Fig 1.
Overview of patient enrollment and adjudicated treatment outcomes at 24 weeks.
DS: drug susceptible; IR: INH resistant, MDR: multi drug-resistant.
Fig 2.
Time to negative results Xpert MTB/RIF Ct relative to MGIT culture.
Kaplan-Meyer curves are shown for the time to TB negativity as indicated by two successive negative MGIT cultures versus TB negativity as indicated by a single negative Xpert test. * Xpert testing was not performed at the week 12 time point.
Fig 3.
Baseline Xpert and MGIT results and time to culture negativity.
The mean, upper and lower mid quartiles, standard deviations, and outliers of the baseline Xpert Ct (top) and baseline MGIT TTP (bottom) are shown for subjects who converted their culture at the indicated time points. All culture negative samples reported as negative at day 42 by MGIT instrument were assigned a TTP value of 45 and all negative Xpert assays were assigned a Ct value of 40. It should be noted that approximately one third of MGIT TTP findings were excluded from the analysis either due to culture contamination or sputum aliquot inconsistencies.
Table 1.
Baseline Xpert Ct value to predict time to negativity status at different time points.
Fig 4.
ROC curve for direct Xpert CT, delta CT and percent closing relative to culture negativity at the same time point.
Assessment of direct Xpert Ct (A), delta Ct (B) and percent closing (C) values to predict the same-day culture negativity at day 7, week 4, week 8, and week 24. ROC curve results for direct Xpert Ct (optimal cut off approximately ≥30) at Day 7, AUC = 96.7 (95% CI 94.1, 99.3); at week 4, AUC = 91.2 (95% 85.8, 96.6); and week 8 AUC = 86.0 (95% 77.9, 94.2) and at week 24, AUC = 90.2 (95% 75.0, 100.0). ROC curve results for the delta Ct (optimal cut off values of approximately ≥ 10) at day 7, AUC = 97.5 (95% CI 94.5, 100.0); at week 4, AUC = 82.8 (95% 72.5, 93.1); at week 8, AUC = 75.2 (95%CI 64.8, 85.6); and at week 24, AUC = 82.2 (95%CI 67.9, 96.5). ROC curve results for the percent closing of the baseline Ct deficit at day 7, AUC = 96.6 (95% CI 94.0, 99.2); week 4, AUC = 91.6 (95%CI 86.2, 96.9); week 8, AUC = 79.8(70.3, 89.4); and at week 24, AUC = 86.0 (95%CI 71.3, 100.0).
Fig 5.
ROC curve for baseline Ct, direct Ct, delta Ct and percent closing deficit to predict treatment failure at the end of the treatment.
Evaluation of ROC curves for direct Xpert Ct showed AUC = 80.2 (95%CI 69.2, 91.3) at week 8, and AUC = 90.2 (95%CI 75.2, 100.0) at week 24. ROC curves for delta Ct demonstrated AUC = 70.2 (95%CI 54.1, 86.2) at week 8 and AUC = 82.6 (95% CI 68.1, 97.0) at week 24. ROC curves for the percent closing deficits at week 8, AUC = 75.2 (95%CI 62.2, 88.1) and at week 24, AUC = 86.2 (95%CI 71.6, 100.0).
Fig 6.
Combined measures of Ct plus percent closing to predict treatment success versus treatment failure.
This figure represents a model that include only two covariates: baseline Xpert Ct and percent closing at 4, 8 and 24 weeks (for panels a, b and c respectively). Optimal thresholds are then defined using Youden’s index (on the risk scores provided from the logistic regression models), with sensitivity and specificity estimates derived from application of leave-on-out cross-validation. The Xpert Ct and percent closing measure for each cured case (blue dots) and each treatment failure case (red dots) are shown. The optimal thresholds to predict treatment success and treatment failure using both Xpert Ct and percent closing at each time point are indicated by a diagonal black line. Color gradation from black to white indicate increase chances of culture negativity. Cross validated results using combination of a direct Xpert Ct and percent closing using Youden’s index to predict treatment failure showed 71.4% sensitivity and 42.5% specificity at week 4, 62.5% sensitivity and 84.0% specificity at week 8, and 75% sensitivity 88.9% specificity at week 24.
Table 2.
An optimal threshold to forecast culture negativity at the same visit and treatment failure using Xpert Ct at the end of the treatment.