To the Editor—The prevalence of extended-spectrum β-lactamase (ESBL)–producing Escherichia coli is increasing worldwide, and infectious disease caused by ESBL-producing E. coli is associated with increased hospitalization and mortality rates. ^{Reference Karanika, Karantanos, Arvanitis, Grigoras and Mylonakis1,Reference Chong, Shimoda and Yakushiji2} The high consumption of broad-spectrum oral antibiotics is a serious issue because it likely increases the prevalence of ESBL-producing E. coli. ^{Reference Kimura, Fukuda and Hayakawa3} However, in Japan, these oral antibiotics, which include macrolides, quinolones, and third-generation cephalosporins (3GCs), account for >90% of all antibiotic consumption. ^{Reference Tsuzuki, Kimura, Ishikane, Kusama and Ohmagari4} Here, we compared the isolation ratio of ESBL-producing E. coli and oral antibiotic consumption in rural Japan.

Data on the characteristics of participating hospitals, amount and days of oral antibiotic prescriptions for outpatients, number of outpatients, and isolation of E. coli for outpatients from January to December 2018 were retrospectively requested from hospitals via the Consortium Against Health Care-Associated Infection in Niigata. This organization is a voluntary organization composed mainly of large-scale hospitals in Niigata Prefecture that aims to improve infection control and prevention. Niigata Prefecture has a population of ~2.2 million. This study was approved by the Ethics Committee of Kaetsu Hospital (no. 2019-006).

Antibiotics were classified as penicillins, first-generation cephalosporins, 3GCs, macrolides, quinolones, tetracyclines, and other. Antituberculosis agents were excluded. Antibiotics were standardized by the defined daily dose (DDD) per 1,000 patient days and days of therapy (DOT) per 1,000 patient days. The DDD was calculated using the 2018 World Health Organization anatomic therapeutic chemical (ATC)/DDD index. The DDD/DOT ratio was calculated as the DDD value divided by the DOT value. Duplicate isolates of E. coli from the same patient were deleted, and ESBL-producing E. coli were determined according to the screening and confirmation tests suggested by the Clinical and Laboratory Standards Institute. The isolation ratio of ESBL-producing E. coli was calculated as the number of isolated ESBL-producing E. coli divided by the number of all isolated E. coli. ^{Reference Chong, Shimoda and Yakushiji2}

Simple linear regression analysis was performed to examine the relationship between antibiotic consumption and the ESBL-producing E. coli isolation ratio. The factors that demonstrated a certain degree of association (P < 0.3) were then included as explanatory variables in multiple linear regression analysis. Antibiotics classified as “other” were not included in the multivariate analysis. All analyses were performed with R version 3.4.1 software (R Foundation for Statistical Computing, Vienna, Austria). Correlation coefficients (R) and P values were calculated in the simple linear regression analysis. The partial regression coefficient (standard error) and adjusted R ² values were calculated in the multiple linear regression analysis. Statistical significance was set at P < .05.

Data were collected from 13 hospitals (41%) of all 32 hospitals. Table 1 shows only the DDD/DOT ratio of 3GC had a significant negative association with the ESBL-producing E. coli isolation ratio (adjusted R² = 0.45). Supplemental Material 1 shows a summary of the data collected from the 13 participating hospitals. Supplemental Material 2 shows antibiotic consumption and the results of simple linear regression analysis between antibiotic consumption and the ESBL-producing E. coli isolation ratio.

Table 1. Multiple Linear Regression Analysis Between Antibiotic Consumption and the ESBL-Producing E. coli Isolation Ratio

Note. ESBL, extended-spectrum beta-lactamase; DDD, defined daily dose; DOT, days of therapy.

^a Values shown are partial regression coefficients (standard error); blank spaces indicate values not selected in the model.

^b Multiple linear regression analysis.

To the best of our knowledge, this is the first report to clarify that the inappropriate dosing of oral 3GCs might be increasing the prevalence of ESBL-producing E. coli. Our results show that only the DDD/DOT ratio of 3GCs had a significant negative association with the isolation ratio of ESBL-producing E. coli, indicating that a low daily dose of a 3GC was associated with an increased ESBL-producing E. coli isolation ratio. In Japan, the approved maintenance dose of antibiotics are often lower than the DDD defined by the World Health Organization. ^{Reference Muraki, Kitamura and Maeda5} In addition, the bioavailability of 3GCs is generally low, and the bioavailability of cefcapene and cefditoren, which were mainly used in this study, was reported as 20%–30%. ^{Reference Tanimura, Uchiyama and Onishi6,Reference Sádaba, Azanza and Quetglas7} Thus, the serum concentration of 3GC can easily become very low when used at a low dose, and the exposure of bacteria to the mutant selection window of the 3GC is likely increased. Even though the mutant selection windows of these oral 3GCs have not been reported, when the dose of cefcapene or cefditoren is reduced, the serum concentration likely reaches the mutant selection window a few hours after administration.

Our study has some limitations, including a small sample size and being a retrospective study conducted in a single region. Additionally, antibiotics for children and adults were assessed without distinction, and elderly patients with suspected renal dysfunction might be administered a reduced dose of antibiotics, which may have influenced the low DDD and DDD/DOT values. Therefore, a larger study is warranted to confirm our findings. Nevertheless, our findings suggest that the inappropriate use of oral 3GC might be increasing the prevalence of ESBL-producing E. coli.