1 Using Prognostic Biomarkers to Target TB Preventive TherapyIDM 5th Annual Disease Modeling Symposium Seattle 2017 Mark Hatherill South African Tuberculosis Vaccine Initiative (SATVI) University of Cape Town, South Africa Mark
2 The global TB epidemic has peaked TB incidence rate falling at 1.5% per year Not fast enough to meet TB control goals
3 Would need 1,000-fold reduction in 33 years Target for TB elimination: Annual incidence <1 case per million population by 2050 Would need 1,000-fold reduction in 33 years 20% annual reduction is >2x faster than historical examples Dye, Annu Rev Public Heath 2013
4 Genetic, Age, HIV, Diabetes, SmokingAdapted from Barry 3rd et al,. Nat Rev Micro 2009 Risk Factors Genetic, Age, HIV, Diabetes, Smoking Continuum of Infection & Disease Quiescent infection Subclinical TB Disease Clinical Interferon-gamma release assay (IGRA) + MTB exposure Symptoms M.tuberculosis
5 High TB Incidence Settings60-80% of adult South Africans infected with M. tuberculosis IGRA+ IGRA- TB cases Mass preventive therapy for IGRA+ would treat most people unnecessarily Require massive resources Incomplete/slow coverage Shift focus from ‘treatment of latent infection’ to ‘targeted preventive therapy’
6 IGRA and Risk for Progression to TB Disease in AdolescentsAndrews, AJRCCM 2016 Recent IGRA+ converters are at highest risk of progression to TB disease (8x) Higher quantitative IGRA conversion value? (TB Ag – Nil Interferon γ concentration (IU/mL) IGRA reversion on serial testing…. MARK COMMUNITY-BASED IMPLEMENTATION SCREEN & TREAT STRATEGY DATA FROM CORTIS PREVALENT TB INCIDENT TB HIV UNINFECTD MAJORITY INCIDENT TB HIV INFECTED MAJORITY PREVALENT TB (25% NO ART/IPT) Andrews, unpublished
7 IGRA and Risk for Progression to TB Disease in InfantsAndrews, Lancet Respiratory Medicine 2017 MARK COMMUNITY-BASED IMPLEMENTATION SCREEN & TREAT STRATEGY DATA FROM CORTIS PREVALENT TB INCIDENT TB HIV UNINFECTD MAJORITY INCIDENT TB HIV INFECTED MAJORITY PREVALENT TB (25% NO ART/IPT) IGRA converters with highest quantitative interferon γ concentration (>4IU/mL) Less likely to revert on subsequent test 40 x higher TB disease risk than IGRA- Incipient TB disease?
8 Adapted from Barry 3rd et al,. Nat Rev Micro 2009How to identify incipient TB before progression to infectious active disease? Continuum of Infection & Disease Quiescent Infection Interferon-gamma release assay (IGRA) + MTB exposure Incipient TB Subclinical TB Disease Clinical TB Disease Symptoms M.tuberculosis
9 2,515 genes differentially expressed TB progressors vs controlsCases Controls Cases Figure 1: Whole blood gene expression is different in cases and controls. Over 2000 significantly differentially expressed genes were identified between cases and matched controls in the training set samples across time approaching TB disease diagnosis. These genes were used in constructing models to classify cases and controls using support vector machine learning algorithms and junction-junction pair selection in various analysis windows spanning 3 years before TB diagnosis to time of TB diagnosis. These genes are about equally split between those with higher expression in cases and those with lower expression in cases, compared with controls. Shown are expression differences between cases and controls in a variety of analysis windows. Red/blue on the heat map indicates enhanced/impaired expression in cases compared with controls. The colour scale above the heat map is indicative of 6 month windows of time approaching TB diseased diagnosis (yellow to red for cases, and blue for matching controls) Courtesy: Dan Zak. Red = Higher expression Blue = Lower expression *Also validation in household contacts (Walzl, GC6)
10 TB risk classification model based on transcript ‘voting pairs’Red edges vote as progressors Green edges vote as non-progressors 247 transcript pairs 47 PCR primers 16 genes A sample predicted to be a progressor A sample predicted to be a non-progressor
11 Y N RNA-Seq not practical for screening in high TB burden countriesBiomark Fluidigm qRT-PCR platform (test 96 genes, 46 duplicate) COR result in 2 days Maximum capacity approximately 400 samples per week Y N
12 16-gene COR prognostic performance Discriminates TB cases from controls up to 18 months before diagnosis COR 12-month Prognostic Performance 70% Sensitivity and 88% Specificity HIV uninfected South African adults (unpublished)
13 NDWG Stop TB Partnership Target Product Profile – Prognostic Biomarker for TB disease Optimal Test Profile Predict progression to TB disease Rule out active disease Revert to negative with Rx >90% sensitivity* >90% specificity* PPV >16% NNT <13 *Sens/Spec may be adjusted to achieve PPV / NNT target profile Minimal Acceptable Test Profile Predict progression to TB disease Need to screen out active disease >75% sensitivity* >90% specificity* PPV >6% NNT <40 Goal: “Better than IGRA” 78% sensitivity 58% specificity PPV 3.7% NNT 69 RR 2-3
14 Goletti et al, ERJ 2016
15 NDWG prognostic test TPPvs 16-gene COR 10 Optimum TPP - NNT: ~13 Optimum 20 30 COR COR signature - NNT: ~37 40 50 Minimum Minimum TPP - NNT: ~40 75 100 150 TST / IGRA - NNT: ~250 / ~85 TOM FIELD ANY QUERIES ABOUT COR PERFORMANCE, VALIDATION MARK FIND/NDWG TARGET PRODUCT PROFILE FOR NEW PROGNOSTIC TESTS FOR TB MINIMAL GOAL (75/75%) TO DOUBLE PVV AND HALVE NNT (8040) COMPARED TO IGRA ‘OPTIMAL’ 90/90 LIKELY UNACHIEVABLE FOR PROGNOSTIC TEST CONTINUUM OF TB RUNS IN BOTH DIRECTIONS SERIAL SCREENING TO ASSES CONVERSION, REVERSION, EFFECT OF THERAPY 250 TST IGRA Cumulative 2 year incidence: 2% Effectiveness of IPT: 50%
16 The landscape of TB preventive therapy has changedHIV infected Effectiveness 3HP non-inferior to 9H Higher treatment completion Similar tolerability HIV uninfected Effectiveness 3HP non-inferior to 9H Higher treatment completion Lower hepatotoxicity A 12-dose, once-weekly, 3-month DOT preventive therapy regimen
17 16-gene COR predicts incipient TB disease >1 year before diagnosis Performance maximal at diagnosis CoR genes Interferon module Inflammation module TOM RE-EMPHASIS 16-GENE COR PROGNOSTIC PERFORMANCE Scriba, Penn-Nicholson, Hanekom, Suliman, Kimbung & many others Gartland, Self & others Can the 16-gene Correlate of Risk be used as a triage test to target curative and preventive therapy?
18 COR Diagnostic PerformanceDoes the 16-gene COR have diagnostic potential as a triage test for undiagnosed TB? COR Diagnostic Performance 87% sensitivity 97% specificity (re-parameterized to published microarrary data) AUC 0.909 0.862 0.985 0.937 0.997 0.991 0.945 0.911
19 A 16-gene COR-targeted Screen & Treat Strategy for TB?Manipulate COR sensitivity % specificity for different purposes 80% COR vote threshold Investigation for Active TB Curative/Preventive Therapy 60-80% COR vote threshold Preventive Therapy <60% COR vote threshold No intervention MARK COMMUNITY-BASED IMPLEMENTATION SCREEN & TREAT STRATEGY DATA FROM CORTIS PREVALENT TB INCIDENT TB HIV UNINFECTD MAJORITY INCIDENT TB HIV INFECTED MAJORITY PREVALENT TB (25% NO ART/IPT) Scriba, Penn-Nicholson, Suliman, Darboe, Kimbung, many others
20 The Correlate of Risk Targeted Intervention Study (CORTIS)ClinicalTrials.gov NCT A Randomized, Partially-blinded, Clinical Trial of Isoniazid and Rifapentine (3HP) Therapy to Prevent Pulmonary Tuberculosis in High-risk Individuals Identified by a Transcriptomic Correlate of Risk Screen >10,000 HIV uninfected SA adult volunteers TB endemic communities Randomize on 16-gene COR+/- result Enrol 3,200 participants (1,500 COR+ and 1,700 COR-) Primary Aims 1: Test whether preventive therapy (3HP) reduces the rate of incident TB disease, compared to standard of care (active surveillance), in COR+ persons 2: Test whether COR status differentiates persons with cumulative prevalent or incident TB disease from persons without TB disease
21 The Correlate of Risk Targeted Intervention Study (CORTIS)ClinicalTrials.gov NCT COR+ Treatment Arm: open-label high dose INH and Rifapentine (12 weekly DOT doses) Blinded COR+ and COR- Observation Arm: active TB surveillance Investigation for TB disease at baseline, symptom-triggered, end of 15 months follow up
22 The Correlate of Risk Targeted Intervention Study (CORTIS)ClinicalTrials.gov NCT COR+ Treatment Arm: open-label high dose INH and Rifapentine (12 weekly DOT doses) Blinded COR+ and COR- Observation Arm: active TB surveillance Investigation for TB disease at baseline, symptom-triggered, end of 15 months follow up Group A COR+ Treated (open label) Group C COR- Observation (blinded) Group B COR+ Observation (blinded)
23 The Correlate of Risk Targeted Intervention Study (CORTIS)TE = Treatment Efficacy 3HP (a) vs (b) RR COR = Relative Risk for incident TB in COR+ vs COR- (b) vs (c) 80% power to reject the null-hypothesis, H0: TE(15) <= 20% under the simulated design hypothesis, TE(15) = 80%, with a one-sided alpha of 0.05 90% power to reject the null-hypothesis, H0: RRCOR(15) <= 2, with a one-sided alpha of
24 Validation of Correlates of Risk of TB Disease in High Risk Populations (CORTIS-HR)Modeling of COR targeted ‘screen & treat’ strategy in South African TB epidemic suggests 5-year impact optimal only if includes HIV infected persons Figure: Epidemic trajectories with screen and treat strategies. Left panel: comparison of impact of strategies on overall TB incidence. Median and 95% range of model predictions for: Red – baseline; Blue – COR strategy in HIV uninfected population only; Green – COR strategy in HIV uninfected and HIV infected population; Cyan – COR strategy in HIV infected population only. Black lines show WHO estimates for South Africa. Right panel: predicted impact of COR HR strategy on HIV+ TB. Sumner, White, unpublished
25 Does the 16-gene COR have diagnostic potential in HIV infected people?No prognostic data yet Risk for TB (CoR) TB disease Healthy LTBI Scriba, Penn-Nicholson, Suliman, Darboe, Kimbung, many others N=60 N=40
26 Association between HIV viral load and 16-gene COR % voteAll samples Samples with only detectable pVL Scriba, Penn-Nicholson, Suliman, Darboe, Kimbung, many others
27 Validation of Correlates of Risk of TB Disease in High Risk Populations(CORTIS-HR) Observational companion study of CORTIS Starting March 2017 Enrol 860 HIV infected SA adults Same TB endemic communities Diagnostic performance 16-gene COR for undiagnosed prevalent TB disease Prognostic performance 16-gene COR for incident TB disease over 15 months follow up Examine effect of IPT, cART and changes in CD4 count, HIV viral load on COR performance 80% power to detect TE > 20%, one-sided alpha 0.05 (alternative hypothesis TE=80%) 90% power to detect RR > 2, one-sided alpha (alternative hypothesis RR=5)
28 Translation to Point of Care: Battery-operated qRT-PCRTranslation to Point of Care: Battery-operated qRT-PCR? POC TB testing critical for mass screen & treat campaigns
29 Success of a COR-targeted Screen & Treat Strategy Dependent on…Success of a COR-targeted Screen & Treat Strategy Dependent on….. Speed and Extent of Coverage TB mortality (per 100k) TB disease incidence (per 100k) Adult latent prevalence (proportion) 10% pop / yr 30 % pop / yr 50% pop / yr Rapid initial decline in TB incidence Depending on % coverage, incidence declines toward 2025 global targets Rebound TB incidence after program end Reduction in prevalence of M.tb (latent) infection 80% power to detect TE > 20%, one-sided alpha 0.05 (alternative hypothesis TE=80%) 90% power to detect RR > 2, one-sided alpha (alternative hypothesis RR=5) Data Courtesy of Wagner, Chang, unpublished
30 Rebound in TB incidence related to size of M.tb infected poolSuccess of a COR-targeted Screen & Treat Strategy Dependent on…Duration of Program Adult latent prev (proportion) TB disease incidence (per 100k) Rebound in TB incidence related to size of M.tb infected pool Prolonged, sustained program >10 years needed to minimize rebound at program end 80% power to detect TE > 20%, one-sided alpha 0.05 (alternative hypothesis TE=80%) 90% power to detect RR > 2, one-sided alpha (alternative hypothesis RR=5) Blue line = baseline Dashed line = 10-years Solid line = 40 years Data Courtesy of Wagner, Chang, unpublished
31 RNA diagnostic and prognostic biomarker = test of incipient TB diseaseNew, short-course TB preventive therapy regimens Next steps… CORTIS (HIV uninfected) performance and efficacy data 2019 Validate smaller 6-gene COR model...ongoing Transfer COR to point of care, handheld PCR device...ongoing CORTIS-HR (HIV infected) performance data 2019 Test COR-targeted Screen & Treat Strategy for HIV infected patients? Trigger cycles of short-course preventive therapy? MARK IMPORTANCE OF INCLUDING HIV INFECTED PEOPLE IN SCREEN & TREAT STRATEGY CHALLENGES OF TESTING AND IMPLEMENTING 3HP/1HP IN SA (CONTINUOUS IPT) WHO SHORTER TREATMENT GUIDELINES ROLE OF COR FOR TARGETING AND TRIGGERING ROLE OF POC TEST NEED FOR LOCAL CLINIC-BASE TESTING USING HIV STRUCTURES
32 COR & CORTIS & CORTIS-HRAcknowledgements COR & CORTIS & CORTIS-HR Participants Collaborators Funders Tom Scriba Adam Penn-Nicholson Willem Hanekom Elisa Nemes Sara Suliman Mbandi Kimbung Fatoumatta Darboe Chris Hikuam Kate Hadley Dan Zak Ethan Thompson Lynn Amon Gerhard Walzl Andre Loxton Jayne Sutherland Alan Aderem GC6 Team Gavin Churchyard Kogie Naidoo Andrew Gartland Steve Self Richard White Tom Sumner Bradley Wagner Stewart Chang Alex Sette and Denise Baker – La Jolla Institute of allergy & immunology Holden Maecker and Mark Davis - Stanford