Safety of reduced antibiotic prescribing for self limiting respiratory tract infections in primary care: cohort study using electronic health records
BMJ 2016; 354 doi: https://doi.org/10.1136/bmj.i3410 (Published 04 July 2016) Cite this as: BMJ 2016;354:i3410
Chinese translation
该文章的中文翻译
All rapid responses
Our recent paper on the safety of reduced antibiotic prescribing for respiratory tract infections found evidence that the incidence of pneumonia might be slightly higher at general practices that prescribe fewer antibiotics. (1 ) We were asked by Public Health England to investigate whether there was evidence of non-linearity of association between the level of antibiotic prescribing and risk of pneumonia. We conducted additional statistical modelling to explore this question.
We fitted fractional polynomial models using the ‘mfp’ command in Stata version 14. The outcome was the occurrence of pneumonia. The proportion (%) of respiratory consultations with antibiotics prescribed was fitted as a predictor after dividing by 10 to obtain the relative risk associated with a 10-unit change in prescribing. The model was a Poisson model including the same covariates as reported in the paper, but for convenience we used robust variance estimates to account for general practice, rather than the random effects models reported in the paper. We compared the goodness-of-fit, parameter estimates, and fitted values, for a model fitted with the antibiotic prescribing proportion as a linear predictor, and for first and second order fractional polynomial (FP) models. (2) In the first order FP model, the square root transformation was selected as the best fitting transformation. In the second order FP, model cubic transformations were selected.
Data presented in the Table, show evidence of slightly improved goodness of fit from fractional polynomial models as compared with the linear model. Incidence rate ratios and their confidence intervals are also presented, but these are less readily interpretable after transformation of the predictor. The Figure presents the predictions for pneumonia from the three models. These reveal similar estimates from the three models. In the range of antibiotic prescribing that is frequent in the UK (30% to 70% of respiratory consultations with antibiotics prescribed), there is no evidence of meaningful non-linearity in pneumonia risk. The shape of the curve for the second order FP model, suggests that there may be no further reduction in pneumonia risk when antibiotic prescribing exceeds about 70% of respiratory consultations with antibiotics prescribed. However, the shaded area represents the confidence intervals for the three lines, these overlap and are wider at either low or high levels of antibiotic prescribing. We caution that the extremes of the curves are informed by fewer data points making it difficult to draw firm conclusions.
We conclude that modelling possible non-linearity of the association between antibiotic prescribing and pneumonia risk offers weak evidence of improved goodness of fit but the coefficients associated with transformed predictors are less straightforward to interpret. Graphical presentation of the fitted values from models of varying complexity reveals no evidence of meaningful non-linearity in pneumonia risk. We caution that these analyses were ‘post-hoc’ and were suggested through initial inspection of the main study results.
Martin Gulliford, Professor of Public Health, King’s College London
Dorota Juszczyk, Research Associate, King’s College London
Toby Prevost, Professor of Medical Statistics, Imperial College London
Table Legend: Comparison of goodness of fit and parameter estimates from fractional polynomial models.
Figure Legend: Graph showing predictions from a linear model compared with non-linear, fractional polynomial (FP) models. Navy line, linear model; red line, first order FP; green line, second order FP; shaded area, confidence intervals for three lines.
References
1. Gulliford MC, Moore MV, Little P, et al. Safety of reduced antibiotic prescribing for self limiting respiratory tract infections in primary care: cohort study using electronic health records. BMJ 2016;354:i3410.
2. Royston P, Ambler G, Sauerbrei W. The use of fractional polynomials to model continuous risk variables in epidemiology. Int J Epidemiol 1999;28:964-74.
Competing interests: No competing interests
Pouwel and colleagues engage in modelling. Not surprising as it is their bread and butter at Public Health England. That disembodied body called the "government" makes a political decision to reduce the use of antibiotics by x% by year ......
A patient presenting to a clinical doctor wants to be treated according to the doctor's clinical judgement. The best treatment for HIS condition at that time.
Neither the Minister in Whitehall, nor the Chief Medical Officer, nor their sub ordinates have examined the patient.
Some of the discussants talk of self-limiting conditions and of viral infections. There is also talk of new methods of determining the susceptibility of the organisms to antibiotics. All fine and dandy. Will Public Health England please tell us which practices in which towns are equipped for these tests? Will they also tell us whether ANY clinician is equipped with foresight to diagnose " Self-limiting"? Will Public Health England tell us whether No viral infections deserve antibiotic cover, regardless of the patient's immunity? There is, afterall something called " opportunist infection" by bacteria.
I acknowledge my lack of knowledge. If Public Health England, or the Minister or the CMO would care to demolish the above, point by point, I would be grateful.
Competing interests: No competing interests
In light of the increasing threat of antibiotic resistance, the UK government recently announced its ambition to reduce inappropriate antibiotic prescribing by 50% by 2020.1 When considering alternative strategies to achieve this ambition it is crucial to obtain information on the safety of potential reductions in antibiotic prescribing. It is reassuring that Gulliford et al. found that even a large reduction in antibiotic prescribing for self-limiting respiratory tract infections was predicted to be associated with only a small increase in numbers of infective complications (1.1 additional case of pneumonia each year for a 10% reduction in antibiotic prescribing for an average sized practice of 7000 registered patients).2
Intuitively, one might expect that it may be safer, in terms of complication rates, to reduce antibiotic prescribing among high-prescribing rather than among already low-prescribing practices. The former supposedly prescribe antibiotics unnecessarily more often, e.g. for viral or self-limiting conditions, whereas extreme low-prescribers might already miss several cases for which an antibiotic prescription would have been indicated. Hence, the association between antibiotic prescribing levels and infective complication rates may be non-linear.
Gulliford et al. suggest that this association is instead linear, as adding a quadratic term did not improve the goodness of fit of their model.2 However, because quadratic terms impose a simple global structure on the non-linear function of the antibiotic prescribing level (i.e. applying the same simple function to the whole range of antibiotic prescribing levels),3 it may be a poor approximation of potential true non-linearity. Figure 4 in Gulliford et al. indeed suggests that the association might be non-linear.2 Very similar incidence rate ratios were found for antibiotic prescribing proportions of <44% and 44-50% (1.00 vs. 0.98 [95% CI 0.83 to 1.15]) and 51-57% and ≥58% (0.74 [95% CI 0.63 to 0.87] vs. 0.70 [95% CI 0.59 to 0.82]), but with a relatively big step between the 44-50% and 51-57% categories.2
To better capture potential non-linearity, without having to impose arbitrary breakpoints or enforcing a too simple global structure, (penalized) spline regression or fractional polynomials could be used.3-5 Dependent on the true shape of the data, the sample size and the correlation between covariates, either (penalized) spline regression or fractional polynomials may perform better.4-5
Given 1) the political commitment to reduce prescribing, 2) the policy implications of whether there exists (relatively) safe zones which could be targets for antibiotic prescribing reductions, 3) it seems reasonable that reductions in antibiotic prescribing may be safer in high-prescribing rather than in low-prescribing practices, and 4) the apparent step-function in Figure 4 that suggests that the association between levels of antibiotic prescribing and infective complications may not be linear, we would be extremely interested to see further evaluations investigating whether this association is truly linear or whether a model using (penalized) spline regression or fractional polynomials would result in a better fit to the data.
1. G7 2016 in Japan: PM press statement. https://www.gov.uk/government/speeches/g7-2016-in-japan-pm-press-statement.
2. Gulliford MC, Moore MV, Little P, Hay AD, Fox R, Prevost AT, Juszczyk D, Charlton J, Asworth M. Safety of reduced antibiotic prescribing for self limiting respiratory tract infections in primary care: cohort study using electronic health records. BMJ 2016;354:i3410.
3. James G, Witten D, Hastie T, Tibshirani R. An introduction to statistical learning: with application in R. New York: Springer, 2014.
4. Strasak AM, Umlauf N, Pfeiffer RM, Lang S. Comparing penalized splines and fractional polynomials for flexible modelling of the effects of continuous predictor variables. Comput Stat Data Analysis 2011;55:1540-1551.
5. Binder H, Sauerbrei W, Royston P. Comparison between splines and fractional polynomials for multivariable model building with continuous covariates: a simulation study with continuous response. Stat Med 2013;32:2262-2277.
Competing interests: No competing interests
Dear Editor,
Nice to read the article of immense public health.
However, in regions of higher prevalence of streptococcal sore throat and rheumatic fever ( particularly in resource-limited settings) ; the prescription of an antibiotic is important to prevent acute rheumatic fever.Therefore the rules of the games may be different in different regions of the world. Moreover, in malnourished children with impaired immunity, it may be important to prescribe antibiotics, even in viral infections as they may become complicated with bacterial infections.
Competing interests: No competing interests
I am always interested in reading research about the complications of withholding antibiotics. The problem is for those of us who man the oars of the NHS each day seeing patients, if a GP doesn't prescribe an antibiotic and the patient is perceived to suffer harm that Gp is in trouble. We can all safety net and maybe on site CRP or procalcitonin may help but what is needed is a change in the law where is harm is caused or perceived to be caused there is no blame to the clinican trying to protect the general population by not prescribing. Otherwise each of us rows too close to the reef and one day will be smashed.
Competing interests: No competing interests
Self-limiting respiratory tract infections are extremely common, with most people experiencing one or more episodes each year. Antibiotic treatment is generally unnecessary, offering minimal or no benefit, while increasing the risks of drug side-effects and increasing antimicrobial drug resistance.[1] Complications are infrequent but when they occur require early detection and treatment.[2,3] Mathioudakis and colleagues [4] draw attention to the potential role of point-of-care testing in the early diagnosis of bacterial infections. This is consistent with the O’Neill review on antimicrobial resistance, [5] which identified the use of rapid diagnostic tests to confirm whether antibiotics are needed as one of seven strategies to reduce the overall demand for antibiotic treatment. However, as van den Bruel and Turner [6] have observed, there are practical obstacles to introducing these testing methods into clinical practice and, like all new health technologies, these testing methods require careful evaluation to ensure that intended benefits are realised.
[1] National Institute for Health and Care Excellence. Prescribing of antibiotics for self-limiting respiratory tract infections in adults and children in primary care. London: National Institute for Health and Clinical Excellence, 2008.
[2] Gulliford MC, Moore MV, Little P, et al. Safety of reduced antibiotic prescribing for self limiting respiratory tract infections in primary care: cohort study using electronic health records. BMJ. 2016;354:i3410.
[3] Petersen I, Johnson AM, Islam A, et al. Protective effect of antibiotics against serious complications of common respiratory tract infections: retrospective cohort study with the UK General Practice Research Database. BMJ 2007:335: 982.
[4]. Mathioudakis AG, Chatzimavridou-Grigoriadou V, Evangelopoulou E, Mathioudakis G. A. Serum procalcitonin to guide antibiotic administration for respiratory tract infections in primary care. BMJ [rapid response]
[5] The review on antimicrobial resistance chaired by Jim O’Neill. Tackling drug-resistant infections globally: final report and recommendations. London: Review on Antimicrobial resistance, 2016. Source: http://amr-review.org/sites/default/files/160525_Final%20paper_with%20co... accessed 21st July 2016.
[6] Ann Van den Bruel, Philip Turner. Forecast diagnostics for antimicrobial resistance (AMR). Oxford: NIHR Diagnostic Evidence Cooperative Oxford, University of Oxford, 2016. Source: http://amr-review.org/sites/default/files/Forecast%20diagnostics_AMR_1%2... accessed 20th July 2016.
Competing interests: No competing interests
Dear Editors,
Hundreds of thousands of patients hospitalized in various Departments of Greek public hospitals, every year, acquire severe nosocomial infections.
Greece ranks first in the EU in hospital-acquired infections. [3][1][10][13][14][15]
Up to 50% of microorganisms implicated appear resistant to all available antibiotics. [5]
Such elevated antibiotic resistance from dangerous superbugs was declared a 'national threat' to NHS systems abroad, which receive tourists or patients treated in Greek public hospitals and return, carrying and spreading, deadly infections. [6][7][8][9][11][12][14][15][16]
These multidrug resistant superbugs are already a national threat in Greece.
Medical tourism providers, patients, European healthcare officials, visitors, summer tourists, should be aware of these possible health hazards, acquired from Greek hospitals.
References
[1] http://cid.oxfordjournals.org/content/53/2/177.long
[2] http://www.bloomberg.com/news/2012-02-09/greek-doctors-battle-hospital-s...
[3] http://www.grreporter.info/en/greece_ranks_first_eu_hospitalacquired_inf...
[4] http://www2.keelpno.gr/blog/?p=1166&lang=en
[5] http://journals.lww.com/ccmjournal/Abstract/2012/12001/491___Hospital_Ac...
[6] http://www.telegraph.co.uk/health/healthnews/10681645/Antibiotics-nation...
[7] http://www.independent.co.uk/life-style/health-and-families/health-news/...
[8] https://www.mja.com.au/journal/2014/200/2/growing-burden-multidrug-resis...
[9] http://www.upi.com/Health_News/2014/03/07/Britain-to-address-increasing-...
[10] http://www.irishhealth.com/article.html?id=22009
[11] http://www.cbc.ca/news/canada/british-columbia/b-c-hospitals-on-watch-fo...
[12] http://www.theglobeandmail.com/news/british-columbia/hospital-beats-back...
[13] http://greece.greekreporter.com/2014/08/11/greece-first-in-antibiotics-u...
[14] http://www.abc.net.au/news/2015-06-16/hospitals-warned-over-new-antibiot...
[15] http://www.pharmatimes.com/Article/15-11-19/Post-antibiotic_era_draws_cl...
[16] http://www.theguardian.com/society/2015/dec/22/almost-too-late-fears-of-...
Competing interests: No competing interests
The research findings of Gulliford et al are very encouraging towards the reduction of antibiotic prescribing and its benefits in terms of less risk of emergence of antibiotic resistance and collateral damage. This evidence helps in increasing the threshold of using antibiotics in respiratory infections.
However, in order to optimise the use of antibiotics, it is important to look for recent or previous history of colonisation or infection with alert organisms i.e. bacteria that are known to cause healthcare-associated infections that are selected because of the use of antibiotics and in particular the broad-spectrum ones such as meticillin-resistant Staphylococcus aureus (MRSA), extended spectrum beta-lactamase producing organisms (ESBLs) and Clostridium difficile.(1) Patients with a history of such organisms should not be prescribed with antibiotics to treat self limiting respiratory infections.
The other point we wish to make is that once antibiotics are prescribed, take upper respiratory virology samples for rapid molecular biology testing such as polymerase chain reaction (PCR) to detect viral nucleic acid.(2) This investigation is helpful in stopping antibiotics once the test confirms the presence of a respiratory virus as a cause of the infection.
1. Public Health England. Health Matters: Antimicrobial Resistance; Dec 2015.
2. Public Health England.UK Standards for Microbiology Investigations, Respiratory Viruses:2014;1-43
Competing interests: No competing interests
In their interesting study, Gulliford and colleagues assessed safety of reduced antibiotic prescribing for self-limiting respiratory tract infections (RTIs) in primary care, in an extensive cohort study of 45.5 million person-years of follow-up(1). Investigators recommended policies aiming to reduce antibiotic prescription in patients without clinical features suggestive of bacterial infections.
However, the study demonstrated an 11% increase in risk of pneumonia associated with 10% decrease in the proportion of people receiving antibiotics, which was considered acceptable by the investigators given the low absolute event rates. Also, this study did not evaluate the risk of treatment failure leading to additional healthcare visits, hospitalisations, days off-work/study and prolonged symptoms, which may be significant, especially in cases of lower respiratory tract infections (LRTIs)(2). Moreover, it is well established that distinction of bacterial infections requiring antibiotics based on clinical features is challenging and inaccurate(3). Thus, alternative strategies should be investigated.
Procalcitonin, the prohormone of calcitonin, is released in different tissues in response to bacterial but not viral infections(4). A recent Cochrane review of 14 randomized controlled trials with 4221 participants assessed procalcitonin guided protocols to initiate antibiotics in acute respiratory tract infections and found procalcitonin could significantly decrease initial prescription rates (adjusted odds ratio 0.10 95% CI: 0.07-0.14), without affecting rates of treatment failure, mortality and recurrent infection(5). Furthermore, a diagnostic guidance recently published by NICE supports procalcitonin as a promising and cost-effective biomarker that might be used in clinical practice to differentiate bacterial versus non-bacterial infections in intensive care units or in patients presenting to the emergency department(6). Point-of-care procalcitonin assays are now available and their use to guide antibiotic administration for respiratory tract infections in primary care should be rigorously assessed with the aim to be introduced to clinical practice in the near future, as they may contribute to safe and significant reduction of antibiotic prescribing in respiratory tract infections in primary care.
References:
1. Gulliford MC, Moore MV, Little P, et al. Safety of reduced antibiotic prescribing for self limiting respiratory tract infections in primary care: cohort study using electronic health records. BMJ. 2016;354:i3410.
2. Currie CJ, Berni E, Jenkins-Jones S, et al. Antibiotic treatment failure in four common infections in UK primary care 1991-2012: longitudinal analysis. BMJ. 2014;349:g5493.
3. Holm A, Nexoe J, Bistrup LA, et al. Aetiology and prediction of pneumonia in lower respiratory tract infection in primary care. Br J Gen Pract. 2007 Jul;57(540):547-54.
4. Biju PG, Garg S, Wang W, et al. Procalcitonin as a predictive biomarker for total body irradiation-induced bacterial load and lethality in mice. Shock. 2012;38(2):170-6
5. Schuetz P, Müller B, Christ-Crain M, et al. Procalcitonin to initiate or discontinue antibiotics in acute respiratory tract infections. Cochrane Database of Systematic Reviews 2012, Issue9. Art.No.:CD007498
6. Diagnostic guidance 18. Procalcitonin testing for diagnosing and monitoring sepsis (ADVIA Centaur BRAHMS PCT assay, BRAHMS PCT Sensitive Kryptor assay, Elecsys BRAHMS PCT assay, LIAISON BRAHMS PCT assay and VIDAS BRAHMS PCT assay (DG18). NICE, 2015.
Competing interests: No competing interests
Re: Safety of reduced antibiotic prescribing for self limiting respiratory tract infections in primary care: cohort study using electronic health records
RAPID RESPONSE SUBMISSION TO THE BMJ REGARDING THIS PUBLICATION
Research
Safety of reduced antibiotic prescribing for self limiting respiratory tract infections in primary care: cohort study using electronic health records
BMJ 2016; 354 doi: https://doi.org/10.1136/bmj.i3410 (Published 04 July 2016) Cite this as: BMJ 2016;354:i3410
This paper is presented as evidence that reduction of antibiotic use in respiratory infections is safe. The reality is however completely different and the paper could be seen as a massive misrepresentation of the results of the research that had been done. This paper should in my opinion be used with medical students as an exercise in critical reading.
The title and stated objective should raise suspicions of bias: “Safety of reduced antibiotic prescribing for self limiting respiratory tract infections in primary care: cohort study using electronic health records” The introduction of the words “self limiting” presupposes the research and demonstrates that the paper was written with a purpose to “prove something”. The research that is presented demonstrates exactly the opposite to what was being looked for and still has been reported as demonstrating safety.
Research findings appear to be purposefully presented the wrong way around, i.e. negative findings first, positive findings later. “No increase is likely in mastoiditis, empyema, bacterial meningitis, intracranial abscess, or Lemierre’s syndrome”.
What follows is the main finding of the research “The adjusted relative risk increases for a 10% reduction in antibiotic prescribing were 12.8% (95% confidence interval 7.8% to 17.5%, P<0.001) for pneumonia and 9.9% (5.6% to 14.0%, P<0.001) for peritonsillar abscess. If a general practice with an average list size of 7000 patients reduces the proportion of RTI consultations with antibiotics prescribed by 10%, then it might observe 1.1 (95% confidence interval 0.6 to 1.5) more cases of pneumonia each year and 0.9 (0.5 to 1.3) more cases of peritonsillar abscess each decade”
The choice of the way to quote the increase in pneumonia and peritonsillar abscess in relation to “an average list size of 7000 patients” is curious in the extreme. Is it not more usual to quote findings “per 100,000 patients”? Quoted like this, the result is: “A 10% reduction in UK GP antibiotic prescribing for respiratory infections will result in 15.7 extra cases of pneumonia per 100,000 patients per year P<0.001”.
If the result is multiplied up by the population of the UK, this would read “A 10% reduction in GP antibiotic prescribing for respiratory infections will be expected to result in about 10,500 more cases of pneumonia each year across the UK”
Pneumonia is portrayed in the article by Dr Gulliford as though it were a minor inconvenience. (I quote) “General practices that adopt a policy to reduce antibiotic prescribing for RTIs might expect a slight increase in the incidence of treatable pneumonia”
Pneumonia is certainly not a minor illness. It carries a significant short and long term mortality (1,2) Estimates of death rates vary but appear to be between 1 and 5%. If the 5% figure is taken, Dr Gulliford would appear to have shown that a 10% reduction in GP prescribing is likely to result in 525 extra deaths in the UK each year from pneumonia. This could arguably be stated the other way around as “A 10% increase in GP antibiotic prescribing for respiratory infection in the UK might save 525 deaths each year”.
Looking at the British Lung Foundation website (3): I see that the UK is 21st from the top in the world for deaths from pneumonia. Of EU Countries, only Romania and Slovakia are higher. The UK is quoted as having an age standardised death rate from pneumonia of 213.9 per million. To put this into context, France has a rate of 65.6, Spain 75.9, Australia 61.5, Canada 77.6. It would seem to me to be highly likely that these Countries are achieving lower pneumonia death rates by early antibiotic prescribing.
I see that the funding body behind this research was “UK National Institute for Health Research Health Technology Assessment programme initiative on antimicrobial drug resistance” My understanding is this body was set up by Government with the specific purpose of reducing antibiotic use and might itself have unrecognised bias. Maybe it should join the tobacco, alcohol and formula milk lobbies on the “suspect funder’s list”
1) EurichDT, MarrieTJ, Minhas-SandhuJK, MajumdarSR. Ten-Year Mortality after Community-acquired Pneumonia. A Prospective Cohort. Am J Respir Crit Care Med 2015;192:597-604. doi:10.1164/rccm.201501-0140OC.26067221
2) EurichDT, MarrieTJ SandhuJK MajumdarSR.Risk of heart failure after community acquired pneumonia: prospective controlled study with 10 years of follow-up BMJ 2017;356:j413 doi: 10.1136/bmj.j413 (Published 13 February 2017)
3) https://statistics.blf.org.uk/pneumonia?_ga=2.28799868.1398801130.155652...
Competing interests: No competing interests