Measuring mortality due to HIV-associated TB Verbal and

Measuring mortality due to HIV-associated TB Verbal and

Measuring mortality due to HIV-associated TB Verbal and minimally-invasive autopsies conducted in South Africa Sanj Karat on behalf of the Lesedi Kamoso team: Mpho Tlali, Salome Charalambous, Violet Chihota, Gavin Churchyard, Yasmeen Hanifa, Suzanne Johnson, Kathleen Kahn, Neil Martinson, Kerrigan McCarthy, Tanvier Omar, Katherine Fielding, and Alison Grant Improving health worldwide www.lshtm.ac.uk Counting TB deaths 1.5 million TB deaths

390,000 among HIV-positive people 95% reduction in TB deaths by 2035 WHO (2015) Global tuberculosis report WHO (2015) The END TB strategy Background Vital registration and cause of death Glaziou et al. Technical appendix to global TB report 2015

ICD-coding of TB deaths human immunodeficiency [HIV] disease resulting in tuberculosis (B20.0) International classification of diseases (ICD), 10 th revision ICD-coding of HIV-associated TB International classification of diseases (ICD), 10 th revision HIV+ TB deaths counted as HIV International classification of diseases (ICD), 10 th revision

Modelling for mortality Estimates of mortality due to HIV-associated TB Modelling TB case-fatality ratios + Pooled research data HIV+ TB incidence

Modelling Prevalence surveys Prevalence surveys Modelling Prevalence surveys Glaziou P, et al. (2015) WHO Global TB Report 2015: Technical appendix UNAIDS (2015) Methods for deriving UNAIDS estimates

Straetemans M, et al. (2011) PLoS One 6(6): e20755 + Casenotification Tissue autopsy: the gold standard Cain 2015 Kenya 11% TB Lucas 1993 Cote dIvoire 38% TB

Nelson 1993 Zaire 41% TB Omar 2015 South Africa 32% TB Ansari 2002 Botswana 40% TB Murray 2007 South Africa 47% TB Martinson 2007

South Africa 79% TB Rana 2000 Kenya 51% TB Cox 2012 Uganda 46% TB Siika 2012 Kenya 34% TB Ngwalali 2005

Tanzania 19% TB Bates 2015 Zambia 65% TB Wong 2012 South Africa 64% TB Cohen 2010 South Africa 47% TB

Menendez 2008 Mozambique 18% TB Garcia-Jardon 2011 South Africa 51% TB Verbal autopsy: useful but limited Setting The TB Fast Track study TB Fast Track a reminder

Population 24 PHCs in semi-rural and peri-urban South Africa HIV+ adults (age 18) CD4 150 cells/lll No ART (6 months) No TB treatment (3 months) Ambulatory at enrolment Lesedi kamoso: research questions Among those who died after enrolment to TB Fast Track 1. What was the prevalence of active TB disease at autopsy? 2. What were the causes of death as assigned by clinical

methods and verbal autopsy? 3. How did verbal autopsy perform in assigning causes of death? Methods HOSPITAL Parent study Hospitals & clinics NHLS database MIA Autopsy data

Verbal autopsy Clinical and research data VA data Methods HOSPITAL Parent study Hospitals & clinics NHLS database

MIA Autopsy data Verbal autopsy Clinical and research data Clinico-pathological panel VA data PCVA Gold standard CoD Autopsy prevalence

Comparison CCVA VA CoD Demographics (n=259) Female 116/259 (55%) Median age

39 (IQR 33-46) years Median CD4 count at enrolment 44 (IQR 20-88) cells/lL Median time from enrolment to death 79 (IQR 34-170) days Death in hospital 145/259 (68%)

Median time from death to VA 140 (IQR 76-288) days Objective 1 Autopsy prevalence Those with MIA (n=34) Female 18/34 (53%) Median age

39 (IQR 33-44) years Median CD4 count at enrolment 34 (IQR 17-66) cells/lL Median time from enrolment to death 60 (IQR 21-175) days Death in hospital 25/34 (74%)

Median time from death to MIA 5 (IQR 3-6) days Autopsy prevalence (n=34) Active TB disease 16/34 (47%) Individuals with TB (n=16) Pulmonary TB 13/16 (81%) Individuals with TB (n=16) Extra pulmonary TB

14/16 (89%) Individuals with TB (n=16) EPTB only 3/16 (19%) PTB + EPTB 11/16 (69%) PTB only, 2/16 (13%) Autopsy prevalence (n=34) Active TB disease 16/34 (47%)

Autopsy prevalence (n=34) Other bacterial infections 23/34 (68%) TB + bacterial infection 9/16 (56%) Autopsy prevalence (n=34) Bacterial pneumonia 11/34 (32%) TB + pneumonia

5/16 (31%) Autopsy prevalence (n=34) NTM disease 3/34 (9%) TB + NTM 1/16 (6%) Autopsy prevalence (n=34) Cryptococcal disease 4/34 (13%) TB + Crypto 2/16 (13%)

Autopsy prevalence (n=34) Pneumocystis pneumonia 2/34 (6%) TB + PCP 1/16 (6%) Autopsy prevalence (n=34) CMV disease 2/34 (9%) TB + CMV 1/16 (6%) Autopsy prevalence (n=34)

Disease Active TB Other bacterial Bact. pneumonia NTM disease Crypto. Disease PCP Prevalence, n (%/34)

16 (47%) 23 (68%) 11 (32%) 3 (9%) 4 (13%) 2 (6%) Objective 2 Clinical causes of death Separating HIV from HIV/TB (n=259) Gold standard (ICD-10) HIV/AIDS 65%

Other 35% Separating HIV from HIV/TB (n=259) Gold standard (ICD-10) HIV/AIDS 65% HIV/AIDS 38% Other 35%

HIV-assoc. TB 27% Other 35% Gold standard (showing HIV-associated TB) Objective 3 Comparison to VA Comparison to VA CoD (n=259) Gold standard (ICD-10)

HIV/AIDS 65% Other 35% 10 Comparison to VA CoD (n=259) Gold standard (ICD-10) HIV/AIDS 65% HIV/AIDS 76%

Other 35% PTB (HIV-) Other 1% 23% VA: Physician-certified (ICD-10) Comparison to VA CoD (n=259) Gold standard (ICD-10)

Performance of VA method (n=259) IndividualHIV/AIDS agreement 65% Cohens kappa PCVA Other 35% 0.04 Overall Chance-corrected concordance 0.06

HIV/AIDS 76% PTB (HIV-) Other 1% 23% VA: Physician-certified (ICD-10) Comparison to VA CoD (n=259) Gold standard (ICD-10)

Performance of VA method (n=259) IndividualHIV/AIDS agreement 65% Cohens kappa PCVA Other 35% 0.04 Overall Chance-corrected concordance 0.06

Proportional agreement PTB Lins concordance correlation coefficient (HIV-) 0.94Other HIV/AIDS 1% CSMF accuracy 76% 0.75 23% VA: Physician-certified (ICD-10) CSMF=cause-specific mortality fraction

Comparison to VA CoD (n=259) Gold standard (ICD-10) HIV/AIDS 65% HIV/AIDS 50% Other 35% PTB (HIV-)

16% Other 34% VA: InterVA-4 software (ICD-10) Comparison to VA CoD (n=259) Gold standard (ICD-10) Performance of VA method (n=259) IndividualHIV/AIDS agreement 65% Cohens kappa

Overall Chance-corrected concordance HIV/AIDS 50% PTB (HIV-) 16% InterVA-4 Other 35% 0.08 0.13

Other 34% VA: InterVA-4 software (ICD-10) CSMF=cause-specific mortality fraction Comparison to VA CoD (n=259) Gold standard (ICD-10) Performance of VA method (n=259) IndividualHIV/AIDS agreement 65% Cohens kappa

Overall Chance-corrected concordance InterVA-4 Other 35% 0.08 0.13 Proportional agreement PTB LinsHIV/AIDS concordance correlation coefficient (HIV-) 50%

CSMF accuracy 16% 0.76 Other 34% 0.62 VA: InterVA-4 software (ICD-10) CSMF=cause-specific mortality fraction Comparison to VA CoD (n=259) Gold standard (ICD-10) Performance of VA methods (n=259)

HIV/AIDS Individual agreement 65% Cohens kappa PCVA InterVA-4 Other 35% 0.04 0.08 Overall Chance-corrected concordance

0.06 0.13 Proportional agreement PTB Lins concordance HIV/AIDS correlation coefficient (HIV-) 50% CSMF accuracy 16%

0.94 Other 0.76 0.75 34%0.62 VA: InterVA-4 software (ICD-10) CSMF=cause-specific mortality fraction Summary The autopsy prevalence of TB was high among individuals enrolled in the community, similar to findings from previous hospital-based studies Majority of those with TB had extrapulmonary and/lor disseminated disease Bacterial and other infections were common; several

individuals had multiple infections TB was the main cause of death in a over a quarter of decedents, but this was not clear when ICD-10 codes were used VA performed poorly in assigning individual CoD, but provided more accurate population-level estimates Recommendations A need to directly measure mortality due to HIV-associated TB Consider amendments to ICD coding of HIV deaths VA currently unsuitable for individual CoD assignment in areas of high HIV prevalence Implement MIA at sentinel sites to allow direct

estimation of TB prevalence at death Acknowledgments We owe an enormous debt to the generous individuals and their families who gave their consent and time for this study Funding The Bill and Melinda Gates Foundation Collaborators South African Department of Health: provincial and district ethics committees Clinic staff at all the TB Fast Track, XPHACTOR, and XTEND clinics Management and staff at Dilokong, Dr George Mukhari, Jane Furse, Jubilee, Mecklenberg, Odi, and Tembisa hospitals The parent study teams TB Fast Track, XPHACTOR, and XTEND

Reviewing physicians Prof Lucille Blumberg and Drs Kim Roberg, Sarah Stacey, Michelle Venter, Sarah Stacey, Evan Shoul, and Dave Spencer The Lesedi Kamoso team Bongani Nkaqa, Khethekile Ntsontso, Monde Phasha, Mphonyana Motsapi, Thabo Setimela, and Zanele Nthebe

Recently Viewed Presentations

  • Lesson 3 Begin With The End in Mind:

    Lesson 3 Begin With The End in Mind:

    Classic Envelopment. Classic Envelopment (almost) Schlieffen Plan (1905, 1914) Classic Envelopment. Napoleonic Wars. Guest Speaker. General Norman Schwarzkopf. Briefing on the Conduct of the Gulf War. Riyadh, Saudi Arabia. February 27, 1991. Gen. Schwarzkopf Briefing
  • Recomn in Bacteria and Viruses - University of Vermont

    Recomn in Bacteria and Viruses - University of Vermont

    30. Genetics and recombination in bacteria Lecture Outline 11/16/05 Replication in bacteria Types of recombination in bacteria Transduction by phage Conjugation ("mating") F+ plasmids Hfr strains Transformation of raw DNA Evidence for recombination in nature Resistance plasmids The Bacterial Genome...
  • Poisoning and Substance Abuse

    Poisoning and Substance Abuse

    Some toxic substances (carbon monoxide) are very poisonous but not irritating. Other toxic gases (chlorine gas and ammonia) are very irritating to the respiratory tract. IV. Inhaled Poisons. A. Poisoning by inhalation occurs if a toxic substance is breathed in...
  • Oliver Is Taken to the Workhouse GUIDED ANALYSIS

    Oliver Is Taken to the Workhouse GUIDED ANALYSIS

    To the child Mr Bumble represents the power of the institutions and he is awed by him; however, Oliver lies to him, after seeing Mrs Mann threatening him with her fist in the background, as he fears her reaction. A...
  • groovy - faculty.etsu.edu

    groovy - faculty.etsu.edu

    Modifications for Students with Learning Disabilities NET-AT Workshop #2 January 20, 2006
  • Detection Algorithms for Biosurveillance: A tutorial Andrew Moore

    Detection Algorithms for Biosurveillance: A tutorial Andrew Moore

    Arial Tahoma Wingdings Math1 Times Times New Roman Letter Gothic MT Courier New MS Mincho Symbol Blends Microsoft Excel Worksheet Microsoft Word Picture Picture (Metafile) Bitmap Image Detection Algorithms for Biosurveillance: A tutorial Slide 2 What you'll learn about What...
  • Da Kmayl (The Supplication of Kumayl ibn Ziyad)

    Da Kmayl (The Supplication of Kumayl ibn Ziyad)

    Merits. Kumayl. Ibn. Ziyad. Nakha'i (r.a) was a confidant amongst the companions of Imam Ali . Ibn. Abi. Talib (a.s) and this sublime . Du'a. was first heard from the beautiful, though anguished, voice of Imam Ali
  • PowerPoint Presentation

    PowerPoint Presentation

    IS400: Development of Business Applications on the Internet Fall 2004 Instructor: Dr. Boris Jukic JavaScript: Arrays Introduction Arrays Data structures of related items Each element has a position number Dynamic Size of an array in JavaScript can be changed (increased)...