These elephants aren’t causal, either

It’s been a while since I’ve fired up the blog, and longer since I did a journal club, but this article from BMJ has gotten right up my nose – partially because of the article, and partially because it’s been doing the rounds in the news media (and as an aside, include the link, you broekens).

I really think this article is being quoted for entirely the wrong reasons (although to noble ends) and – because I enjoy being devil’s advocate even to myself, I’m going to discuss why. What the article is, however, is an interesting application of Big Data – and a salutary lesson in the importance of antimicrobial stewardship.

We are well aware than antimicrobial resistance is increasing – in Australia as well as overseas. This study has very commonly been widely reported in the media as being due to rising resistance (generally using the term “SUPERBUGS!!”)

So lets look at it.


Antibiotic prescription data was collected from the Clinical Practice Research Datalink – data from 700 primary care practices, containing records from over 14,000,000 individuals.  Prescription data is collected from practice software.

Episodes of treatment with a single antibiotic were collected between 1991 and 2012, and linked with indications for prescription recorded in the database.  Four conditions were selected, to represent common presentations to general practice – they were:

  • Upper Respiratory Tract Infections (URTI)
  • Lower Respiratory Tract Infections (LRTI)
  • Skin and Soft Tissue Infections (SSTI)
  • Otitis Media (OM)

Prescribing data was then interrogated for “treatment failure” – which for the purposes of the study was defined as (my emphasis):

  • Prescription of a different antibiotic within 30 days
  • Admission to hospital with an infection-related diagnosis within 30 days
  • Presentation to the emergency department within 3 days of prescription
  • GP referral to an infection-related specialist within 30 days of prescription

(Anybody who has listed to me talk for more than 5 minutes on anything to do with antibiotics can probably see where I’m going with this).


The results include some impressively bignums:

  • More than 4 million prescriptions for URTIs (38.6%)
  • More than 3 million for LRTIs (28.7%)
  • 2.5 million for SSTIs (23.4%)
  • 1 million for Otitis media (9.2%)

The headline figure is an antibiotic “failure” rate of 14.7% – by diagnosis:

  • 12% for URTIs
  • 16.9% for LRTIs
  • 12.8% for SSTIs
  • 12% for OM

And 94.4% of these “failures” were the result of a repeat prescription rather than a referral or ED presentation.

There’s also quite a bit of data of interest to people like me on which antibiotic was prescribed.

Interestingly, some of the media reports have focused on the finding that failure rates for second-line antibiotic prescription were higher than first line – more of that shortly.


Antibiotic resistance has increased.  This we know, data tells us so.

1 / 7 (yes, by the way, media, 14.7% expressed as a fraction is closer to 1/7 than 1/10) general practice prescriptions resulted in a repeat prescription.

Superbugs cause GP prescriptions to fail, right?

Whoa, Nellie!

Correlation is not causation.

xkcd: Correlation. By Randall Monroe. CC-BY-NC

Interpreting antibiotic use data is quite challenging.  Part of this reason is because documentation is terrible – one of the benefits of this study at least is that the data seems to be quite robust in terms of diagnoses recorded.

The data I mainly work with in my hospital stewardship program is pharmacy dispensing data – the number of tablets (or ampoules) that leave the pharmacy shelves.  There isn’t any reference to what the antibiotics are actually being used for. This is why our clinical audits are such an important part of a good stewardship program.

This study at least looks at a restricted set of diagnosis groups, so we have some idea about the why as well as the how much.

What it doesn’t cover is the appropriateness of treatment. We know that around one quarter to half of prescriptions are not in keeping with best practice guidelines for the treatment of the condition being treated.  That’s not to say the treatment is ineffective – indeed, I rarely find any patients not receiving effective treatment, but just not the most optimum treatment.

That last sentence is actually correct – the number of times I see a treatment failure due to antibiotic resistance is small. (I can’t quantify this yet, ask me in five years).

So what else could cause it?

Obviously there are limitations to the data, so that a patient could represent with a different infection and receive a second course of antibiotics for a new issue and be  considered a failure from the point of the trial.

In this study, however, I can actually see two elephants in the room.

Need For Treatment

Of the diagnostic groups selected, there is quite good evidence that two-and-a-bit of them do not actually require antimicrobial treatment.

From Australia’s Therapeutic Guidelines: Antibiotic:

The benefits of antibiotic therapy in pharyngitis, tonsillitis, nonsuppurative otitis media and sinusitis are limited. Routine use of antibiotics in these conditions should be avoided, to limit potential adverse effects and to reduce selection of bacterial resistance, both in individuals and in the community.

(pharyngitis, tonsillitis and sinusitis are all in the umbrella group of URTIs, non-suppurative OM is the less severe – and more common – end of the otitis spectrum).

It continues:

If aiming to prevent… complications or to shorten the illness, antibiotics are not mandatory. Valuable symptomatic relief is provided by paracetamol or nonsteroidal anti-inflammatory drugs. Antibiotics shorten the illness when bacterial causes are likely, by less than one day, and slightly reduce the risk of acute otitis media as a complication (a Cochrane review suggested that nearly 200 people would require treatment to prevent one case of otitis media). Therefore antibiotics may be indicated by balancing these modest benefits against possible harms (eg rashes and other adverse effects, and bacterial resistance).

Further more, a large and well-designed randomised-controlled trial late in The Lancet Infectious Diseases in 2012 found that amoxicillin (the most common antibiotic prescribed in the BMJ trial) for LRTIs other than pneumonia (and certainly a more common LRTI diagnosis than pneumonia) was no more effective than placebo.

So.  If antibiotics are not generally indicated or have been shown to be ineffective to treat a goodly number of these conditions, why are we then surprised than antibiotics are failing to improve things?

Diagnostic Error

Antimicrobial guidelines are diagnosis-based – if you suspect this problem, use these antibiotics.  It is worth noting that second-line antibiotics in this paper were more likely to result in “failure” – suggesting that where it is possible to do so, prescribing according to guidelines is a good idea.

What it doesn’t cover is the case when the diagnosis is actually incorrect.  This can be simple – “I think your sore throat is probably bacterial” (note that as above, this doesn’t mandate antibiotic use anyway) or more complex.

I’ll use the Skin/Soft Tissue Infections as an example.

The most common SSTI I see as an infection doctor is cellulitis – bacterial infection of the skin / subcutaneous tissues. Usually caused by Group A Streptococcus or Staphylococcus aureus.

Cellulitis – from – believed to be fair use

There are however, other causes of red legs –

Commonly, we see changes of varicose eczema (also known as lipodematosclerosis)in people who have swollen legs due to heart disease:

Varicose eczema – from – believed to be fair use

Clots in legs can also present similarly –

DVT – from WebMD – believed to be fair use

Or more simply, antibiotics could be given for a complication rather than the primary problem – and in the case of recurrent lower limb cellulitis, often we see that a patient will keep getting recurrent infections until the primary cause (usually tinea pedis) is treated.


I don’t have any firm figures for these speculations of mine.   But I’m not telling you that “1/10 prescriptions fail because of superbugs”.  What I am telling you is that a retrospective study of the represcription of antibiotics with a nebulous definition of failure, for a group of conditions that are generally self-limiting might not be the Superbugs are Destroying the World article it is being portrayed as.

What the article is, however, is a great pointer of why we need better antimicrobial stewardship – in the community as well as in hospital.

~ by Trent on September 29, 2014.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: