DST/NRF Centre of Excellence in
Epidemiological Modelling and
Analysis
(SACEMA)
What can SACEMA do for Africa?
John Hargrove
The South African Centre for
Epidemiological Modelling and Analysis
• SACEMA is a new Centre of Excellence, with
core funding from the South African
Government. Housed at the University of
Stellenbosch in the Western Cape, it is a
national centre – with a mandate to promote
epidemiological modelling and analysis in
South Africa.
Philosophy (I)
• “Understanding many economic, health and
environmental processes requires the development
of dynamical mathematical models and thus a
reasonably high level of mathematical expertise.
• South Africa, like many other developing countries,
is relatively weak in this area.
• Intention is to bring together people interested in
addressing such problems while providing training
opportunities for young mathematicians who are
interested in applying their skills to address critical
problems in South Africa”.
Philosophy (II)
• “Although the Centre operates on a worldwide basis and
contributes to the general advancement of epidemiology, it
must also be considered in the context of South African
epidemiology. Each programme will benefit the South
African epidemiology community ..”
• “If South Africa is strong in the field, South African
scientists will play a major part in the programme; if
relatively weak, the programme should help to raise South
African standards. Instructional courses, aimed at younger
researchers and research students, will play a vital role.”
[Adapted from the website of the Isaac Newton Institute]
Five-year goals
1. Complete work in the field of mathematical
epidemiology that contributes substantially to
the alleviation of the effects of major diseases,
and other problems, currently affecting people
in South Africa, and in Africa as a whole.
2. Strengthen sustainable capacity in South
Africa, the Region and Africa generally, to
continue, and build on, this work.
Strategy (I)
• Use international and local scientists, including graduate
students, to achieve the above goals.
• Ensure flow of high-quality workers through SACEMA;
develop strong international links – Berkeley, Paris, Ghent,
WHO, CDC, Stats Canada, Johns Hopkins, NIH, Rutgers....
• Fund short/medium term fellowships for scientists to work
at SACEMA; and buy-out of time for scientists working
elsewhere.
• Use summer schools/meetings to collect small numbers of
interested, talented individuals to focus intensively on a
given modelling problem.
• Visiting fellows to contribute to capacity building as well as
research wherever appropriate.
Taking AIMS
• SACEMA will encourage suitable students from
AIMS, and from South African universities and
institutes, to take on projects in epidemiology.
• This can then be used as a selection process that can
identify the most promising students for potential
recruitment as junior fellows at SACEMA.
• Through the SACEMA/StIAS/AIMS network, link
good students with other institutions and people with
good projects.
Strategy (II)
• Focus initially on HIV-AIDS, TB and associated
diseases.
• Seek major international funding to further support
these efforts – NIH, PEPFAR, Global Fund.
• Use this funding to facilitate expansion into other
areas –malaria/trypanosomiasis, bovine TB, avian
flu, impact of climate changes, optimal fish
harvesting, …
Strategy (III)
Crucial to the achievement of Goal I are:
1.
2.
3.
4.
5.
Sound understanding of biology/epidemiology.
Innovative mathematical modelling.
Access to the best possible data.
Interdisciplinary collaboration.
Good communications with policy makers.
Develop TB/HIV database
Seebregts
HIV/TB model
Williams/Corbett/Lauer
HIV male circ. model
Auvert/Williams/L-Smith
Host-viral dynamics project
Witten
Vaccine modelling
Welte
Malaria/tryps
Torr/Hargrove/Vale
in Botswana [Lungu]
Analysis of health etc data.
Malaria/warming/GIS
Freeman/Marijani
Stanford-SA bio-informatics
Seioghe
Modelling HIV/TB
HIV prevalence/incidence
Marinda/Hargrove
Superspreading
Lloyd-Smith
HIV population models
Matthews
HIV/TB in W. Cape
Wood/Lawn
Microsimulation modelling
Pretorius/Welte
Bovine TB/Kruger
Getz/Geoghan
DE modelling
CD4/mortality [Ouifki]
Male circumcision
[MC]
•
•
Auvert’s MC trial at Orange Farm [Gauteng]
indicated that MC reduces sexual transmission
of HIV from female to male by 60%.
SACEMA associates Williams, Lloyd-Smith
and others modelled this situation and
estimated the effects [on HIV infections,
prevalence and deaths] of promoting MC as a
public health policy.
 im
sm
sf
 if s m
 im s f
im
if
 if
The simplest model for men and women
No MC    2
100% MC    5
If MC reduces transmission in one
direction by a factor of  this is
equivalent to a reduction in both
directions by a factor of 1-√(1- )
MC is equivalent to a vaccine which
reduces transmission by 37%
MC
c
I(%)
(1-c)P/10
 = 0.6
HIV
P
I(k)
I(%)N
Elimination of HIV?
• In South Africa R0 ≈ 5.
• Introducing MC could reduce R0 to 2.
• A further reduction of 2 would then be
sufficient to eliminate HIV as a public
health problem
Superspreading and the effect of individual
variation on disease emergence
•
•
Quantitative study of epidemic dynamics centres on
the basic reproductive number, R0
Yet real epidemics (e.g. SARS in 2003) feature
“superspreading” individuals who infect far more
people than the average case.
How to incorporate superspreading in outbreak models?
How prevalent is superspreading for different diseases?
How does individual variation affect outbreak dynamics?
Superspreading and the effect of individual variation on
disease emergence
• “Normal” SARS cases infected
0 to 3 others, but
“superspreaders” infected 10,
20 or more.
• Is superspreading an
exceptional property of SARS,
or common to all infectious
diseases?
• How can this individual-level
variation be modelled, and
how does it affect outbreak
dynamics?
Lloyd-Smith, Schreiber, Kopp & Getz Nature 438: 355-359.
Quantifying individual variation in transmission
1. Collect detailed transmission data for many diseases.
2. Apply maximum-likelihood estimation and Akaike’s
Information Criterion to select best statistical model for
transmission data.
3. Compare diseases using model estimates.
HIV infection modelling
– within host
• Host viral dynamics. Intracellular delays,
drug treatment and immune response.
[Witten/Ouifki]
• HIV Strain Dynamics
[Welte/Pretorius/Mwanga]
New campaigns (I)
Modelling changes in HIV prevalence and
incidence.
• There are fundamental problems in the modelling
of HIV epidemics that on occasion lead to major
agencies, responsible for advising African
governments, arriving at inappropriate
conclusions about trends in the epidemic.
• Well illustrated by the situation in Zimbabwe.
36
32
ANC prevalence
Greater Harare
In 2004 MoH working
with UNAIDS, WHO and
CDC suggested no change
in HIV prevalence after
about 1994/1996.
Prevalence (%)
28
24
20
16
Now clear prevalence has
been declining since at
least 1998 – and incidence
perhaps as early as 1994.
12
Double logistic fit
ZVITAMBO estimates
MoH estimates 2000-2004
Mahomed et al. (1991)
Mbizvo et al. (1996)
MoH projection (2004)
8
4
1984
1988
1992
1996
Year
2000
2004
2008
Why were recent estimates so incorrect?
•
Inadequate data.
•
Poor/inappropriate modelling.
•
Imperfect understanding of the
biological and mathematical problems.
•
Measures of incidence very rare.
Modelling the HIV epidemic: tactics
•
•
•
•
•
Data available on the behaviours of individuals –
and on the [consequent] changes in HIV prevalence
and incidence at the population level.
Mathematical challenge is to produce models that
marry the two.
At the individual level use branching processes.
At population level use compartmental dynamical
simulations as currently for male circumcision.
But how to combine the two? New mathematical
and modelling approaches are needed. Network
theory? Micro simulation?
HIV incidence via cross-sectional surveys
• The rate of acquisition of new infections
defines the development of the HIV epidemic.
• Previously measured via longitudinal studies.
• CDC have used ZVITAMBO samples to
validate a “detuned” ELISA technique that
allows estimation of HIV incidence via crosssectional surveys.
1.6
Theoretical graph of sqrt(OD-n) vs ln(ti, j)
1.2
Selected OD cut-off (B)
Square root of OD-n
0.8
0.4
Negative baseline (A)
.
.
. .
.
Window (Wi )
0.0
-0.4
Slope = b1,i
-0.8
-1.2
Intercept = b0,i
-1.6
0
1
2
3
4
5
Log time (ti, j days) since last negative
6
7
Optical density vs time since last HIV negative test
HIV incidence via cross-sectional surveys
• Unfortunately the method over-estimates
incidence by a factor of 2-3 or even more.
• Work at SACEMA has suggested ways of
adjusting the estimates such that the BED
could be used to estimate HIV incidence
from cross-sectional survey data.
Integrating incidence measures into ANC
sentinel-site monitoring
•
Once technique is perfected, prevalence and incidence
can be estimated from the same cross-sectional survey.
•
Greatly enhances ability to model the epidemic.
•
Presents opportunities to identify problem situations and
estimate the early effect of the roll-out of ARV therapy.
•
Indications, from ZVITAMBO and elsewhere, that the
prospect of treatment has greatly increased levels of
VCT and reduced stigma.
•
Will it also lead to a decrease in incidence?
New campaigns (II)
Modelling the HIV-TB epidemics.
• Pretorius/Ouifki working with Wood team, modelling
HIV/TB situation at Masiphumelele
• May well be necessary to involve micro-simulation.
• And the problem may be so large that may also be
necessary to use distributed computing techniques.
The importance of good data
Modelling the HIV-TB epidemics.
• Ensuring access to the best TB and HIV data available in South
Africa will be crucial to this project.
• The sine qua non of success of all SACEMA projects will be
the requirement of access to good data.
• Reference has been made above to three projects that have
utilised excellent data from the ZVITAMBO project.
• There is much more available ….
In ZVITAMBO 14,110 post partum women and their babies
followed for two years, three monthly intervals – providing
blood and breast milk samples at each follow-up visit.
Analyses of the data already produced important insights into
HIV – but still 600,000 samples not yet analysed.
What can SACEMA do for Africa?
• Attract a strong team of mathematicians and modellers
to improve understanding of diseases affecting Africa.
• Work with AIMS, and other institutes, to encourage
young African mathematicians in the pursuit of careers
in mathematical epidemiology.
• Work with African scientists in accessing, and making
available for analysis, the best data available - in South
Africa and elsewhere in Africa.
• Funding promising African students to attend
appropriate workshops.
• Source donor funds to facilitate the above.