Well after a little bit of a hiatus (exam-based once again) we’re hopefully back for some regular blogging. When having a look at things to write about it looks like the topic of allergy will be a bit of a theme over the coming few months, for 2 big reasons. Firstly, the 6th National Audit Project (NAP6) from the Royal College of Anaesthetists (RCOA) is due to report on the 14th May. This time NAP is looking at perioperative anaphylaxis and will no doubt provide a wealth of new information on this (fortunately) rare complication. Whilst we will look at that report in more detail when it comes out, the current focus is on the upcoming DALES project. DALES stands for Drug Allergy Labelling in the Elective Surgical population, and is the 3rd annual project from the Research and Audit Federation of Trainees (RAFT). Us lot at NWRAG will be helping to deliver the project in the North West and we are currently continuing to recruit interested trainees (get in touch here if you are interested). This study is looking at better understanding the problems of drug ‘allergy’ labelling in patients, with the inverted commas and slight raise of the eyebrows being of particular interest here. This is because the application of an allergy label is frequently applied to non-allergy reactions (for instance well recognised side effects) and can interfere with clinical practice and good patient care. The problems that this is causing, and how anaesthetists interact with such labels, still has some unknown quantities, hence the role of this study. As such, I wanted to use this blog post and follow up ones to explore the background of this topic in a bit more detail.
This week we will look at a paper that starts by aiming to answer one of the ‘so what?’ questions that can sometimes be thrown at research problems. In today’s paper the question relates specifically to penicillin allergy, and the challenge of “So what? We have other antibiotics available.” To me this seems a not completely unreasonable question, and I have clearly seen cases where a low threshold for giving another class of antibiotic was used because of a feeling that this is the safest thing to do. Why risk the horror of anaphylaxis when you can just give a macrolide? Just practice saying “Yes I know he said he was allergic to penicillin but I gave him a penicillin anyway, your honour”. This particular paper was cited by the RAFT team as an example of why this clinical question might have a further reach of impact that at first glance. The paper is entitled:
Health care use and serious infection prevalence associated with penicillin “allergy” in hospitalised patients: a cohort study.
The full paper is available for free here: https://www.ncbi.nlm.nih.gov/pubmed/24188976
What is it all about?
We have touched a little bit on the background problem above. The authors of this paper had also noted the high rate of dubious ‘allergic to penicillin’ labels that their patients had, quoting a group of their patients in whom only 2% actually had a true allergy when they underwent testing. The logic behind their concern was that a lot of these patients were subsequently receiving antibiotics such as fluoroquinolones or third generation cephalosporins, with a recognised link to healthcare associated infections, particularly C. Difficile. As such, they wanted to have a closer look to see if there was actually a demonstrable problem here.
What did they do?
The study was a retrospective analysis of patient ‘cases’ who were admitted over a 2 year period to hospitals using a specific health plan record system (Kaiser Permanente Healthplan) in Southern California (the exact number of hospitals isn’t stated). All patients with an active ‘penicillin allergy’ label at the time of admission were selected as the observation group. They were then matched to 2 control groups by means of identifying the admission category (e.g. cancer, orthopaedic), gender, age (+/-1 year), and admission time (+/- 1 month of each other). They only used the first admission with an allergy label in this period as the index case.
Once the cohorts had been created, information from the record system was used to find the outcome measures of:
Once the cohorts had been created, information from the record system was used to find the outcome measures of:
- Hospital length of stay (primary outcome)
- Antibiotic courses used
- Rates of C. Difficile, MRSA and VRE infection
What did they find?
Their total patient database included 462,225 patients, of which 51,807 (11.2%) had a penicillin ‘allergy’ label. They were able to match 51,582 (99.6%) of them to 2 matched controls, giving a total of 103,164 control patients.
When looking at their primary outcome, having a penicillin allergy label added 0.59 (95% CI 0.47-0.71) days to the hospital stay, being slightly more if you were female (0.68) compared to male (0.35).
The analysis of antibiotic prescription was a little harder to unpick. The ‘top 10’ list of antibiotics used showed that there was an increased use of ciprofloxacin, vancomycin and clindamycin in the allergy-label group compared to the matched controls. The increased rate of fluoroquinolones was quite notably significant at 25.3% vs 14.3% (p<0.00001). However, there wasn’t a huge use of penicillins in the control group, with other differences seeming to be made up by an increased rate of first-generation cephalosporins in the control group, and a smaller increased rate of third-generation cephalosporins.
Finally, the results for the prevalence of the specific infections were perhaps the most intriguing. In their results they noted that having a penicillin allergy label resulted in a risk ratio for C. Difficile of 1.234 (95% CI 1.156-1.317), for MRSA of 1.141 (1.071-1.317) and VRE of 1.301 (1.125-1.504). Interesting.
When looking at their primary outcome, having a penicillin allergy label added 0.59 (95% CI 0.47-0.71) days to the hospital stay, being slightly more if you were female (0.68) compared to male (0.35).
The analysis of antibiotic prescription was a little harder to unpick. The ‘top 10’ list of antibiotics used showed that there was an increased use of ciprofloxacin, vancomycin and clindamycin in the allergy-label group compared to the matched controls. The increased rate of fluoroquinolones was quite notably significant at 25.3% vs 14.3% (p<0.00001). However, there wasn’t a huge use of penicillins in the control group, with other differences seeming to be made up by an increased rate of first-generation cephalosporins in the control group, and a smaller increased rate of third-generation cephalosporins.
Finally, the results for the prevalence of the specific infections were perhaps the most intriguing. In their results they noted that having a penicillin allergy label resulted in a risk ratio for C. Difficile of 1.234 (95% CI 1.156-1.317), for MRSA of 1.141 (1.071-1.317) and VRE of 1.301 (1.125-1.504). Interesting.
Is it any good?
As is often the case with anything non-RCT, it is always easier to start with the criticisms, but I think there are a few good points to this study. Its size is one, as it looked at a large number of patients over a decent time span. Their choice of outcomes seems reasonable to answer their clinical question, and most of the study design seemed nice and straightforward so as not to add too much confusion as to how they got their results. I also think that there isn’t really a better way to run such a study, as clearly a trait such as an allergy label is not something that can be randomised or blinded. As such, my usual criticism about lack of randomisation etc. would seem a bit harsh.
In terms of problems with the study, I always have a little bit of concern about the patient selection in this type of study. The database they have worked with is from a single insurer, and I wonder what possible biases might arise from this starting point, especially as this American healthcare model is less well known to me. Is there a selection bias for patients here? Are there subtle financial incentives that might be unconsciously present, for example, around antibiotic prescribing? I don’t feel that I know the system well enough to understand how representative this cohort will therefore be. I feel similarly about the whole process of matching patients to a control, and especially when in this case they decided to match each case to two control patients without actually saying why. Did a one-to-one match not give as exciting results? If this matching process isn’t really explicit in its methods then it leaves me a little suspicious about possible statistical trickery.
In terms of problems with the study, I always have a little bit of concern about the patient selection in this type of study. The database they have worked with is from a single insurer, and I wonder what possible biases might arise from this starting point, especially as this American healthcare model is less well known to me. Is there a selection bias for patients here? Are there subtle financial incentives that might be unconsciously present, for example, around antibiotic prescribing? I don’t feel that I know the system well enough to understand how representative this cohort will therefore be. I feel similarly about the whole process of matching patients to a control, and especially when in this case they decided to match each case to two control patients without actually saying why. Did a one-to-one match not give as exciting results? If this matching process isn’t really explicit in its methods then it leaves me a little suspicious about possible statistical trickery.
Final Thoughts
Overall I think the study raises some very interesting questions, though I am not sure we can be completely confident about the conclusions that the authors have drawn. There does seem to be a difference in some quite important clinical outcomes, and the data would point towards the allergy label being a possible cause. This seems plausible, as there does seem to be a difference in prescribing practices of antibiotics between the two groups, and it would seem that the choice of antibiotics in the ‘allergy’ cohort C.Difficile-ogenic, for instance. However, as the authors note, the explanation for the MRSA incidence doesn’t have a clear mechanism, and perhaps the conclusions the authors are trying to draw aren’t any better supported than the conclusion that being allergic to stuff increases your risk of medical complications. That being said, the difference between reported and actual penicillin allergy that the authors describe (just 2% actually having a true allergy!) suggests that in may be the label in itself, rather than an actual impact of allergy. As such, I am more drawn to their conclusion that the widespread acceptance of ‘allergy’ labels could be resulting in harm to patients.
Well thanks for reading. I hope this article has, at the very least, got you even more interested in the upcoming DALES project. This paper has really got me wondering more about how the difference between allergy and intolerance may be more important to elucidate than it might first appear. Given how common this is in our daily practice, I am intrigued to find out more. So please let me know your thoughts and check out the websites below. Hopefully have another post coming soon to follow up on this theme further.
BW
Tom Heaton
RAFT - www.raftrainees.com
NWRAG - www.nwrag.co.uk
Image courtesy of unsplash.com
NWRAG - www.nwrag.co.uk
Image courtesy of unsplash.com
RSS Feed
