Robotic bronchoscopy versus CT-guided biopsy: same yield, far fewer chest drains, weak evidence

Design

Systematic review and meta-analysis of direct comparisons, ROBINS-I and GRADE, with a cohort-genealogy step to exclude overlapping patients

Population

Three mutually independent retrospective cohorts from one US health system, adults with peripheral pulmonary lesions

Primary outcome

Strict 2024 ATS/CHEST diagnostic yield

Effect

Yield risk ratio 0.99 (95 per cent CI 0.93 to 1.06, I-squared 24 per cent); pneumothorax requiring drain or admission 0.25 (0.14 to 0.46, I-squared 0)

Robotic-assisted bronchoscopy and CT-guided transthoracic needle biopsy compete for the same peripheral pulmonary lesion, and almost all published comparisons are indirect. This review took only direct comparisons, and then did something unusual: a cohort-genealogy step to find how many of the eligible cohorts actually contained different patients.

The answer was three. Five retrospective studies were eligible, all from a single US health system, four of them sharing patients. Across the three independent cohorts, diagnostic yield under strict 2024 American Thoracic Society and American College of Chest Physicians criteria was comparable (risk ratio 0.99, 95 per cent CI 0.93 to 1.06, I-squared 24 per cent), with absolute yield falling from 88 per cent under looser definitions to 74 to 84 per cent under strict ones - a useful reminder of how much definition inflates published yield. Pneumothorax needing a chest drain or admission was about three-quarters less common with the robotic approach (0.25, 0.14 to 0.46). Procedure time was about 50 minutes longer for robotic bronchoscopy where same-session staging endobronchial ultrasound was counted in that time, and only about 8 minutes longer where it was not.

Certainty was low for the pleural complication finding and very low for everything else, all the data are retrospective and from one health system, and lesion size and bronchus sign - the two things that most determine which approach should be chosen - were almost never reported. The signal on pleural complications is plausible and mechanistically expected, but the authors are right that this needs a multicentre randomised trial rather than adoption.

• Do not read equal yield as established equivalence; the study says equivalence was not formally tested.
• Note how much diagnostic yield falls under strict criteria - 88 per cent becomes 74 to 84.
• The pneumothorax advantage is the most robust finding, and it is still low certainty.
• Choose by lesion size, bronchus sign and location, none of which this evidence stratifies by.
• Count staging endobronchial ultrasound separately when comparing procedure times.

The statistics, in plain English

The cohort-genealogy step matters: pooling five studies that share patients would have counted the same people several times and produced false precision, so restricting to three independent cohorts is the honest choice even though it weakens the result. A risk ratio of 0.99 with an interval of 0.93 to 1.06 shows no detectable difference, which is not the same as demonstrating equivalence - that needs a prespecified margin. Two outcomes reported by only two cohorts each did not survive the small-study correction, so they should be disregarded.

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.