Update on IGRA Predictive Value Sandra Kik, PhD Molebogeng Rangaka, MD, PhD Madhukar Pai, MD, PhD McGill International TB Centre, McGill University University of Cape Town & London School of Hygiene and Trop Med sandra.kik@mail.mcgill.ca
Latent tuberculosis infection (LTBI) Hugh reservoir of people with a latent tuberculosis infection 1/3 rd of the world population is estimated to be infected with M. tuberculosis (approximately 2.3 billion people) Yearly about 8.7 million new tuberculosis cases (2011) 2
Simple distinction between LTBI and active TB No infection CONTAC T Infection(LTBI) No disease progression (90%) Early (5%) Progression to active disease Late (5%) = reactivation Estimated 10% of LTBI cases will develop active TB Highest risk of disease progression 2-5 years after infection
Diagnostics in use for diagnosis of LTBI Tuberculin skin test (TST) Interferon-gamma release assays (IGRA) QuantiFERON TB Gold In Tube? T-SPOT.TB In house RD1 ELISA or ELISPOT 4
Guidelines on IGRAs IGRAs not recommended for diagnosis of active TB Most low incidence countries recommend IGRAs for contact investigation in adults As a single test After a positive TST Or together with TST Denkinger et al. CMI 2011 5
Visits Immune response LTBI vs active disease TST or IGRA? TST or IGRA? Multiple visits needed for TST, single visit for blood draw IGRA Both measure evidence of TB infection. TST likely to remain positive, IGRA more likely to revert over time Both cannot distinguish between LTBI and active disease Sensitivity Similar, in the order of 70-80% Specificity HIV TST low specificity since impacted by previous BCG vaccination IGRA high specificity (~95%) Both are impacted by HIV (less sensitive), but IGRA less so 6
Systematic review and meta-analysis published in 2012 New studies have been published since this time 7
We are currently updating this systematic review Objectives: Can IGRA predict the development of incident active TB? Is predictive ability of IGRA higher than TST? What is the incidence rate of TB in IGRA+ and IGRA- subjects? What is cumulative risk ratio for pos vs neg IGRA by incorporation bias? What is incidence rate ratio of IGRA+ vs IGRA- by type of IGRA? 8
Methods Search was repeated for publications since Jun 2011-Jan 2013, using same terms and in- and exclusion criteria Inclusion criteria: Longitudinal studies (pro/retrospective), at least 1 yr follow-up Adults /children, with/without HIV, without TB at baseline Any IGRA (in house or commercial) with RD1 antigens Reported any incidence of (clinical) TB, no restrictions on how TB was diagnosed Exclusion criteria: Studies that had zero incident TB cases Data extraction: 1 reviewer extracted data, other reviewer verified all data extracted 9
Methods: quality assessment Modified Newcastle-Ottawa quality assessment scale for observational studies* Selection of study groups Whether methods to exclude TB included smear and/or culture Comparability of study groups Ascertainment of exposure or outcome Microbiological confirmation of TB outcomes Assessment of whether IGRA results were incorporated into the diagnosis of active TB (incorporation bias) Assessment of whether IGRA+ people where more actively investigated for TB disease than IGRA- people (differential work-up bias) *http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp 10
Example of study with high likelihood of incorporation & differential work-up bias 11
Work-up bias: Example of study with high likelihood of incorporation & differential work-up bias Incorporation bias: Study staff not blinded to IGRA results 12
Example of study with high likelihood of incorporation & differential work-up bias Discussion section of Aichelburg paper, acknowledge possibility of bias: Incorporation bias / work-up bias may lead to: Overestimation of the TB incidence rate in IGRA positives Overestimation of the risk ratio of IGRA positives vs IGRA negatives Overestimation of the incidence rate ratio of IGRA positives vs IGRA negatives 13
Overview previous and new studies (N=20) 14
Study characteristics of included studies (15 + 5 new) Study area: 11 High income countries (Austria, Netherlands, Norway, Japan, Germany, Portugal, Korea, UK, France) 5 Middle income countries (Turkey, Colombia, China, India, South Africa) 4 Low income countries (Ethiopia, The Gambia, Senegal, Kenya) Type of study population TB contacts: 10 studies Outpatients: 3x HIV+, 2x other immuno-compromised Other: renal transplant recipients, health care workers, adolescents with no prior TB, asylum seekers, HIV exposed infants IGRA s assessed: ELISPOT: 8* (3x in-house, 5x T-SPOT.TB) WBA: 13* (2x in-house, 11x QFT-GIT) * 1 study assessed both T-SPOT.TB and QFT-GIT 15
Study characteristics of included studies (15 + 5 new) TST assessed: 12/20 All 20 studies used one IGRA at baseline (no repeat testing for conversions) Median follow-up: 3 yr IPT: given to all/some of the study participants in 11/20 16
Studies included in our analysis What is the incidence rate of TB in IGRA+ and IGRAsubjects? 12/20 studies What is cumulative risk ratio for pos vs neg IGRA by incorporation bias? 20/20 studies What is incidence rate ratio of IGRA+ vs IGRA- by type of IGRA? 9/20 studies (only those without incorporation bias) 17
New New Study ID IGRA positive Hill (2008) Bakir (2008) Kik (2009), WBA Kik (2009), ELISPOT del Corral (2009) Lienhardt (2010) Leung (2010) Jonnalagadda (2010), mothers Jonnalagadda (2010), infants Joshi (2011) Mahomed (2011) Kim-SH (2011) Lange (2012) Kim-Y (2012). IGRA negative Hill (2008) Bakir (2008) Kik (2009), WBA Kik (2009), ELISPOT del Corral (2009) Lienhardt (2010) Leung (2010) Jonnalagadda (2010), mothers Jonnalagadda (2010), infants Joshi (2011) Mahomed (2011) Kim-Y (2012). NOTE: Weights are from random effects analysis TB incidence rates (per 1000) Silicotic men HIV-infected moms and infants HIV-infected outpatients, untreated Incidence Rate (95% CI) 9.24 (3.80, 14.60) 20.50 (10.00, 36.00) 17.00 (5.50, 39.00) 19.00 (7.30, 43.40) 7.70 (5.00, 11.30) 14.40 (8.40, 23.00) 32.00 (17.80, 52.30) 42.00 (15.00, 91.00) 48.00 (16.00, 112.00) 3.69 (1.66, 8.00) 6.40 (4.50, 8.00) 25.70 (8.17, 62.10) 12.00 (0.00, 33.00) 84.50 (34.00, 176.80) 5.00 (1.90, 8.10) 6.00 (1.60, 15.40) 12.10 (2.50, 35.00) 10.40 (1.30, 37.50) 4.00 (1.70, 8.50) 6.90 (2.50, 14.90) 7.10 (0.90, 25.50) 16.00 (3.00, 47.00) 24.00 (5.00, 69.00) 3.38 (1.69, 6.76) 2.00 (1.20, 3.80) 32.00 (11.80, 71.50) 12/20 publications reported TB incidence rates IGRA-positive: Incidence ranges from 0.4 to 8% IGRA-negative: Incidence ranges from 0.2 to 3% -25 0 25 50 75 100 125 18
New studies New studies Association between IGRA and incident TB: Study ID RR, stratified by potential incorporation/work-up bias None Doherty (2002) Hill (2008) Bakir (2008) Kik (2009) Kik (2009) del Corral (2009) Lienhardt (2010) Leung (2010) Jonnalaggada (2010) Jonnalaggada (2010) Joshi (2011) Kim-SH (2011) Kim-Y (2012) Subtotal (I-squared = 0.0%, p = 0.721). Possible Aichelburg (2009) Yoshiyama (2010) Harstad (2010) Diel (2011) Mahomed (2011) Costa (2011) Haldar (2012) Lange (2012) Bergot (2012) Subtotal (I-squared = 55.2%, p = 0.022). Overall (I-squared = 47.6%, p = 0.007) NOTE: Weights are from random effects analysis Risk Ratio (95% CI) 10.00 (1.42, 70.22) 1.84 (0.79, 4.31) 3.80 (1.22, 11.86) 1.96 (0.40, 9.53) 1.40 (0.34, 5.74) 1.89 (0.83, 4.34) 2.13 (0.85, 5.34) 3.82 (0.89, 16.40) 2.69 (0.69, 10.52) 2.23 (0.54, 9.12) 1.09 (0.38, 3.10) 17.70 (0.96, 325.14) 3.32 (1.09, 10.08) 2.29 (1.65, 3.17) 136.13 (7.16, 2588.46) 6.74 (3.63, 12.52) 18.76 (2.36, 149.22) 148.36 (9.00, 2446.45) 2.89 (1.55, 5.41) 18.38 (0.99, 341.04) 6.52 (2.54, 16.76) 27.77 (1.15, 670.11) 3.55 (0.22, 56.48) 9.02 (4.32, 18.84) 3.71 (2.53, 5.43) 20/20 publications reported RR No bias: Overall RR 2.29 (1.65-3.17) Possible bias: Overall RR 9.02 (4.32-18.84).1 1 100 19
Association between IGRA and incident TB: IRR, stratified by type of assay (only studies without incorporation bias) Study ID Incidence Rate Ratio (95% CI) New ELISPOT Hill (2008) Bakir (2008) Kik (2009), ELISPOT Lienhardt (2010) Leung (2010) Jonnalagadda (2010), mothers Jonnalagadda (2010), infants Kim-Y (2012) Subtotal (I-squared = 0.0%, p = 0.997). WBA Kik (2009), WBA del Corral (2009) Joshi (2011) Subtotal (I-squared = 0.0%, p = 0.784). Overall (I-squared = 0.0%, p = 0.996) 1.90 (0.80, 4.50) 3.41 (1.08, 10.70) 1.41 (0.32, 7.15) 2.00 (0.84, 5.41) 4.50 (1.03, 19.68) 2.60 (0.70, 10.30) 2.00 (0.40, 13.10) 2.62 (0.76, 9.33) 2.09 (0.94, 3.25) 1.91 (0.41, 13.78) 1.82 (0.80, 4.20) 1.00 (0.31, 3.58) 1.41 (0.25, 2.57) 1.75 (0.93, 2.57) ELISPOT: IRR: 2.09 (0.94-3.25) WBA: IRR: 1.75 (0.93-2.57) NOTE: Weights are from random effects analysis -5 1 10 20 20
IGRA vs TST comparison: which has greater predictive value? (studies that did a head-to-head) None of the new studies qualified to be included in this analysis; thus results of Ranganka et al still hold. 21
Incidence rate of discordant results: (studies that did a head-to-head) Incidence rate per 1000 person years Gambia (Hill et al) Turkey (Bakir et al) Senegal (Lienhardt et al) India (Joshi et al) TST+ / IGRA+ 8.9 22.2 14.8 3.9 TST+ / IGRA - 9.6 7.4 4.9 2.7 TST- / IGRA+ 12.4 11.7 5.9 2.9 TST- / IGRA- 4.0 5.1 9.9 3.5 None of the new studies qualified to be included in this analysis; thus results of Ranganka et al still hold. 22
Work in progress Final search to be updated till July 2013. We know of at least 2 studies that are expected to be published soon. Secondary analysis still to be performed: Stratification by preventive therapy & HIV-status. 23
Conclusions thus far of this update Incidence rates of TB, even in IGRA positive individuals, are low, suggesting that a vast majority (>95%) of IGRA+ individuals do not progress to TB disease during follow-up. This is similar to the TST. Based on the evidence thus far, IGRAs appear to have similar predictive value as the TST (perhaps slightly higher, but statistically not significant). All existing LTBI tests (TST and IGRAs) have only modest predictive value and may not help identify those who are at highest risk of progression to disease. In some settings (mostly low TB incidence), the % IGRA+ will be less than % TST+, reducing the number needed for preventive therapy. 24
TST can be positive in all states except the last IGRAs are likely to be positive in all states as well? Or spontaneous convert/revert? Nat Rev Micro 2009 25
Are there ways to increase the predictive value for a LTBI test? 1. Only test those who are at high risk 2. Incorporate biomarkers with other known risk factors (age, recent conversion, HIV etc.) into a composite scoring system to generate multivariable risk prediction models 3. Use a higher cut-off for prediction (as compared to diagnosis) 4. Use serial testing to resolve underlying phenotypes (e.g. stable conversions) 5. Identify new biomarkers and measure an array of biomarkers (biosignature) 26
Bottom line IGRAs are an incremental advance; not transformational* We are still looking for a highly predictive LTBI test that can help target preventive therapy We need biomarkers and composite risk prediction models that can help resolve the various phases of LTBI spectrum *LoBue P & Castro K. JAMA 2012 27
Acknowledgement LSHTM / University of Cape Town Lele Rangaka for the hard work/support McGill University Madhukar Pai Dick Menzies