Design: Secondary analysis of a randomised trial in 60 primary care practices
Population: 345 adults with lower respiratory tract infection in Switzerland
Primary outcome: Time to recovery by nasopharyngeal bacterial PCR detection
Effect: Median 7 days in both groups; aHR 1.13 (95% CI 0.84 to 1.50)

This BJGP study was a secondary analysis of a trial in 60 Swiss general practices. Nasopharyngeal swabs from 345 adults with lower respiratory tract infection were tested afterwards by PCR for bacteria and viruses, and recovery was compared by whether bacteria were found.

Bacteria were detected in 41%. Median time to recovery was seven days with or without them (adjusted hazard ratio 1.13, 95% CI 0.84 to 1.50). Antibiotic prescribing did not change that relationship (interaction hazard ratio 1.10, 0.64 to 1.90), and cough, sputum and breathlessness lasted the same. Cough of any severity lasted a median of 16 days.

Syndromic PCR panels are increasingly offered in private laboratories, including in India, and a positive bacterial result is easily read as a reason to prescribe. This study suggests that for chest infections in primary care the result neither predicts the course nor identifies who benefits from antibiotics. Clinical assessment for pneumonia remains the deciding step. This was a single secondary analysis, so it is strong evidence against routine use rather than proof of no value in every case.

Avoid relying on nasal PCR panels to decide on antibiotics for chest infection in primary care.
Base antibiotic decisions on clinical signs of pneumonia and patient risk.
Expect bacteria on a nasal swab in about four in ten patients; it often means carriage.
Use the money saved on tests for follow-up of patients who are not improving.

Why it matters

A positive lab result can push a prescription that the patient does not need.

The statistics, in plain English

A hazard ratio of 1.13 with an interval from 0.84 to 1.50 means no detectable difference in recovery speed. The interaction test asks whether antibiotics helped more in patients with bacteria; its interval (0.64 to 1.90) also shows no such effect. The study was not designed for this question originally, which limits its power for small effects.