no competing interests
References
1. Priest P, Yudkin P, McNulty C, Mant D, Wise R.
Antibacterial prescribing and antibacterial resistance in English general
practice: cross sectional
study. BMJ 2001; 323: 1037-1041
2. Richards DA, Toop LJ, Chambers ST, Sutherland MG, Harris BH, Ikram
RB, et al. Antibiotic resistance in uncomplicated urinary tract infection:
problems with interpreting cumulative resistance rates from local
community Laboratories. New Zealand Medical Journal 2001 (in press).
3. Livermore DM, Macgowan AP, Wale MCJ. Surveillance of antimicrobial
resistance. BMJ 1998;317:614-615.
Competing interests: No competing interests
Dear Sir,
there is increasing resistance following the suboptimal therapy
(erythromycin against streptococci).
But I see a much greater danger in not-treating streptococci. The
reemergence of post-streptococcal-reactive disease is induced by
horizontal-gene-transfer of pathogenic factors. But not treating a strep-
throat means a three percent probability of inducing post-streptococcal-
reactive disease like nephritis or carditis.
We should not laugh about flu, cold, influenza like illness. Prof. Dr. W.
H. Veil, Jena described the problems 62 years ago and I see his patients
in my practice. And thanks to Prof. Fleming I am able to help!!
1. W. H. Veil: Der Rheumatismus und die streptomykotische Symbiose.
F. Enke, Stuttgart, 1939
Sincerely Yours
Friedrich Flachsbart
Eisenacher Str. 6,
37085 Göttingen,
Friedrich.Flachsbart@t-online.de
Competing interests: No competing interests
EDITOR - Antibiotics are prescribed in primary care for a variety of
indications in different patient populations. Urinary tract infections
are relatively unusual because they are more liable to occur in younger
women, whereas the other main area of usage of beta-lactam antibiotics is
for chest infections (often in older men). The data presented by Priest
et al do not distinguish the target populations for the antibiotics
prescribed, and along with the 30 fold range in the sending of
microbiological specimens for analysis and the cross-sectional design of
the study it is hard to be sure what these results mean.
We audited the resistance of urine specimens sent for analysis in our
practice in 1994 and 1995 and at that time found substantial levels of
resistance to amoxicillin (20% to 50%). Although amoxicillin remains a
popular choice of antibiotic for other conditions (such as chest
infections) we have long since abandoned it for suspected urinary tract
infection. Total practice amoxicillin prescribing is therefore a very poor
marker for the antibiotic load (both type and quantity) on younger women
with urinary tract infections and it is not too surprising that the
correlation with amoxicillin resistance in urinary coliforms is poor.
The authors conclude that this is a very unreliable way to assess the
impact of changes in antibiotic prescribing on resistance levels in
general practice. What puzzles me is how they can then go on to suggest
that “ trying to reduce the overall level of antibiotic prescribing in UK
general practice may not be the most effective strategy for reducing
antibiotic resistance in the community”. They appear to be able to make a
conclusion on effective strategies without presenting any data at all on
the results of this strategy or any other comparative strategy.
No interventions are assessed in this paper and the study type (cross
-sectional survey) is unsuitable for drawing conclusions about causation
in either direction. The tendency for higher antibiotic prescribing to
induce resistance might be balanced by the tendency for reported
resistance in the specimens to reduce the likelihood of that antibiotic to
be prescribed!
Do the authors have other data from prospective studies comparing the
results of different approaches to reducing antibiotic resistance in the
community to support their final conclusion? I would be particularly
interested to know the evidence to support the use of 3 day courses of
antibiotics for urinary infections as opposed to 5 day courses for
example.
Competing interests: No competing interests
In their study of antibacterial prescribing and antibacterial
resistance in English general practice, Priest et al(1) conclude that
routine microbiological isolates should not be used for surveillance of
antibacterial resistance in the community. I would agree, since as Wise
reports in his accompanying commentary (2), the dynamics of the interplay
between prescribing and resistance are exceedingly complicated. In their
study Priest at al have laudably tried to keep things simple, but they
have not considered all of the factors that may account for antibacterial
resistance in uropathogens.
The first part of their study examined the relationship between
amoxycillin/ampicillin and trimethoprim resistance rates in coliforms and
the prescribing rates of these antibiotics. In urinary tract infections
the majority of coliform isolates are Escherichia coli and the
determinants of urovirulence in E. coli are well understood e.g. adhesive
fimbriae, haemolysins and siderophores such as aerobactin (3). Many of
these urovirulence factors are plasmid mediated and these plasmids
commonly also have determinants for antibiotic resistance (4). This
linkage of urovirulence genes to antibiotic resistance genes may result in
artificially high rates of antibiotic resistance that are not related to
selection following antibiotic usage. To clarify the situation what we
also need to know is whether the antibiotic resistant rates are the same
in isolates of E. coli that do not harbour plasmids with urovirulence
genes. Priest et al also state in their study design that since resistance
to ampicillin and amoxycillin can be due to production of beta lactamase,
the use of any beta lactam antibacterial could potentially select for or
induce this resistance. Consequently they also examined the association
between ampicillin and amoxycillin resistance and prescribing of all beta
lactam antibiotics. Their assertion is only partly true.
Most urinary
isolates of amoxycillin or ampicillin resistant E. coli produce
relatively narrow spectrum beta lactamases such as TEM-1 and will therefore
be susceptible to most of the other commonly used beta lactams in general
practice. What Priest at al should have also eaxmined is the association
of amoxycillin or ampicillin resistance with the use of non beta lactam
antibacterials. This is because plasmids from urinary isolates of E. coli
encoding resistance to amoxycillin and trimethoprim commonly also have
resistance determinants for other non â lactam antibiotics such as
tetracycline, chloramphenicol and sulphonamides (4). Use of these non-beta
lactams antibiotics may consequently select for resistance to amoxycillin
and ampicillin. Although the use of these other antibacterials may not be
that common in general practice, use in agriculture may be higher (5).
Finally I would make a plea for the use of the term coliform to be
discouraged. It is unclear from the study of Priest at al what exactly
they mean by the term coliform. The term is commonly used (almost
exclusively in Britain) to describe a group of bacteria that share some
basic phenotypic properties that are easily and cheaply recognised in a
routine diagnostic microbiology laboratory. The term coliform is not
synonymous with E. coli and the use of the term probably varies from one
laboratory to another. Coliforms are made up of a highly heterogeneous
population of different bacterial species. Isolates of E. coli, Klebsiella
sp., Citrobacter sp. Enterobacter sp. May all be described as coliforms.
Amongst these different bacterial species, rates of antibiotic resistance
may vary greatly, for example virtually all isolates of Klebsiella sp.
are resistant to amoxycillin, and isolates of Enterobacter sp. will almost
always be resistant to amoxycillin co-amoxiclav and cephalosporins.
Although E. coli is likely to be the commonest coliform isolated from
urinary tract infection, changes in isolation rates of these other
species may have significant effects on resistance rates of “coliforms” .
In the study of Priest at al it would be impossible to tell whether there
are unusual rates of isolation of these more commonly antibacterial-
resistant organisms.
References
1. Priest P, Yudkin P, McNulty C, Mant D. Antibacterial prescribing
and antibacterial resistance in English general practice: cross sectional
study. BMJ 2001; 323: 1037-1041.
2. Wise R. Commentary: antibiotic resistance is a dynamic process. BMJ
2001; 323: 1041.
3. Johnson JR. Virulence factors in Escherichia coli urinary tract
infections. Clin Microbiol Rev 1991; 4: 80-128.
4. Riley PA, Threlfall EJ, Cheasty T, Wooldridge KG, Willimas PH,
Phillips I. Occurrence of FIme plasmids in multiply antimicrobial-
resistant Escherichia coli isolated from urinary tract infection.
Epidemiol Infect 1993;110:459-68.
5. Standing Medical Advisory Committee on Antimicrobial Resistance. The
path of least resistance. London: Stationery Office, 1998.
Competing interests: No competing interests
Surveillance of Antimicrobial Resistance - does a broad brush paint a better picture?
Dear Sir
In their paper on "Antibacterial prescribing and antibacterial
resistance in English general practice: cross sectional study" (BMJ 232,
3rd November 2001) Priest et al conclude that routine microbiological
isolates should neither be used for the surveillance of antimicrobial
resistance in the community, nor for monitoring the outcome of changes in
antibacterial prescribing by general practitioners.
Whilst we agree that approaching individual laboratories for data on a few
antibiotic/organism combinations and linking this data to that on
prescribing would not be a cost-effective approach to surveillance, their
conclusion is over generalised, based as it was on a short duration study.
The key question is "does improved prudent prescribing slow down or
reverse the effects of antimicrobial resistance and how can this be
measured?" To answer this, surveillance has to involve sustained
susceptibility data collection for a given population over a substantial
time period, allowing the monitoring of trends and an investigation of the
relationship with prescribing over several years.
The automatic electronic mapping of laboratory sensitivity data onto
populations would enable better monitoring of levels of resistance for
both community and hospitals from local to national level. Complete
coverage across a given geographical area overcomes difficulties with ill-
defined laboratory catchment areas, and enables the use of population
denominators and comparison with existing public health data sets e.g.
prescribing data. A system has been developed (and piloted in Trent
Region) by the Public Health Laboratory Service Antimicrobial
Susceptibility Surveillance Unit to collect all routinely generated
susceptibility data from hospital microbiology laboratories with a view to
facilitating the above.
Such "broad-brush" information can then be used to identify areas that
need further investigation and it is at this point that sentinel studies
are required. The dynamics between hospital and community resistance
patterns mandate the need for better sampling protocols, and effects of
interventions in prescribing can be studied using this approach.
To solely initiate a series of sentinel GP practices would introduce bias
in terms of geographical area, environmental factors, caseload and
enthusiasm of participants, thus extrapolation of interventions carried
out in these different populations could be seriously misleading.
However, a sentinel practice is a valuable tool to focus on an area or to
investigate a given problem.
Yours faithfully
Dr Martin CJ Wale
Consultant Regional Epidemiologist and Head of the PHLS Antimicrobial
Susceptibility Surveillance Unit (AmSSU).
Mr Peter R Ridout
Manager, AmSSU
Dr Joan A Birkin
Scientist, AmSSU
PHLS Antimicrobial Susceptibility Surveillance Unit (AmSSU).
CDSC Trent
Public Health Laboratory
Queens Medical Centre
Nottingham
NG7 2UH
Correspondence to Dr Joan A Birkin at jbirkin@cdsctrent.phls.nhs.uk
Competing interests - None
Competing interests: No competing interests