Results:
A total of 235 patients were recruited into the study, (healthy volunteers (n=28). In ALISTER: active TB (n=30), latent TB (n=88), NTM infection (n=25). In SAEFRIF: HIV-TB coinfection (n=64)). Demographic and clinical characteristics for patients within ALISTER and SAEFRIF are presented in Table 1. The healthy volunteers were predominantly white, with a median age of 27 [IQR: 24-30] years. There was no significant difference in miR-122 (P=0.09) between the groups. There was a statistically significant but clinically minor difference in K18 between groups (P=0.03) (Table 2 & Figure 1). Serial samples were collected in 65 patients in the ALISTER study (Figure 2) . The time patients had spent on treatment was (median [IQR]: active TB 3.0 [2.4-6.9]; latent TB 3.9 [2.4-12.2]; NTM infection 4.0 [2.2-6.9] weeks). Commencing treatment was associated with a statistically significant, clinically insignificant increase in ALT (median [IQR]: baseline 15 [12-24]; on treatment 17 [13-27] IU/L; P=0.03) and miR-122 (median [IQR]: baseline 3.12 [1.20-5.63]; on treatment 3.95 [1.75-7.98] fM; P=0.01). There was also an increase in K18 which did not reach statistical significance (median [IQR]: baseline 150 [103-224]; on treatment 167 [110-246] U/L; P=0.4). Serial samples were also collected in 50 SAEFRIF patients (Figure 3). In the SAEFRIF patients, starting treatment did not lead to a significant change in ALT (median [IQR]: week 0: 22 [14-31]; week 2: 24 [14-32] IU/L; P=0.62) or K18 (median [IQR]: week 0: 204 [130-344]; week 2: 169 [117-262] U/L; P=0.3).
When all samples from all time points were included there was a significant correlation between ALT and miR-122 (n=251, Spearman rank r = 0.52, 95%CI = 0.42-0.61, P<0.0001) (Figure 4A).There was a significant, but less tight, correlation between ALT and K18 (n=491, Spearman rank r = 0.42, 95% CI = 0.34-0.49, P<0.0001) (Figure 4B). miR-122 correlated with K18 (n=252, Spearman rank r = 0.32, 95% CI = 0.20-0.43, P<0.0001) (Figure 4C).
MiR-122 was increased 8.0-fold in samples with elevated ALT (>50 IU/L) compared to samples with normal ALT (predefined as ≤50 IU/L) (median [IQR]: elevated ALT 23.9 [11.5-60.4]; normal ALT 3.00 [1.31-4.82] fM; P<0.0001) (Figure 5A). K18 was increased 2.3-fold in those samples with elevated ALT (>50 IU/L) compared to samples with normal ALT (≤50 IU/L) (median [IQR]: elevated ALT 395 [217-683]; normal ALT 170 [120-250] U/L; P<0.0001) (Figure 5B). ROC analysis was performed on these grouped samples (Figure 5C & D). MiR-122 identified elevated ALT (>50 IU/L) with high accuracy (ROC-AUC = 0.93, 95 % CI = 0.88-0.98, P<0.001). K18 identified elevated ALT (>50 IU/L) with lower accuracy (ROC-AUC = 0.80, 95% CI = 0.72-0.87, P<0.0001).
The inter-individual variability in miR-122, K18 and ALT was compared in the different patient groups (Table 2) . Inter-individual variability was higher for both miR-122 and K18 than ALT. Sequential samples from the SAEFRIF trial (patients with normal ALT (<50IU/L) and three or more samples collected) were analysed to determine the intra-individual variability over time (CV median [IQR]: ALT 23.9 [16.6-36.1] %; K18 35.4 [24.9-41.9] %; P=0.02). Intra-individual variability was higher for K18 than ALT.
In this study there were two cases of DILI (as pre-defined in study protocols), both cases were for patients receiving isoniazid alone for the treatment of latent TB within the ALISTER study. The first case was of a 51-year-old white Scottish male patient, who experienced peak ALT activity of 431 IU/L, (Figures 6A, C & E) . Before starting treatment, he had normal ALT (18 IU/L), miR-122 was 4 fM and his K18 was 58 U/L. 3 months into treatment his ALT activity increased to 431 IU/L, miR-122 rose to 60 fM and K18 rose to 1248 U/L. Drug treatment was halted, ALT returned to within normal limits 18 weeks later. Formal causality between isoniazid and liver injury was determined using the RUCAM scale and was determined as probable (RUCAM scale = 6).
The second case was of a 71-year-old white British female patient,(Figures 6B, D & F). At baseline her ALT activity was 28 IU/L, miR-122 was 17 fM and K18 was 219 U/L. 2 weeks into treatment her ALT activity had not increased (25 IU/L) but K18 and miR-122 had risen (307 U/L and 63 fM, respectively). DILI was present 5 weeks into treatment with the presence of a drug rash and elevated ALT at 194 IU/L. At this time K18 activity had risen to 3490 U/L and miR-122 concentration had increased further to 336 fM. Drug treatment was halted, and the patient was discharged and lost to further follow up. RUCAM causality assessment indicated isoniazid was the probable cause of liver injury (RUCAM scale = 7).