has hiv prevalence decreased in india and what caused the decreases? peter vickerman
TRANSCRIPT
Has HIV prevalence decreased in India and what caused the decreases?
Peter Vickerman
HAS HIV PREVALENCE DECREASED?
Indian sentinel surveillance• Suggests decline in HIV prevalence in all groups for 2003 to 20101:
• Similar relative declines in FSW, MSM and ANC, BUT smaller in IDU
• Used to produce modelled estimates – suggests ↓ by third from 2000-112
• These estimates average trends over contrasting states:• How do trends vary at the state level?
1. HIV sentinel surveillance 2010-11: a technical brief2. Technical report India HIV estimates 2012
R² = 0.9729
-50%
-40%
-30%
-20%
-10%
0%
10%
0% 10% 20% 30% 40% 50% 60%
Chan
ge in
HIV
pre
vale
nce
from
200
3 to
201
1
FSW HIV prevalence in 2003
R² = 0.8576
-1.4%
-1.2%
-1.0%
-0.8%
-0.6%
-0.4%
-0.2%
0.0%
0.2%
0.4%
0.6%
0.0% 0.5% 1.0% 1.5% 2.0%
Chan
ge in
HIV
pre
vale
nce
from
200
3 to
201
1
ANC HIV prevalence in 2003
Change in HIV prevalence at state level highly correlated to baseline HIV prevalence
• All high prevalence states have decreased BUT some low prevalence states have increased
• Larger decreases in states with higher HIV prevalence in 2003:• Epidemic stage may be playing a role?
ANC HIV prevalence FSW HIV prevalence
Why has HIV prevalence decreased?
• Number of analyses have explored why HIV prevalence has declined in FSW and ANC or general population:
• Statistical analyses to determine if related to intervention parameters for key populations
• Model analyses
STATISTICAL ANALYSES
Analysis 1: Decreases in FSW HIV prevalence• Cluster randomised sampling in 24 districts in 4 southern states: 2 to 3
rounds of IBBA from 2004 to 20091
• FSW HIV prevalence declined in districts related to program coverage1
• Also large increases in condom use and intervention outcomes over these periods, with condom use related to program exposure2
Round 1 % Round 3 %Adjusted OR
(95% CI) P value
HIV-1 infection 24.8 13.3 0.54 (0.40-0.74) < 0.001
Syphilis 12.7 8.8 0.57 (0.32-1.02) 0.06
High-titre syphilis 6.6 3.2 0.44 (0.25-0.77) 0.004
Has not been con-
tacted
Less than one year
One year Two years Three years0
20
40
60
80
100
Duration since first contacted by intervention staff
% F
SW
s u
sin
g c
on
do
ms
con
sist
en
tly
1. Alary STD 20142. Deering BMC PH 2011
Analysis 2: Ng Lancet 2011 • Considered whether the decline in ANC HIV prevalence in a district is
related to intervention investment (per HIV infected) for key populations
• Found negative association –greater declines in HIV prevalence if more investment in key population interventions
• Did not consider possible causal pathway
Analysis 3: Banandur, JAIDS 2013
• Did not consider possible causal pathway
• Considered whether decline in ANC HIV prevalence for young women (<25 years – incidence proxy) is related to being in a high intensity ‘Avahan’ intervention district for key populations
• Found negative association – Higher intervention intensity Avahan districts observed greater declines in HIV prevalence, 2003-2008
• However, decline in HIV prevalence also in ‘control’ non-Avahan districts and decline catches up
Analysis 4. Arora BMJ Open 2013• Considered if coverage of or spending on FSW interventions is related
to HIV and Syphilis prevalence in young ANC women (<25 years)
• Only number of STIs treated in FSWs is related to HIV prevalence in ANC women but all attributes are related to syphilis risk for ANC women
• Although effects not so strong for HIV, they do consider one step further along causal pathway and all interventions (not just Avahan)
MODEL ANALYSES
Modelling evidence – collection of analyses• HIV transmission shown to be driven by FSWs in India1
• Modelling within a Bayesian framework showed declines in HIV prevalence amongst FSWs can not be explained by natural HIV dynamics, but are likely due to ‘additional’ increases in condom use amongst FSWs2,3
• These models predict substantial (42%) decrease in HIV transmission in whole population2,3 and,
• Prevalence declines in general population consistent with ANC trends1
1. Vickerman AIDS 20102. Boily AIDS 20133. Pickles LGH 2013
Modelling evidence• HIV transmission is driven by FSWs in India1
• Modelling within a Bayesian framework showed declines in HIV prevalence amongst FSWs can not be explained by natural HIV dynamics, but are likely due to ‘extra’ increases in condom use amongst FSWs2,3
• These models predict substantial decrease in HIV transmission in whole population2,3 and,
• Prevalence declines in general population consistent with ANC trends1
1. Vickerman AIDS 20102. Boily AIDS 20133. Pickles LGH 2013
Mysore Belgaum
Discussion• Evidence suggests HIV prevalence has decreased in most
states and population sub-groups
• Statistical analyses suggest some of these declines may be due to scale up of key population targeted interventions
• Although prevalence may have been decreasing anyway
• Modelling suggests HIV transmission is driven by FSW and that recent scale up of FSW/MSM targeted interventions has decreased overall HIV transmission.
• Limitations: 1. Most evidence is based on Avahan districts in 4 southern states; 2. Lack of understanding why some states have increasing prevalence
Rest of the CHARME Team• Centre hospitalier affilié universitaire
de Québec (Laval University): lead institution: direction and coordination• Michel Alary (principal investigator)
• Imperial College of Science, Technology and Medicine• Marie-Claude Boily• Mike Pickles • Anna Phillips• Sharmistha Misra
• London School of Hygiene and Tropical Medicine• Anna Vassall• Sudha Chandrashekar • Charlotte Watts• Mike Pickles• Kate Mitchell
• University of Manitoba / Karnataka Health Promotion Trust• Stephen Moses; James Blanchard; B.M.
Ramesh; Reynold Washington; Sushena Reza Paul; Shajy Isac; Jan Bradley
• Field data collection; project infrastructure in India
• INDIAN PARTNERS AND COLLABORATING INSTITUTIONS• St John’s Medical College (IPHCR),
Bangalore• Tata Institute of Social Sciences, Mumbai• Academic Staff College of India, Hyderabad• Centre for Media Studies, Hyderabad
• University of British Columbia• Kathleen Deering
• Dept. HIV&STIs, PHE, London, UK• Catherine Lowndes
• University of Montreal• Annie-Claude Labbé
Hypothesis-testing
What does the data suggest- sentinel surveillance data
• Main data used to obtain trends in HIV prevalence
How collected:
• ANC: Women consecutively sampled – 400 in each clinic
• STD clinic attendees are consecutively sampled amongst those with STD syndromes– 250 from each site
• High-risk groups sampled from service points (de-addiction center, drop-in centers, clinics) – 250 at each
• Testing is anonymous, unlinked and quality assurance is undertaken
HIV sentinel surveillance operational guidelines
How has the sentinel sites scaled up• Number of ANC sites scaled up massively until 2006 but then
remained stable till 2011 (506 urban and 182 rural)
• Number of HRG sites very low before 2003, and still increasing – IDU (79), FSW (261), MSM (96) in 2011
• Over 250,000 ANC women and 100,000 HRG tested
Annual HIV sentinel surveillance country report 2006
Avahan AIM: “To slow rate of HIV transmission to general population by raising coverage of targeted interventions across large areas.”
General population
Impact evaluation
High-risk groups
General population
Targeted interventions-Use mathematical modelling within a Bayesian framework to examine whether trends in HIV prevalence in different high risk groups:
Could be due to natural disease dynamics and pre-existing interventions?OR are the trends more likely due to the possible increases in condom use after Avahan intervention.Estimate impact associated with increases in condom use that account for parameter uncertaintyCan increases in condom use be attributed to Avahan
Bridging infectionsBridging infections
ANC HIV prevalence at state level 2004-2011
• HIV prevalence decreased at national level and in high prevalence (HP) states and some low prevalence (LP) states, BUT
HIV sentinel surveillance 2010-11: a technical brief
• Stable or slowly rising in other LP states
• Number of ANC sites with HIV prevalence>1%:
• ↓135 to 65 in HP states,
• ↑ 8 to 24 in LP states
• Similar trends for FSW and MSM, BUT IDU stable or slowly increasing
Modelled estimates using methods recommended by UNAIDS/WHO Global Reference Group on Estimations, Projections and Modelling
• Based on the spectrum model involving the estimation and projection package using• Epidemiological data from 12 rounds of HIV sentinel surveillance (ANC, FSW,
MSM and IDU) for 1998-2011, IBBA and size estimations
Technical report India HIV estimates 2012
0.00%
0.10%
0.20%
0.30%
0.40%
0.50%
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
• Curves produced for 34 states for each risk group
• Suggests HIV prevalence in India has decreased by a third from 0.40% to 0.27% during 2000 to 2011
ANC HIV prevalence trends at state levelDecreasing in high prevalence states
Stable or increasing in moderate HIV prevalence states
Increasing in some low prevalence states
HIV sentinel surveillance 2010-11: a technical brief
ANC 2006 ANC 2011
Where has HIV decreased or increased
FSW HIV prevalence trends at state levelDecreasing or stable in high prevalence states
Increasing or stable in low HIV prevalence states
HIV sentinel surveillance 2010-11: a technical brief
• Similar trends found for MSM
• For IDU trends less clear with trends generally being stable or increasing