The effectiveness of interventions that support penicillin allergy assessment and de-labelling of adult and paediatric patients by non-allergy specialists: A systematic review and meta-analysis

Abstract


Introduction
Approximately 6% of the general population (West, Smith et al. 2019) and 15% of hospital inpatients have a record of penicillin allergy (penA).(Macy and Contreras 2014, Trubiano, Smibert et al. 2018, Powell, Honeyford et al. 2020) Penicillin-based antibiotics are first-line treatment for many infections but patients with penA labels are usually treated with second line antibiotics (Powell, Honeyford et al. 2020) which are often more costly, can be less effective in certain clinical circumstances, more toxic, and often broader spectrum, potentially increasing a patient's risk of future infections with resistant bacteria.(Krah, Jones et al. 2021) More than 95% of individuals with a penA label can tolerate penicillin. (Shenoy, Macy   et al. 2019, DesBiens, Scalia et al. 2020)   Assessment of patients with reported penAs has been the role of allergists, but allergy services are limited.(Krishna, Huissoon et al. 2017) Traditional penA testing requires skin testing prior to drug provocation testing, which remains the main testing method in Europe, making penA testing resource intense.(Mirakian, Leech et al. 2015, Romano, Atanaskovic-Markovic et al. 2020) Direct drug provocation testing (DPT), an oral challenge test, in patients with a low risk allergy history is less resource intense.Two systematic reviews have confirmed the safety and efficacy of DPT (without prior skin testing) as a method of delabelling adults, delivered both by allergists and non-allergists.(DesBiens, Scalia et al. 2020, Cooper, Harbour et al. 2021) Skin testing prior to DPT has also been successfully delivered by non-allergists.(Wall, Peters et al. 2004, Englert andWeeks 2019) The American Academy of Allergy Asthma and Immunology with the Infectious Diseases Society of America wrote to the Centers for Medicare and Medicaid Services to urge US hospitals to include verification of penA as part of its mandatory antibiotic stewardship programs.(Immunology2020) The World Health Organisation has since recommended antibiotic de-labelling as an effective antimicrobial stewardship strategy.(Europe2021 ) Enablement of the wider healthcare workforce to de-label eligible patients is required to deliver penA assessment and de-labelling at scale.Understanding the wider frameworks that enable non-allergists to safely de-label is required, enabling development of effective interventions that facilitate penA de-labelling by non-allergy specialists.
We systematically reviewed the literature to determine the proportion of patients with a reported penA who were safely de-labelled by non-allergy healthcare workers (HCWs), categorising components of interventions using the Effective Practice and Organisation of Care (EPOC) taxonomy of health interventions,((EPOC) 2016) and report any measured antimicrobial stewardship and health-system impact.

Methods
This systematic review and meta-analysis was conducted in accordance with the Joanna Briggs Institute (JBI) methodology for systematic reviews of effectiveness, (Tufanaru C 2020) and is reported using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) checklist.(Liberati, Altman et al. 2009) Inclusion/exclusion criteria (i) Any patient (adult/child) with a penA record, in any healthcare context; (ii) having undergone penicillin allergy de-labelling (PADL) using any method; (iii) by non-allergy specialists; defined as a medical professional whose primary specialisation is not in allergy, or who has not trained in allergy as part of their specialty.(Savic, Khan et al. 2019) (iv) penA assessment and de-labelling interventions delivered by immunologists, or allergy specialists were excluded.All study designs were included, except case reports.

Search strategy
The following databases were searched from inception to 21/01/2022 (NP) EMBASE (Ovid), MEDLINE (Ovid), CINAHL (Ovid), PsycInfo, Web of Science and Cochrane CENTRAL as was the grey literature.Known experts in the topic were contacted to ensure we have not overlooked relevant literature.The search strategy was reviewed by an experienced information specialist (KO).Only studies published in English were included due to a lack of funding for translation services.(Appendix 1) Titles and abstracts were screened by two independent reviewers (NP, SA, DK, RO, JS) against the inclusion criteria (RAYYAN software).(Ouzzani, Hammady et al. 2016) Full text citations were assessed against the inclusion criteria by two independent reviewers (NP, RO) using RAYYAN software.(Ouzzani, Hammady et al. 2016) (Appendix 2 & 3).Disagreements were resolved through discussion.

Assessment of methodological quality
Eligible studies were critically appraised by two reviewers (NP, BK) using critical appraisal instruments from the JBI.(Tufanaru C 2020) Authors were contacted to request additional data, where required.Studies were not excluded on the grounds of their risk of bias.

Data extraction
Data were extracted by one reviewer (NP), using a purpose built extraction tool in Excel (Corporation 2018) and included study design, country, setting, population age, gender, inclusion criteria, exclusion criteria, allergy testing method(s), HCW(s) delivering

Definitions
See appendix 4 for definitions for de-labelling, skin testing/ drug provocation testing (ST/DPT), direct drug provocation testing (DDPT) and direct de-labelling on history alone (DDL), successful de-label and definitions of harm.

Data analysis
Population-weighted proportional meta-analysis was conducted on studies with a low/moderate risk of bias to determine the proportion of participants successfully delabelled and the proportion with a positive penA test by de-label method (DDL, DDPT and ST/DPT), using the R package meta v 5.2.0.(Schwarzer 2022) Statistical heterogeneity was assessed using Chi-square test (threshold P<0.1) and the I 2 statistic (I 2 values <25%, 25-75% and >75% considered to represent low, moderate and high heterogeneity, respectively).
Overall estimates were obtained using random effects models.(Tufanaru, Munn et al. 2015) A funnel plot was generated to assess publication bias with funnel plot asymmetry tested using the Egger test.(Egger, Smith et al. 1997) We used the studentized residual to identify studies that contributed most to heterogeneity.(Viechtbauer and Cheung 2010) Studies with z absolute values >1.96 (Viechtbauer and Cheung 2010) were excluded from the analysis to assess their influence on the overall estimates.The remaining data are presented in narrative form.

Study inclusion
In total, 11,545 papers were identified, of which 3411 were excluded due to duplication.

Methodological quality
Of fifty-six case series studies, six, 19, and 31 had a high, moderate and low risk of bias, respectively.Risk of bias assessments are shown in appendix 5.
Heterogeneity remained unchanged after sensitivity analysis except for the proportion of patients de-labeled on history alone.(Appendix 11)

Discussion
Rates of PADL varied from 14-41% depending on PenA assessment method.Less intensive methods that targeted the smaller population of lowest risk patients de-labelled a smaller proportion than those employing more formal testing and included higher risk patients.
Once patients were assessed as suitable for de-labelling, rates of PADL were high (≥ 95%), PADL was commonly delivered by a small team, or an individual HCW, as an outreach service and always in the hospital setting.Less commonly, the responsible medical team delabeled patients.Individual HCW, or small teams, limits the reach of PADL across a hospital.
The advantage of small teams, or individual delivery of PADL, is a greater likelihood of the requisite knowledge and motivation, but delivery of PADL by the wider workforce may enable a broader reach across the hospital.Adequate knowledge, motivation and competing demands may hinder the delivery of PADL by the wider workforce.Quality improvement methodology (Bauer, MacBrayne et al. 2021, Louden, Hansen et al. 2021) and financial incentives (Bauer, MacBrayne et al. 2021) have been used to motivate staff, but this adds further expense and time resource to PADL.Whether PADL is safer and more effective as a small team/individual or delivered by the wider workforce needs further study, and the barriers/enablers to the delivery of PADL at scale need exploration.Given the safety of direct DPT in low-risk patients there is potential to extend this to healthcare settings outside of the hospital but this requires further study.
There was high heterogeneity between studies, with several possible explanations.Risk stratification prior to testing was done on both patient factors and allergy history which varied between studies.Route of DPT administration, location of testing and HCW(s) undertaking testing also varied.Others have reported oral challenges to be better tolerated than IV challenges; challenges in the inpatient setting more likely to be tolerated than in the ambulatory setting; tolerance in children reported to be higher than in adults; although tolerance was reported to be similar between those with and without infection.(Harandian, Pham et al. 2016, DesBiens, Scalia et al. 2020) Some studies only assessed using one method and some studies used all three assessment methods introducing further potential for heterogeneity.Optimisation of testing protocols requires further study and harmonisation.
We found low heterogeneity between studies assessing the proportion of tested patients who were successfully de-labelled, and the proportion harmed, by DDPT.There was high heterogeneity between studies looking at PADL in those identified suitable for DDL, but after the sensitivity analysis and removal of one study the recalculated heterogeneity was low.A similar systematic review of the literature, not restricted to non-allergists, reported the successful de-label of 595 (97%) patients using DDPT and comparable to our findings providing external validity to these data.(DesBiens, Scalia et al. 2020) We report harm post DDPT to be 2%, comparable to the expected 0.5-2% ADR rate in patients without a history of penA but low when compared to other direct DPT studies.(Shenoy,Macy et al. 2019, DesBiens, Scalia et al. 2020) We found low heterogeneity between ST/DPT studies when looking at harm from de-labelling but heterogeneity was high between studies looking at the proportion of tested patients de-labelled by ST/DPT.We found the rate of harm to be low in our study when compared to other studies reporting penicillin tolerability following ST/OC (1% verves 6%) which may be explained by allergists testing higher risk patients or higher rates of false positive skin in some studies, or differing definitions of harm.(DesBiens, Scalia et al. 2020)

Limitations
All the studies are from high income countries (70% from USA), therefore the findings may not be generalisable to LMICs.However, the proportion of tested patients de-labelled and adverse event rates are similar across studies with data from eight countries.
Most studies were case series, with inherent patient selection bias, and the inclusion of conference abstracts limited the review of methodology.Conference abstracts are limited by the extent of reporting and quality.(Schererand Saldanha 2019) However, the inclusion of abstracts gives a wider and more representative view of the non-allergist de-label activity, particularly important as full paper publication of conference abstracts is reported to be low.(Scherer and Saldanha 2019) The high heterogeneity between studies limits the certainty of our findings.
To reduce publication bias we searched trial registries, unpublished studies, the bibliographies of included studies and asked known experts in the field for missing studies.
Despite this, 5 of 8 funnel plots identified high risk of publication bias.
The side effect rate was reported in those de-labeled on history alone.Given the background rate for a penicillin reaction is 0.5-2%,(Shenoy, Macy et al. 2019)we would expect to see some evidence of harm in the 812 patients de-labelled on history alone upon subsequent penicillin re-exposure.It was not clear how many patients went on to receive a penicillin post de-label.The rate of harm in this patient population requires further study.
The statistical power of I 2 test is limited in meta-analyses with <20 studies and / or the average study sample size was <80, with all the meta-analyses in this study below this threshold.(Huedo-Medina, Sánchez-Meca et al. 2006)

Conclusions
Non-allergists have used several approaches to assess and PADL, all of which appear to be effective and safe.More comprehensive testing capability allowed a greater proportion of assessed patients to be de-labelled.A diverse workforce has delivered PenA assessment services outside of allergy/Immunology services.The consequences of PADL were reported to be increased use of penicillin, and other beta-lactams, with a subsequent reduction in non-beta-lactam antibiotic use and a reduced antibiotic drug spend.PADL is often limited to individual HCWs or small groups of HCWs within a hospital, predominantly delivered as an outreach service, which limits the impact of PADL.Delivery of PADL by the primary healthcare provider and extending PADL to health-care settings outside the hospital, will broader the impact of PADL.A small number of studies showed provider delivered PADL to be safe and effective but further studies are required on whole of hospital implementation of PADL delivered by primary provider teams.Studies were from high-income countries, data are also needed from low-and middle-income countries.
PADL, components of the PADL interventions, details about education and training, number of assessed patients, number tested, number that experienced unintended harm, and any reported antibiotic stewardship or healthcare system impact.Extraction of data from seven studies (10%) was validated by a second reviewer.Intervention components were categorised using the EPOC taxonomy of health interventions, enabling grouping of health system interventions by conceptual or practical similarities.((EPOC)2016) Studies that utilised a risk stratification protocol for allergy testing were categorised in the "packages of care" subcategory.Complex interventions were categorised into the "care pathways" subcategory.(Skivington,Matthews et al. 2021) Governance arrangements were categorised as "Authority and accountability for quality of practice".
indicating good acceptability of testing and results.PenA assessment by non-allergists was delivered by a diverse workforce to a diverse patient population and demonstrated the significant opportunity to reduce erroneous penA labels in line with global antibiotic stewardship ambitions.(Shenoy,Macy et al. 2019, Committee 2020, Jeimy, Ben-Shoshan et al. 2020, Sneddon, Cooper et al. 2021, Europe 2021    ) This review found that penA assessment by non-allergists was safe: of tested patients 1.7% had a subsequent reaction, but none were serious.PADL increased penicillin use and reduced non-penicillin use, e.g., quinolones and aztreonam, with associated reduced antibiotic spend.HCW time taken to de-label varied depending on the testing method.Local PADL interventions might need to balance the staff resource available with the potential impact on patient care by prioritising patients according to greatest need, or where PADL has the greatest potential for improved patient care or health system impact.(Macy and Contreras 2014) The potential antibiotic cost savings are likely to offset the HCW and the skin testing costs,(Macy and Contreras 2014) but the HCW costs are often not/poorly described.PADL is delivered by HCWs and their time has an inherent cost that needs adequately describing to enable appropriate healtheconomic analysis.The wider and longer-term impact of PADL, due not only to reduced drug acquisition costs but also savings in terms of potential reductions in length of stay and mortality are estimated to have been ten times the cost of allergy testing.(MacyandContreras 2014, Macy andShu 2017) The longer-term impact of PADL on patient, health systems and AMR requires further study.Most interventions protocolised PenA assessment with allergists contributing to the development of protocols.The low number of studies reporting direct access to an allergy expert during the day-to-running of PADL provides reassurance of the effectiveness/safety of these protocols without an allergist present.Education was a key theme supporting the appropriate use of the testing protocols.