Letter to the Editor Regarding “Safety of Short‐Term Trimethoprim–Sulfamethoxazole Use for Uncomplicated Cystitis: A Nationwide Retrospective Cohort Study”

We read with great interest the article by Taniguchi et al., which investigated the short-term safety of trimethoprim–sulfamethoxazole (TMP–SMX) compared with fluoroquinolones in patients with uncomplicated cystitis using a large Japanese claims database 1. We thank the authors for addressing this clinically important question, which has direct implications for antibiotic stewardship in Japan. While we appreciate this valuable contribution, we believe that the conclusion that “TMP-SMX may represent a reasonable first-line option for acute uncomplicated cystitis” is inadequately supported by the study design or data, particularly with respect to the evaluation of rare but serious hypersensitivity reactions. We have two major methodological concerns. First, the characterization of prior work requires clarification. Taniguchi et al. state that “previous retrospective pharmacovigilance and observational studies did not adjust for confounding factors” 1, citing our study 2 among others. However, our study used a case-crossover design 2, which inherently controls for time-invariant confounders, including genetic predisposition such as human leukocyte antigen alleles 3, 4, and was chosen to address confounding as rigorously as possible within the constraints of the limited sample size. Second, and more fundamentally, the study by Taniguchi et al. 1 targets a different estimand from that of our prior work 2 and is not well suited to evaluating the risk of Stevens-Johnson syndrome (SJS)/toxic epidermal necrolysis (TEN). The 7-day follow-up is inadequate for evaluating drug-induced hypersensitivity 1. As the authors acknowledge 1, exanthematous drug eruptions typically occur 4–14 days after treatment initiation, and SJS/TEN generally develops within 4–28 days after exposure 5, 6. Consistent with this latency, our study and others used hazard periods of 28–56 days 2, 7, 8. Although treatment for uncomplicated cystitis is brief, the latency of drug-induced SJS/TEN extends beyond the treatment period itself. Accordingly, a minimum follow-up of 28 days would have been more appropriate 2, 7, 8, and the rationale for selecting a 7-day risk window should be clarified in the Methods. The study is also fundamentally underpowered for the safety question it addresses 1. Given the extremely low incidence of SJS/TEN, the sample of only 1071 TMP-SMX users, compared with 49 702 fluoroquinolone users was insufficient to detect this outcome. Based on the 56-day cumulative incidence of SJS/TEN in our previous study, 67.4 per 100 000 for TMP-SMX and 1.4 per 100 000 for quinolones 2, the expected number of events would have been less than one in either group even with longer follow-up, and effectively zero within the 7-day window used here. Consistently, no events were observed in either group 1. Moreover, because both TMP-SMX and fluoroquinolones are associated with elevated SJS/TEN risk 2, a head-to-head comparison would be expected to yield a smaller difference in risk than comparison with an unexposed reference, further reducing the ability to detect a meaningful difference. This limitation is compounded by the identification of SJS/TEN using ICD-10 diagnosis codes alone (table S3 in Taniguchi et al. 1), an approach with lower sensitivity (61.5%) than validated algorithms incorporating clinical course and mucocutaneous care (76.9%) 9. Taken together, these characteristics indicate that Taniguchi et al. and our prior study do not provide directly comparable evidence 1, 2. Rather than being contradictory, the two studies address different populations, follow-up windows, and underlying safety questions: a selected population of women with uncomplicated cystitis over 7 days 1 versus a broader population of antibiotic users over 28–56 days 2. In this context, the absence of a detectable difference is more likely to reflect limited follow-up and insufficient power than evidence of comparable safety. We therefore suggest that the conclusions be interpreted cautiously and that further studies with longer follow-up, larger sample sizes, and validated ascertainment of SJS/TEN are needed before these findings can inform prescribing decisions. The authors have nothing to report. Toshiki Fukasawa was previously employed by the Department of Digital Health and Epidemiology at Kyoto University with support from Eisai Co., Ltd. and Kyowa Kirin Co., Ltd.; he has received research funding from AstraZeneca K.K. and the Pfizer Health Research Foundation, consulting fees from Asahi Kasei Pharma Corporation, JMDC Inc., MSD K.K., and Real World Data Co., Ltd., and honoraria from DeSC Healthcare Inc., the Foundation for Biomedical Research and Innovation at Kobe, MeDiCU Inc., and the Research Institute of Healthcare Data Science; and his spouse is employed by MSD K.K. Hisashi Urushihara has received research funding from EPS Corporation, Senju Pharmaceutical Co., Ltd., and Shionogi he is also an editor/editorial board member of Pharmacoepidemiology and Drug Safety and a co-author of this article. He was excluded from editorial decision-making related to the acceptance of this article for publication in the journal. Takayuki Okura declares no conflicts of interest. Koji Kawakami has received research funding from AstraZeneca K.K., Eisai Co., Ltd., Kyowa Kirin Co., Ltd., OMRON Corporation, and Toppan Inc.; consulting fees from Advanced Medical Care Inc., JMDC Inc., LEBER Inc., Santen Pharmaceutical Co., Ltd., Shin Nippon Biomedical Laboratories Ltd., and Ubicom Holdings Inc.; executive compensation from Cancer Intelligence Care Systems Inc. and Mytreya Inc.; and honoraria from Kyoto University Original Co., Ltd., Pharma Business Academy Co., Ltd., SANSHO Co., Ltd., Shionogi & Co., Ltd., and Taisho Pharmaceutical Co., Ltd.