First-generation cephalosporins did as well as later ones for outpatient pyelonephritis

Design

Multicentre retrospective observational cohort

Population

1147 adults discharged from 11 US EDs with pyelonephritis on oral cephalosporins

Primary outcome

14-day treatment failure

Effect

19.6% vs 17.6% vs 18.0% (first vs second vs third generation); no significant difference

DOGMA-ED, a retrospective cohort across 11 US academic emergency departments, included 1147 adults (over 85% women) discharged with pyelonephritis on an oral cephalosporin between 2023 and 2025. 54% received a first-generation agent (mostly cephalexin), 8% second-generation (cefuroxime) and 38% third-generation (mostly cefdinir). Published in Annals of Emergency Medicine on 6 October 2026.

Treatment failure at 14 days, meaning recurrence, a return visit, a new antibiotic or admission, was 19.6% with first-generation, 17.6% with second-generation and 18.0% with third-generation agents. None of the differences was significant.

Third-generation oral cephalosporins are often chosen on the assumption that broader is better. This suggests that, for patients well enough to go home, the generation may matter less than whether the organism is susceptible.

This was observational, and US resistance patterns differ sharply from India's, where ESBL-producing E coli is common and can defeat every oral cephalosporin. Guidelines have not changed on the basis of this study. Culture before treating and choose by local resistance.

• Send a urine culture before starting antibiotics in every patient with pyelonephritis
• Consider a first-generation cephalosporin where local resistance data support it
• Check local ESBL rates; where they are high, oral cephalosporins of any generation may fail
• Review culture results at 48 to 72 hours and switch if the organism is resistant
• Give clear return advice; about one in five patients failed treatment at 14 days

Why it matters

It challenges the reflex to reach for a third-generation agent when a narrower one may do.

Don't overread it

Observational US data; it does not establish equivalence where ESBL resistance is common.

The statistics, in plain English

The confidence intervals for the differences (for example -3.4 to 6.5 points for first versus third generation) include zero, so no difference was detected. The second-generation group was small, so its interval is wide.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.