The Clinical Impact of Antibiotic Allergy Labels on One‐Year Outcomes of Solid Organ Transplant Recipients

TL;DR

This study shows that incorrect antibiotic allergy labels may lead to worse outcomes for organ transplant patients, including longer hospital stays and more infections.

Contribution

The study is the first to analyze the impact of antibiotic allergy labels on post-transplant outcomes in a large cohort.

Findings

Patients with a penicillin allergy label were more likely to test positive for Clostridioides difficile.

Allergy-labeled patients used more alternative antibiotics and had longer hospital stays.

Antibiotic allergy labels may be a modifiable risk factor for poor transplant outcomes.

Abstract

Antibiotic allergy labels (AALs) are common and often incorrect. They have many potential impacts, including the use of broader‐spectrum antibiotics and suboptimal treatment of infections. The impact of inaccurate allergy labels on post‐transplant outcomes in the solid organ transplant population is not well described. We performed a retrospective review of 2,373 consecutive solid organ transplants occurring between 2011 and 2021, to analyze the impact of AALs, specifically penicillin, on outcomes in the first year after transplantation. Three hundred and twenty‐two patients (13.6%) had a penicillin allergy label, while 572 patients (24%) had at least one antibiotic allergy label. Patients with a penicillin allergy label were more likely to have a positive Clostridioides difficile test (p = 0.021). Patients with allergy labels also had significantly more utilization of alternative antimicrobial agents (p < 0.001) and longer inpatient hospital durations (p = 0.013). This study suggests that AALs may be a risk factor for inferior outcomes after solid organ transplantation and could represent a modifiable target for pre‐transplant optimization.

Tables

TABLE 1. Demographics and baseline characteristics with PAL association.

Patient characteristic	Overall (N = 2373)	No‐PAL (N = 2051 86.4% of total)	PAL (N = 322, 13.6% of total)	p‐value
Age, Mean (SD)	54.2 years (13.3)	54.2 years (13.2)	54.8 years (13.5)	0.393
Gender, No. (%)	906‐ Female 38.2% 1467‐ Male 61.8%	752 (36.6%) Female	154 (52.2%) Female	< 0.001
Transplant type, No. (% of total)
Kidney	1080 (45.5%)	951 (46.4%)	129 (40.1%)
Liver	599 (25.2%)	520 (25.4%)	79 (24.5%)
Lung	408 (17.2%)	331 (16.1%)	77 (23.9%)	0.014
Heart	216 (9.1%)	187 (9.1%)	29 (9.0%)
Pancreas	70 (2.9%)	62 (3.0%)	8 (2.5%)
Comorbidities, No. (% of total)
Pre‐transplant diabetes	753 (31.7%)	642 (31.3%)	111 (34.5%)	0.274
Pre‐transplant obesity	82 (3.5%)	64 (3.1%)	18 (5.6%)	0.024
Allergy label, No. (% of total)
Penicillin	322 (13.6%)
Quinolone	101 (4.3%)
Cephalosporin	95 (4.0%)
Macrolide	68 (2.9%)
Vancomycin	59 (2.5%)
Sulpha	45 (1.9%)
Tetracycline	32 (1.3%)
Antifungal or antiviral	21 (0.9%)
Other antimicrobials a	98 (4.1%)
More than one antimicrobial allergy	572 (24.1%)

TABLE 2. Univariate and multivariate analysis of penicillin allergy and one‐year primary and secondary outcomes.

			Univariate analysis			Multivariate analysis
Clinical outcomes	No‐PAL (N = 2051)	PAL (N = 322)	OR	95% CI	p‐value	OR	95% CI	p‐value
Positive C. Difficle PCR	170 (8.2%)	43 (13%)	1.71	1.19–2.44	0.005	1.54	1.07–2.22	0.021
Death	127 (6.2%)	25 (7.7%)	1.28	0.82–1.99	0.272	1.16	0.73–1.85	0.519
Graft failure	144 (7%)	29 (9%)	1.31	0.86–1.99	0.205	1.22	0.79–1.87	0.371

Full text

Introduction

1

Antibiotic allergy labels (AALs), specifically penicillin allergy labels (PALs), are prevalent yet frequently inaccurate. The prevalence of PALs ranges from 6%–25%, with the highest rates recorded in patients admitted to the hospital [1, 2]. However, greater than 90% of these patients are found to be non‐allergic when verified through skin testing or drug challenge [3]. This is due to a variety of factors, including historical manufacturing practices, decreasing IV penicillin utilization, waning allergenicity over time, and side effect or allergy misattribution [4, 5, 6, 7].

Although PALs naturally lead to penicillin avoidance, they often trigger broader, unnecessary avoidance of all β‐lactams by prescribers, despite low rates of cross reactivity [8]. Compared to β‐lactam antibiotics such as penicillins or cephalosporins which are often first‐line agents, alternative antibiotics such as fluoroquinolones, carbapenems and vancomycin are associated with significant complications. These include the emergence of drug‐resistant organisms, a higher risk of Clostridioides difficile infection (CDI), higher costs and duration of inpatient hospitalizations, and increased failure of antibiotic therapy [8, 9, 10, 11, 12]. These complications can be mitigated by the removal of inaccurate AALs, or ‘de‐labeling’. While the safety and benefits of de‐labeling programs (i.e., cost savings and reduced length of stay) are well documented in the general population, data in specialized populations remain sparse [13].

Solid‐organ transplant (SOT) recipients represent a uniquely high‐risk population due to their immunocompromised state. Although SOT outcomes have improved over time infections, specifically those caused by antibiotic‐resistant organisms and CDI, remain a major cause of morbidity and mortality [14, 15, 16, 17, 18]. SOT recipients utilize antibiotics more frequently and possess higher rates of AALs than the general population [19, 20, 21, 22]. Despite this, the impact of AALs on clinical outcomes in the SOT population remains poorly defined. Existing studies, which have focused on the very early post‐transplant period, suggest that β‐lactam allergy labels increased alternative antibiotic use but were unable to define an impact on clinical outcomes including mortality, CDI, or the duration of hospitalization [21, 22].

CDI is a particular concern in SOT with a 5‐10 percent prevalence and a hospital onset incidence 5 times higher than the general population [23, 24]. AALs exacerbate this risk by necessitating alternative agents which are strongly associated with CDI risk. Given CDI is associated with up to a threefold increase in mortality, prolonged hospital stays, and higher healthcare costs, understanding the role of AALs in this context is critical [25, 26]. In this context we hypothesized that AALs, specifically PAL may represent an independent risk factor for adverse outcomes, particularly CDI. Here, we describe the prevalence of AALs in SOT recipients at our institution and evaluate their impact on one‐year clinical outcomes.

Methods

2

Study Setting, Design and Population

2.1

We performed a retrospective cohort study, including patients undergoing their first single organ transplant at the University of Minnesota over a nine‐year period between 1/1/2012 and 1/1/2021. The University of Minnesota is a multiorgan transplant center that performs and provides pre‐ and post‐ operative care for heart, lung, liver, kidney, pancreas, small bowel and multiorgan transplant candidates and recipients. Data from the first year following transplantation was reviewed. Patients aged less than 18, undergoing a second or combined transplant (other than pancreas along with kidney) or those who opted out of research were excluded. All adult single organ recipients without exclusion criteria were included. This study was approved by the University of Minnesota IRB and adhered to ethical principles of research as outlined in the declaration of Helsinki.

Data Sources, Study Definitions and Outcomes

2.2

Our institutional database consists of major demographic, clinical, and outcomes data and is prospectively maintained by analysts to obtain longitudinal data on patients. Key clinical and quality metrics such as duration and number of hospitalizations, death and graft loss in the first year after transplant is specifically captured. Graft loss was defined as per UNOS criteria [27]. Data on key baseline (pretransplant) co‐morbidity rates of diabetes, ascites, chronic obstructive pulmonary disease, end stage renal disease, obesity and cystic fibrosis were additionally obtained from the database. Allergy labels present on the date of transplantation were obtained from the medical record and categorized into antibiotic classes by an infectious disease trained physician. Data regarding CDI testing was obtained from the medical record and dually sourced from an infection prevention database, and any positive C. difficle polymerase chain reaction (PCR) test occurring in the first year after the date of transplantation was included. C. difficle PCR testing during the entire study period was performed on the Cepheid Xpert platform (Cepheid, Sunnyvale, California). C. difficile testing at our institution has been regulated by institutional practice guidelines and mandatory electronic medical record checks, reserving it for only symptomatic patients who do not have an alternative explanation for diarrhea, such as recent stool softener use (Supplementary Data Appendix A). Inpatient antibiotic usage data for the first year after transplant was also obtained from the medical record, including drug, dose, and duration.

Our primary clinical outcome of interest was CDI in the first year after transplant. We chose C. Difficle PCR test positivity as a surrogate marker for CDI given our institutional protocols limiting testing only to patients with a high pre‐test probability for true CDI. Secondary clinical outcomes included mortality and graft loss in the first year after transplant, the duration of the index hospitalization in which the transplant occurred, the total number of inpatient days in the first year after transplantation, and the total inpatient days of antibiotic therapy for fluoroquinolones, carbapenems, vancomycin and aztreonam.

Statistical Analysis

2.3

Data was analyzed using SPSS version 26.0 (IBM, Armonk, New York). Variables with any missing data were not included in our analysis. The baseline characteristics of the cohort were compared using Pearson's χ^2^ test for categorical variables and Mann‐Whitney‐U test for continuous variables given the non‐normal distribution. To determine associations, bivariate (χ^2^and t‐test) and multivariate (logistic and linear) regression analysis were performed on prespecified categorical and continuous variables respectively. Co‐variates used for all multivariable analysis were transplant type, age at time of transplant, gender, pre‐transplant diabetes and pre‐transplant obesity. These covariates were selected based on prior literature, clinical justification, and/or a p value of less than 0.05 in the bivariate analysis [28, 29]. All models were checked for multicollinearity using variance inflation factors. We performed survival analysis using the Kaplan–Meier method to estimate time‐to‐event outcomes. A p‐value of less than 0.05 was considered significant.

Results

3

Study Population and Demographics

3.1

We obtained data from 2,373 consecutive SOT recipients between 2012 and 2021. The baseline characteristics and demographics of this cohort are shown in Table 1. The most common AAL was penicillin, seen in 322 (13.6%) patients. In our cohort 572 (24.1%) patients had 1 AAL and 161 (6.8%) had >1 AAL (Table 1). Patients with a PAL were statistically more likely to be female, have received a lung transplant, and be obese pre‐transplant. There was no significant difference between rates of diabetes or age for patients with or without a PAL.

Associations Between Penicillin Allergy Label and Clinical Outcomes

3.2

Primary Outcome: Clostridioides difficile PCR Positivity as a Surrogate for CDI

3.2.1

The results of bivariate and multivariate analyses for the association between pre‐transplant PAL and recipient outcomes are shown in Table 2. Having a PAL was strongly associated with C. difficile PCR positivity in the post‐transplant period. The significance of the association between a positive C. difficile PCR and the presence of a PAL persisted in the multivariate model. Those with a PAL had an odds ratio of 1.54 for having a positive C. difficile PCR in this model (95% CI 1.07–2.22, p‐value = 0.021). C. difficile PCR positivity by itself was not associated with a statistically significant increase in death (OR 1.30, CI of 0.77–2.2, p = 0.30) or graft loss (OR 1.36, CI 0.83–2.21, p = 0.21) in the first year.

Secondary Outcomes

3.2.2

Death and Graft Failure

3.2.2.1

Death and graft failure trended higher in those with a PAL compared with those without, but it was not statistically significant in either the univariate or multivariate analysis (Table 2). The presence or absence of PAL resulted in no difference in the timing of first year death or graft failure among those patients in which this outcome occurred.

Antibiotic Usage

3.2.2.2

Measured alternative antibiotic use was significant in this cohort, with 1222 (51.5%) receiving vancomycin, 1040 (43.8%) receiving a fluoroquinolone, 509 (21.4%) a carbapenem and 33 (1.4%) Aztreonam. Among the 322 PAL patients there was significantly increased inpatient utilization of alternative antibiotics in the first year after transplant. In the univariate and multivariate analysis, we reviewed quinolone, carbapenem, vancomycin, and aztreonam days of therapy (DOT) (Table 2). In the univariate and multivariate analyses quinolone, carbapenem, and vancomycin had statistically significant longer DOT in those with a PAL. Among the antibiotic groups reviewed, carbapenems were associated with the most significant increase in DOT with a mean increase of 3.05 days (95% CI 2.0–4.08 days) in those with a PAL compared to those without. This result remained consistent in the multivariable analysis (OR 2.99, 95% CI 1.95–4.03, p‐value < 0.001).

Hospitalization Within the First Year After Transplant

3.2.2.3

PAL was associated with a statistically significant longer index hospitalization and a greater amount of time spent in the hospital within the first year after transplant in the univariate and multivariate analyses (Table 2). The effect was strongest on the duration of the index hospitalization, accounting for a mean increase of 5.2 additional days in our multivariable model (OR 3.48, 95% CI 2.30–8.27, p = 0.001). This effect persisted for the entire first year with an additional 2.5 days spent in the hospital on average (95% CI 1.07–9.15, p value = 0.013).

Supplementary Analyses

3.3

Analysis of Other Allergy Categories

3.3.1

Comparable results were seen when analyzing patients with a β‐lactam allergy (penicillin and/or cephalosporin and/or carbapenem), or any antibiotic allergy label versus no antibiotic allergy label (Supplementary Tables 1–4). In these supplementary analyses, a higher prevalence of allergies was seen in the lung transplant population, with significantly increased C. difficile PCR positivity, and alternative antibiotic usage rates and a numerical trend to higher mortality and graft loss in the first year using the same multivariable model. The effect on the duration of hospitalization persisted.

Analysis of Trends in Variables Between 2012–2021

3.3.2

There were no significant changes to baseline cohort variables, antibiotic usage, CDI PCR positivity rates, hospital durations, or PAL rates during the study period (Supplementary Table 5)

Discussion

4

Our study adds to the growing body of literature by highlighting the potential negative impact of AALs on transplant outcomes. To our knowledge this is the largest study with the longest follow‐up to assess the cumulative impact of multiple antibiotic courses during the highest‐risk first year after transplant. By considering the entirety of this critical period, we were able to demonstrate several associations with negative clinical impacts on C. difficle PCR positivity, antibiotic utilization and the duration of hospital stays. Our longer follow‐up period may have enabled us to capture the downstream effects of penicillin avoidance that were not evident in prior studies with shorter follow up.

The majority of prior studies with SOT patients have had either small or mixed cohorts, some including non‐transplant patients. First in a retrospective cohort study of 313 liver transplant recipients, Khumra, et al. reported higher rates of CDI and resistant gram‐negative infections among those with antibiotic allergies [30]. In a large retrospective study utilizing de‐identified ICD‐9 codes from the national inpatient sample, involving 15,489 kidney transplant recipients Nelson, et al. reported a higher cost of hospitalization but did not find an impact on clinical outcomes [20]. The two largest prior studies, by Imlay, et al. and Zhang, et al. failed to show a significant impact of β‐lactam allergy labels on clinical outcomes such as CDI or mortality within the first 100‐180 days but highlighted similar significant impacts on antibiotic prescribing [21, 22]. These prior studies did not report data on or showed no effect of, β‐Lactam allergy labels on the duration of initial or early hospitalization. In contrast, our larger study with longer follow up and a more homogenous population was able to demonstrate an effect. Our findings could be due to the duration of follow‐up, as well as differing institutional practices, transplant protocols, and patient populations.

The high rates of AAL among our population align with this existing literature showing an increased prevalence of allergy labels in the SOT population. Similarly, our data on increased alternative antibiotic usage is in line with previously published data from other centers. Both Imlay, et al. and Zhang, et al. showed an increase in inpatient usage of alternative antibiotics in their patients with β‐lactam allergies. While these studies showed the strongest effect on aminoglycoside use with a modest effect on carbapenems, our study was slightly different, perhaps because of differing institutional practices. The significantly increased use of carbapenems in PAL patients was particularly interesting given that carbapenems themselves are also β‐lactams, highlighting the non‐evidence‐based avoidance and use of certain classes of antibiotics in patients with AALs. Our weak association between PAL and aztreonam use can perhaps be attributed to extremely low utilization of this antibiotic at our center. While we were unable to obtain specific costs, we infer that the cost would have been much larger for our PAL patients due to increased duration of hospitalization, utilization of alternative antibiotics and possible higher CDI rates. This highlights the powerful potential impact allergy assessment can have in antimicrobial stewardship programs.

We acknowledge that our study is not without limitations. First, AALs may serve as a surrogate marker of overall infectious and post operative risk. Conceivably, higher risk patients with complex pre‐transplant infectious histories have had a higher chance of accumulating AALs, compared with lower risk patients. Second, the retrospective nature of our database review restricted our analysis to discrete variables present in the electronic medical record, precluding the use of patient level datapoints. For example, despite our institutional diagnostic stewardship efforts not all positive C. Difficle PCR tests equate to a true infection, some would have reflected colonization. Additionally, while our penicillin allergic and nonallergic cohorts were well matched based on our available metrics, confounding variables may still remain. Furthermore given this was a single center study the findings may not be fully generalizable due to variations in practices and patient populations across institutions among other confounders. Finally, although we observed while no major shifts in antibiotic usage or baseline demographics over the study timeframe, the effect of evolving clinical practices and guidelines on individual cases across the various transplant teams could not be quantified.

Conclusion

5

Our findings suggest several negative clinical impacts associated with PALs in the SOT population including higher rates of C. difficile PCR positivity, longer hospitalizations, and increased use and duration of alternate antibiotics. While the retrospective nature of this study precludes a definitive causal link, this data highlights the clinical burden associated with AALs which are often inaccurate. In an era where there is continued focus on improving the outcomes of SOT patients our results propose that antibiotic allergy delabeling, specifically targeting penicillin, should be key component of the pretransplant evaluation and optimization process. Future multi‐center studies with prospective data collection, or investigation of the effect of allergy delabeling programs could validate our findings and guide efforts at improving outcomes in this high‐risk patient population.

Funding

The authors have nothing to report.

Disclosure

The Authors have no disclosures or conflicts of interest.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Demographics and baseline characteristics with any AAL association. Table S2: Univariate and multivariate analysis of any antibiotic allergy and one‐year primary and secondary outcomes. Table S3: Demographics and baseline characteristics with a BLAL association. Table S4: Univariate and multivariate analysis of beta‐lactam allergy label and one‐year primary and secondary outcomes. Table S5: Trends in cohort factors over time. Appendix A: University of Minnesota Diarrhea/C. Difficile clinical decision support. Appendix B: Details of “other antimicrobial” allergies.