Biography
I am a respiratory physician and clinical researcher at University College London. I use data science to improve the diagnosis and prognosis of respiratory infections, by integrating epidemiological, machine learning and ‘omics methods. I work on tuberculosis and respiratory viral infections in particular and use a range of methods spanning epidemiology, individual participant data meta-analysis, multivariable risk prediction and RNA sequencing. A particular focus of my work is to develop and evaluate tools to better target preventive treatment for tuberculosis towards those most likely to benefit.
Key Impacts
Blood RNA biomarkers to predict treatment failure or relapse after truncated treatment of pulmonary tuberculosis
Biomarkers indicating cure at end of treatment could optimise implementation of shorter treatment strategies. We assessed candidate blood RNA biomarkers to predict treatment failure or relapse (TFR) following 8 weeks’ treatment in the TRUNCATE-TB trial.Methods: TRUNCATE-TB randomised adults (18–65 years) with rifampicin-susceptible pulmonary TB (NCT03474198) to either 24-week standard treatment; or to the TRUNCATE strategy, comprising initial 8-week treatment with re-treatment where needed. We performed blood RNA sequencing of samples from baseline, weeks 4/8, end of treatment (if treatment was extended), and week 24. We evaluated the accuracy of 12 candidate RNA biomarkers (from systematic review) for 96-week TFR.Results: RNA data were available from 166 standard treatment and 446 TRUNCATE strategy participants (median 4 samples per participant). TRUNCATE strategy participants included 80 experiencing TFR and 309 without TFR. Candidate RNA biomarkers were highly correlated (Spearman ρ 0.76-0.98). FCGR1A showed the highest mean area under the receiver operating characteristic curve (AUROC) for TFR across sampling timepoints. Biomarker scores were highest at baseline. These resolved over time with similar trajectory by treatment strategy (Figure), but median scores remained >2 standard deviations above the mean of a healthy control population after 8 weeks’ treatment. Biomarker scores continued to decline after treatment cessation in TRUNCATE strategy participants. Discrimination of FCGR1A for TFR among TRUNCATE strategy participants was moderate from baseline to end of treatment (AUROCs 0.65-0.67); but improved at week 24 (AUROC 0.71; 95% CI 0.65-0.78).Conclusions: Residual elevation of RNA biomarkers at 8-weeks and similar trajectories among those receiving truncated or standard treatment suggests ongoing bacterial clearance or gradual immune homeostasis after 8 weeks. Among TRUNCATE strategy participants, RNA biomarkers provide limited prediction of long-term TFR when measured at baseline through treatment end. Divergent biomarker trajectories thereafter suggests a potential role in predicting TFR after treatment cessation.Figure:
Source: Conference 2024