High frequencies of antimicrobial drug resistance were observed in O157 and non-O157 Shiga toxin-producing E. coli strains recovered from patients in Michigan during 2010-2014. Resistance was more common in non-O157 strains and independently associated with hospitalization, indicating that resistance could contribute to more severe disease outcomes.
Shiga toxin-producing Escherichia coli (STEC) contributes to 265,000 cases of foodborne illness annually in the United States (1). Most infections are caused by O157 strains; however, non-O157 STEC infections have increased (2). Antimicrobial drug resistance among STEC has been reported (3-5) but is probably underestimated. Given the importance of resistance in E. coli pathotypes, we sought to determine the prevalence of resistant STEC infections and assess the effects of resistance on disease.We obtained 358 STEC isolates from the Michigan Department of Health and Human Services (MDHHS) Reference Laboratory (Lansing, MI, USA), collected during 2010-2014. Of these, 14 were outbreak associated. We examined 1 strain per outbreak using protocols approved by Michigan State University (MSU; Lansing, MI, USA; IRB #10-736SM) and MDHHS (842-PHALAB). Overall, 31 (8.8%) strains (23 non-O157, 8 O157) were resistant to antimicrobial drugs (Table). Resistance to ampicillin (7.4%) was most common, followed by trimethoprim/sulfamethoxazole (SXT) (4.0%) and ciprofloxacin (0.3%). Compared with national rates, resistance to ampicillin and SXT was higher, but not significantly different, for O157 isolates from Michigan (online Technical Appendix Figure 1, https://wwwnc. cdc.gov/EID/article/23/9/17-0523-Techapp1.pdf) (6). One strain was resistant to all drugs, and all resistant strains had high MICs (ampicillin, >64 μg/mL; ciprofloxacin, >32 μg/ mL; SXT, in 1:19 ratio, >32/608 μg/mL). Notably, resistance was twice as common for non-O157 (11.1%) than for O157 (5.5%) strains. O111 strains (n = 7) had significantly higher resistance frequencies (24.1%) than other non-O157 serogroups (p = 0.03). We found variation by year and season; resistance frequencies were highest in 2012 (online Technical Appendix, Figure 2) and during winter/spring (online Technical Appendix Table 1), but neither trend was significant. We also observed a strong but nonsignificant association between resistance and hospitalization but no association for urban versus rural residence (7) or county after stratifying by prescription rates (8) in the univariate analyses.We conducted a multivariate analysis using logistic regression, with hospitalization as the dependent variable; we included variables with significant (p<0.05) and strong (p<0.20) associations from the univariate analysis as independent variables. Forward selection indicated that hospitalized patients were more likely to have resistant infections (odds ratio [OR] 2.4, 95% CI 1.00-5.82) and less likely to have non-O157 infections (OR 0.4, 95% CI 0.21-0.61) (online Technical Appendix Table 2), suggesting that resistant infections or O157 infections may cause ...