Two studies assessed both patients with index allergy to penicillins tested for cefazolin and patients with index allergy to cefazolin tested for penicillins.
CrI indicates credible interval; ED, emergency department. The width of the diamonds corresponds to the length of the 95% CrI for the meta-analytical pooled result. Thin lines indicate the length of the 95% CrI for each primary study. Thick lines indicate the length of the 68% CrI for each primary study.
eTable 1. Search Queries Used in the Different Electronic Bibliographic Databases
eTable 2. Characteristics of Included Cohort Studies
eTable 3. Characteristics of Case Reports Included
eTable 4. Risk of Bias of Individual Studies
Customize your JAMA Network experience by selecting one or more topics from the list below.
Sousa-Pinto B, Blumenthal KG, Courtney L, Mancini CM, Jeffres MN. Assessment of the Frequency of Dual Allergy to Penicillins and Cefazolin: A Systematic Review and Meta-analysis. JAMA Surg. 2021;156(4):e210021. doi:10.1001/jamasurg.2021.0021
What is the incidence of dual allergy to cefazolin and natural penicillins?
This systematic review and meta-analysis included 6147 patients from 77 studies. Forty-four patients were identified as allergic to a natural penicillin and cefazolin, resulting in a meta-analytical frequency of 0.7%.
The low frequency of penicillin-cefazolin dual allergy suggests that most patients should receive cefazolin regardless of penicillin allergy history.
Cefazolin is the preoperative antibiotic of choice because it is safer and more efficacious than second-line alternatives. Surgical patients labeled as having penicillin allergy are less likely to prophylactically receive cefazolin and more likely to receive clindamycin or vancomycin, which results in higher rates of surgical site infections.
To examine the incidence of dual allergy to cefazolin and natural penicillins.
MEDLINE/PubMed, Web of Science, and Embase were searched without language restrictions for relevant articles published from database inception until July 31, 2020.
In this systematic review and meta-analysis, a search of MEDLINE/PubMed, Web of Science, and Embase was performed for articles published from database inception to July 31, 2020, for studies that included patients who had index allergies to a natural penicillin and were tested for tolerability to cefazolin or that included patients who had index allergies to cefazolin and were tested for tolerability to a natural penicillin. A total of 3228 studies were identified and 2911 were screened for inclusion.
Data Extraction and Synthesis
Data were independently extracted by 2 authors. Bayesian meta-analysis was used to estimate the frequency of allergic reactions.
Main Outcomes and Measures
Dual allergy to cefazolin and a natural penicillin.
Seventy-seven unique studies met the eligibility criteria, yielding 6147 patients. Cefazolin allergy was identified in 44 participants with a history of penicillin allergy, resulting in a dual allergy meta-analytical frequency of 0.7% (95% credible interval [CrI], 0.1%-1.7%; I2 = 74.9%). Such frequency was lower for participants with unconfirmed (0.6%; 95% CrI, 0.1%-1.3%; I2 = 54.3%) than for those with confirmed penicillin allergy (3.0%; 95% CrI, 0.01%-17.0%; I2 = 88.2%). Thirteen studies exclusively assessed surgical patients (n = 3884), among whom 0.7% (95% CrI, 0%-3.3%; I2 = 85.5%) had confirmed allergy to cefazolin. Low heterogeneity was observed for studies of patients with unconfirmed penicillin allergy who had been exposed to perioperative cefazolin (0.1%; 95% CrI, 0.1%-0.3%; I2 = 13.1%). Penicillin allergy was confirmed in 16 participants with a history of cefazolin allergy, resulting in a meta-analytical frequency of 3.7% (95% CrI, 0.03%-13.3%; I2 = 64.4%). The frequency of penicillin allergy was 4.4% (95% CrI, 0%-23.0%; I2 = 75%) for the 8 studies that exclusively assessed surgical patients allergic to cefazolin.
Conclusions and Relevance
These findings suggest that most patients with a penicillin allergy history may safely receive cefazolin. The exception is patients with confirmed penicillin allergy in whom additional care is warranted.
More than 17 million surgical procedures are performed each year in the US.1 The Centers for Disease Control and Prevention health care–associated infection prevalence survey estimated that there were 110 800 surgical site infections associated with inpatient operations.2 Cefazolin, a first-generation cephalosporin, is the guideline-recommended antibiotic for most surgical procedures.3,4 Cefazolin is the first-line treatment because it is the most widely studied, has the appropriate spectrum of activity against organisms commonly encountered in surgical site infections, is well tolerated, and has a low acquisition cost. However, the 10% of patients reporting a penicillin allergy are less likely to receive cefazolin and more likely to receive clindamycin or vancomycin, resulting in increased odds of developing a surgical site infection.5,6
Avoidance of cefazolin in patients with penicillin allergy is grounded in evidence from more than 40 years ago and is being increasingly called into question. Research from the 1960s and 1970s reported a cross-reactivity rate of 8% between penicillins and cephalosporins.7,8 The production process of penicillins included using a cephalosporin mold and has been blamed for the high rates of cross-reactivity observed before synthetic manufacturing of β-lactams became standard in the middle 1980s.9 Regardless of manufacturing improvements, the current US Food and Drug Administration label for all cephalosporins warns of a 10% cross-reactivity between penicillins and cephalosporins.9,10 Contemporary data have established that cross-reactivity between penicillins and cephalosporins is primarily based on the R1 side chain, not the shared β-lactam ring structure.11,12 Cefazolin does not share an R1 side chain with other penicillins or cephalosporins.9,13,14 This difference in R1 side chains should make cefazolin safe to administer in patients with a penicillin allergy. However, few studies have been published evaluating the incidence of allergy, or hypersensitivity, to natural penicillins (ie, penicillin, amoxicillin, and ampicillin) and cefazolin.
The primary outcome of this systematic review and meta-analysis is the incidence of dual allergy to cefazolin and natural penicillins. In addition, we aimed to identify variables associated with between-study differences in dual allergy between penicillins and cefazolin. The term cross-reactivity was not used as the primary outcome because it implies a biological reaction based on a common chemical structure that is unlikely between cefazolin and penicillins.
We included observational studies in which the primary outcome of dual allergy to a natural penicillin (penicillin, amoxicillin, or ampicillin) and cefazolin could be assessed. Included studies consisted of (1) those assessing patients with self-reported or confirmed allergy to a natural penicillin who were exposed to cefazolin or tested for cefazolin allergy and (2) those assessing patients with self-reported or confirmed cefazolin allergy who were exposed to, or tested for, penicillin allergy. We did not apply exclusion criteria based on publication date, status, or language. This systematic review with meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline and Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guideline.
Two of us (B.S.-P. and M.N.J.) searched 3 electronic bibliographic databases (MEDLINE/PubMed, Web of Science, and Embase) for articles published from database inception to July 31, 2020. Search queries are listed in eTable 1 in the Supplement. Searches on electronic bibliographic databases were complemented by reviewing the references of included primary studies and by contacting authors.
After duplicates were removed, each study was independently assessed by 2 reviewers (all authors in pairs), first by title and abstract screening and second by full-text reading. Non–English-language articles were translated before full-text reading. Data from included primary studies were independently extracted by 2 reviewers (all authors in pairs) using purposely built online forms. For each study, we retrieved information on variables that potentially explained differences in the results, including publication year, country, sampling method (consecutive or convenience), clinical setting, participants’ age group, and characteristics of the index reactions. Index reaction details included culprit drugs, diagnostic tests, clinical manifestations, and timing of allergic reaction. Immediate reactions were defined as those occurring during the first hour after exposure to the culprit drug, with the remaining reactions classified as nonimmediate.
For studies in which the index allergic reaction was a penicillin, we retrieved data on the number of patients with self-reported or confirmed penicillin allergy as well as the number of those who also had a cefazolin allergy, gathering information about whether the latter was confirmed by skin testing, in vitro testing (ie, specific IgE test), drug challenges (ie, a controlled or stepwise drug exposure), and/or drug exposure for prophylactic or therapeutic purposes. Likewise, for studies in which the index allergic reaction was cefazolin, we retrieved information on the number of patients with cefazolin allergy and those who also had a penicillin allergy (as well as on how the latter was confirmed).
Disagreements between reviewers in data selection or extraction were solved by consensus. Authors of primary studies were contacted to provide missing information, including from unpublished studies.
The quality of included primary studies was independently assessed by 2 researchers (B.S.-P., K.G.B., and/or M.N.J.) using an adaptation from the checklist developed by Hoy et al15 for prevalence studies. Of the 11 items, we used 5 that were deemed adequate for this study, namely (1) if the sample was representative of the target population; (2) if random or consecutive sampling methods were used; (3) if the likelihood of nonresponse bias was minimal (ie, if >75% participants with index allergy to penicillins underwent testing and/or exposure to cefazolin or vice versa); (4) if an acceptable or sufficiently complete definition of allergic or hypersensitivity reaction was used in the study; and (5) if the same methods of assessment and data collection were used for all study participants.
We performed bayesian meta-analysis following a random-effects model based on a binomial likelihood to estimate the frequency of allergic reactions to cefazolin in patients with history of penicillin allergy and the frequency of allergic reactions to penicillins in patients with history of cefazolin allergy.16 We opted for this approach because it can more adequately handle proportions equal to 0, which were frequent in included primary studies. By contrast, a frequentist approach would require adding a continuity correction (eg, 0.5) to the proportions equal to 0, which might bias meta-analytical results.
Bayesian methods yield probability distributions of the parameters of interest (posterior probabilities) based on prior probability distributions and on the observed data. In this study, we used meta-analytical methods to estimate the posterior probabilities of the frequency of developing allergic reactions. Of those posterior probabilities, we retrieved the median value and the respective 95% credible intervals (CrIs) (range of values within which, with 95% probability, the true frequency of allergic reactions lies). For building bayesian models, we used uninformative prior distributions for the effect size measure and for the tau parameter (dnorm [0,0.00001] and dgamma [0.00001,0.00001], respectively).
We assessed heterogeneity (existence of differences beyond those that would be expected by random sampling only) by computing estimates of the I2 statistic, with I2 greater than 50% indicative of severe heterogeneity. Heterogeneity sources were explored by means of meta-regression methods and subgroup analysis. Exponentials of meta-regression coefficients were interpreted as odds ratios (ORs).
Meta-analysis was performed using the rjags package of R software, version 4.0.0 (R Foundation for Statistical Computing). For each analysis, we ran at least 30 000 iterations with a burn-in of 15 000 sample iterations.
The search yielded 3218 records, of which 307 were duplicates (Figure 1). After the screening phase, 317 articles remained. A total of 77 met the inclusion criteria and were included in this systematic review. Of note, the full texts of 9 articles were not available even after contacting the authors. Eight articles were not included despite meeting the eligibility criteria of this systematic review because the participants had also been assessed in included publications with larger samples.
Of the 77 included primary studies, 43 were cohort studies (eTable 2 in the Supplement), which were quantitatively assessed by meta-analysis. The remainder were case reports, whose description is available in eTable 3 in the Supplement.
We included 30 primary studies17-46 that assessed the frequency of cefazolin allergy in series of patients with self-reported or confirmed penicillin allergy (Table 1). The included primary studies17-46 had been published between 1976 and 2020, assessing a total of 6001 participants. In 9 studies,18,20,25,34,36,38-40,44 penicillin allergy was confirmed by diagnostic testing (skin, in vitro, and/or drug challenges). In 13 studies,18,20,22,23,25,31,32,34,37-40,44 cefazolin allergy was assessed by means of skin tests or drug challenges, whereas prophylactic or therapeutic introduction of cefazolin occurred in 18 studies.17,19,21,24,26-30,33,35-37,41-43,45,46
Cefazolin allergy was identified in 44 participants, resulting in a meta-analytical frequency of 0.7% (95% CrI, 0.1%-1.7%; I2 = 74.9%) (Figure 2A). Such frequency was lower for participants with self-reported penicillin allergy (0.6%; 95% CrI, 0.1%-1.3%; I2 = 54.3%) than for those with confirmed allergic reactions (3.0%; 95% CrI, 0.01%-17.0%; I2 = 88.2%) (OR, 0.4; 95% CrI, 0.02-1.8; 92% probability of OR <1).
Thirteen studies17,19-21,25,29-31,33,35,41,42,46 exclusively assessed surgical patients (n = 3884), among whom 0.7% had confirmed allergy to cefazolin (95% CrI, 0%-3.3%; I2 = 85.5%). Unconfirmed penicillin allergy was associated with lower risk of cefazolin allergy (OR, 0.01; 95% CrI, <0.01-0.1). Ten studies17,19,21,29,30,33,35,41,42,46 described patients with unconfirmed penicillin allergy who had been exposed to perioperative cefazolin, resulting in a dual allergy frequency of 0.1% (95% CrI, 0.1%-0.3%; I2 = 13.1%).
We included 15 primary studies25,39,47-59 that assessed the frequency of penicillin allergy in series of patients with cefazolin allergy (Table 2). The included primary studies25,39,47-59 assessed a total of 146 participants. All patients had confirmed cefazolin allergy. Seven studies47,49,50,52,53,57,58 included only patients who had had anaphylactic reactions to cefazolin. Penicillin allergy was assessed by means of drug challenge or exposure in 9 studies.47,49-54,57,59 Thirteen studies25,39,48-51,53-59 assessed penicillin allergy status by skin tests.
Penicillin allergy was confirmed in 16 participants, corresponding to a meta-analytical frequency of 3.7% (95% CrI, 0.03%-13.3%; I2 = 64.9%) (Figure 2B). Lower heterogeneity was observed when assessing studies that only assessed immediate reactions (subgroup frequency, 2.8%; 95% CrI, 0.8%-4.9%; I2 = 3.4%).
Eight studies25,47,49,50,52,53,57,58 exclusively assessed surgical patients (n = 104), among whom 4.4% had confirmed allergy to penicillins (95% CrI, 0%-23.0%; I2 = 75.5%). In all surgical studies25,47,49,50,52,53,57,58 that exclusively assessed immediate reactions to cefazolin, all participants had anaphylaxis as their index reaction. Among these patients, 2.7% (95% CrI, 0.8%-3.9%; I2 = 0.9%) had confirmed allergy to penicillin.
A risk of bias graph is presented in Figure 3, whereas the risk of bias summary for each included primary study is presented in eTable 4 in the Supplement. Overall, the items with the largest number of studies classified as having a high risk of bias concerned the sampling methods and the possibility of nonassessment bias. Eight studies38,40,44,50,52,53,57,59 had all items classified as low risk of bias.
This systematic review and meta-analysis assessed the frequency of patient allergies to a natural penicillin and cefazolin. The study assessed 6001 patients with an allergy to a natural penicillin who subsequently received or were tested for an allergy to cefazolin. Forty-four of these patients were identified as allergic to cefazolin, resulting in a meta-analytical frequency of 0.7%. The frequency of dual allergy was higher among patients with confirmed penicillin allergies (3.0%) compared with patients with self-reported or unconfirmed penicillin allergies (0.6%). The meta-analytical frequency of dual allergy for surgical patients was 0.7%, with surgical patients with unconfirmed penicillin allergies having an even lower rate of dual allergy (0.1%). A total of 146 patients had an index allergy to cefazolin and subsequently received or were tested for an allergy to a natural penicillin; in this cohort, the frequency of dual allergy was 3.7% and 4.4% considering surgical patients only. These results are consistent with those from Picard et al,60 who identified a dual allergy frequency for penicillin-cefazolin of 1.3% based on 3 primary studies that involved a total of 75 patients.
Debate about cross-reactivity vs dual allergy among β-lactam antibiotics remains a contemporary issue.61 It is possible that some patients experience a cross-reactivity to the β-lactam ring structure based on the higher frequency of dual penicillin and cefazolin hypersensitivity from studies12,18,25,62 that used skin testing as their diagnostic method. However, it may be that this higher risk is only observed in patients with IgE-mediated allergies or that it may mirror false-positive results because the positive predictive value of skin tests appears limited.63 An alternative explanation of a higher dual allergy rate for patients with a confirmed penicillin allergy is multiple drug hypersensitivity syndrome. The prevalence of multiple drug hypersensitivity syndrome was 1.2%, according to a study by Blumenthal et al64 that assessed nearly 800 000 patients.
One of the most intriguing studies to assess the controversy of cross-reactivity and multiple drug allergy syndrome is from Strom et al,65 who analyzed the United Kingdom General Practice Research Database. This large case-control study found that patients with no history of drug allergy had a lower frequency of allergic reactions to sulfonamide nonantibiotics (1.1%) compared with patients with history of allergic reactions to sulfonamide antibiotics (9.9%) and patients with a history of penicillin allergy (14.2%). Because penicillins and sulfa drugs are not chemically similar, the authors concluded that the high rate of allergic reactions between sulfa antibiotics and sulfa nonantibiotics was attributable to a greater predisposition to allergic reactions in general among patients with a history of medication allergies.66,67 Although the current meta-analysis did not assess the frequency of allergic reactions to a non–β-lactam, we were able to identify 19 primary studies (among the 240 articles that were fully read in the selection stage of this systematic review) that reported on the frequency of cefazolin allergic reactions in patients with no registration of penicillin allergy. On the basis of data from 13 149 patients, we obtained a meta-analytical frequency of 0.6% (95% CrI, 0.1%-1.2%; I2 = 59.0%), close to that observed for patients with penicillin allergy. In addition, our observed low rate of dual allergy should increase practitioner confidence in using cefazolin in patients allergic to penicillins, particularly when data suggest cefazolin is the optimal antibiotic, such as perioperative prophylaxis.
In the performed meta-analyses, severe heterogeneity was observed. A meta-regression and subgroup analyses were performed to identify potential variables that could explain differences across primary studies. Higher rates of dual allergy were seen with studies from the 1970s and 1980s18,20,37 and those with small samples.25,48,49,54 We also observed that having an unconfirmed penicillin allergy and being prophylactically or therapeutically exposed to cefazolin without previous testing was associated with lower risk of developing a cefazolin hypersensitivity reaction. This finding may be attributable to a selection bias in the real-world observational studies, with practitioners selecting lower-risk patients with penicillin allergy to prophylactically or therapeutically receive cefazolin. Notably, severe heterogeneity was not identified when performing a meta-analysis limited to surgical patients who self-reported penicillin allergy and received cefazolin without previous testing. Unfortunately, there was no information available in the primary studies on several variables that may also contribute to explain heterogeneity, such as the proportion of patients with allergy to other medications or the presence of other immunologic comorbidities.
This study has strengths. To our knowledge, this is the most extensive systematic review and meta-analysis conducted to assess the frequency of cross-reactivity between natural penicillins and cefazolin and the first fully dedicated to this research question with the aim of improving perioperative antibiotic prophylaxis. One prior study60 assessed the cross-reactivity rate for penicillin-cefazolin in 3 primary studies that involved a total of 75 patients, finding a 1.3% reaction rate. Our bayesian methods allowed us to appropriately summarize data that included 0 cells. We also performed meta-regression and subgroup analyses to explore sources of heterogeneity, identifying variables that explained differences across studies. To minimize publication and information bias, we searched 3 different electronic bibliographic databases without applying exclusion criteria based on the date or language of publication and contacted authors whenever relevant information was missing. Relevance to clinical practice is the final strength of this study because the results could lead to the standardization of preoperative cefazolin in patients with unconfirmed penicillin allergies, thus decreasing surgical site infections.
The study also has limitations, mostly related to the methods of the included primary studies. The possibility of selection bias exists. It is likely that only a small percentage of patients with a natural penicillin allergy who get tested for and/or exposed to cefazolin are described in the literature. In addition, in many primary studies, not all patients reporting a penicillin allergy were prophylactically or therapeutically exposed to cefazolin, which may have biased our sample. Unfortunately, information was lacking on whether those patients who did not receive cefazolin tended to be different (eg, regarding the symptoms or severity of their index allergy) from those who were exposed to this drug. A possibility of information bias also exists. For example, surgery studies reporting only reactions in the surgery setting would miss delayed hypersensitivity reactions. The most significant limitation, however, is the lack of specific information about each allergic reaction because cross-reactivity is not plausible in patients who experience different physiologic reactions to each antibiotic (eg, anaphylaxis [IgE mediated] to cefazolin and contact dermatitis [T-cell mediated] to penicillin). Without individual-level data describing risk factors for medication allergies, including existing immunologic diseases and history of other medication allergies, we could not clarify cross-study differences or distinguish between probability of cross-reactivity and multiple drug hypersensitivity syndrome. Finally, the small number of patients with index cefazolin allergy limited meta-regression analyses.
This systematic review and meta-analysis found that cross-reactivity between penicillins and cefazolin was rare, with hypersensitivity reactions to cefazolin occurring in less than 1% of patients with unconfirmed penicillin allergy and in 3% of patients with allergy confirmation. Similar results were observed when specifically analyzing the surgical setting, with only 1 hypersensitivity reaction in each 1000 patients with unconfirmed penicillin allergy receiving cefazolin. These results suggest that most patients with a penicillin allergy history undergoing surgical prophylaxis may safely receive cefazolin. The exception may be those patients with confirmed penicillin allergy or a history of severe reactions, in whom additional care is warranted. The increased incidence of dual allergy may not be specific to cefazolin and may extend to other perioperative antibiotics. Future primary studies and systematic reviews should assess the frequency of allergic reactions to other antibiotics in surgical patients with confirmed penicillin allergy. These reassuring findings are broadly relevant to surgery because there are clear patient care consequences of avoiding cefazolin for surgical prophylaxis.
Accepted for Publication: December 17, 2020.
Published Online: March 17, 2021. doi:10.1001/jamasurg.2021.0021
Corresponding Author: Meghan N. Jeffres, PharmD, BCIDP, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of Colorado Anschutz Medical Campus, 12850 E Montview Blvd, Mail Stop C 238, V20-1212, Aurora, CO 80045 (meghan.jeffres@CUAnschutz.edu).
Author Contributions: Drs Sousa-Pinto and Blumenthal contributed equally to this work. Drs Sousa-Pinto and Jeffres had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Sousa-Pinto, Blumenthal, Courtney, Jeffres.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Sousa-Pinto, Blumenthal, Courtney, Jeffres.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Sousa-Pinto, Jeffres.
Administrative, technical, or material support: Blumenthal, Mancini, Jeffres.
Supervision: Blumenthal, Jeffres.
Conflict of Interest Disclosures: Dr Blumenthal reported receiving grants from the National Institute of Allergy and Infectious Diseases and the American Academy of Allergy, Asthma, and Immunology Foundation during the conduct of the study and grants from Crico, the risk management foundation of Harvard Medical School, outside the submitted work. In addition, Dr Blumenthal has a patent for the Beta-Lactam Allergy Clinical Decision Support Tool licensed by Persistent Systems. No other disclosures were reported.