Dear Fellow Doctors,
I would like to take a moment of your time to discuss a point of medical practice which seems to remain controversial despite it being really common and a well-researched problem. I would like to tell you about the modern management of the humble throat infection. Yes, it is a really common, dare I say dull topic. However, it would appear that our profession has the wrong end of the proverbial stick when it comes to this quotidian diagnosis.
Firstly, let me introduce myself and tell you why my perspective ought to be relevant and enlightening to you. I am a small town doctor working in primary care and Emergency Medicine. I work in a beautiful part of remote northern Australia. I have an interest in paediatrics and specifically the acute care of Aboriginal kids. I see a lot of children with common throat infections – don’t we all! The reason I feel somewhat qualified to discuss this topic is because I work in a population where the complications of acute streptococcal infections still produce massive morbidity.
Our population have a burden of disease rarely seen in the developed world. Last year I saw 3 new cases of Sydenham’s chorea. I see children with acute rheumatic fever and carditis on a weekly basis. We are currently in the throes of an epidemic of post-streptococcal glomerulonephritis. You might say that pus is our bread and butter. The incidence of those dreaded complications of the common strept throat is astronomical in my neck of the woods. So that is why I want to chat to you about a few things – namely: throat swabs and antibiotics.
So, how do we decide who has a true “strept throat” and not just a virus?
Let me start on the practice of throat swabbing children with acute tonsillo-pharyngitis.
So, when do I swab a child’s throat? Easy – never. I can honestly state that I have not swabbed a child’s throat in order to make the diagnosis of “strept throat” in the last decade.
I have certainly swabbed a number of throats in that time – but only in order to satisfy the diagnostic criteria of the diseases which we worry about i.e. acute rheumatic fever or post-strept GN. To be clear, I really only swab them after the fact – that is after the diagnosis is clinically apparent and we really just need a microbiological sample to prove the case, satisfy the diagnostic criteria and provide information to the folk in the Public Health office and a few Microbiologists whom might be interested. This is not a tool for making routine management decisions.
I hear that in much of North America that the rapid throat swab test is used to decide which children require antibiotics. Well, that opens another whole can of worms – do they really need antibiotics? But before we get onto that vexed question – lets look briefly at the diagnostic strategy that is “throat swabbing” and why it just doesn’t make much sense.
Before we discuss the “diagnosis” of strept throat – please recall that many children have asymptomatic colonisation with group A streptococci – they are well. The background rate is between 9 and 14% (Link). This seems to surprise a lot of doctors when you point this out to them. We do prefer the binary answers i.e. positive swab equals disease, negative equals no-disease. To be clear; having a positive swab for GAS is NOT an illness. This is not a patient-oriented outcome. What we really want to know is which patient will get symptomatic relief if they have an infection that may respond to antibiotics – that is not the same as asking “who has strept?”.
So can we use clinical findings to work out who is likely to have a true strept throat, rather than a viral infection? This is well studied. There are a number of ‘clinical decision rules’ that we can utilise. For example, the modified Centor Criteria, McIsaac or the Fever PAIN score are easy to use – just a few questions and a cursory exam give you all the data you need to get a score. Here is how it pans out from the studies:
Lets look at the Modified Centor Score:
Kids get +1 point just for being kids[ aged 3 – 14 yrs] . Then if you have any of the other criteria – tender anterior nodes, exudate or swollen tonsils, fever > 100.4 (38 C) or the absence of a cough you are up to 2 points and you have an 11 – 17% chance of having a positive throat culture. Or basically – the same as our pretest guess based on what we know about carriage rates. So having a Centor score of 0, 1 or 2 is pretty good evidence that you are wasting your, and your patient’s time by doing a swab. If they had a score of 0 or 1 – then you probably wouldn’t be asking the question. The 3 year old with a runny nose and low-grade fever has a virus 99% of the time. Don’t apply this rule to those kids!
Incidentally if you have 3 Centor points then your rate of swab positive is about 1/3 and if you have 4 or 5 points – you are up to 50% – even money as they say. So a kid with a full house of clinical features suggesting strept pharyngitis has a 50:50 chance of having a + swab. Save your cash – and toss a coin instead.
So do swabs help predict response to antibiotics? This is a better question, which is also tougher to answer. Are they any better than simple clinical exam and history?
There is a recent large clinical trial out of the UK – the PRISM trial in the UK. [Little et al, BMJ Oct 2013] This trial compared 3 arms – each with a different strategy to target AB prescribing
(1) Delayed prescription advice – ie. given a script and told to wait 5 days to see if symptoms resolved
(2) Use of a Clinical score (FeverPAIN) – which is quite similar to the Centor score
(3) Using a RADT in pts with a clinical score of 3 or more (55 % strept positive rate).
The second and third groups had significantly improved symptoms, shorter duration of sxs, AND less antibiotics prescribed. The FeverPAin score seemed to help target AB prescribing and reduced symptoms. The use of RADT tests also did the same thing. BUT adding a RADT did NOT add anything to Fever PAIN clinical scoring when used in tandem. Therefore it would seem like a waste of resources to use a test where simple clinical exam will do the same job.
So we can use this to identify patients whom are unlikely to benefit from having a swab – i.e. the Centor 0 – 2 group will more likely have commensal strept than actual disease – so do not swab them. So what about the others…
This is where it gets fun. Having a high Centor score makes it more likely that you will have a positive swab test. It DOES NOT predict that you will have a more serious clinical course or be at higher risk of the rare but nasty sequelae [Reference]. However I have had in numerous discussions with folk who genuinely get extra-worried if their patient has a Centor score of 5. I have even seen a kid get a whack of IM ceftriaxone – purely because he had a high Centor score – not because he looked particularly septic!
Summary of Swabbing: It is:
– Test characteristics from Journ of PHC 2012 = +LR of 2.9 and a -LR of 0.04. So much better test for excluding disease in kids. In a low risk kid a positive may get you up to ~ 30% Strept + rate. recall that a swab result does not differentiate between a colonised throat and a true infection.
(Edit: the test characteristics vary widely depending on the type of assay you are using – newer ones are more sensitive, less specific)
– does not predict a better outcome +/- treatment any better than a simple clinical scoring tool and
– costs a lot more. [ I found prices from $40 – $100 on commercial sites.] The clinical score costs what a minute of your time is worth !
Even in the kid with all the features of Strept throat – there is not much evidence to suggest swabs can help predict who should get antibiotics. The question of whether or not antibiotics actually help is to be covered next….
So lets assume you have taken a look at a kid and clocked up a full suite of symptoms suggestive of strept (eg. large, kissing exudative tonsils, tender nodes and a fever without a preceding viral prodrome, coryza or cough.) The Centor scores suggest 50% rate of swab positivity. It certainly is not 100%!. Now we come to the big issue – to treat with antibiotics or not…. I am going to break this down into 3 areas for discussion.
(1) patient-oriented benefits i.e. pain and suffering,
(2) preventing suppurative complications : Quinsy, otitis media, sinusitis, cellulitis / impetigo
(3) preventing non-suppurative complications [i.e rheumatic fever, PSGN].
Let’s start with the most likely actual potential benefit of antibiotics.
What do antibiotics do for kids with sore throats in terms of symptoms?
This question should be easy enough to answer – search the Cochrane database. The reviews there are huge – and they include papers from the last 50 years. So a heterogeneous group of patients, actually there are surprisingly inadequate trials in kids specifically. There are not many recent trials – so we may be seeing a bias – with antibiotic resistance emerging and improved hygiene etc in the last 50 years.
The simple summary of the data is:
– antibiotics reduce pain after 3 days. This effect was seen in subjects whom were swab positive. About half of non-treated patients were improved by day 3, more in the AB groups. The NNT was about 6 for “pain reduction at 3 days”.
– by the 7 day mark the difference was much less with an NNT of 21 for pain resolution by day 7. That is most of the untreated kids were better, the gap between the treated and untreated cohorts was smaller at one week.
OK, excellent… so what were the downsides? Unfortunately a lot of these trials do not report harms accurately. So if you want to know exact rates – it is tricky. The common adverse effects attributed to penicillin and similar antibiotics commonly used for sore throat are: diarrhoea, nappy rash, cadidiasis, skin reactions and the big one – anaphylaxis. Studies looking at the common, minor reactions would put the NNH [number needed to harm] between 6 and 9. Anaphylaxis is a bit more bothersome – though rarer – maybe 1 in 1,000 to 1 in 40,000 depending on the paper you believe. It is rare.
To be honest – if you are having a risk vs. benefit chat with the parents about ABs – then anaphylactic reaction does seem a bit over the top. I just don’t think that it really weighs on a parent’s mind enough to change the way they feel about “fixing” their sick child ASAP.
So here is how I would discuss the risks and benefits of antibiotics in a kid with a high Centor score / clinical impression of strept:
“Jimmy has a nasty throat infection. Based on how he looks, his age and other symptoms I would estimate that there is a 50:50 chance that he has a “true strept throat” which antibiotics have been shown to help. When we say ‘help’ – we mean that he is more likely to have less pain etc in 3 days. For every six kids we treat with antibiotics, one will get this benefit. The other 5 will not. However, there are a number of other ways we can reduce Jimmy’s pain without using antibiotics – and they work a lot faster. On the downside, roughly the same number of kids that we treat with antibiotics will get a side effect – like diarrhoea or a rash from the antibiotics. So I reckon it is a close balance between risks and harms. Would you like to hear more about antibiotics or I can give you a plan for managing Jimmy’s pain and fever?”
I would not even delve into the murky waters of the benefits of antibiotics for more serious sequelae – these are so rare and the evidence is so sparse that it would not be a meaningful discussion. If the family want to discuss problems like quinsy or rheumatic fever then I would imagine they are the type of folk whom are going to insist on at least a “wait and see prescription” – which seems fair enough. I would however make it clear that the ABs will do very little for their child’s symptoms in the short term – and insist they use a good analgesia plan. My favourite line in this scenario: “So are you specifically wanting antibiotics or for your son’s symptoms to get better?” It really helps parents think about what they want for their kid. We need to separate the concept of ‘reducing suffering’ from ‘antibiotic prescription’.
There are a few papers out there that suggest the simple NSAIDs give more rapid and significant relief of throat pain. This review of the data from the Brit Journ of GP in 2000 suggests that ibuprofen and paracetamol are both efficacious, quick and well tolerated. This seems like a reasonable option – paracetamol and ibuprofen are omnipresent in parent’s medicine cupboards – so I am sure most will have already dosed their munchkins – just need advice on how to do this safely and effectively.
I am a fan of a “dose of Dex” for the patient with a nasty sore, swollen throat – as a Paeds Anaesthetic doc – we give a lot of Dexamethasone for prophylaxis of swelling, nausea etc – it seems to be good. So is there much data on using steroids for sore throat? Well – yes… and no. There have been a series of trials looking at steroids – but they all gave antibiotics concurrently – so tough to say if they are testing steroids, or a combination effect of ABs and steroids. The Cochrane folk did a review in 2012 – and concluded that the addition of steroids (oral, IM..) roughly tripled your chanced of being symptom-free in 24 hours. So I do use these, but would love to see the trial of Dex vs placebo vs dex + ABs vs placebo. Dex is free in Australian EDs – it tastes better than prednisolone – but do what you like!
Can we prevent local purulent complications of pharyngitis / tonsillitis?
Lets look at the dreaded local suppurative complications of tonsillitis – quinsy, otits media, sinusitis, cellulitis / impetigo. Do we have any good data on this ? Yep – the Brits recently published a huge prospective clinical cohort study with over 14,000 patients! Check it out here: Predictors of suppurative complications from acute sore throat in primary care Brit Med Journ Nov 2013. And here are some raw numbers for you…. have a guess – how common are these?
Of the 13,288 patients they have data on for complications: 47 developed quinsy (0.35%), 38 got sinusitis (0.29 %), 69 had acute otitis media (0.52 %) and only 20 (0.15 %)had cellulitis / impetigo at follow up. So if you take all comers – about 1.3 % will get a suppurative complication – regardless of the antibiotic prescription. The numbers are tiny! I am sure most doctors would guess these rates would be closer to 10 or 20 % if asked.
So any benefit is going to be tiny in absolute terms. The NNT for preventing OM or sinusitis was 193 in a Lancet Infectious Dis 2014 paper from the same data set. Can you sell a reduction in otitis media from 0.6 % down to 0.5 %?? It is just not worth discussing.
Actually if you look at the raw data the rate of quinsy actually went up in the antibiotic groups – from about 0.2 % to roughly 0.4 % [a 100% increase! – see recent blog post on the evils of relative risks.] I am sure this was biased by the GPs tending to prescribe ABs to folk with really bad looking tonsils that were ‘near quinsy’. In this older paper from the Brit Journ of GP 2007 – only about a third of patients diagnosed with quinsy actually presented with a sore throat prior to being diagnosed with a quinsy. So even if antibiotics were effective [maybe they are not?] – you will not see two-thirds of the patients anyway – they will just rock up with a hot, red, angry quinsy de novo – and you should drain it – not so much use antibiotics then! [Oh, use an ultrasound to drain it – very cool.]
So – for suppurative complications the summary discussion ought to be – the risk of another local infection is very low – about 1 in a hundred. Antibiotics may reduce this a bit, but it is not really clinically significant. The preferred strategy here should be to safety net and ask patients to come back for review if they develop new symptoms. Even then – are we going to treat otitis media or sinusitis any more so with antibiotics?
What about rheumatic fever and glomerulonephritis? Can we prevent these?
OK – now onto the really rare complications – the “non-suppurative” ones. Namely acute rheumatic fever and post-strept glomerulonephritis. As I mentioned earlier – we here in tropical Australia are sadly world experts on these diseases. They remain relatively common despite ours being a very rich country with a decent, socialised health care system. It just is not too hard to get some antibiotics in most places, and for free! This is the part where I might get a little controversial as there is a lot of political and medical debate about how we should deal with these diseases. But to cut a long story short – I am not convinced that antibiotics are going to prevent these problems.
First lets look at a real, first world population – western Scotland . The incidence of rheumatic fever in the developed world is low, very low, super-super low! In 1985 (30 years ago) this paper in Scotland [Howie, Journ RCGP, 1985] estimated it would take 12 GPs working a lifetime to identify one single new case of acute rheumatic fever. That is cool: in a medium-sized town – only one single case of ARF would have been diagnosed since that paper was published!
Now in my part of the world acute rheumatic fever is common. We drill the Jones criteria into our RMOs and medical students. They need to be able to pick it clinically. It is common enough. We have a set of National Guidelines which you can read. The 2012 version is here.
We throw a lot of antibiotics around – the unofficial policy is to treat any Aboriginal kid with a febrile, sore throat, ears ache or skin infection aggressively. There is a really proactive immunisation system – with great rates of coverage – over 95% in most communities. We have been doing this for a few decades now… and yet we are still seeing these complications of streptococcal disease. Why is this? Why are these kids still getting sick despite the marvels of modern medicine?
Well – lets take a quick history lesson in streptococcal disease. These diseases were common in Australia and other first world countries a hundred years ago. In the early 20th century the rates of rheumatic fever (or scarlet fever) began to decline in urban, developed countries. Here is a typical graph (this one from the UK). Note that the steep part of the decline occurred between 1860 and 1900.
So -this makes it unlikely that antibiotics are the cause of this dramatic decline. In fact – antibiotics had very little impact of the rates of disease or morbidity when they were introduced towards the middle of the 20th century.
McKinlay & McKinlay out of Harvard wrote this paper: The questionable contribution of medical measures to the decline of mortality in the United States in the twentieth century. – it included streptococcal disease. I have taken this table from the text and underlined the streptococcal stats for you: as you can see the impact of penicillin between 1946 and 1973 was almost exactly ZERO.
OK. So why did all those nasty diseases go away? Well, the prevailing wisdom would suggest that a change in the “social determinants of health” occurred. Rheumatic disease is the result of a chronic exposure to strept antigen in a susceptible host. Streptococci thrive in overcrowded, malnourished people with poor access to basic hygiene (running water, washing facilities, refuse disposal). This is what we see in the north of Australia – these are diseases of poverty – it is unusual to see rheumatic disease in well nourished, suburban Aboriginal kids.
What happened in that 50 years from 1860 – 1910? People got running water, toilets and the average size of the household declined. Folk were able to wash regularly, nutrition generally improved. So I would say that the decline of these diseases was more the work of plumbers, builders and town planners rather than doctors and microbiologists.
So should we really be worrying about all of those kids with sore throats?
There is a large body of evidence to suggest that rheumatic heart disease are associated with pyoderma (skin infections) rather than pharyngitis. [McDonald, Clinical Infect Diseases 2006] and as such – treating children with throat infections ay be a wildly misguided strategy in the first instance. Currie et al also found a strong relationship with impetigo and subsequent post-strept GN disease. Pharyngitis just does not seem to be particularly common in these kids. In fact in my time in the tropics – I cannot recall a single case of “non-suppurative” disease in which the doctors were able to identify a clear, antecedent sore throat. We just do not see this! Although it is common to do swabs and return a “positive” – but as we know – that is expected in a good number of these kids – based on th known carriage rates.
The majority of the antecedent infections that cause non-suppurative complications fly under our radar – they are not seen by doctors, and hence do not get treated.
For example during the outbreak of ARF in Salt Lake County, Utah in the mid-80s it was noted that at least two thirds of the cases reported no clear antecedent sore throat. [NEJM, Veasy et al 1987] And there was no change in the community rate of acute pharyngitis to suggest an epidemic. They didn’t mention the rate of antecedent pyoderma!
The data around primary prevention of non-suppurative complications is scant and of low quality. The guidelines and protocols all link together into a mess of counter citation – the best original data comes from the US Air Force trial in 1949 by Denny et al in JAMA 1950. The NNT from the only data set I can find is 42 for ARF.
1634 US airforce trainees were included. The trial took place at a camp in Wyoming 1949. Airmen with acute sore throat AND exudative tonsils were randomised by their dog tag numbers (odd vs even). The intervention was IM depot penicillin (dose was changed mid-trial) in peanut oil! Outcome was the development of ARF as per the traditional criteria (Jones,1944). So roughly 800 in each group The rate of rheumatic fever in the non-treated group was 23/804 (including proabale ARF diagnoses) = 2.86% That is close to the 3 – 5% that gets quoted in most Aboriginal surveillance papers. So a high risk cohort! 4/798 in the treated group got ARF = 0.5%. So an absolute risk reduction of 2.36 % or an NNT of ~ 42 (magic number – same as tPA for STEMI).
A metanalysis of trials looking at primary prevention in high risk groups was published by Kerdemelidis in Paed Inf Dis Journal 2009 – it included 6 low quality trials, spread over a large time period with only 1 using penicillin. Not exactly the stuff we would accept as proof for any other intervention in modern medicine.
So – in summary – for preventing acute rheumatic fever and PSGN…
– You’ll need to wait a long time to see a case in any first world practice – it is just really rare.
– Antibiotics seems to reduce the rate of these complications – but we are talking about a near-infinite NNT!
– There is a good amount of data that the infections that lead to ARF etc are (1) subclinical / not coming to our attention, (2) more likely skin infections than throat infections and (3) the result of chronic host-antigen interactions as a result of chronic exposure.
– This is a disease of poverty. If we want to prevent it we need primordial measures – better housing, running water and access to good food. Doctors and antibiotics (i.e primary prevention) are unlikely to help.
[NB: secondary prevention with long-term antibiotics in children whom have confirmed ARF does benefit that high risk cohort. The Australian Heart Foundation makes it clear: “Secondary prophylaxis with regular benzathine penicillin G (BPG) is the only RHD control strategy shown to be effective and cost-effective at both community and population levels.”]
How do I practice?
I live and work amongst some of the sickest kids in the developed world. I would really love to be able to prevent as much of the morbidity that I see every day. And yet… I do not believe that aggressive investigation or antibiotic use will achieve this goal.
I have two strategies for two populations:
(1) For most well kids from an affluent background (i.e. the vast majority of kids you will all be seeing). This group does include a large number of Aboriginal children whom live in modern homes with good general health and no chronic diseases.
In children with the usual background, “developed world” risk there appears to be no real basis to prescribing antibiotics with the aim of preventing suppurative or non-suppurative complications.
There is clinical equipoise when it comes to using antibiotics for “symptom reduction”. The NNT was ~6 on day 3. However, there are more effective and less harmful medications which will achieve this goal faster. I only prescribe antibiotics after a realistic discussion about the likely risk and benefit has been had, and failed to convince the parents. I routinely tell them to stop the antibiotics as soon as their child is improved. Why run the risk of more adverse effects?
I give a written analgesia program – with doses and suggested times. I see a lot of under-dosing of pain meds in this scenario.
I offer a stat dose of Dexamethasone (0.15 – 0.3 mg/kg) to kids with nasty swelling, bad pain or swallowing difficulty.
(2) For kids from communities where there is a well documented high incidence of rheumatic disease, poor living conditions, strept skin disease.
Many of my colleagues opt to give antibiotics to all Aboriginal kids with a sore throat – and I think this is reasonable. If you can identify a “high risk” cohort – then the equation tilts in favour of treatment. I would argue that even within this group – we can further define the “at risk” kids by looking for signs of chronic illness, malnutrition and recurrent skin disease [scabies, impetigo etc]. Asking about the household – how many live there, diet, and the access to hygiene facilities etc. Indeed most guidelines make recommendations based on ethnicity – which is a pretty blunt tool for picking ‘at risk’ children.The evidence here is almost non-existent. We are making major extrapolations about the efficacy of primary prevention with antibiotics.
Given the equipoise for symptom relief – I reckon the balance is tipped towards treating in this group – as there is also a theoretical chance that we might actually reduce the incidence of rheumatic disease – not a lot of evidence though. Swabbing these kids is not pragmatic – most of these communities have lab times that run into days or weeks- so the call needs to be made on the risk and clinical score – not on swabs.
So that is why I do not swab, treat symptomatically and rarely prescribe antibiotics for an acute sore throat. In the very high risk children I treat empirically – based on the background risk and a very remote evidence basis. After all penicillin is not too toxic!
If we really want to improve the lot of our patients – and significantly reduce the burden of streptococcal disease – then we need a systematic approach to “primordial prevention”. We need better houses, schools, nutrition and access to the basic services that we all take for granted in the First world. We need plumbers, not doctors!
As always – I would be very happy to hear your thoughts. There is a huge diversity of practice all over the world when it comes to treating kids with sore throats. I hope that my perspective from the coal-face of post-strept sickness will give you a new perspective in your own practice.