Antimicrobial anaphylaxis: the changing face of severe antimicrobial allergy

Abstract

Objectives

The epidemiology, clinical characteristics and outcomes of antimicrobial-associated anaphylaxis remain ill-defined. We sought to examine antimicrobial anaphylaxis with regard to: (i) the frequency of implicated antimicrobials; (ii) attributable mortality; and (iii) referral for definitive allergy assessment.

Methods

This was conducted through a national retrospective multicentre cohort study at five Australian tertiary hospitals (January 2010 to December 2015). Cases of antimicrobial anaphylaxis were identified from ICD-10 coding and adverse drug reaction committee databases.

Results

There were 293 participants meeting the case definition of antimicrobial anaphylaxis and 310 antimicrobial anaphylaxis episodes. Of 336 implicated antimicrobials, aminopenicillins (62/336, 18.5%) and aminocephalosporins (57/336, 17%) were implicated most frequently. ICU admission occurred in 43/310 (13.9%) episodes; however, attributable mortality was low (3/310, 1%). The rate of anaphylaxis to IV antibiotics was 3.5 (95% CI=2.9–4.3) per 100 000 DDDs and the rate of hospital-acquired anaphylaxis was 1.9 (95% CI=2.1–3.3) per 100 000 occupied bed-days. We observed overall low rates of hospital discharge documentation (222/310, 71.6%) and follow-up by specialist allergy services (73/310, 23.5%), which may compromise medication safety and antimicrobial prescribing in future.

Conclusions

This study demonstrated that a high proportion of severe immediate hypersensitivity reactions presenting or acquired in Australian hospitals are secondary to aminopenicillins and aminocephalosporins. Overall rates of hospital-acquired anaphylaxis, predominantly secondary to cephalosporins, are low, and also associated with low inpatient mortality.

Introduction

Anaphylaxis is a severe IgE-mediated hypersensitivity reaction that is rapid in onset and potentially fatal.^1–3 Recently published international studies demonstrated that anaphylaxis fatality and admission rates are increasing, in particular medication-associated anaphylaxis.^4–6 The rate of documented antibiotic allergies, including presumed anaphylaxis, in Australian inpatients receiving antibiotic therapy is estimated to be 18%.⁶ An Australian study by Liew et al.⁷ demonstrated that of all anaphylaxis fatalities over time, 57% have been attributed as drug-related, with penicillins and cephalosporins being the culprit drugs in 15.6% of drug-related anaphylaxis fatalities. The rate of hospital-acquired anaphylaxis to antimicrobials in Australian hospitals is unknown and the resultant impact on inpatient mortality, post-discharge allergy assessment and event recurrence is ill-defined.

The epidemiology of antimicrobial-associated anaphylaxis in a modern era of increasing utilization of broad-spectrum penicillins and cephalosporins also remains poorly described. On review of available literature, there are observable trends in antimicrobial prescribing and consumption in the community in Australia over the past 25 years.^8–10 It is evident that from 2013 onwards the three most commonly prescribed antimicrobials in the community have remained the same and now account for more than 50% of prescriptions; namely, cefalexin, amoxicillin and amoxicillin/clavulanate. In 2017, cefalexin was the most frequently prescribed antimicrobial in the community (20.1% of all antimicrobials prescribed; amoxicillin 19.7%; amoxicillin/clavulanate 17.5%). Overall, it has been observed that there has been a substantial increase in consumption over the last 20 years (1994–2014) of amoxicillin/clavulanate and cefalexin, whereas amoxicillin consumption has remained relatively similar. Of note, dispensing of β-lactamase-sensitive penicillins has remained low (<5%) when compared with systemic antimicrobial dispensing by class in these recent national reports (2013–17). Available data from the Australian Statistics on Medicines also reflects this, where prescriptions for phenoxymethylpenicillin (penicillin V or penicillin VK), for example, in 500 mg capsule form, have decreased from 230 981 prescriptions in 1997 to 11 288 prescriptions in 2010.^11,12 The potential impact of this change in prescribing on implicated antimicrobials in the epidemiology of anaphylaxis is unknown.

We sought to define the causality of antimicrobial anaphylaxis in Australia. Furthermore, we sought to examine antimicrobial anaphylaxis with regard to: (i) attributable morbidity (i.e. ICU admission); (ii) attributable (inpatient) mortality; and (iii) referral for definitive allergy assessment.

Methods

Setting and participants

An Australian retrospective multicentre cohort study was conducted at five university-affiliated tertiary hospitals in three different states between January 2010 and December 2015. Cases of antimicrobial-associated anaphylaxis were identified at each site via ICD-10 coding (at all hospitals), an approach that has been previously evaluated with high positive and negative predictive value for case identification of anaphylaxis.¹³ Adverse drug reaction (ADR) committee databases were also utilized. ADR committees receive referrals from hospital pharmacists from any referral source; this includes anaphylaxis and drug allergy (at Austin Health, Alfred Health and Sir Charles Gairdner Hospital only). Search terms for ADR databases included: anaphylaxis; anaphylactic shock; urticaria; angioedema; face/lip/oral mucosal swelling; wheals or hives; bronchospasm; wheeze; asthma; hypotension; and collapse. These search terms were used to help identify potential cases of antimicrobial anaphylaxis, which then underwent further assessment via predetermined inclusion and exclusion criteria to fulfil case definition (see below). Search terms from ICD-10 coding included T88.6 (anaphylactic shock due to adverse effect of correct drug or medication properly administered) and L50.0 (allergic urticaria). Medical records review of identified participants was undertaken to ensure that antimicrobials were implicated in each anaphylaxis episode.

Cases of antimicrobial anaphylaxis were defined as per the following criteria, adapted from the revised nomenclature for allergy from the World Allergy Organization definition:¹

• Moderate to severe immediate IgE-mediated reactions: (a) eligible participant in study period who is over 18 years of age; (b) development of itchy widespread rash, within 1–2 h of ingestion/administration of culprit antimicrobial, with or without development of associated hypotension, tachycardia, orofacial angioedema and wheezing; and (c) need of at least one dose of adrenaline, with or without concurrent administration of fluids and/or antihistamines [any of these treatments can be initiated en route to hospital—by GP, Emergency Department (ED) or patient—or at hospital].

• Moderate to severe accelerated IgE-mediated reactions: (a) eligible participant in study period who is over 18 years of age; (b) development of itchy widespread rash, between 2 and 6 h after ingestion/administration of culprit antimicrobial, with or without development of associated hypotension, tachycardia, orofacial angioedema and wheezing; and (c) need of at least one dose of adrenaline, with or without concurrent administration of fluids and/or antihistamines [any of these treatments can be initiated en route to hospital (by GP, ED or patient) or at hospital].

Exclusion criteria: (i) cutaneous-only reactions; (ii) delayed cutaneous reactions; (iii) patients with a history of idiopathic urticaria, angioedema or anaphylaxis; and (iv) patients with systemic mastocytosis.

Data collection

Patients identified with antimicrobial anaphylaxis meeting inclusion criteria had their demographics including age, gender, history of atopy and/or asthma, history of non-antibiotic-drug or antibiotic-drug allergies, comorbidities and immunosuppression status collected. The antimicrobial anaphylaxis episode was further characterized, including presenting clinical syndrome (rash, urticaria, angioedema, hypotension, respiratory and/or gastrointestinal symptoms); therapeutics (antihistamine, adrenaline or steroid therapy); and outcome data (length of hospital stay, ICU admission and inpatient attributable mortality). This was collated via a standardized case-report data collection form (see the Supplementary data available at JAC Online) by participating-site principal investigators. At selected participating sites (Austin Health, Alfred Health, Sir Charles Gairdner Hospital and John Hunter Hospital) a standardized cross-sectional patient survey post-antimicrobial anaphylaxis was performed via a phone questionnaire (see the Supplementary data available at JAC Online). To determine the rate of anaphylaxis in hospital-acquired cases only we used data from the participating hospitals on the antibiotic volumes, expressed as DDDs. We also obtained data on occupied bed-days (OBDs) as a commonly used denominator.

Definitions

Anaphylaxis was defined in the aforementioned inclusion criteria. Antibiotics were classified using the following terminology: penicillins encompassed penicillin VK, penicillin G, penicillin unspecified, flucloxacillin, dicloxacillin, amoxicillin, ampicillin, amoxicillin/clavulanate, piperacillin/tazobactam and ticarcillin/clavulanate; aminopenicillins included amoxicillin and ampicillin; and aminocephalosporins included cefalexin and cefaclor. Narrow-spectrum penicillins were further defined to include penicillin VK, penicillin G, flucloxacillin, dicloxacillin, amoxicillin and ampicillin. Narrow-spectrum cephalosporins included cefalexin, cefaclor, cefazolin and cefalotin. Narrow-spectrum β-lactams included both narrow-spectrum penicillins and narrow-spectrum cephalosporins. Atopy was classified as having an active or inactive history of allergic disease including allergic rhinitis, asthma and atopic dermatitis. Participants were deemed immunocompromised if they were on immunosuppressant medications, prednisolone dose ≥10 mg daily for >1 month biological therapy (within 6 months; 12 months if rituximab), or if they had been a solid organ transplant or HSCT recipient, or had active haematological malignancy, any cancer or asplenia. Attributable morbidity, specifically ICU admission, was defined as any patient who was admitted to the ICU as a direct consequence of their anaphylaxis episode. Attributable mortality was defined as death due to anaphylaxis or complications related to it during index hospital admission.

Statistical analysis

Basic statistical analysis via Stata was performed by principal and secondary investigators, with biostatistician support. Categorical variables were summarized using frequency and percentage and compared using a χ² test or Fisher’s exact test. Continuous variables were assessed for the normality of distribution using a Shapiro–Wilk test. They were then summarized using mean (SD) or median (IQR), as appropriate, and compared using a paired t-test or the Wilcoxon rank sum test. For statistical analysis to estimate a denominator for antimicrobial use, and rate of antimicrobial anaphylaxis in hospital, exact CIs were calculated using Stata 16.0 assuming Poisson distribution. A multivariable logistic regression model was used to identify predictors of ICU admission using R version 3.5.1. Multisite ethics approval was granted by the Austin Hospital Ethics Committee (HREC/16/Austin/144).

Results

Demographics

In total, 293 study participants were included, with a female predominance (178/293, 61%) and median age of 51 years (IQR 36–67 years). Age-adjusted Charlson comorbidity index (CCI) median score was 1 (IQR=0–3). History of asthma was recorded in 31 participants (31/293, 10.6%) and atopy in 57 participants (57/293, 19.5%). History of non-antibiotic drug allergy was recorded in 41 participants (41/293, 14%). Previous history of antibiotic drug allergy is elaborated on further in this results section. Twelve (4.1%) participants were immunocompromised.

A total of 310 antimicrobial anaphylaxis episodes were recorded—eight participants were noted to have more than 1 episode recorded (seven participants with 2 episodes and one participant with 3 episodes). Of the 310 episodes, 336 antimicrobials were implicated—there were 18 episodes in which 2 antimicrobials were implicated, 1 episode in which 3 antimicrobials were implicated and 2 episodes in which 4 antimicrobials were implicated.

Frequency of implicated antimicrobials

Figure 1(a) shows the frequency of antimicrobials implicated in antimicrobial anaphylaxis and Figure 1(b) shows the number of anaphylaxis episodes according to each antimicrobial class. Overall, amoxicillin or ampicillin was the most frequently implicated antimicrobial (62/336 episodes, 18.5%).

Figure 1.

(a) Anaphylaxis episodes per implicated individual antimicrobial. All implicated antimicrobials (n=336) for the 310 episodes of anaphylaxis are indicated. ‘Other’ includes cefuroxime, cefalotin, ticarcillin/clavulanate, meropenem, clarithromycin, famciclovir, norfloxacin, rifampicin, fusidic acid and fluconazole, all implicated in one anaphylaxis episode. (b) Anaphylaxis episodes per antimicrobial class. For definitions of antimicrobial class see the text. For the 310 episodes of anaphylaxis, implicated antimicrobials have been grouped with regard to their pharmacological class. Trimethoprim (eight episodes) is included in ‘sulphonamides’. ‘Other’ includes tetracycline (three), lincosamide (three), famciclovir (one), rifampicin (one), fusidic acid (one) and fluconazole (one).

Community- versus hospital-onset antimicrobial anaphylaxis episodes

Antimicrobial anaphylaxis occurred more frequently in the community than in hospitals [60% (185/310) versus 40% (124/310), P<0.0001]. Per participant, however, adrenaline was administered more frequently in the patients with hospital-onset compared with community-onset anaphylaxis [67.6% (198/293) versus 27.6% (81/293), P<0.0001]. Adrenaline was administered by ambulance officers in 74 (23.9%) episodes and by medical officers in 176 (56.8%) episodes. In the community-onset episodes, amoxicillin (50/185, 27%), cefalexin (40/185, 21.6%) and amoxicillin/clavulanate (27/185, 14.6%) were the most frequently implicated antimicrobials. In contrast, cefazolin (32/124, 25.8%) and ceftriaxone (20/124, 16.1%) predominated in hospital-onset episodes.

Multiple medications (including multiple antimicrobials or antimicrobial and non-antimicrobial drugs) were implicated in 71/310 episodes (23%). Of these, 13/310 episodes (4.2%) included non-antimicrobial drugs (anaesthetic agents, opioid analgesia, non-steroidal anti-inflammatory drugs or other). In these 13 episodes, cefazolin was identified as the implicated antimicrobial in 2/13 episodes (15.4%).

Rate of in-hospital IV antimicrobial anaphylaxis

We estimate that the rate of anaphylaxis to IV antimicrobials was 3.5 (95% CI=2.9–4.3) per 100 000 DDDs or roughly 1 in 28 000 daily doses of IV antimicrobials. The rate of hospital-acquired anaphylaxis was 1.9 (95% CI=2.1–3.3) per 100 000 OBDs. In examination of the two most frequently implicated antimicrobials, the rate of anaphylaxis to cefazolin was 7.3 (95% CI=5.0–10.3) per 100 000 DDDs or 0.6 (95% CI=0.4–0.8) per 100 000 OBDs. The rate of anaphylaxis to ceftriaxone was 14.2 (95% CI=9.0–21.3) per 100 000 DDDs or 0.4 (95% CI=0.3–0.7) per 100 000 OBDs.

Morbidity and mortality associated with antimicrobial anaphylaxis episodes

The median length of hospital stay was <24 h (IQR=0 to <48 h). Of 310 episodes, there were 135 admissions requiring hospitalization (135/310, 43.5%) and 43 admissions to the ICU (43/310, 13.9%). Thirty-nine (12.6%) episodes required adrenaline infusion, of which 23 (7.4%) warranted further management in ICU. Attributable mortality was recorded in 3 out of 310 episodes (1.0%). All three who died from antimicrobial anaphylaxis were immunocompetent and had no recorded history of antibiotic allergy, atopy or asthma. Two of the three episodes were hospital onset, where cefazolin was administered intra-operatively.

Predictors of ICU admission

Utilizing a multivariable logistic regression model, the major predictors for ICU admission per antimicrobial anaphylaxis episode were found to be multiple implicated drugs, requirement for adrenaline infusion and administration of antimicrobial by IV route only (Table 1).

Table 1.

Predictors of ICU admission

Predictor/variable	ICU admission		Adjusted OR (95% CI)	Pa
	no (n=267)	yes (n=43)
Age >65 years	74 (28%)	11 (26%)	0.77 (0.29–1.91)	0.582
History of atopy	53 (20%)	7 (16%)	0.53 (0.15–1.64)	0.301
Multiple implicated drugs	48 (18%)	23 (53%)	3.20 (1.36–7.57)	0.008
Hospital onset	88 (33%)	36 (84%)	1.83 (0.58–5.95)	0.304
Adrenaline infusion	16 (6%)	23 (53%)	10.48 (4.15–27.75)	<0.001
IV antimicrobial only	53 (20%)	31 (72%)	3.41 (1.15–10.38)	0.028
Cefazolin	17 (6%)	16 (37%)	1.55 (0.51–4.73)	0.434

Adjusted covariates included age, atopy history, multiple implicated drugs, onset location of anaphylaxis, adrenaline infusion and mode of administration of antimicrobial.

a

Statistically significant P values (<0.05) are shown in bold.

Prior antimicrobial allergy history in antimicrobial anaphylaxis

In total, 111 of 310 (35.8%) episodes had a reported prior antimicrobial allergy (IgE mediated or non-IgE mediated) or any ADR. Thirty-two episodes (32/310, 10.3%) had a reported previous anaphylaxis episode. Prior history of antimicrobial allergy to penicillins [66 episodes (66/310, 21.3%; 13/310, 4.2% aminopenicillins)] and cephalosporins [13 episodes (13/310, 4.2%; 5/310, 1.6% aminocephalosporins)] predominated.

Of 32 episodes in which a history of previous antimicrobial anaphylaxis was recorded, 5 had subsequent anaphylaxis in the study period to the same antimicrobial (15.6%). Nineteen participants had anaphylaxis to a β-lactam antibiotic in both episodes (19/32, 59.4%). The specific β-lactam antibiotics that were culprits for previous anaphylaxis and anaphylaxis during the study period are included in Table S1 (available as Supplementary data at JAC Online). Specifically, there were four episodes (4/32, 12.5%) that had a listed prior penicillin allergy (may have been penicillin G, penicillin VK or amoxicillin) and then subsequent anaphylaxis to cefalexin in the study period.

Hospital documentation and follow-up

In discharge summaries, on manual chart review from the hospital record, 222 out of 310 episodes (71.6%) were recorded explicitly as an anaphylaxis episode. Seventy-three out of 310 (23.5%) received a specialist allergy clinic referral. Of 222 episodes of anaphylaxis, 97 were formally referred to and reviewed by respective ADR committees (43.7%).

Patient phone follow-up questionnaire

Four participating sites contributed to the phone-call follow-up questionnaire, with a total of 258 participants. Within a maximum 10 year period, 10 participants (3.9%) received the same antibiotic since the index antimicrobial anaphylaxis episode and 6 (2.3%) participants had a further episode of anaphylaxis despite the majority of these patients (12/18, 66.7%) receiving written information from the hospital on discharge. Twenty-five (9.7%) participants had medical alert jewellery. Although not routinely indicated for sole drug allergy, 16 participants (6%) were provided with an adrenaline auto-injector (EpiPen^®).

Discussion

Our study explored the recent epidemiology and outcomes of antimicrobial anaphylaxis in Australia and demonstrated a ‘change in face’ of antimicrobial anaphylaxis. A key finding was that of aminopenicillins and aminocephalosporins being the most frequently implicated antimicrobial class in antimicrobial anaphylaxis. Despite a significant proportion of patients with antimicrobial anaphylaxis requiring ICU admission, there was a low attributable mortality rate. Hospital discharge documentation, ADR committee referral and specialist allergy referral rates overall were low and a significant rate of avoidable re-exposure to the offending drug noted.

The frequency of aminopenicillins implicated in this study is comparable to previous reports where penicillin allergy prevalence was at 9%–10%, a rate higher than that of other antimicrobial classes.^14–16 This likely reflects a change in the prescribing patterns, whereby aminopenicillins and aminocephalosporins have become more frequently prescribed antimicrobials in Australia.⁸ The observed changing prescribing patterns in Australia, and to some extent other parts of the world, may continue to have a major influence on observed trends in the epidemiology of antimicrobial anaphylaxis.

In an attempt to measure the rate of in-hospital anaphylaxis, including to the most commonly implicated antimicrobials cefazolin and ceftriaxone, we demonstrated overall a low rate of hospital-acquired anaphylaxis. Whilst the per-dose risk of anaphylaxis appears higher with ceftriaxone than cefazolin we caution against overinterpretation of these data. We provide these data as an indicative rate standardized to a measure of antibiotic use, DDD (cefazolin 3 g; ceftriaxone 2 g), which may not reflect the actual doses received.

Despite low attributable mortality, there was high morbidity associated with antimicrobial anaphylaxis; a significant proportion of participants required ICU admission and adrenaline infusion. Low mortality rates may be explained by a younger population with low age-adjusted CCI included in this study, in contrast to the previous finding of older age and prevailing cardiovascular disease being important risk factors for fatal drug anaphylaxis.² Significantly associated predictors for ICU admission included multiple implicated drugs, requirement for adrenaline infusion and IV administration of antimicrobials. Necessary precautions should thus be taken in patients with complex comorbidities who develop hospital-onset antimicrobial anaphylaxis, who are also administered multiple causal drugs simultaneously, to provide an early escalation of management in a more appropriate setting such as ICU.

The lack of documentation in participants’ discharge information and low rates of referral to an ADR committee and/or to a specialist allergy service is concerning. Increased training and knowledge of drug allergy is urgently required in specialist training programmes to address this deficiency.¹⁷ The absence of follow-up allergy assessment has implications for post-antimicrobial anaphylaxis medication safety and antibiotic appropriateness. For example, post-event assessment should scrutinize episodes of antimicrobial anaphylaxis for potential non-IgE mast-cell activation, especially in the 5% of cases where glycopeptides or fluoroquinolones are implicated—drugs recently identified to activate the mast cell receptor MRGPRX2 (15/293, 5.1%).^18,19 Three percent of our antimicrobial anaphylaxis cohort on follow-up via the phone questionnaire were re-exposed to the offending drug—this is completely avoidable with appropriate post-antimicrobial anaphylaxis assessment, testing and documentation. This highlights the need for an antimicrobial anaphylaxis model of care or a streamlined clinical pathway, incorporating a referral to specialist allergy services and a coordinated centralized risk communication approach to the patient and all care providers by linking hospital and community medical records.^20,21

It is widely known that cefazolin shares no R1 side chains with other penicillins and cephalosporins,²² and that this cohort frequently tolerates all other penicillins and cephalosporins.²³ Our study observed that cefazolin was implicated in one quarter of hospital-onset anaphylaxis episodes, due to its widespread utilization in perioperative prophylaxis, yet at a rate of only 0.6 (95% CI=0.4–0.8) per 100 000 OBDs. In this perioperative setting where cefazolin is implicated, frequently neuromuscular blocking agents may also be implicated; thus combined β-lactam and anaesthetic allergy testing is often recommended to be undertaken to establish a definitive culprit medication.^24,25 In an effort to apply this to our cohort, interestingly, we found that there were only 2 out of 13 episodes in which cefazolin was recorded as the implicated antimicrobial with a non-antimicrobial drug (anaesthetic agent, opioid analgesia or non-steroidal anti-inflammatory drug).^8,26,27

Some limitations exist in this study. The use of retrospective data and medical chart review are both subject to inherent information biases. However, there was a stringent effort to capture all possible cases at each participating site with broad ICD-10 classification, careful chart review and cross-referencing with ADR committee records to reduce such biases. Additionally, despite application of an explicit anaphylaxis definition, some of these cases may reflect non-IgE-mediated mast-cell activation. Inclusion criteria specifying ‘treatment with adrenaline’ may have led to some cases of anaphylaxis being excluded where treatment other than adrenaline might have been provided, despite being severe in nature. The strength of this definition, however, meant an increased confidence in case inclusion and anaphylaxis definition.

This unique national study demonstrates predominance of aminopenicillins and aminocephalosporins in overall antimicrobial anaphylaxis causality presenting or occurring in hospital. In hospital-acquired cases, cefazolin and ceftriaxone prevailed, albeit at a low overall rate (1.9 per 100000 OBDs) and associated with low inpatient mortality. The data are potentially generalizable to the wider Australian population and those in high-income countries. The recent change in the trend of antimicrobial anaphylaxis may be driven by the changes in antibiotic consumption patterns, in particular, with increase in the use of aminopenicillins and aminocephalosporins.⁸ The suboptimal rates of thorough documentation and specialist follow-up and/or testing is cause for concern and places patients at risk for further anaphylaxis episodes and morbidity. The findings from this study should act as a call for an increase in resources to support both a standardized reporting system for antimicrobial-associated anaphylaxis and mandatory review by specialist allergy services post-event. The utilization of a coordinated national registry for serious ADRs linked to hospital electronic medical records and with automated adverse report cards forwarded to treating community practices and patients would be a step forward in preventing unnecessary drug re-exposure.

Acknowledgements

This research has been previously published in a provisional form as a poster publication at the Australian Infectious Diseases Society Conference (New South Wales, 2017; poster 72) and at the Twenty-Ninth European Congress of Clinical Microbiology and Infectious Diseases (Amsterdam, The Netherlands, 2019; poster P2009).

We would like to acknowledge the following people for their contribution to this manuscript: Ms Kelly Cairns (Department of Infectious Diseases, Alfred Health, Melbourne, VIC, Australia), Ms Sharmila Khumra, Ms Misha Devchand and Ms Jane Booth (Department of Pharmacy, Austin Hospital, Melbourne, VIC, Australia), Ms Alicia Neels (Department of Pharmacy, University Hospital Geelong Barwon Health, Geelong, VIC, Australia) and Mr Jason Seet (Department of Pharmacy, Sir Charles Gairdner Hospital, Perth, WA, Australia).

Funding

This study was supported by a National Health and Medical Research Council (NHMRC) postgraduate scholarship (GNT 1139902 to J.A.T.) and a postgraduate scholarship from The National Centre for Infections in Cancer, National Health and Medical Research Council, Centre for Research Excellence (GNT 1116876 to J.A.T.).

Transparency declarations

None to declare.