Penicillin allergy management strategies relevant for clinical practice – a narrative review

Penicillin-induced anaphylaxis was first reported in 1945, and in 1968, the World Health Organization noted a mortality rate of 0.002% associated with anaphylaxis.¹ To date, there has been no increase in the incidence of allergic reactions to penicillins, and evidence indicates that penicillin sensitization diminishes over time with approximately 50% of penicillin-allergic patients losing their sensitivity within 5 years and about 80% within 10 years.^2,3 Nevertheless, penicillin allergy is the most commonly reported drug allergy, with prevalence estimates ranging from 6% to 31% across different populations and geographic areas.^{4, 5,6,7} Despite this, research indicates that between 77% and 99% of individuals labeled as penicillin-allergic can tolerate penicillins.^6,8,9

Penicillin allergy diagnosis typically involves skin tests, blood tests for the identification of specific immunoglobulins E (IgEs), and drug provocation tests (DPT) which are usually conducted by allergy specialists.^10,11 However, a global shortage of trained allergists limits access to these evaluations, leading to many individuals being inaccurately classified as penicillin-allergic based on self-reports.¹²

Patients with a documented β-lactam allergy are significantly less likely to receive cefazolin, with a likelihood nine times lower, and are more frequently prescribed non-β-lactam antibiotics perioperatively by non-allergists.¹³ A study from Thailand revealed that allergists are more inclined to confirm penicillin allergies and show less concern about the potential cross-reactivity with third- and fourth-generation cephalosporins compared to non-allergists. The study also found that nonallergists often prescribe alternative antibiotics without confirming penicillin allergy status due to concerns over medicolegal issues related to penicillin re-exposure.¹⁴ Additionally, some practitioners remain hesitant to prescribe penicillins even after a negative DPT to penicillins.¹⁵ While avoiding penicillin is warranted in cases of severe, documented reactions like anaphylaxis, many patients with a self-reported penicillin allergy may not have an immunological sensitivity at all, and therefore should not avoid penicillins.¹⁶ Mislabeling often stems from the misclassification of mild reactions, leading to unnecessary penicillin avoidance.

Penicillin avoidance is associated with poor clinical outcomes, including prolonged hospitalization,¹⁷ and higher readmission rates.¹⁸ Moreover, subjects labeled as allergic to penicillin are at a higher risk for increased mortality and morbidity¹⁹ due to the use of less effective and broader-spectrum antibiotics, such as vancomycin, fluoroquinolones, clindamycin, and macrolides, which carry higher risks of adverse effects like increased rates of gastrointestinal side effects, and risk of QT prolongation (for the macrolides²⁰), and contribute to antimicrobial resistance.¹⁹ Subjects with methicillin-susceptible Staphylococcus aureus (MSSA) bloodstream infections labeled as penicillin-allergic are often treated with vancomycin which is associated with inferior survival rates, lower cure rates and higher mortality compared to beta-lactam (BL) antibiotics.²¹ The prevalence of penicillin allergy labels in MSSA bacteremia patients is concerning, with up to 25% reporting an allergy.²² Furthermore, non-BL alternatives have shown inferior efficacy in treating Gram-negative bacilli infections, contributing to higher rates of clinical failure.²³ In addition to infection outcomes, inaccurate penicillin allergy labeling contributes to increased rates of Clostridioides difficile infections²⁰ and surgical site infections.²⁴

Accurate assessment of penicillin allergy is therefore crucial for optimizing antibiotic use, enhancing patient care, and minimizing the risk of complications. The following sections will discuss the current strategies that focus on clinical history elements which aim to enhance allergy documentation and risk stratification in clinical practice.

Based on their role in decision-making, disease management strategies or models can be classified as assistive in the sense that they generate probabilities without issuing recommendations or directive in the sense that they issue recommendations based on the estimated risks to guide physicians in managing clinical situations.^{25, 26,27} In recent years, various penicillin allergy management strategies have emerged to distinguish between individuals with hypersensitivity reactions to penicillins and those without, and to allow administration of penicillins in non-allergist departments. Several comprehensive reviews^{28, 29,30,31,32} have been published that explore these strategies but it is currently difficult to choose the most efficient one. Nonetheless, a number of penicillin allergy management strategies emerge as particularly pertinent for clinical application. These primarily rely on risk classification, which is often based exclusively on the severity of the reported clinical history. They stem from research conducted in the United States (the Blumenthal strategy³³), Australia (the PEN-FAST strategy/score³⁴), and Europe (the European Network on Drug Allergy (ENDA) strategy,³⁵ the Chiriac-strategy/score⁶ with its retrained version – the RChiriac-score³⁶ and the Schrüfer strategy³⁷). This review will also examine the Combined-score,³⁶ a tool developed by Ghiordanescu et al. in a study comparing the efficacy of four established strategies – PEN-FAST, Blumenthal, ENDA, and the Chiriac-score. The Combined-score was designed to explore the potential for enhancing the performance of penicillin management strategies, making it interesting for assessing potential improvements in the efficacy of existing approaches.

When selecting an appropriate clinical model, it is important to carefully consider critical parameters such as the classification of low-risk patients, sensitivity, specificity, and rates of misclassification. By calculating local prevalence, clinicians can determine positive and negative predictive values (PPVs and NPVs), which decisively inform their assessment of acceptable risk levels – a threshold that varies significantly across specialties and available medical resources. Lower true allergy prevalence can undermine the reliability of high NPVs while the positive PPVs decline alongside reduced penicillin allergy prevalence. Therefore, diagnostic tools must be evaluated on a sufficiently large cohort of confirmed allergic patients to ensure accurate NPVs results.

The ENDA³⁵ and Blumenthal³³ strategies underwent external validation within a retrospective cohort comprising 1884 subjects with a history of hypersensitivity reactions to penicillins presenting at the Allergy Unit of the University Hospital of Montpellier, where the prevalence of positive allergy tests (skin tests or drug provocation tests) was recorded at 20.3%.³⁶ This prevalence aligns with previous findings from the same research group,⁶ yet it is notably higher than results reported in studies conducted in the United States³ and Australia.^34,39 In this study, both strategies showed considerable overlap in their classification of low-risk subjects, with 44.5% of cases falling into ENDA class 3 and 38.0% into Blumenthal class 3. The sensitivity rates were also similar, recorded at 76.0% for ENDA and 77.0% for Blumenthal, although their specificities were significantly lower, with ENDA at 50.0% and Blumenthal at 42.0%. The NPVs for the ENDA and Blumenthal methods were found to be 89.0% and 88.0%, respectively, while the PPVs were 28.0% for ENDA and 25.0% for Blumenthal. By design, these strategies were safe, as none of the subjects with positive outcomes and a history of anaphylaxis, SCAR, or other severe non-immediate reactions (such MPE with fever alone, severe/generalised MPE with extensive desquamation, mucosal lesions alone, vasculitis, serum sickness-like disease, purpura, isolated cytopenia, isolated hepatic involvement) were misclassified. Nevertheless, the ENDA strategy has resulted in unnecessary avoidance for 79.7% of subjects. Conversely, the Blumenthal strategy leads to unnecessary avoidance in 46.4% of the cases. The authors comment that, although the ENDA and Blumenthal strategies may not classify as many patients as low-risk, their high sensitivity renders them safe for clinical application by non-allergists. However, by increased unnecessary avoidance they can compromise individual and public health outcomes.³⁶

The directive strategy developed by Reichel et al. was applied to a cohort of 200 patients presenting at the Allergy Department of the University Hospital Würzburg with previous reactions to various antibiotics, including BL antibiotics. However, the study design raises several questions regarding its methodological approach. Notably, not all patients were investigated for their hypersensitivity status, and in cases where investigations were conducted, some patients were re-exposed to chemically related antibiotics rather than the actual culprit drug, which questions the clinical relevance of the outcome and external validity of these findings. The strategy was later refined by Schrüfer et al.³⁷ and applied to a cohort of 800 subjects from Germany with suspected hypersensitivity to BL antibiotics in which the positive outcome prevalence (positive skin tests, specific IgEs or DPT) was 25.6% which is similar to that of the Montpellier cohort.³⁶ The refined strategy recommended delabeling in 41.0% of the cases and exhibited a sensitivity of 90.0%, and a specificity of 56.0% which were comparable to the other two directive strategies. Of the subjects with positive outcome, 10.0% (n=21) have been misclassified. Of these 8 (representing 2.6% of the total subjects receiving a recommendation for delabeling) had a clinical history of anaphylaxis. This raises concerns about the sensitivity of question 5 in accurately identifying anaphylaxis within the clinical history. The NPV and PPV were 94.0% and 41.0%, respectively. Unnecessary avoidance occurred in 44.5% of cases. Schrüfer’s research found that among the 195 patients who answered affirmatively to question 4, 106 (54.4%) had a positive allergy outcome. However, the authors relied on allergy test results as a proxy for actual allergy status, which may have led to an overestimation of clinically relevant sensitization in some low-risk patients with MPE who only underwent skin testing. This issue was similarly noted as a limitation in the study by Ghiordanescu et al.³⁶ Given that MPE is usually a mild, manageable reaction, it is questionable whether the Schrüfer strategy advising penicillin avoidance and referral for additional allergy testing is appropriate, particularly considering the strain it places on an already overwhelmed specialty for a condition that is generally manageable.

The PEN-FAST assistive strategy (score) was derived using a prospective cohort from Melbourne of 622 patients with reported allergy to penicillins, where the outcome prevalence (positive skin tests or DPT) was 9.3%.³⁴ The original strategy, which uses a threshold of less than 3, demonstrated a sensitivity of 70.7% and a specificity of 78.5%, along with a NPV of 96.3% and a PPV of 25.3%. At this threshold, 17 out of 460 low-risk patients (3.7%) have been misclassified. The derived cohort in Melbourne was validated on three cohorts from Nashville, Sydney, and Perth³⁴ which had outcome prevalences of 19.0%, 27.0%, and 48.0%, respectively. The sensitivity and specificity for these cohorts were 73.7% for Nashville, 70.4% for Sydney, and 87.5% for Perth. The NPVs varied across these cohorts, ranging from 84.9% to 98.4%. Except for the PALACE study,⁴⁰ which is a randomized controlled trial, the validation of the PEN-FAST algorithm was performed in retrospective cohorts.^{41,42, 43,44, 45} In the cohorts from Su et al.,⁴² Piotin et al.,⁴¹ Castagna et al.,⁴⁴ and Gonzalez-Estrada et al.,⁴³ outcome prevalences ranged from 2.0% (Gonzalez-Estrada’s Mayo non-enriched test) to 66.2% (Piotin). Sensitivity varied across these cohorts, from 23.3% (Gonzalez-Estrada’s Mayo enriched test) to 100% (Su et al.), while specificities ranged from 56.0% (Piotin) to 81.6% (Gonzalez-Estrada’s Mayo enriched training). The NPVs for these studies varied between 75.4% (Gonzalez Estrada’s Mayo enriched test) and 100% (Su et al.). PEN-FAST was found to be unreliable in pediatric populations under 12 years of age, showing a sensitivity of 57.0%, a specificity of 45.7%, and an NPV of 95.0%. This conclusion was drawn from a study involving 2031 allergy labels, which had an outcome prevalence of 5.3%.⁴⁵ In the study by Ghiordanescu et al.,³⁶ PEN-FAST strategy demonstrated sensitivity and specificity rates of 66.0% and 73.0%, respectively, which are consistent with the original publication’s findings.³⁴ In this particular study, PEN-FAST identified 65.0% of the subjects as low-risk for allergies. However, 33 patients with a severe index reaction (anaphylaxis, other severe non-immediate reactions) were incorrectly classified as low-risk, representing 3.0% of all low-risk patients. The NPV for a PEN-FAST score of less than 3 was 90.0%, suggesting its usefulness as a rule-out test, though this was slightly lower than reported in previous studies.^{40, 41,42,43} This discrepancy may be due to the higher prevalence of positive outcomes (20.3%) in Ghiordanescu et al.’s cohort,³⁶ compared to the 9.3% in the original study.³⁴ Additionally, differences in participant characteristics could have influenced the algorithm’s performance. While the original study excluded cases categorized by Ghiordanescu’s study as other severe non-immediate reactions, excluding these cases from the analysis, did not notably affect the NPV, which remained at 90.0%. For comparison, the retrospective cohorts from Piotin et al.⁴¹ and Su et al.⁴² focused only on subjects with immediate reactions, while the PALACE study⁴⁰ excluded participants with anaphylaxis and severe non-immediate reactions.

The original Chiriac-score⁶ was derived using a retrospective cohort of 1991 subjects and externally validated on a prospective cohort of 200 subjects from Southern France, undergoing investigations for hypersensitivity to BL (penicillins and cephalosporins). Subjects with SCAR were excluded from the cohort used to develop the model. The prevalence of outcomes (positive skin tests or DPT) differed between the retrospective and prospective cohorts, at 23.6% and 31.0%, respectively. The PPVs were 40.0% in the retrospective cohort versus 57.0% in the prospective cohort, while the NPVs were 83.0% and 82.0%, respectively. The misclassification rates were similar between the two cohorts, at 19.4% for the retrospective cohort and 21.0% for the prospective cohort. Among the misclassified patients, 13 (3.1%) experienced anaphylaxis during allergy testing. The RChiriac-score³⁶ demonstrated an AUC of 0.77 with a sensitivity of 55.0% and a specificity of 85.0% at Younden’s index. The NPV and PPV for these coordinates were 88.0% and 49.0% respectively. The authors assessed various thresholds for predicted probabilities to classify patients as low risk, resulting in 46.0% to 77.0% being categorized in this way, with sensitivity ranging from 55.0% to 80.0% and specificity from 53.0% to 85.0%. The model misclassified between 14 and 53 patients who experienced severe reactions, representing 2.0% to 4.0% of those identified as low risk.

The Combined-score had an AUC of 0.80 with a sensitivity of 57.0% and a specificity of 91.0% at Youden’s index.³⁶ The NPV and PPV for these coordinates were 89.0% and 62.0% respectively. As for the RChiriac-score,³⁶ the authors assessed various thresholds for predicted probabilities to classify patients as low risk, resulting in 50.0% to 82.0% being categorized in this way, with sensitivity ranging from 57.0% to 80.0% and specificity from 58.0% to 91.0%. Five to 28 patients with severe index reactions were misclassified by the model, which represents less than 1.0% to 2.0% of the low-risk cohort. In the comparison conducted by Ghiordanescu et al.,³⁶ the performance of the Combined-score was largely comparable to that of the PEN-FAST and RChiriac scores. This indicates that enhancements to penicillin allergy risk stratification tools, even when customized for particular populations or local penicillin allergy incidences are probably not feasible. Overall, the scoring strategies demonstrate effective delabeling capabilities; however, this comes at the cost of an increased number of misclassified positive subjects exhibiting high-risk index reactions.³⁶ A detailed comparative analysis of the PEN-FAST and Chiriac- score derivation and validation cohorts is available in the supplementary materials of the publication by Ghiordanescu et al.³⁶

Delabeling of penicillin allergy is influenced by the patient’s risk profile.^30,46 In high-risk subjects, a comprehensive approach combining clinical history, laboratory tests, and drug provocation challenges is essential for an accurate diagnosis of penicillin allergy.⁴⁶ While testing for penicillin-specific IgE antibodies can assist in diagnosing penicillin hypersensitivity, the method has limited sensitivity and specificity.⁴⁷ In vitro and in vivo (skin tests) to penicillins are usually performed by allergy specialists.⁴⁶ In contrast, in low-risk populations, graded oral or intravenous drug challenges are considered the diagnostic gold standard.⁴⁸ These are currently mainly performed in inpatient settings which have the best evidence for optimizing antibiotic use.⁴⁹ Nevertheless, their safety of use in outpatient settings has also been recently demonstrated.^32,46 Research shows that 96% of low-risk patients can be effectively delabeled through graded challenges,⁵⁰ whereas skin testing achieves a success rate of only 88%, particularly in low-risk groups with non-immediate reactions where the predictive value is decreased.⁵¹ Additionally, detectable serum IgE levels against penicillins may be found in subjects whose clinical histories do not indicate an IgE-mediated allergic response, therefore their use in a population with low-pretest probability is questionable.⁴⁹ However, in clinical practice, low-risk non-immediate reactions often present challenges in classification across different strategies, as well as in recommendations for exposure avoidance. For example, the authors of the Schrüfer strategy comment that although MPE is generally regarded as low-risk for certain phenotypes there is a considerable chance for a positive challenge. Conversely, the ENDA and Blumenthal strategies classify MPE as low-risk with recommendations for moderate or limited avoidance, respectively. It is nevertheless important to note that the ENDA strategy considers a particular phenotype of MPE – the severe MPE as high-risk. Still, the literature lacks clear stratification of severity for MPE, aside from references to body surface coverage greater than 50%, which is associated with MPE in Drug Rash Eosinophilia with Systemic Symptoms syndrome, or the presence of bullae.⁵² Furthermore, patients may struggle to accurately describe papular eruptions, leading to potential diagnostic difficulties. In addition, clinical recognition of these reactions can be challenging, and healthcare providers themselves might require proper training.

Implementing penicillin allergy delabeling strategies, which use clinical predictors to stratify risk, is an important step toward optimizing both individual patient care and public health outcomes. Clinical history predictors play a crucial role in structuring the allergy assessment and ensuring that best practices are followed throughout the testing process. However, despite advancements over the past decade, none of the chosen strategies that are particularly significant for clinical practice appear to achieve an ideal equilibrium between safety and unnecessary avoidance. Moreover, significant improvements in the accuracy of these tools based solely on clinical history are unlikely as shown by the Combined-score performance.³⁶ Therefore, when choosing a tool for a specific clinical setting, it is essential to consider factors like the type of clinic, which influences the prevalence of penicillin allergies and symptom distribution.

Since none of the strategies has proven perfectly balanced, the following pathway involving two filters can be followed by non-allergists in cases where emergency treatment with BL antibiotics is necessary. The directive strategies can be applied as a first step to assess the risk class to which a patient belongs. The ENDA and Blumenthal strategies, unlike the Schrüfer strategy, which recommends alternative antibiotics without providing details on the potential use of cephalosporins, carbapenems/aztreonam, or even penicillins with different side chains, have the advantage of offering embedded guidance based on the risk class with available antibiotic replacements. High-risk patients with medium or extensive avoidance recommendations can later be referred for allergy testing. Nevertheless, in emergencies involving high-risk patients where cross-reactive antibiotics may be required, it is advisable to seek telephone guidance or direct consultation from an allergist, and possibly conduct urgent skin testing before administering antibiotics. For low-risk patients, any of the scoring strategies (PEN-FAST, Chiriac/RChiriac-score, or the Combined-score), which are highly efficient for delabeling, can be employed to verify if the patient is indeed low-risk also according to these strategies. If this is the case, the patient can be safely delabeled and administered the antibiotic according to their needs, under surveillance. A potential pathway to follow by non-allergists is depicted in Figure 2. Evidence is accumulating that, regardless of the tool or definition of low-risk, direct challenges to confirm penicillin allergy are increasingly safe and accessible.⁵³ However, it should be noted that the existing tools are not flawless, and there is no absolute guarantee that patients will not react upon re-exposure to penicillins. Non-allergist practitioners must therefore remain aware of the potential risks and be prepared to manage both severe and non-severe hypersensitivity reactions in collaboration with their teams.

Figure 2.

Suggested pathway for the management of patients labeled as allergic to penicillins in non-allergist settings.

Penicillin allergy management strategies can increase the use of penicillin antibiotics in patients with a history of reactions to penicillins. However, there is still uncertainty about whether these strategies can effectively reverse the negative outcomes associated with penicillin allergy labels.

Critical questions persist about the optimal strategies for specific clinical contexts, as well as the logistical considerations necessary for their successful implementation in clinical practice. Currently, the strategies analyzed in this review do not appear to be ready for widespread application on their own, and further advancements relying solely on clinical history are unlikely to yield substantial improvements. Despite ongoing systematic reviews evaluating their efficiency, a shift in research focus may be necessary, emphasizing the identification of novel biomarkers. Therefore, it is crucial to pursue a thorough investigation of in vitro and ex vivo methods aimed at discovering new allergy biomarkers, particularly those related to BL antigenic determinants.