The same clinical signs data as above represented in two dimensions using Multiple Correspondence Analysis (MCA). The figure shows that patient clinical phenotype observed from the bedside, and classified by an unsupervised clustering algorithm, corresponds closely to blood bacillary load measures and risk of mortality. Each participant's 12-week outcome status and MTBBSI quantification results from day of recruitment are also shown on a navy-yellow colour scale to indicate level of correspondence with the clustered clinical signs data. Participants (rows) and clinical signs (columns) are hierarchically clustered (indicated by row dendrogram), with two highest level clusters of participants indicated by a colour key (cluster A, top ten rows, green cluster B bottom 18 rows, blue). Heatmap of binary clinical signs from baseline assessment at time of recruitment (black tile = present, white tile = absent).
The motivating assumption for this ordinal regression approach was that each of the three quantification methods could be treated as indicator variables reporting on an unobserved latent variable, namely blood bacilli load.Ī. Random effects for intercept and slope by participant were used. Interactions between time and method, and time and outcome, were also included, allowing both intercepts and slopes to vary, to assess if overall bacilli load and/or rate of change in bacilli load on treatment varied across levels of these categorical variables. The three quantitative measures of blood bacilli (DMN-Tre microscopy count, MFL culture TTP, and blood Xpert-ultra Ct) were each converted to an ordinal scale ranging from 1 for a negative test, to 10 for higher-load observed values, and regressed on variables: time in days from start of treatment, quantification method (DMN-Tre microscopy, MFL culture, or blood Xpert-ultra), and 12-week outcome (survived or died). We investigated the relationship between dynamics of blood bacillary load and mortality in a combined pharmacodynamic model as follows.
Filters were then sealed inside 20 mm glass bottom cell culture dishes mounted in an aluminium holder custom-made for the microscope stage. The sample was then placed in a 13 mm sterile centrifuge ultrafiltration holder pre-loaded with a 13 mm black 0.6 μm polycarbonate membrane filter, which was then centrifuged at 1800 x g for 10 min. This lysis step was established empirically from permutations of published lysis methods and resulted in a filterable lysate without >10% loss of mycobacterial cells in spiked samples. After incubation, the sample was mixed with 25 mL of a detergent and enzymatic lysis buffer pre-warmed to 37 ☌, and incubated at 37 ☌ for 25 min with inversions, before pelleting (3000 x g, 25 min) and resuspension in 2 mL 0.22 µm-filtered water. The pellet from the 3 mL blood aliquot was resuspended in 5 mL of Myco/F Lytic broth containing DMN-Tre to a final concentration of 154 µM, and the sample incubated in the dark and under agitation for 18 h at 37 ☌. Half the pellet from the 10 mL blood aliquot was inoculated into a second Myco/F Lytic bottle, and half was stored at -20 ☌ for downstream Xpert-ultra testing.