In July 2016, the FDA updated its boxed warning for fluoroquinolones (FQ) in light of ongoing safety concerns. Our objective was to examine trends in FQ and other antibiotic prescribing before and after the warning.
We analyzed electronic health records for inpatient, outpatient, and emergency department encounters during April 2014-August 2017 in the University of North Carolina Health Care System (N=14,524,758 encounters). Patients under age 2 years were excluded. We estimated FQ prescribing rates per 1000 encounters (henceforth, /1000). Using an interrupted time series approach, we fit segmented linear regression models to assess trends before and after the FDA warning, accounting for seasonality and autocorrelated errors. We evaluated trends by gender, age, and point of care, and explored potential replacement of FQs by other antibiotics after the warning.
The average FQ prescribing rate was 9.2/1000 (95% CI 3.8, 14.7). Before the warning (April 2014-July 2016), the FQ prescribing rate decreased by 0.6/1000/year (95% CI 0.3, 0.9) (Figure). At the time of the warning, the slope for the FQ prescribing rate steepened by 0.8/1000/year (95% CI -0.1, 1.7); as a result, after the warning (July 2016-August 2017), the rate decreased by 1.4/1000/year (95% CI 0.7, 2.1). FQ prescribing was most common among adults age ≥65 (mean rate 12.2/1000), but relative trends were similar across age and gender. Average FQ prescribing rates were highest in the inpatient setting (40.5/1000) compared to emergency (18.3/1000) and outpatient (6.2/1000) encounters; relative trends were similar across settings. Rates for other common antibiotics did not increase after the warning, either overall (Figure) or in any subgroup. Results were robust to sensitivity analysis for lagged effects. Additionally, the interpretation of results is anchored by trends for inhaled corticosteroids, which we analyzed as a negative control (Figure).
The July 2016 FDA warning on FQs was associated with a small deceleration in FQ prescribing. There was no evidence of replacement by antibiotics with similar indications. We observed no evidence of heterogeneity across subgroups defined by gender, age, and point of care. Future research should assess the potential impact of the warning in clinically defined subgroups in various settings.
D. J. Anderson, None
N. Pajewski, None
A. Kinlaw, None