MIGRATION, MINES AND MORES: THE HIV EPIDEMIC IN SOUTHERN AFRICA

Transcription

1 MIGRATION, MINES AND MORES: THE HIV EPIDEMIC IN SOUTHERN AFRICA John Hargrove Introduction to the Applications of Mathematics in Biology and Medicine AIMS, Muizenberg 2 July 2012

2 The colonial map of Africa is broadly dominated by French colonies in west and central Africa and by the British in southern and east Africa

3 In the postcolonial map of Africa, country names have changed but the colour distribution is broadly similar

4 But this is not a political map. This map shows the peak HIV prevalence by country. However, there are striking similarities with the political map Data from UNAIDS: Epidemiology Fact sheets for urban sites only

5

6 Finer gradations of the prevalence figures highlight southern Africa as the epicentre of the pandemic. With East Africa generally higher than west and central regions.

7 HIV prevalences have been compared between countries at a single time point. This can be misleading. Uganda Swaziland % 5% % 37% Should compare peak values [cf Drain et al 2004, 2006]

8 In fact adjusting for former colonial power only accounts for 17% of the variance; the more important factor seems to be geographical location which accounts for 66% of the variance. The peak HIV prevalence 18-26% percentage points higher in southern region. Why?

9 Male circumcision (MC) has been identified as a key factor in accounting for variance in prevalence. Overall, fraction of men circumcised accounts for ca 35% of variance in HIV prevalence. Effect less clear within regions

10 In every region there is a significant effect on peak HIV prevalence of the proportion of Muslims in a country. But P < 0.05 only in east Africa for the effect of male circumcision. Need to separate the regional effects from those of being circumcised, and of being Muslim

11 Multivariate analysis shows all effects are significant. Strong protective effect of male circumcision Added protection associated with being Muslim Large residual regional effect. r 2 = 0.80

12 Even after adjusting for the proportion of males circumcised, there is an additional protective effect of being Muslim. And there is a large additional risk attached to living in a southern African country. WHY? Is there some common factor tending to reduce the risk of HIV infection in Muslim communities and, at the same time, increase the risk among people in southern and, to as lesser extent, eastern Africa?

13 In univariate analyses of this type Drain et al (2004, 2006) found a large number of factors that significantly affected HIV prevalence. But in multivariate analyses they only saw the effects of male circumcision, region, fraction < 25 years old, female literacy, immunisation and number of doctors per 100k of population. Various shortcomings in analysis: 1. Treated Africa as a single unit. 2. Compared prevalence at one point in time. 3. Did not consider importance of family. 4. Did not consider migration.

14 Early analyses of HIV in Africa concentrated overwhelmingly on sexual behaviour patterns in Africa and assumptions about these practices. Sanders and Sambo (1991) single out the destruction of the family unit through the process of migratory labour patterns as an important social and economic factors underlying the spread of HIV infection in Africa

15 Family coherence and migration patterns perhaps not included in modelling because difficult to measure/little data. But HIV prevalence and incidence in ZVITAMBO study in Harare, Zimbabwe associated with: financial insecurity and weak psychosocial support

16 Spatial epidemiology of HIV in a small world Simple model to illustrate spatial effects Doubling time = 1 yr; Life expectancy = 10 yrs Number partners = 4; Size of population = 16,000 Next slide shows two populations (left and right) Left Sex with neighbours only Right 90% of sex with neighbours 10% with people at random

17

18 Past tendency to identify truckers for (spatial) spread of HIV; parallels were drawn between good road systems and the rapid spread of the virus, via commercial sex workers at truck stops.

19 The problem is not so much roads, as Rhodes. [Cecil John and company] Who developed mines [and much else] in southern Africa. The gold was plentiful but deep; cheap (migrant) labour was essential if the mines were to be profitable.

20 Sanders and Sambo (1991) pointed out that all over southern Africa, particularly, authorities had introduced poll, hut and cattle taxes which forced the population to seek employment on farms, in cities on the mines. But for the mines only men were recruited

21 Whereas men were recruited to work in these situations, their wives and families were not allowed to accompany them. Hundreds of thousands men were [and still are] accommodated in single sex hostels. Labour for the south African mines was/is recruited from all over the southern African region. And similar situation existed anyway in other mining areas, and in cities in the region. [In Harare in the 1960s illegal to have an African man s wife living with him on a suburban property].

22 More lately there has also been mass migration to cities in southern and eastern Africa. Point is that it is oscillating migration. Women [generally] maintain(ed) a family base in a homeland, native reserve, Bantustan, communal area

23 We see the scale of the problem in SA. Have suggested the ultimate roots - maybe right or wrong but question is what can we do? Long term solution, if analysis is correct, lies with an entire change of ethos - changes in law regarding housing and employment, rural development of all aspects including general health care and education. Is there any indication that there will be improvement in the HIV situation in the absence of such changes?

24 There seem to be some things going on in Zimbabwe what can we learn from them?? First law of statistics? Look at your data. Second law of statistics? Play with your data.

25 . In a Trial in Harare, between Oct 1997 and Jan 2000, 14,000 women were recruited over a 27-mo period and tested for HIV. HIV prevalence initially increases with age peaking at a horrendous level of 50% for women aged about 30. Then declines sharply.. Why the increase with age? Why the decline? Prevalence (percent) Age

26 Prevalence at recruitment M e a n p r e v a le n c e vs m o n t h o f r e c r u it m e n t M o n t h o f re c r u it m e n t Now pool on age and see whether there is any relationship between HIV prevalence and time. Is there any trend in the prevalence with date of recruitment??

27 40 Prevalence at recruitment Mean prevalence vs month of recruitment N J M M J S N J M M J S N J Month and year of recruitment. For the ZVITAMBO Trial, HIV prevalence increased significantly during 1998, thereafter it declined significantly. Prevalence Nov %. Dec %. Jan %.

28 0.40 HIV prevalence Year When the ZVITAMBO data are amalgamated with other data from Harare ANC sites, prevalence appears to have peaked at the end of 1998 and seems to have been declining ever since. Can we model these changes?

29 Mortality per Mortality in Harare Zimbabwe Mortality in Harare. With the end of the war in Zimbabwe in 1980 there was a large influx of foreign aid, jobs were created, and health and education services were improved Year Mortality in Harare declined until the effects of the HIV- AIDS epidemic made themselves felt.

30 β N λsi/n δi S I µs I Normal (Weibull 2) β N λ = birth rate = S + I = infection rate P(surviving) Exponential (Weibull 1) Time (years) δi = Weibull mortality

31 Prevalence Incidence/mortality Year 0.00

32 No matter how we change the parameter values we cannot fit the data. This raises an important problem in mathematical epidemiology. It is not sufficient to come up with a clever mathematical model: it must also make sense biologically. We need to understand the biology. And when we have a model that cannot fit the observed data we need to ask Why? Are the data the problem? Or is it our model?

33 We have a model where HIV prevalence, incidence and mortality all rise to unrealistically high levels. What does our model currently say about the human element, and about variability of behaviour? Is that reasonable? How could we change that in the model?

34 β N λsi/n δi S I µs I Normal (Weibull 2) β N λ = birth rate = S + I = infection rate P(surviving) Exponential (Weibull 1) Time (years) δi = Weibull mortality

35 We change so that the rate at which people become infected declines as the prevalence increases? Is that reasonable? Perhaps reasonable if, as prevalence increases, the remaining susceptibles are uninfected specifically because they exhibit lower average risk??

36 ~ βn λsi/n δi S I µ S µ I 1.0 β ~ = birth rate N = population λ = λe αp δi = Weibull mort. Relative transmission e αp Prevalence (%) Heterogeneity in sexual behaviour

37 Year 0.00 Prevalence Incidence/mortality

38 Now we have a model which at least peaks at appropriate levels of HIV prevalence, incidence and mortality. But the prediction is that all of these stay very high. Why does the model predict no decline? What does our model assume about changes in human behaviour with time? How could we change that in the model?

39 ~ ~ βn λsi/n δi S I µs I β ~ N ~ λ = birth rate = population = λc(t) δi = mortality Relative transmission C(t) Year Including control

40 Year 0.00 Prevalence Incidence/mortality

41 Now we have a model which fits all of the HIV prevalence, incidence and mortality quite nicely. But notice that we have simply assumed that the transmission rate declines with time? What makes it do that? Consider a model where we assume human behaviour is affected by the number of deaths observed.

42 ~ βn λsi/n * δi S I µs I β ~ N λ = birth rate = population * αm = λe δi = mortality Relative transmission e αm Annual mortality (%) Mortality leads to behaviour change

43

44

45 Year 0.00 Prevalence Incidence/mortality

46 The new model fits the increases in HIV prevalence, incidence and mortality quite nicely. But then the functions start to oscillate? Why do they do that? Think about what the model says about how human behaviour is affected by observed deaths. What does this say about memory in the model?

47 In South Africa, while HIV incidence appears to be declining, little evidence so far of any major, decline in HIV prevalence. Contrast this with the Zimbabwe situation.

48 Year 0.00 Prevalence Incidence/mortality

49 What can be done about the situation in South Africa? Consider more proactive medical interventions? May well be years away from delivering even a partially effective vaccine or microbicide Randomised control trials show male circumcision provides protection to men against HIV infection. Now being offered free to all males in Orange Farm.

50 The Potential Impact of Male Circumcision on HIV in sub-saharan Africa Modelling study suggests that: 1. Over the next ten years in sub-saharan Africa, male circumcision could avert 2.0 ( ) million new HIV infections and 0.3 ( ) million deaths 2. In the ten years thereafter, a further 3.7 ( ) million new HIV infections and 2.7 ( ) million deaths..

51 SACEMA has taken a lead in pushing an initiative using medical intervention with ARTs as an offensive rather than defensive weapon Based on observations that a policy of initiating ART at CD4 count <200 has serious shortcomings: 1. Up to 66% will already have already have had TB by this stage. 2. CD4 count does not fully recover in some patients. 3. Even after 3 years on ART the probability of infection with TB is 8-10 times higher than in HIV negative people. 4. Suggestions that prolonged infection with HIV leads to accelerated senescence of immune system. 5. Data from Cape study shows that young (<30) HIV positive patients not enrolled into ART therapy. [Fails to break transmission]

52 Conclusions The roots of the catastrophic HIV epidemic in southern Africa are deeply imbedded in the colonial past. A more aggressive, imaginative medical approach to the HIV epidemic in South Africa is urgently indicated. However, unless the inherited structural deficiencies are addressed they will continue to provide the fuel not only for the HIV epidemic but for a host of other social ills. We do not under-estimate sociological, political, financial difficulties of addressing the social or medical issues. Examples of Zimbabwe and Uganda suggest that the power of the people should not be under-estimated in effecting behavioural change leading to reductions in HIV incidence.