Comparative safety of different antibiotic regimens for the treatment of outpatient community-acquired pneumonia among otherwise healthy adults

Abstract

Background: Evidence is limited about the comparative safety of antibiotic regimens for treatment of community-acquired pneumonia (CAP). We compared the risk of adverse drug events (ADEs) associated with antibiotic regimens for CAP treatment among otherwise healthy, non-elderly adults.

Methods: We conducted an active comparator new-user cohort study (2007-2019) of commercially-insured adults 18-64 years diagnosed with outpatient CAP, evaluated via chest x-ray, and dispensed a same-day CAP-related oral antibiotic regimen. ADE follow-up duration ranged from 2-90 days (e.g., renal failure [14 days]). We estimated risk differences [RD] per 1000 treatment episodes and risk ratios using propensity score weighted Kaplan-Meier functions. Ankle/knee sprain and influenza vaccination were considered as negative control outcomes.

Results: Of 145 137 otherwise healthy CAP patients without comorbidities, 52% received narrow-spectrum regimens (44% macrolide, 8% doxycycline) and 48% received broad-spectrum regimens (39% fluoroquinolone, 7% β-lactam, 3% β-lactam + macrolide). Compared with macrolide monotherapy, each broad-spectrum antibiotic regimen was associated with increased risk of several ADEs (eg, β-lactam: nausea/vomiting/abdominal pain [RD per 1000, 3.20; 95% CI, 0.99–5.73]; non–Clostridioides difficile diarrhea [RD per 1000, 4.61; 95% CI, 2.47–6.82]; vulvovaginal candidiasis/vaginitis [RD per 1000, 3.57; 95% CI, 0.87, 6.88]). Narrow-spectrum antibiotic regimens largely conferred similar risk of ADEs. We generally observed similar risks of each negative control outcome, indicating minimal confounding.

Conclusions: Broad-spectrum antibiotics were associated with increased risk of ADEs among otherwise healthy adults treated for CAP in the outpatient setting. Antimicrobial stewardship is needed to promote judicious use of broad-spectrum antibiotics and ultimately decrease antibiotic-related ADEs.

Keywords: Community-acquired pneumonia; administrative data; adverse drug events; antibiotic safety; antimicrobial stewardship.

Figure 1.

Derivation of the cohort of adults with outpatient community-acquired pneumonia in the MarketScan Commercial Database (index events 1 July 2007 to 30 November 2019). Abbreviation: CAP, community-acquired pneumonia.

Figure 2.

Weighted RDs per 1000 of adverse drug events by antibiotic regimen versus macrolide (reference) among adults treated for community-acquired pneumonia in the outpatient setting. Standardized mortality ratio–weighted RD was calculated on the last day of outcome-specific follow-up (14 days for nausea/vomiting/abdominal pain, rash/urticaria, unspecified allergy, renal failure; 30 days for non–C difficile diarrhea, vulvovaginal candidiasis; 90 days for tendinopathy). 95% CIs were calculated via nonparametric bootstrap sampling with replacement (N = 1000). We required ≥8 adverse drug event cases in both the reference category (ie, macrolide) and the comparator group to ensure stability of the effect estimate. Abbreviations: C. difficile, Clostridioides difficile; CI, confidence interval; NE, not estimable; RD, risk difference.

Figure 3.

Weighted cumulative incidence curves of select adverse drug events by antibiotic agent among adults treated for community-acquired pneumonia in the outpatient setting. The lines represent the standardized mortality ratio-weighted cumulative incidence estimate; shaded bands represent corresponding 95% confidence intervals. Abbreviation: C. difficile, Clostridioides difficile.

Figure 4.

Weighted RDs per 1000 negative-control outcomes by antibiotic regimen versus macrolide (reference) among adults treated for community-acquired pneumonia in the outpatient setting. Standardized mortality ratio–weighted risk difference represents the risk difference on the last day of follow-up (day 30). 95% CIs were calculated via nonparametric bootstrap sampling with replacement (N = 1000). We required ≥8 cases in both the reference category (ie, macrolide) and the comparator group to ensure stability of the effect estimate. Abbreviations: CI, confidence interval; NE, not estimable; RD, risk difference.