• No results found

Time since previous TB diagnosis, time on ART and number of previous TB episodes were not associated with signature scores !

In document University of Cape Town (Page 141-147)

Chapter 4: Predicting recurrent TB disease in HIV-infected persons on ART

4.4: Results

4.4.9: Time since previous TB diagnosis, time on ART and number of previous TB episodes were not associated with signature scores !

In the Malherbe et al study of TB treatment, inflammation detected in the form of PET-hot lesions measured by PET-CT was observed in HIV-uninfected persons up to a year after end of successful TB therapy (Malherbe et al., 2016). We investigated the effect of time since previous TB diagnosis and number of previous TB episodes on on-going inflammation measured by signature scores. Since effective ART results in lower risk of active TB, decreased immune activation and inflammation in HIV infection (Klatt et al., 2013), the effect of time on ART on signature scores was

measured to determine the effect of ART on on-going inflammation. Time since previous TB and time on ART were calculated by subtracting the dates of ART start and TB treatment start (within seven days of diagnosis) from the sampling date.

Some participants had previous TB episodes at the start of primary TB study (i.e.

START/SAPiT) but the number of previous TB episodes was not recorded. Thus, we stratified the number of previous TB episodes into one or more than one in our analyses of the TRuTH study. The median time from previous TB to sampling time was 1,325 days (IQR: 1,042 to 1,605 days). No association was observed between time since previous TB and ACS-11 gene or rPSVM.1 signature scores. Time on ART (median = 1,089 days, IQR: 791 to 1,436 days) was also not associated with signature scores at the corresponding sampling time points. Signature scores were not different between persons with one previous TB episode and those with more than one previous TB episode, p=0.49 and p=0.43 for the ACS 11-gene and rPSVM.1 signatures,respectively. This disproves our hypothesis that signature scores will be higher in persons with more than one episode of previous TB.

Figure 15: Longitudinal kinetics of signature scores in the TRuTH cohort. (A&B) Longitudinal kinetics of the ACS 11-gene (A) and rPSVM.1 (B) signature scores in progressors and non- progressors. Median and interquartile ranges are shown. (C&D) Longitudinal kinetics of the ACS 11- gene (C) and rPSVM.1 (D) signature scores in individual progressors and non-progressors. Dots represent time points and connecting lines represent individuals. (E&F) Association between the ACS 11-gene (E) and rPSVM.1 (F) signature scores and bacterial burden measured by days to culture positivity in sputum. Spearman r values are shown.

!

0 20 40 60 80 100

Signature score (%)

0.0070

Time to TB diagnosis (Days) 91- 180 181- 360 361- 540

>540 0-90

ACS 11-gene

A rPSVM.1

0.0049

0.0035

0 20 40 60 80 100

Signature score (%)

Time to TB diagnosis (Days) 91- 180 181- 360 361- 540

>540 0-90

B

Progressors Non-progressors

0 20 40 60 80 100

Time to TB diagnosis (Days)

Signature score (%)

91-180 181-360 361- 540

>540 0-90

C

0 20 40 60 80 100

Time to TB diagnosis (Days)

Signature score (%)

Progressors Non-progressors

91- 180 181- 361- 360 540

>540 0-90

D

0 5 10 15 20 25

0 20 40 60 80 100

Time to culture positivity (Days)

Signature score (%)

r = -0.27 p = 0.28 E

0 5 10 15 20 25

0 20 40 60 80 100

Time to culture positivity (Days)

Signature score (%)

r = 0.29 p = 0.24 F

4.4.11: The ACS 11-gene and rPSVM.1 signatures could not differentiate Mtb infection from subclinical TB in ART naïve HIV-infected persons!

The TRuTH study was designed to detect recurrent TB disease using very intensive case finding methods, including performing induced sputum collection at every study visit. It is important to note that most study participants who were diagnosed with recurrent TB were asymptomatic at the time of TB diagnosis (n=25), suggesting that many of the progressors had subclinical TB, rather than active TB disease. Our previous work with the ACS 16-gene signature showed that subclinical TB could be detected in asymptomatic HIV-uninfected progressors up to a year before active TB diagnosis (Zak et al. 2016). These data support our interpretation that the ACS 11- gene and rPSVM.1 signatures are detecting subclinical disease in the TRuTH cohort, within three months of recurrent TB disease diagnosis.

To address this further, we aimed to determine if the signatures can differentiate between subclinical TB and Mtb-infected controls and active TB cases in HIV- infected persons. We hypothesized that the signatures can differentiate subclinical TB from Mtb infection in HIV-infected persons.

In order to address this, the ACS 11-gene and rPSVM.1 signatures were applied to a small cross-sectional cohort of HIV-infected individuals established by Esmail and Wilkinson, wherein subclinical TB was diagnosed by PET-CT in 10 asymptomatic individuals with Mtb infection who were identified to have active pulmonary lesions (Esmail et al., 2016, 2018). A group of 21 asymptomatic HIV-infected individuals with Mtb infection and no active pulmonary lesions were included as controls while 15 patients with microbiologically-confirmed active TB were included as cases. As

observed in the pilot cohort in Chapter 3, the signatures could readily differentiate active TB from Mtb-infected controls (Figure 16, Table 12). The ACS 11-gene signature could also differentiate subclinical TB from active TB, whilst the rPSVM.1 signature did not significantly differentiate between these two groups (Figure 16A &

B, Table 12). Neither signature could differentiate subclinical TB from Mtb infection. It was very notable that signature scores were generally very high in all three groups, including the Mtb-infected controls, perhaps suggesting very high HIV loads.

Regardless, median ACS 11-gene signature scores were higher in active TB cases than in subclinical TB and Mtb-infected controls (Figure 16C). The rPSVM.1 signature scores from the active TB cohort were only higher than that of the Mtb- infected control group (Figure 16D).

Table 12: Performance of the signatures in distinguishing subclinical TB from Mtb infection and active TB disease in ART naïve HIV-infected persons from the Esmail cohort (Esmail et al., 2016, 2018)

ACS 11-gene rPSVM.1

ROC AUC (95% CI) p-value ROC AUC (95% CI) p-value Subclinical vs

Mtb infection

0.55 (0.33-0.76) 0.69 0.64 (0.42-0.86) 0.20 Subclinical vs

Active TB

0.77 (0.56-0.98) 0.03 0.71 (0.49-0.92) 0.09 Active TB vs

Mtb infection

0.80 (0.65-0.95) 0.0027 0.85 (0.72-0.98) 0.0004

Figure 16: Ability of the signatures to distinguish subclinical TB from Mtb infection and active TB disease in the Esmail cohort. (A and B) ROC curves depict classification potential of the ACS 11-gene (A) and rPSVM.1 (B) signatures in distinguishing subclinical TB from Mtb infection and active TB disease in HIV-infected persons. (C and D) Distribution and differences in ACS 11- gene (C) and rPSVM.1 (D) signature scores between the three classes of TB.

Subclinical TB vs Active TB Mtb infection vs active TB Subclinical TB vs Mtb infection

0 20 40 60 80 100

0 20 40 60 80 100

100 - Specificity (%)

Sensitivity (%)

ACS 11-gene

A rPSVM.1

Subclinical TB vs Active TB Mtb infection vs active TB Subclinical TB vs Mtb infection

0 20 40 60 80 100

0 20 40 60 80 100

100 - Specificity (%)

Sensitivity (%)

B

0 20 40 60 80 100

Signature score (%)

Mtb infected Subclinical TB Active TB 0.7 0.02

C 0.0019 D

0 20 40 60 80 100

Signature score (%)

Mtb infected sub-clinical TB Active TB 0.09 0.21

0.0002

In document University of Cape Town (Page 141-147)