PEN-FAST: Penicillin Allergy Clinical Decision Rule.

Development and Validation of a Penicillin Allergy Clinical Decision Rule.

Trubiano JA, Vogrin S, Chua KYL, et al.

JAMA Intern Med. 2020 May 1;180(5):745-752. [Full text]

Patients report a history of penicillin (PCN) allergies though less than 10% of penicillin allergies may be confirmed by formal testing [1]. Simply the reporting of a PCN allergy can affect the prescribing of anti0biotics by caregivers. As an antimicrobial stewardship (AMS) intervention antibiotic allergy testing (AAT) can be implemented. But AAT is labor and time intensive, and given 90% or more of reported allergies are not confirmed, most are probably unnecessary. A method to risk stratify those patients with reported PCN allergy who are low-risk and it is safe to receive PCN without AAT would be a major step in AMS. The authors developed PEN-FAST, an acronym of factors that predict PCN allergy, derived from a prospective set of patients with reported PCN allergy, and then implemented PEN-FAST on an independent external validation set.

Patient population and Design

The derivation of PEN-FAST was obtained from a prospective penicillin allergy–tested cohort from 2 sites in Melbourne, Australia. Later external validation used 3 retrospective cohorts from Sydney and Perth, Australia, and Nashville, Tennessee. All of these patients were 16 years or older and underwent AAT.

In all of these cohorts a positive skin prick test was a wheal 3 mm more than control and flare 5 mm more than control after 15 minutes. An positive intradermal test was positive if there was a 3 mm or greater increase in inoculation site (0.02 mL) with >5 mm flare after 15 minutes. A positive oral challenge only included patients reporting an immune-mediated reaction (e.g. rash).

Univariate analysis considered variables patient and allergy characteristics. The patient characteristics included age, sex, ethnicity, history of psychiatric illness, age-adjusted Charlson comorbidity index, and immunocompromised status. Antibiotic allergy characteristics included number of patient-reported antibiotic allergy labels, time since the last reported antibiotic allergy (in years), allergy phenotype, and treatment of the index antibiotic allergy episode.

Univariate logistic regression examined associations between characteristics outlined above with any penicillin-positive allergy test result. Only variables at least 5% ‘correlation’ were included in the multivariable logistic regression model. A backward stepwise bootstrapping procedure was used eliminating variables with low predictive ‘association’ (2-sided P >0.10) and reincluding variables with higher ‘assocaiton’ (2-sided P 0< 0.05). This model was internally validated.

External validation of the model was performed in the 3 separate retrospective data sets. Risk score was calculated for each patient, and the model performance was evaluated by comparing AUC, sensitivity, specificity, NPV, and PPV.


The internal derivation and validation cohort included 622 patients (367 female [59.0%] and 255 male [41.0%]; median age, 60 [IQR, 48-71] years); the external validation cohort consisted of 945 patients (662 female [70.1%] and 283 male [29.9%]; median age, 55 [IQR, 38-68] years). The prevalence of a positive PCN allergy test was 9.3% (95% CI, 7.2%-11.9%).

The 4 features associated with a positive PCN allergy test result on multivariable analysis were summarized by the mnemonic PEN-FAST (Penicillin allergy, Five or fewer years ago, Anaphylaxis/angioedema or Severe cutaneous adverse reaction (SCAR) [2], and Treatment required for allergy episode). The treatment required category was a minor criteria worth 1 point, all others were major and worth 2 points. The following 4 risk groups were developed: very low risk (0 points), with a risk of allergy of 0.6%; low risk (1 or 2 points), with a risk of 5%; moderate risk (3 points), with a risk of 20%; and high risk (4 or 5 points), with 50% probability of having a positive PCN test result

Overall, a PEN-FAST score of 3 points had good discrimination scores in external validation: 0.81 (95% CI, 0.75-0.86) for Melbourne, 0.81 (95% CI, 0.71-0.91) for Sydney, 0.73 (95% CI, 0.66-0.80) for Perth, and 0.85 (95% CI, 0.74-0.96) for Nashville. The overall NPV of a PEN-FAST score of < 3 had a NPV of 96.3% (95% CI, 94.1%-97.8%) for a positive AAT.


Using the PEN-FAST model, a score of <3 was able to exclude severe PCN allergy with a NPV of 96.3% in an independent validation set of patients with a prevalence of positive PCN allergy tests around 9%. This excludes non-penicillin β-lactam allergies and was derived mostly from Australian based patients that were mostly inpatient, so extrapolation to other populations may be limited. Given the large number of patients with reported PCN allergy, AAT is not a feasible risk stratification tool. With further independent external validation, PEN-FAST could effectively identify low risk patients that could at proceed to use of β-lactam antibiotics or an oral antibiotic challenge, which is easier to scale than AAT.

F: Follow upYes
R: RandomizationN/A
I: Intention to treatN/A
S: Similar at baselineYes
B: BlindingN/A
E: Equal treatmentGrant from the Austin Medical Research Foundation (AMRF); postgraduate scholarship GNT 1139902 from the NHMRC; a postgraduate scholarship from the National Centre for Infections in Cancer; grants 1P50GM115305-01, R21AI139021, R34AI136815, and 1R01HG01086301 from the NIH; the NHMRC; and grant 1K12HS026395-01 from the AHRQ
S: Source (funding)

  1. Blumenthal KG, Peter JG, Trubiano JA, Phillips EJ. Antibiotic allergy. Lancet. 2019 Jan 12;393(10167):183-198. doi: 10.1016/S0140-6736(18)32218-9.
  2. Duong TA, Valeyrie-Allanore L, Wolkenstein P, Chosidow O. Severe cutaneous adverse reactions to drugs. Lancet. 2017 Oct 28;390(10106):1996-2011.