Previous antibiotic exposure and antimicrobial resistance in invasive pneumococcal disease: results from prospective surveillance.

BACKGROUND Estimating the risk of antibiotic resistance is important in selecting empiric antibiotics. We asked how the timing, number of courses, and duration of antibiotic therapy in the previous 3 months affected antibiotic resistance in isolates causing invasive pneumococcal disease (IPD). METHODS We conducted prospective surveillance for IPD in Toronto, Canada, from 2002 to 2011. Antimicrobial susceptibility was measured by broth microdilution. Clinical information, including prior antibiotic use, was collected by chart review and interview with patients and prescribers. RESULTS Clinical information and antimicrobial susceptibility were available for 4062 (90%) episodes; 1193 (29%) of episodes were associated with receipt of 1782 antibiotic courses in the prior 3 months. Selection for antibiotic resistance was class specific. Time elapsed since most recent antibiotic was inversely associated with resistance (cephalosporins: adjusted odds ratio [OR] per day, 0.98; 95% confidence interval [CI], .96-1.00; P = .02; macrolides: OR, 0.98; 95% CI, .96-.99; P = .005; penicillins: OR [log(days)], 0.62; 95% CI, .44-.89; P = .009; fluoroquinolones: profile penalized-likelihood OR [log(days)], 0.62; 95% CI, .39-1.04; P = .07). Risk of resistance after exposure declined most rapidly for fluoroquinolones and penicillins and reached baseline in 2-3 months. The decline in resistance was slowest for macrolides, and in particular for azithromycin. There was no significant association between duration of therapy and resistance for any antibiotic class. Too few patients received multiple courses of the same antibiotic class to assess the significance of repeat courses. CONCLUSIONS Time elapsed since last exposure to a class of antibiotics is the most important factor predicting antimicrobial resistance in pneumococci. The duration of effect is longer for macrolides than other classes.