Slides - Gender, Sexuality, and Health

Sexual networks and HIV infection
Partnership concurrency and (some of) its
possible implications
Stéphane Helleringer
PopFam
HIV Prevention in eastern/southern Africa
• Generalized HIV epidemics: >1% of adults
– Prevalence of HIV over 12% of adults in 9 southern
African countries, “hyperendemic settings”
• Common approaches to HIV prevention in eastern
and southern African countries:
– ABC (Abstain, Be faithful, Use condoms)
• Incidence of HIV remains high in Southern and
Eastern Africa
“New” Ideas in HIV Prevention
• Reducing infectivity of HIV in relationships:
– Vaccines
– Male circumcision
– Microbicides (compounds that can be applied inside the
vagina to protected against HIV)
– Increased access to Anti-retroviral treatment: “Treatment is
prevention”
• Reducing exposure to HIV in populations: Behavior
change interventions to address key features of sexual
networks
– Multiple and Concurrent Partnerships (“MCP”)
MCP and HIV epidemics
• Concurrent partnerships:
Two or more sexual
partnerships that overlap in
t
time
A
B
C
Individual 1 (Concurrent)
• Opposed to “serial
monogamy”: partnerships
that follow each other
sequentially
A
B
C
t
Individual 2 (Serial)
MCP and HIV epidemics
• Does not present specific risk
for indiv. 1 vs 2.
A
B
C
• Higher risk of HIV infection for
partners of an index case.
Individual 1 (Concurrent)
• Possible interaction with
trends in viral loads:
– Acute infection may amplify
effects of partnership concurrency
A
B
C
Individual 2 (Serial)
• Morris and
Kretzschmar
(1997)
• The more
concurrency in a
population:
– The faster the
spread of HIV
– The larger the
final epidemic
size
MCP and HIV spread in Africa
• Heated debate (recently in Lancet): are concurrent
partnerships the “key driver” of generalized HIV epidemics?
– Does concurrency explain uneven spread of HIV between
countries/regions?
• Yes, because:
• Models predict it;
• More concurrency in
Uganda than in
Thailand and the US
• No, because:
• More concurrency in
west Africa than in
east Africa, but less
HIV (WHO 4 cities)
MCP and HIV spread in Africa
• Debate largely focused on population-level effects
of concurrency:
– Polygyny and differences in HIV prevalence across the
continent
• Difficult empirical question: concurrency is one of
many properties of sexual networks
• Separate question: is concurrency an independent
individual-level risk factor for HIV in sub-Saharan
populations?
Measures of partnership concurrency
•
Different ways to measure concurrency:
–
•
Direct vs. indirect.
Direct method: ask respondent about X more
recent partners; then ask whether had other
partners had the time.
–
“At any time during your relationship with _____, did
you have sex with someone else?”
–
Differentiates well between serial and concurrent multiple
partnerships
Measures of partnership concurrency
•
Indirect method: ask respondent about X more
recent partners;
–
Date of first sex with _____?
–
–
Date of last sex with _______?
Expect to have sex again with _______?
•
Concurrency reconstructed from overlap between
dates of consecutive partnerships
•
Different measures available in LNS data
Data on MCP
• Ego-centric network data
(e.g., DHS): only index
case interviewed and
tested
• Network data including
partner tracing: partners
elicited and contacted
A
B
C
A
B
C
MCP and HIV prevention in Africa
• Interventions addressing concurrency currently
rolled-out in several countries (e.g., OneLove
campaign in SA)
• Focus of the debate among prevention community
has been on behavioral change exclusively: are BCC
campaigns warranted given the evidence?
• Other implications?
– Concurrency and biomedical prevention?
– Concurrency and HIV treatment?
Objectives
1. Describe a network study conducted on Likoma
Island, Malawi
2. Assess association of partnership concurrency
and HIV infection at the individual-level on
Likoma
3. Draw implications of the impact of sexual
networks on HIV spread for HIV testing services
CONTEXT
Likoma Network Study
•Longitudinal study
investigating populationlevel sexual networks and
their impact on HIV
epidemics
•First data collection in
2005/06
•Follow-up in 2007/08
Likoma
Likoma Island
From
Mozambique
From Malawi
Likoma Island
• 7,000 inhabitants over 11
sq. miles
• Fishing community
• Importance of remittances
from mainland migrants
• Limited market activity
• Isolated but frequent
movement in/out of the
island
Likoma Island: HIV Services
• Testing and counseling
services available.
• ARV treatment available
on the island since
December 2005.
• Very limited prevention
activities
– Occasional ABY programs
in schools
DATA
Likoma Network Study Process
(1)
Enroll all 18-49 years old
ACASI
SEXUAL
NETWORK
SURVEY
HOUSEHOLD
CENSUS
(Detailed info on
Names, nicknames,
Age, socioeconomic
Characteristics)
LISTS OF
5 MOST RECENT
SEXUAL
PARTNERS
(Names, nicknames
Age, occupation
Residence)
Name generator
(3)
(Retrospective info
On partnerships)
(2)
LNS: HIV testing
• Home-based HIV testing with all eligible individuals
• Team of 20 health workers, resident of the
mainland
• Pre- and post test counseling
• Two parallel rapid test kits used for HIV diagnosis:
– Concordant results: communicated to the respondent
– Discordant results: referred to hospital
RESULTS
LNS: Participation
• 11 villages of the island included in the study
population:
– 2,433 eligible respondents aged 18-49,
• 2,176 conducted sexual network survey
– 95 made no reports of relationships
• 1,684 tested for HIV infection:
– 199 with known HIV infection (detected either in
2005/06 or in 2007/08)
Partnership concurrency among women
100
90
80
70
60
50
40
30
20
10
0
Monogamous
relationship
Serial Multiple
Relationships
Concurrent
Multiple
Relationships
<30 years old 30-39 years 40+ years old
old
Partnership concurrency among men
100
90
80
70
60
50
40
30
20
10
0
Monogamous
relationship
Serial Multiple
Relationships
Concurrent
Multiple
Relationships
<30 years old 30-39 years 40+ years old
old
Reliability of concurrency data
• 90 respondents (43 men, 37 women) reinterviewed on average 10 days after initial
interview
• 81% of men made similar/consistent reports of
concurrency during a given relationship
• 73% of women made similar /consistent reports
of concurrency during a given relationship
Concurrency and HIV risk
N(%)
aOR (95% ci)
Monogamous
relationship
776 (11.3%)
1
Serial multiple
partnerships
339 (15.3%)
2.01 (1.35, 2.99)
Concurrent
multiple
partnerships
342 (12.3%)
1.75 (1.14, 2.69)
Index case engaged
in…
Casual Partners
Tested and counselled
Absent at time of visit(s)
Partner not identified
Marital/regular partners
Refused HTC
Partner died
Partner not initiated
Concurrency and HIV risk among partners
N(%)
aOR (95% ci)
Monogamous
relationship
37 (51.3%)
1
Serial multiple
partnerships
53 (58.5%)
1.33 (0.50, 3.54)
Index case engaged
in…
Concurrent
multiple
partnerships
34 (76.5%)
3.76 (1.12, 12.6)
Limitations
• Small sample sizes
– Does association vary with relation type, index gender…
• Retrospective data: assessing association between
concurrency and HIV among partners of HIV cases.
– Not causation, need data on incidence among partners
• Sample selection bias due to:
– Selective partner tracing / Selective participation in HIV testing
• Does not allow calculating duration of overlap
between concurrent partnerships
Summary
• Concurrency widespread among both men and
women on Likoma Island
• HIV transmission occurring in sexual networks:
– Concurrent relationships may have an additional effect
on transmission
Sexual networks and HIV testing
• Key problem for scaling-up of HIV services:
– Limited case finding because low uptake of HIV testing
and counseling
• Patients present late for treatment (compromises
treatment outcomes)
• Limited population-level effect of secondary
prevention (“prevention for positives”)
Sexual networks and HIV testing
• Large costs associated with increasing case finding:
– Screening mechanisms
• Opt-out testing and counseling in hospitals and
other clinical settings (e.g., Antenatal care)
• Door-to-door VCT:
– Yearly visit to all households
– Identifies co-residing partners of HIV index cases
Sexual networks: contact tracing
• Limited partner notification:
– responsibility of the patient
• Contact tracing: elicit partners from HIV index case
and make plans to inform partners of possible
exposure:
– Client-initiated/Provider-initiated
• Effective if:
– High prevalence of HIV among partners (high yield)
– Low costs of tracing partners
100
60
20
40
• No significant
differences in
HIV prevalence
between
spouses and
other (possibly)
concurrent
partners
0
Prevalence of HIV infection
80
• High
prevalence of
HIV among
partners
Among marital partners
Among non-marital partners
Non-marital relations
No other partner nominated
Two or more other partners nominated
Marital relations
One other partner nominated
Contact tracing: possible benefits
• High prevalence of HIV among partners of HIV index
cases:
– Potential for detection of infection clusters; particularly
through tracing of non-marital partners
– Potential to prevent further transmission in sexual
networks
• Cost-effective if:
Pr( Partners) / Pr( Pop)  C (CT ) / C ( Screening )
Contact tracing: possible benefits
• Save on case finding  increase (or maintain)
spending on treatment
• Earlier initiation of treatment:
– Improved clinical outcomes
– Reduced transmission in sexual networks
Contact tracing: possible benefits
• Possible externalities on the strengthening of health
systems
• Tracing capacity useful for:
– Tracing of patients lost to follow-up in ARV treatment
programs and other chronic care programs
– Tracing of outbreaks of other infectious
diseases/surveillance
• Development of health information systems:
– Data confidentiality
Conclusion
• Concurrency may play a key role in HIV transmission
on Likoma
• More studies based on longitudinal designs are
needed
• Structures of Sexual networks (serial or concurrent)
are a key input in planning delivery of health
services