Clinical Question

In patients reporting a penicillin allergy, is it possible to determine the likelihood of true allergy without formal testing?

Evidence Summary

Penicillin allergy is the most commonly documented drug allergy in medical records, with a prevalence of approximately 10% of all patients.¹ However, in up to 90% of patients with a reported allergy, penicillins are tolerated on allergy testing.² As a consequence of patients being labeled as having a penicillin allergy, alternative (typically broader-spectrum) antibiotic classes are often used, with potentially poorer efficacy and safety profiles. This leads to increased multidrug-resistant organisms, treatment failure, and health care costs and prolonged hospitalizations.³

A formal drug challenge is the preferred test for investigating immunoglobulin E–mediated penicillin hypersensitivity (type I hypersensitivity)⁴; skin testing is also a commonly used validated tool.¹ A number of studies have been performed to develop clinical prediction rules using features of the allergy history to help determine which patients labeled as having a penicillin allergy can safely be given a beta-lactam antibiotic. Despite the high prevalence of patients labeled as having a penicillin allergy, there is an international shortage of those proficient in conducting formal drug challenges.⁵

A study in a tertiary referral center in the United Kingdom used multivariable logistic regression to identify patients at low risk of type I beta-lactam allergy.⁶ This included patients who had no history of anaphylaxis, who had a reaction more than one year before referral, and who could not recall what the index drug was. Only 1.6% of patients with all of these traits had type I hypersensitivity. Another study derived and validated two algorithms in a retrospective cohort of individuals evaluated for beta-lactam allergy.⁷ Similarly, anaphylaxis, shorter time since the reaction occurred, and reaction onset less than one hour after most recent drug intake were identified as predictors of a true allergy. However, both algorithms had low sensitivity (51% and 60%), precluding their clinical use.⁷

A more recent study derived and validated a clinical prediction rule for penicillin allergy.⁸ The PEN-FAST (penicillin allergy, five or fewer years ago, anaphylaxis/angioedema, severe, treatment) rule was derived from a prospective cohort of 622 allergy-tested patients in Melbourne, Australia, using a multivariable logistic regression model. This identified five independent predictors of an oral challenge positive for penicillin allergy. Discrimination (the ability to separate those with a positive allergy result from those without) was shown to be good in the derivation and internal validation cohort (area under the curve = 0.81). Calibration (agreement between predicted and observed rates of allergy) was also good across multiple measures.

External validation was performed in multi-center retrospective cohorts in Australia (Perth, n = 334; Sydney, n = 80) and the United States (Nashville, Tenn., n = 531), with 94% of participants undergoing an oral penicillin challenge and the remainder undergoing a scratch test or intradermal test alone.⁸

The study identified three risk factors to include in the clinical prediction rule⁸:

• Allergy event occurring five or fewer years ago (2 points)

• Anaphylaxis/angioedema or severe cutaneous adverse reaction (2 points)

• Treatment required for the episode (1 point)

Total scoring ranged from 0 to 5, allowing for stratification of patients based on risk of a positive result on penicillin allergy testing⁸:

• Very low (0 points): 0.6% risk (1 out of 164)

• Low (1 or 2 points): 5% risk (16 out of 296)

• Moderate (3 points): 19% risk (25 out of 132)

• High (4 or 5 points): 53% risk (16 out of 30)

The investigators chose a cutoff score of less than 3 points, which yielded a sensitivity of 70.7% (95% CI, 57.3% to 81.9%) and specificity of 78.5% (95% CI, 74.9% to 81.9%) in the derivation cohort. A total of 74% of the cohort had a score of less than 3, and, of these patients, only 3.7% (17 out of 460) had a positive test result.⁸ However, different cutoff scores may be considered for risk stratification depending on clinician preference, the setting, and capacity for allergy testing.