Predictive model quantifies risk of antibiotic resistance in uUTI

Investigators have identified key risk factors for non-susceptibility to antibiotics among patients with Escherichia coli–caused uncomplicated urinary tract infection (uUTI) and developed a novel framework to contextualize risk of non-susceptibility to these treatments.^1,2

"Future studies can provide additional insights by validating our models in different clinical settings and in ex-US populations and recalibrating them to include population-level features such as local antibiotic nonsusceptibility rates," wrote the authors.

The data were published in Clinical Infectious Diseases.¹

For the study, the team developed 4 predictive models to estimate the likelihood of non-susceptibility to 4 commonly prescribed classes of antibiotics (nitrofurantoin, trimethoprim-sulfamethoxazole, β-lactams, and fluoroquinolones). The models were based on electronic health record data from 87,487 female patients in the United States.

Overall, the predictive models identified the number of previous UTI episodes, prior β-lactam nonsusceptibility, prior fluoroquinolone treatment, Census Bureau region (particularly in the South), and race as key predictors of having a non-susceptible isolate to 3 or more antibiotic classes. Race and number of previous UTI episodes were key predictors among all 4 antibiotic classes.

Additionally, prior non-susceptibility to an antibiotic class was associated with significantly higher odds of non-susceptibility to that antibiotic class. Prior non-susceptibility to a β-lactam agent was shown to be a significant predictor for non-susceptibility to trimethoprim-sulfamethoxazole (OR, 1.40; 95% CI, 1.23–1.61), β-lactams (OR, 4.09; 95% CI, 3.56–4.70), and fluoroquinolones (OR, 1.33; 95% CI, 1.15–1.52; all P < .05). Similarly, prior treatment with fluoroquinolones was shown to be a significant predictor of nonsusceptibility to trimethoprim-sulfamethoxazole (OR, 1.22; 95% CI, 1.10–1.35), β-lactams (OR, 1.43; 95% CI, 1.29–1.59), and fluoroquinolones (OR, 2.54; 95% CI, 2.29–2.82; all P < .05).

The average area under the receiver operating characteristic curve (AUROC) for the final models in the test cohorts was 0.67 (standard error [SE], 0.010) for nitrofurantoin, 0.66 (SE, 0.004) for trimethoprim-sulfamethoxazole, 0.66 (SE, 0.005) for β-lactams, and 0.72 (SE, 0.005) for fluoroquinolones.

The investigators also developed a risk categorization framework to classify patients’ isolates as having a low, moderate, or high risk of nonsusceptibility to each antibiotic class. Data showed a high risk of non-susceptibility to nitrofurantoin, trimethoprim-sulfamethoxazole, β-lactams, and fluoroquinolones among 8.1%, 14.4%, 17.4%, and 6.3% of patients, respectively.

Across classes, the proportion of patients with high-risk isolates was 3 to 12 times higher among those with non-susceptible isolates vs those with susceptible isolates. Similarly, the proportion of patients with true non-susceptible isolates was 3 to 10 times higher among patients with high-risk isolates vs those with low-risk isolates.

In total, the study included 87,487 patients with an average age of 49.7 to 50.4 years across the training and test cohorts for each antibiotic class. Among all patients, 85.1% to 85.5% were White, and 63.8% to 66.3% resided in the Midwest.

Overall, the authors concluded,¹ “Findings from our study provide valuable insight on key patient characteristics to consider when assessing risk of antibiotic nonsusceptibility, thereby advancing the understanding of antibiotic nonsusceptibility in uUTI, and potentially informing optimal treatment strategies in this population. Future studies can provide additional insights by validating our models in different clinical settings and in ex-US populations and recalibrating them to include population-level features such as local antibiotic nonsusceptibility rates.”

References

1. Shields RK, Cheng WY, Kponee-Shovien K, et al. Development of Predictive Models to Inform a Novel Risk Categorization Framework for Antibiotic Resistance in E. coli-Causing Uncomplicated Urinary Tract Infection. Clin Infect Dis. 2024:ciae171. doi:10.1093/cid/ciae171

2. Analysis Group researchers develop a novel framework to help clinicians predict the likelihood of antibiotic-resistant bacteria in patients with uncomplicated urinary tract infections. News release. Analysis Group. May 22, 2024. Accessed May 24, 2024. https://www.prnewswire.com/news-releases/analysis-group-researchers-develop-a-novel-framework-to-help-clinicians-predict-the-likelihood-of-antibiotic-resistant-bacteria-in-patients-with-uncomplicated-urinary-tract-infections-302151513.html