Clostridium difficile Infection Risk with Important Antibiotic Classes: An Analysis of the FDA Adverse Event Reporting System.

Introduction: Antibiotic use is an important risk factor for Clostridium difficile infection (CDI). Prior meta-analyses have identified antibiotics and antibiotic classes that pose the greatest risk for CDI; however, CDI epidemiology is constantly changing and contemporary analyses are needed.
Objectives: The objective of this study was to evaluate the association between CDI and important antibiotic classes in recent years using the FDA Adverse Event Report System (FAERS).
Methods: FAERS reports from January 1, 2015 to December 31, 2017 were analyzed. The Medical Dictionary for Regulatory Activities (MedDRA) was used to identify CDI cases. We computed the Reporting Odds Ratios (RORs) and corresponding 95% confidence intervals (95%CI) for the association between antibiotics and CDI. An association was considered statistically significant when the lower limit of the 95%CI was greater than 1.
Results: A total of 2,042,801 reports (including 5,187 CDI reports) were considered, after inclusion criteria were applied. Lincosamides (e.g., clindamycin) had the greatest proportion of CDI reports, representing 10.4% of all lincosamide reports. CDI RORs (95%CI) for the antibiotic classes were (in descending order): lincosamides 46.95 (39.49-55.82), monobactams 29.97 (14.60-61.54), penicillin combinations 20.05 (17.39-23.12), carbapenems 19.16 (15.52-23.67), cephalosporins/ monobactams/carbapenems 17.28 (14.95-19.97), cephalosporins 15.33 (12.60-18.65), tetracyclines 7.54 (5.42-10.50), macrolides 5.80 (4.48-7.51), fluoroquinolones 4.94 (4.20-5.81), and trimethoprim-sulfonamides 3.32 (2.03-5.43).
Conclusion: All antibiotic classes included in the study were significantly associated with CDI. Lincosamides (e.g., clindamycin) had the highest CDI ROR among the antibiotics evaluated in this study.

Introduction

Clostridium difficile infection (CDI) is a great public health concern in hospital and community settings. In the first decade of the twenty-first century, United States hospitals noted a profound increase in CDI incidence 1. Since then, national standards required hospitals to implement effective infection control interventions and antimicrobial stewardship programs to prevent CDI. Nationally-representative studies now indicate that CDI rates among hospitalized patients might be declining 2. With the decline in CDI incidence in hospitals, there appears to have been a concurrent shift to community-onset CDI 3.

A rich and diverse intestinal microbiota prevents CDI; disruption of microbiota, especially due to antibiotic use, can lead to loss of colonization resistance and proliferation of C. difficile 4,5. Antibiotic exposure is the most important risk factor in both hospital and community-onset CDI 6-8. In previous meta-analyses conducted between 1988 and 2009, clindamycin, fluoroquinolones, and cephalosporins had the highest CDI risks 6-8.

Given the change in CDI epidemiology in recent years, more recent data are needed to evaluate the current CDI associations with various antibiotics. The FDA Adverse Event Reporting System (FAERS) provides recent data on CDI and antibiotics 9. The objective of this study is to evaluate CDI associations with antibiotics using FAERS data from 2015 to 2017.

Methods

Data Source

FAERS is a publicly available database organized into Quarterly Data Files, which contain adverse event reports that were submitted to United States Food and Drug Administration (FDA) 9. FAERS data include patient demographic information (age and sex), drug information (drug name, active ingredient, route of administration, and drug's reported role in the event), and reaction information. Each report lists a primary suspected drug with one or more adverse reactions and may include other drugs. Clinical outcomes, such as death and hospitalization, may also be reported.

Study Design

FAERS data from January 1, 2015 to December 31, 2017 were obtained from the FDA. Some adverse event reports were submitted multiple times with updated information. Therefore, duplicate reports were removed by case number, with the most recent submission included in the study. Reports containing drugs which were administered in oral, subcutaneous, intramuscular, intravenous, and parenteral routes were included in the study, while other routes of administration were excluded.

Drug Exposure Definition

Each antibiotic was identified in the FAERS drug files by generic and brand names listed in the Drugs@FDA Database 10. Only drugs with a reported role coded as “PS” (Primary Suspect Drug) or “SS” (Secondary Suspect Drug) were included in this study 11. Antibiotics with less than three CDI reports were excluded from the data analysis 12.

Adverse Drug Reaction Definition

FAERS defines adverse drug reactions using Preferred Terms from the Medical Dictionary for Regulatory Activities (MedDRA). MedDRA includes a hierarchy of terms, which are (from the highest to the lowest) System Organ Classes (SOC), High Level Group Term (HLGT), High Level Term (HLT), Preferred Term (PT), and Lowest Level Term (LLT). Standardised MedDRA Queries (SMQs) are groupings of MedDRA terms, usually at the PT level, which relate to an adverse drug reaction. Pseudomembranous colitis (SMQ), including Preferred Terms “Clostridial infection”, “Clostridial sepsis”, “Clostridium bacteraemia”, “Clostridium colitis”, “Clostridium difficile colitis”, “Clostridium difficile infection”, “Clostridium test positive”, “Gastroenteritis clostridial”, and “Pseudomembranous colitis” were used to identify CDI cases 13. “Clostridium difficile sepsis”, which is a Lowest Level Term, was also used in the study.

Statistical Analysis

A disproportionality analysis was performed by calculating Reporting Odds Ratios (RORs) and corresponding 95% confidence intervals (95%CI) for the association between CDI and each antibiotic class or individual antibiotic 14. ROR was calculated as the ratio of the odds of reporting CDI versus all other events for a given drug, compared with these reporting odds for other drugs present in FAERS 14. An association was considered to be statistically significant if the 95%CI did not include 1.0 (see Table 1 for the calculation of ROR and CI) 14. A higher ROR suggests a stronger association between the antibiotic and CDI. A subgroup analysis was performed on patients who were 65 years or older and patients less than 65 years old. The Cochran-Armitage Trend Test was used to assess a change in the trend of CDI reports in patients who took fluoroquinolones from 2004 to 2017. Data analysis was performed using Microsoft Access 2016, Microsoft Excel 2016 (Microsoft Corporation, Redmond, WA), SAS 9.4, and JMP Pro 13.2.1 (SAS Institute, Cary, NC).

Table 1

A two by two contingency table for a drug (A) - ADR (X) combination

	ADR (X)	Other ADRs	Total
Drug (A)	a	b	a+b
Other drugs	c	d	c+d
Total	a+c	b+d	a+b+c+d

† ADR = adverse drug reaction; ROR = (a/b)/(c/d); 95% Confidence Interval (CI) = eln(ROR)±1.96√(1/a+1/b+1/c+1/d)

Results

After inclusion and exclusion criteria were applied and duplicate reports were removed, FAERS contained a total of 2,042,801 reports from January 1, 2015 to December 31, 2017. There were 5,187 CDI reports from 2015 to 2017, which were included in the data analysis. Female patients represented 61% of CDI patients who had gender information. CDI patients who had age information had a median age (IQR, interquartile range) of 62 (27) years. Please see Table 2 for the gender and age information of patients who were taking various antibiotics.

Table 2

Gender and age information for patients on antibiotics

Antibiotic Class/Antibiotic	% Female	Median age (IQR)
Lincosamides (clindamycin)	58	58 (28)
Monobactams (aztreonam)	56	55 (35)
Penicillin combinations	48	62 (28)
Piperacillin-tazobactam	39	64 (25)
Amoxicillin-clavulanate	54	60 (32)
Ampicillin-sulbactam	46	67 (33)
Carbapenems	44	63 (29)
Meropenem	44	61 (32)
Ertapenem	44	69 (24)
Imipenem-cilastatin	46	63 (27)
Cephalosporins, monobactams, and carbapenems	47	63 (34)
Third/fourth-generation cephalosporins	49	62 (39)
Cefepime	44	64 (20)
Cefotaxime	39	44 (61)
Ceftriaxone	51	63 (41)
Tetracyclines	60	51 (35)
Tetracycline	60	26 (32)
Doxycycline	60	51 (35)
Macrolides	61	54 (35)
Erythromycin	63	55 (26)
Clarithromycin	61	55 (33)
Azithromycin	59	48 (41)
Fluoroquinolones	58	58 (27)
Ofloxacin	43	68 (24)
Ciprofloxacin	57	57 (28)
Levofloxacin	61	59 (24)
Moxifloxacin	55	55 (26)
Trimethoprim-sulfamethoxazole	45	60 (28)

† IQR = interquartile range

The lincosamide class had the highest CDI ROR (46.95, 95%CI: 39.49-55.82) among all antibiotic classes included in the study (Figure 1). Clindamycin was the only antibiotic in the lincosamide class which met the inclusion criteria. The monobactam class (including aztreonam only) demonstrated the second highest CDI ROR (29.97, 95%CI: 14.60-61.54). The CDI ROR of the trimethoprim-sulfonamides class was the lowest (3.32, 95%CI: 2.03-5.43).

Figure 1

Reporting Odds Ratios (RORs) for † CI = confidence interval; CDI = Clostridium difficile infection

Among patients who took penicillin combinations, carbapenems, cephalosporins, tetracyclines, macrolides, fluoroquinolones, and trimethoprim-sulfamethoxazole, patients who were 65 years or older had a higher CDI ROR than those less than 65 years old (Figure 2). Among patients who took lincosamides, patients who were 65 years or older had a lower CDI ROR than those less than 65 years old.

Figure 2

Reporting Odds Ratios (RORs) for † CI = confidence interval; CDI = Clostridium difficile infection; yrs = years

The Cochran-Armitage Trend Test demonstrated that there was a significant relationship between the proportion of CDI reports in patients who took fluoroquinolones and the year of reporting (p<0.0001). From 2004 to 2010, 2.3% of fluoroquinolone reports had CDI. From 2011 to 2017, 1.7% of fluoroquinolone reports had CDI.

Discussion

Our antibiotic CDI association rank order was similar to previous meta-analyses 6-8. Our results demonstrated significant CDI associations (from strongest to weakest) with lincosamides, monobactams, penicillin combinations, carbapenems, cephalosporins, tetracyclines, macrolides, fluoroquinolones, and trimethoprim-sulfonamides.

In a prior meta-analysis of antibiotics and the risk of community-associated CDI (CA-CDI), the risks from the highest to the lowest were: clindamycin, fluoroquinolones, CMCs, macrolides, trimethoprim- sulfonamides, and penicillins, with no effect of tetracycline on CDI risk 6. In another prior meta-analysis of CA-CDI and antibiotics, the risks from the highest to the lowest were: clindamycin, fluoroquinolones, cephalosporins, penicillins, macrolides, and trimethoprim-sulfonamides, while no association was found between tetracyclines and CDIs 7. Regarding hospital-acquired CDI (HA-CDI), a prior meta- analysis indicated that the associations from the strongest to weakest were: third-generation cephalosporins, clindamycin, second-generation cephalosporins, fourth-generation cephalosporins, carbapenems, trimethoprim-sulfonamides, fluoroquinolones, and penicillin combinations 8. FAERS data do not specify whether CDI is community-associated or hospital-acquired; therefore, our results are likely a mixture of CA-CDI and HA-CDI.

The higher CDI RORs associated with clindamycin, penicillin combinations, and carbapenems may be due to their activity against anaerobes and disruption of gut flora 15. Clindamycin had the highest CDI ROR in our study, which is consistent with the highest CDI risks associated with clindamycin in prior meta-analyses 6,7. Piperacillin- tazobactam had the second highest ROR in our study; the reasons might include the broad-spectrum antimicrobial activity of piperacillin-tazobactam and the great extent of gut flora disruption as a result 16,17. Trimethoprim-sulfonamides had the lowest CDI ROR among the antibiotic classes included in our study. In previous meta-analyses, trimethoprim-sulfonamides also had one of the lowest CDI risks 6-8.

Our results demonstrated that fluoroquinolones had a weaker association with CDI compared with most of the antibiotic classes included in the study, except for trimethoprim-sulfonamides. Prior meta- analyses have implicated fluoroquinolones as one of the highest risk antibiotics for CDI 6,7; however, these studies used data during the CDI epidemic that was associated with the fluoroquinolone-resistant ribotype 027 Clostridium difficile strain 18,19. A more recent meta-analysis by Vardakas et al. did not implicate fluoroquinolones as one of the highest risk antibiotics, which is consistent with our findings 20. Given that ribotype 027 strains are now endemic in healthcare settings, our data suggest that fluoroquinolones might not be as important of a CDI risk factor as before considering the recent changes in CDI epidemiology 21. A recent article published in 2017 demonstrated that a concomitant decline in inpatient fluoroquinolone use and the NAP1/027 strain may have contributed to the decrease in the incidence rate of long-term-care facility-onset CDI from 2011 to 2015 22. Our results from the Cochran-Armitage Trend Test also indicated that there was a trend of decrease in CDI risk with fluoroquinolones from 2004 to 2017.

In the subgroup analysis, the CDI ROR rank order in both subgroups (< 65 years old and ≥ 65 years old) were similar to that in all patients. Our results showed that older patients had a higher CDI ROR among most of the antibiotic classes analyzed (Figure 2). It is known that CDI risk is higher in patients 65 years or older 23.

Knowledge of the CDI risk associated with antibiotic classes has important implications for antimicrobial stewardship. Therapeutic interchanges could be identified, especially for those patients who have a high baseline risk for CDI (e.g., elderly, frequent hospitalizations, and comorbid conditions). For example, to treat non-severe purulent skin and skin structure infections in patients with a high risk of CDI, trimethoprim-sulfamethoxazole could be preferred to clindamycin, considering the much lower CDI ROR of trimethoprim-sulfamethoxazole 24.

Limitations

A causal relationship between a drug and an adverse drug reaction (ADR) cannot be established by FAERS. The spontaneous and voluntary reporting of ADRs may lead to significant bias due to underreporting and lack of overall drug use data 25,26. The association between a drug and an ADR is confounded by concomitant drugs and comorbidities. Media attention and recent drug approval might affect the reporting behaviors. Furthermore, epidemiological shift in the circulating C. difficile strains in the United States might account for the weaker association between fluoroquinolones and CDI in our study; however, the FAERS study design does not permit us to investigate this hypothesis. Therefore, we believe the next step in this line of research will be to confirm these findings in a future case-control or cohort study.

Conclusions

All antibiotic classes evaluated in the study were significantly associated with CDI. Lincosamides (e.g., clindamycin) had the highest CDI ROR and trimethoprim-sulfonamides had the lowest CDI ROR of all the antibiotic classes investigated in this study. Results from FAERS should be interpreted with caution in the context of data limitations. Antibiotic stewardship is needed to prevent CDI and to improve health outcomes.