The prevalence of a self-reported penicillin allergy in this Norwegian population was lower than in most previous studies. A true IgE-type allergy was unlikely in 45% of the self-reported penicillin allergic patients, due to an interview algorithm. These patients were offered a direct oral penicillin challenge, of whom 98% of the tested patients had no adverse events and could be de-labeled from penicillin allergy in their hospital records.
Penicillin allergy de-labeling gave immediate clinical and environmental benefit as the antibiotic treatment regimen could be narrowed from a non-penicillin to a penicillin if the patient suffered from a penicillin-sensitive infection when indicated by the Norwegian national guideline of antibiotic use.
The prevalence of penicillin allergy in the hospital records in our medical department inpatient population was 4.6%. This is lower than expected, as the prevalence of penicillin allergy is commonly reported to be around 10% [9, 12, 16, 17]. Other studies have reported penicillin allergy prevalence ranging from 5.9% in the UK general population  to 15% in hospitalized patients . The variable findings may in part be due to the differences in age and health status in the different study populations. Another possible reason for the low prevalence finding in our study might be that antibiotic use in Norway is among the lowest in the world, thus leading to a lower prevalence of antibiotic-associated adverse drug reactions . A study published in 2006 with a study population of admitted patients in a hospital in Denmark, a country with similar antibiotic use to Norway, reports a prevalence of 5%, which is in line with our study . They state that penicillin G/V is used more often in Denmark than in other European countries, which tend to use more broad-spectrum penicillins. Broad-spectrum penicillins are associated with a higher risk for adverse drug reactions compared to penicillin G&V . To our knowledge, no previous studies have been conducted in a random sample of acutely admitted patients in a Norwegian medical department, and the present study adds to current knowledge on penicillin allergy over-reporting in hospital records.
Our study also shows that a direct oral amoxicillin challenge is an easy and a safe way to rule out penicillin allergy in acutely hospitalized patients, when carefully selected. 98% of the tested patients had no immediate reaction. Previous comparable studies with varying study populations have reported 94–98% tolerance to the oral challenge [7,8,9, 11, 16, 17], and none of them have reported severe immune-mediated reactions due to the provocation test, when inclusion criteria were met.
Fortysix percent of the tested patients were under treatment with antibiotics at the time of testing. For 42% of these patients, their in-hospital antibiotic treatment was changed from a non-penicillin to a penicillin, which shows an immediate effect of inpatient penicillin allergy de-labelling. Patients with a negative allergy test who did not change treatment to a narrow spectrum penicillin were either not currently admitted because of an infection, or they suffered from infections/bacteria not susceptible to penicillin treatment (either based on microbiological findings or based on the empirical Norwegian national guidelines for antibiotic use). In a comparable study, 56% percent of inpatients used beta-lactam antibiotics subsequent to testing .
A downside of testing acutely hospitalized patients is that 36% of the eligible inpatients did not consent to testing because of anxiety or because they felt too sick or frail to participate. In comparable studies with inpatients, 37% and 48% did not consent to testing [16, 17]. The non-consenting patients did not differ from the consenting patients in their baseline characteristics (Table 1). There are, however, other differences. Sixtysix percent of the non-consenting patients reported to have had a benign rash as their index reaction to penicillin, as opposed to only 30% of included patients. In addition, as many as 56% of included patients had no recollection of their reaction; for patients that did not consent, this was 22%. This indicates that patients who remember their index reaction, or had a benign but bothersome rash, and thus may have had a more serious reaction, are less likely to consent to the oral challenge, which may represent a bias. A study by Siew et al. has indeed shown that patients who recall their index reaction to beta-lactam antibiotics are associated with increased probability of a true allergy .
Our study followed quite strict exclusion criteria compared to similar studies [9, 16]—we excluded not only life-threatening reactions, but all patients with a possible IgE-mediated reaction. Only 45% of the screened patients were identified as low-risk. Roughly half (47%) of all admitted patients were female; of the screened patients with a reported penicillin allergy, 69% were female. The even higher proportion of women in the exclusion group might indicate that women are more prone to penicillin allergy than men, but it may also mean that women have had more previous antibiotic exposure, or that they recall and report previous allergic symptoms to a higher extent than men. Female sex as a risk factor for penicillin allergy, and adverse drug reactions in general, has been described in previous studies  23. Among the excluded patients, there might have been some who did not have a true IgE type allergy; thus, our strict screening algorithm might have contributed to exaggerating rather than deflating the estimated number of patients with suspected IgE type allergy.
It might seem excessive to test low-risk patients, and in the future it might be concluded that it is not necessary to expose these patients to testing entirely. However, in the current situation these patients continue to report they are allergic to penicillin. Compared to the costs of using alternative antibiotics, which are associated with longer hospitalisations, higher risk of adverse drug reactions and increasingly resistant bacteria, the benefits of oral challenges outweigh the costs.
Our study has several limitations. Firstly, the study was small and carried out in a single local hospital in Oslo, Norway, with a predominantly ethnic Norwegian population of high socioeconomic status.
Secondly, we only tested to exclude immediate IgE type reactions. However, studies that have tested patients with a multiple-day course of penicillin report a prevalence of mild delayed reactions close to the baseline incidence of the general population, and testing for a longer period of time causes unnecessary exposure to antibiotics [2, 7]. Thirdly, enteral exposures have a reported lower risk of anaphylaxis than intravenous exposure . It might be possible that the rate of allergic reactions would have been higher (and more accurate) if patients were tested with intravenous penicillin rather than through enteral admission. Fourthly, data were collected in two separate periods by three different medical doctors. Fifthly, 42% of screened patients were excluded because an IgE-mediated reaction could not be ruled out. This includes patients who self-reported urticaria, relying on their ability to distinguish urticaria from other rashes. This could contribute to either under- or over-estimation of IgE-mediated reactions. Furthermore, we did not compare findings to placebo, in contrast to another study , but since there were no patients reporting any subjective reactions during testing, there does not seem to be an obvious need for placebo comparison, and we do not expect these factors to influence our findings in a significant way.
Our protocol definition of low-risk allergy  included symptoms of intolerance (e.g. headache, mild GI-symptoms) and consequently these patients where challenged. These patients have a very low probability of having penicillin hypersensitivity, and thus comparable studies using other protocols have de-labelled this group of patients through medical history alone, not offering oral challenges . One could argue that offering these patients oral challenges would yield a falsely high rate of negative tests. However our results are similar to the above mentioned study  suggesting that the rate of negative challenges is not influenced by our selection of patients. For patients with no medical understanding of allergic reactions, undergoing an oral challenge can feel reassuring, and thereby contribute to refraining from reporting a penicillin allergy in the future.
As oral penicillin challenges in low-risk patients have been demonstrated to be safe, future studies might even look into the possibility of by-passing the oral penicillin challenge and de-label penicillin allergy solely based on the specificity of the inclusion criteria. Further research is warranted to develop a standardized algorithm to identify low-risk patients, as previous studies have used different methods.