Interesting point! I did look at higher QFN thresholds but the gains in spec were balanced by losses in sens so PPV stayed at 4%. Only combining both tests got PPV up to 7%.

My understanding is that direct measures of IFN are less dynamic and have a higher limit of detection, esp if unstimulated

Thanks to @rishigupta.bsky.social @mnoursad.bsky.social @tomscriba.bsky.social

7/7
In summary:
1. Single-gene transcripts are just as good as multi-gene signatures - this may aid translation by simplifying assays & encouraging commercial competition
2. They show promise to stratify therapy, either alone in high-burden settings or in combination w/ IGRA in low-burden settings

6/7
We performed decision curve analysis to evaluate the clinical utility of biomarker-stratified therapy
- High-burden settings: RNA biomarker > IGRA
- Low-burden settings: IGRA > RNA biomarker, but combining both tests was superior if aiming to treat <50 people to prevent 1 case

5/7
Importantly, the performance of single-gene transcripts was consistent across high- and low-burden settings

In comparison, IGRA performance was heterogenous (v poor specificity in high-burden settings, as expected)

4/7
Performance was similar for asymptomatic prevalent and incident disease up to 12 months, but dropped after 12 months from sampling to disease

Shown here for BATF2, the best performing single-gene:

3/7
We performed an IPD-MA of >6500 samples from 7 cohorts across high- and low-burden settings, and compared 80 single-genes vs 8 multi-gene signatures

5 single-genes were equivalent to the best-performing multi-gene signature to detect subclinical TB over a 12 month interval

2/7
Previously, we showed that several of these multi-gene signatures perform equivalently, are co-correlated and have common upstream IFN & TNF signalling pathways

So multiple genes might not offer orthogonal value, and simplification of signatures may accelerate translation...

1/7
Detecting and treating subclinical TB is essential to prevent progression to clinical disease and reduce transmission

Several multi-gene RNA signatures have been discovered with promising accuracy to detect clinical TB or predict progression to disease

Agreed, my key takeaway is dalbavancin is non-inferior to SOC for clinical efficacy. So a good option especially if issues with compliance/IV access/self-discharge etc

Optimising the use of currently available treatments through treatment stratification could reduce treatment duration by 2-months for the majority. TB treatment stratification strategies should be prospectively trialled in a phase III RCT.
#TBSky

📌

Nice case, thanks for sharing!

Exactly, we analyse tests like they act in isolation but (most) clinicians interpret them with Bayesian inference.

There’s no such thing as a bad test, just a bad user!

📌

Agreed. But I think the median of 10 days probably reflects the stormy course and source control challenges in sick ICU patients rather than guidelines

The population were patients with sepsis in ICU (50% had septic shock) so I think 10 days is probably reflective of practice in ICU.

Though the estimate for effect size (-0.88 days duration) from PCT is not practice changing…

First wave ISGs (eg MX1) induced rapidly, correlate with VL & infectivity - potential biomarker to stratify patients who'll benefit from antivirals?