Write a comment

This retrospective single-center cohort study evaluates the diagnostic value of microbial cell-free DNA (mcfDNA) sequencing of bronchoalveolar lavage (BAL) fluid (Karius test) in immunocompromised, mechanically ventilated patients with suspected pneumonia. The authors compare KT-BAL results with standard-of-care (SOC) diagnostics and analyze clinical outcomes. The study finds that KT-BAL substantially increases the diagnostic yield - identifying additional bacterial, viral, and fungal pathogens missed by SOC and that detection of these missed pathogens is associated with longer intubation duration, suggesting potential clinical significance.

Strengths

1. Novel and clinically relevant

The preprint addresses an important gap in pneumonia among immunocompromised patients, where conventional tests frequently fail to identify causative pathogens.

1. Methodological rigor

The preprint uses pre-existing prospectively adjudicated cohort, minimizing bias in pneumonia diagnosis and provides a robust foundation for retrospective molecular analyses.

1. Comprehensive comparison of KT-BAL and SOC

It provides specific comparisons by organism category, showing where KT-BAL adds the most value (in fungi, opportunistic pathogens, DNA viruses).

1. Transparency in report/reproducibility

The inclusion of STROBE compliance, data provenance (PhysioNet) and clear funding/COI strengthens its transparency.

Weaknesses

1. Retrospective single-center design

The retrospective approach in the preprint may raise concerns in selection bias and possibly limit casual interpretation of outcome.  

1. Lack of clinical adjudication of KT-BAL findings

The preprint may benefit from prospective validation with clinical correlation to improve diagnostic utility. It is unclear how many of the additional detections were true infections vs colonization/contamination.

1. Absence of plasma mcfDNA comparison

Given Karius’s established plasma test, a matched analysis could clarify whether BAL fluid offers additional diagnostic yield over noninvasive plasma testing.

Some potential suggestions

• While the Karius categories are described clearly, the preprint could include more examples/rationales for these designations (especially Category 3 detections in non-pneumonia controls).

• Figures 2-4 effectively illustrate the key findings, but it could have clearer legends, particularly specifying sample sizes per subgroup

• Detailed organism lists (Supplementary Tables 1-3) could be moved into main text to improve accessibility.

• Reporting exact p-values/effect sizes for major comparisons could improve interpretability. eg) ICU mortality difference is borderline significant → should be interpreted cautiously.

Competing interests

The author declares that they have no competing interests.

Use of Artificial Intelligence (AI)

The author declares that they did not use generative AI to come up with new ideas for their review.