Thursday, February 17, 2011

Disputing external validity of African HIV Studies

This is a copy of a Journal article:

African Journal of AIDS Research 2008, 7(1): 1–8
Long-term population effect of male circumcision in generalised HIV
epidemics in sub-Saharan Africa
Michel Garenne

This paper examines the complex relationship between male circumcision and HIV prevalence and incidence in
sub-Saharan African countries that have generalised epidemics. In South Africa, the mean yearly HIV incidence and
an estimate of the net reproduction rate of the epidemic (R0) (in this case, the ratio of the number of HIV-infected
persons between 1994 and 2004 to the number of persons infected in 1994 from which they were presumed to
have become infected) were computed from antenatal clinic data for the period 1994–2004, and then compared,
by province, to prevailing levels of male circumcision (high, medium and low). In South Africa, mean yearly HIV
incidence and net reproduction rate of the epidemic were not lower in provinces with higher levels of male circumcision.
For thirteen other countries where Demographic and Health Survey data were available, male HIV prevalence
in circumcised and non-circumcised groups was compared. A meta-analysis of that data, contrasting male HIV
seroprevalence according to circumcision status, showed no difference between the two groups (combined risk
ratio [RR] = 0.99, 95% CI = 0.94–1.05). Individual case study analysis of eight of those countries showed no significant
difference in seroprevalence in circumcised and uncircumcised groups, while two countries (Kenya and
Uganda) showed lower HIV prevalence among circumcised groups, and three countries (Cameroon, Lesotho and
Malawi) showed higher HIV prevalence among circumcised groups. In most countries with a complex ethnic fabric,
the relationship between men’s circumcision status and HIV seroprevalence was not straightforward, with the
exception of the Luo in Kenya and a few groups in Uganda. These observations put into question the potential
long-term effect of voluntary circumcision programmes in countries with generalised HIV epidemics.
Keywords: demography, epidemiology, HIV infection, household surveys, meta-analysis, prevention and control, quantitative
data, statistical analysis
Numerous studies have shown a relationship between
male circumcision and HIV infection. Fink (1986) referred
to frequent observations that herpes and syphilis were
more common among uncircumcised men, arguing that
this could be true as well for cases of HIV infection, in
particular because the inner foreskin hosts a large number
of Langerhans cells which are prime targets for HIV. Fink
(1987) noted that HIV first spread in areas of Africa where
most men were not circumcised. Bongaarts, Reining, Way
& Conant (1989) systematically reviewed African ethnic
groups practicing circumcision and found a geographical
correspondence with areas of high HIV seroprevalence, and
a statistical relationship between the proportion of circumcised
males and male HIV prevalence in capital cities of 28
countries. Since then, similar observations at the individual
level were made in case-control studies as well as in cohort
studies. Moses, Bradley, Nagelkerke, Ronald, Ndinya-
Achola & Plummer (1990) and Moses, Plummer, Bradley,
Ndinya-Achola, Nagelkerke & Ronald (1994) reviewed
30 epidemiological studies and concluded that a majority
showed a significant association between male circumcision
and HIV infection. Weiss, Quigley & Hayes (2000)
reviewed 27 studies and, after a meta-analysis, concluded
that male circumcision reduces the rate of female-to-male
HIV transmission by about half, on average.
These ecological and observational studies were recently
confirmed by three randomised controlled trials, conducted
in South Africa, Kenya and Uganda, which found that
circumcised men had a 51–60% lower risk of HIV infection
over a 12- to 24-month period than did uncircumcised
men (Auvert, Taljaard, Lagarde, Sobngwi-Tambekou, Sitta
& Puren, 2005; Bailey, Moses, Parker, Agot, MacLean,
Krieger et al., 2007; Gray, Kigozi, Serwadda, Makumbi,
Watya, Nalugoda et al., 2007).
There seems little doubt that male circumcision reduces
the individual risk of female-to-male transmission of HIV,
other factors controlled for. However, how this observation
may translate into the long-term population dynamics of HIV
epidemics remains an open question. In particular, does a
50% reduction in annual risk at the individual level translate
into a strong reduction of HIV incidence and prevalence at
the population level in a generalised HIV epidemic?
Several mathematical models have attempted to predict
the possible effect of male circumcision (e.g. see Gray,
Li, Wawer, Gange, Serwadda, Sewankambo et al., 2003;
Williams, Lloyd-Smith, Gouws, Hankins, Getz, Hargrove
Downloaded By: [Australian National University Library] At: 23:00 29 January 2011
2 Garenne
et al., 2006; Nagelkerke, Moses, De Vlas & Bailey, 2007;
Orroth, Freeman, Bakker, Buvé, Glynn, Boily et al., 2007).
However, these exercises do not guarantee prediction of
the final impact in populations, especially in situations where
the current dynamics of an HIV epidemic remain poorly
understood. Indeed, no mathematical model can accurately
reproduce observed rates of HIV seroprevalence with realistic
parameters in the case of heterosexual transmission.
Several African countries with large groups of circumcised
and uncircumcised men allow one to study the differential
dynamics of HIV epidemics according to males’
circumcision status. Furthermore, the case of countries
with minority populations, either circumcised or not, also
sheds some light on these complex relationships. These
contrasting situations in generalised epidemics are unique,
as they provide information of what could be the potential
impact of voluntary circumcision programmes. This study
uses two types of data for a quantitative investigation:
data from antenatal clinics and data from population-based
surveys. Since the two methodologies are different, they are
presented sequentially.
Data and methods
The first type of data, used only for South Africa, was the
routine surveillance of HIV prevalence of pregnant women
at antenatal clinics (sentinel sites). Routine surveillance in
South Africa started in 1990 and has been maintained ever
since. These data allow one to study the dynamics of the
HIV epidemic over time, focusing on women of reproductive
age. South Africa is affected by high levels of HIV seroprevalence.
For example, HIV seroprevalence among attendants
at antenatal clinics in South Africa reached 29.5% in 2004
(South African Department of Health, HIV and AIDS statistics,
available online:
Two population-based studies, published by the Human
Sciences Research Council (HSRC) (2002 and 2005), also
found high HIV seroprevalence: estimates for 2002 and
2005, respectively, were 12.8% and 13.3% prevalence
among females of all ages; 9.5% and 9.3% prevalence
among men of all ages; and peaks at 32% and 33%
prevalence for women aged 25–29. These figures were
consistent with the antenatal clinic data.
South Africa is divided into nine provinces, and has nine
large ethnic groups and four racial groups. With respect
to the black African population, some of the provinces are
largely ethnic-specific, others (in particular industrialised and
urbanised areas) are a mixture of several ethnic groups.
Among the major African ethnic groups, three are known to
widely practice male circumcision: the Xhosa (who comprise
83% of the population of the Eastern Cape Province), the Pedi
(52% of the population of Limpopo Province), and the Venda
(16% of Limpopo). Three ethnic groups practice circumcision
moderately: Sotho, Ndebele and Shangaan; and three
groups practice circumcision rarely: Zulu (comprising 81% of
KwaZulu-Natal Province, and 26% of Mpumalanga Province),
Tswana (65% of the North West Province) and Swazi (31%
of Mpumalanga). Although never universal or forbidden in
any ethnic group, circumcision levels are considered to be
about 75% in the first group (Xhosa, Pedi, Venda), around
45% in the second group (Sotho, Ndebele, Shangaan), and
around 15% in the third group (Zulu, Tswana, Swazi) (HSRC,
2002). The 2002 HSRC study found that 35% of men overall
in South Africa were circumcised, which is consistent with the
proportion expected if projecting from the ethnic composition.
A study (RADAR, 2002) conducted among the Pedi in
southern Limpopo Province found that 89% of males aged
14–35 were circumcised. In that study, however, a greater
proportion of circumcised men were HIV infected (7.5%
seroprevalence) than were uncircumcised men (5% seroprevalence)
(pers. comm., Dr Paul Pronik, director of the RADAR
project, November, 2006).
Since 1994, HIV seroprevalence data for attendants
at antenatal clinics in South Africa have been tabulated
and published according to province, by the South African
Ministry of Health, the sole consistent series of such
data available in Africa. Data for HIV seroprevalence in
1994 (P1994) and in 2004 (P2004) allow one to reconstruct
the dynamics of the epidemics in the provinces over
the ten-year-period. Two indices were calculated, these
were: the mean yearly HIV incidence, calculated as
(P2004 – P1994) / 10 ; and an estimate of the net reproduction
rate of the epidemic, calculated as R0 = (P2004 • K – P1994) / P1994 ,
where K is the coefficient of population growth between
1994 and 2004. The K coefficient was calculated for the
female adult population aged 15–49 from data provided
by Statistics South Africa (available online: www.statssa. R0 gives the ratio of the number of HIV-infected
persons between 1994 and 2004 to the number of persons
infected in 1994 from which they were presumed to have
become infected. This assumes a 10-year survival rate
for HIV-infected persons, and ignores those who became
infected and died during the interval, and thus it is obviously
an underestimation of the true value. However, the idea is
not to provide the best estimate of R0, but simply to give a
simple measure of the reproductive capacity of the epidemic
that can be compared among groups.
Demographic and Health Surveys (DHS) are largescale
standardised surveys conducted in many countries
on representative samples of the population. Data from
13 surveys conducted in sub-Saharan Africa provided
information on both male circumcision status and on HIV
seroprevalence in the adult male population (aged 15–54
years). These surveys allow one to compute the ratio of HIV
prevalence after many years in the epidemic according to
male’s circumcision status. The surveys are of high quality
and are free from selection biases otherwise built into the
antenatal clinic data. They are probably the most reliable
source of information currently available on the relationship
between HIV seroprevalence and level of male circumcision
in national populations. The risk ratio (RR) was calculated
as the ratio of HIV seroprevalence among circumcised
males to that of uncircumcised males. Statistical testing of
the differences was done using standard tests for risk ratios.
A combined RR was calculated by meta-analysis, with
weights for each survey being proportional to the inverse of
the standard error.
South Africa
In South Africa, the dynamics of the HIV epidemics in the
Downloaded By: [Australian National University Library] At: 23:00 29 January 2011
African Journal of AIDS Research 2008, 7(1): 1–8 3
nine provinces differed from what was expected based
on the proportions of circumcised men in each province
(Table 1). For the country as a whole, mean yearly HIV
incidence was 2.2% per year, and the estimate of the net
reproduction rate of the epidemic (R0) reached 3.6 between
1994 and 2004 (i.e., 3.6 newly infected persons between
1994 and 2004 per person infected in 1994). Mean HIV
incidence was almost equal for the three groups considered
(i.e., high, medium and low levels of circumcision): 2%
mean incidence per year in provinces with a high level of
circumcision (specifically, 2.3% in the Eastern Cape; 1.6% in
Limpopo); 2.5% mean incidence per year in provinces with a
medium level of circumcision (2% in the Free State; 2.7% in
Gauteng), and 2.1% mean incidence per year in provinces
with a low level of circumcision (1.4% in the Western Cape;
1.6% in the Northern Cape; 2.6% in KwaZulu-Natal, 1.9% in
Mpumalanga, and 2% in the North West).
With respect to R0, the pattern was reversed: R0 was
higher (6.4) in provinces with a high level of circumcision
(6.1 and 6.8 in the Eastern Cape and Limpopo, respectively),
medium (R0 = 4.5) in provinces with a medium
level of circumcision (5.3 and 2.6 in Gauteng and the Free
State, respectively), and lower (R0 = 2.7) in provinces with
a low level of circumcision (2.2 in KwaZulu-Natal; 2.0 in
Mpumalanga; 3.4 in the North West). Moreover, it should be
noted that the estimate of the net reproduction rate of the
epidemic was very high for the Western Cape (R0 = 14.5)
and the Northern Cape (R0 = 9.1); these two provinces have
a somewhat different social fabric and large portion of the
population who belong to the ‘coloured’ racial group. More
importantly, South African provinces with higher values
of R0 were primarily areas with relatively low levels of HIV
seroprevalence in 1994, as may have been anticipated.
(Indeed, the computation of R0 assumed that all cases of
HIV transmission occurred within the given province, an
obvious oversimplification in a country where internal
migration is frequent and accounts for a large share of the
spread of HIV.) Note that different provincial levels of HIV
prevalence in 1994 may be explained by a variety of socioeconomic
factors. In particular, in 1994, areas with the most
cases of HIV infection were the largest industrial centres
and the areas hosting the two largest harbours (Durban and
Richard’s Bay).
In summary, there was no evidence that HIV transmission
over the 1994–2004 period was slower in those South
African provinces with higher levels of male circumcision.
Thirteen countries with Demographic and Health Surveys
Among the thirteen Demographic and Health Surveys
(DHS) investigated, those of eight countries (Burkina
Faso, Cote d’Ivoire, Ethiopia, Ghana, Niger, Rwanda,
Tanzania and Zimbabwe) showed no significant difference
in male HIV seroprevalence in circumcised and uncircumcised
groups; data from two countries showed lower HIV
prevalence among circumcised groups, as expected (Kenya
and Uganda), and three countries (Cameroon, Lesotho and
Malawi) showed higher prevalence among circumcised
groups, contrary to expectations (Table 2).
The combined risk ratio (RR) calculated by metaanalysis
was 0.99, not significantly different from 1 (95%
CI = 0.94–1.05). Therefore, for the continent as a whole, the
analysis found no statistical evidence to support the belief
that male circumcision reduces the level of HIV seroprevalence
in a national population. Of course, there is heterogeneity
among the countries investigated; below I present
further case study analyses. Furthermore, in countries
where a large majority of the population is either circumcised
or not, minorities who have the opposite practice
might also be quite different for a variety of socio-economic
reasons. Two countries, Tanzania and Lesotho, deserve
special mention because they represent a more balanced
situation of the prevalence of circumcision (that is, large
groups are either circumcised or not).
Tanzania has a large number of ethnic groups, with some
110 groups recorded in the Ethnologue database (see www. In this country, some groups practice
male circumcision, others do not, so that about 70% of the
total male population is circumcised, a much higher proportion
than in South Africa. Roughly speaking, groups living
in western regions of Tanzania and closer to Lake Victoria
do not practice male circumcision, whereas groups living in
eastern regions of the country often practice circumcision.
The pattern of relationship between male circumcision
and HIV infection in Tanzania differed from what was
expected. The 2003 DHS for Tanzania showed higher HIV
seroprevalence among circumcised men (6.5%) than among
Table 1: Parameters of the dynamics of HIV infection in provinces of South Africa, by circumcision status, 1994–2004 (R0 = net reproduction
rate of the epidemic)
Level of circumcision1
High Medium Low Total
Number of provinces 2 2 5 9
Population (2001), millions 11.7 11.5 21.6 44.8
Population increase 1994–20042 1.18 1.20 1.14 1.17
HIV seroprevalence in 1994 (%)3 3.9 7.1 9.5 7.5
HIV seroprevalence in 2004 (%) 24.1 32.3 30.7 29.5
Mean yearly HIV incidence (%) 2.0 2.5 2.1 2.2
Estimate of R0 6.4 4.5 2.7 3.6
1 Levels of circumcision — High level: Eastern Cape, Limpopo; Medium level: Gauteng, Free State; Low level: KwaZulu-Natal, Mpumalanga,
North West, Northern Cape, Western Cape
2 Population increase refers to the female population aged 15–49
3 HIV seroprevalence among pregnant women who attended antenatal clinics
Downloaded By: [Australian National University Library] At: 23:00 29 January 2011
4 Garenne
uncircumcised men (5.6%). This was likely partly due to a
correlation with urbanisation, a major factor in determining
HIV prevalence. However, even after controlling for urbanisation,
no correlation between male circumcision and HIV
seroprevalence was found: 9.5% versus 9.7% prevalence
for circumcised versus uncircumcised groups, respectively,
in urban areas, and 4.6% versus 5.2% for those groups
in rural areas, with neither of the differences significant
(p = 0.938 and p = 0.425, respectively).
Further investigation using multivariate methods confirmed
the individual risk of contracting HIV, according to circumcision
status. Using the data from the same DHS survey, a
series of linear logistic regression analyses were performed,
using age, age-square, urbanisation, wealth, and circumcision
status as predictors of HIV seroprevalence for males.
Results showed an odds ratio of 0.66 associated with male
circumcision (95% CI = 0.49–0.88), again consistent with
other findings. This, however, does not change the key
finding — that HIV seroprevalence estimates after 25 years
of transmission opportunity were similar for circumcised and
uncircumcised groups of males, even after controlling for
Lesotho is a much more homogenous country, with a
single ethnic group, low level of urbanisation (about 28%)
and relatively high level of education compared to other
African countries. HIV seroprevalence is also fairly homogenous
throughout the country, with only minor differences
revealed in relation to urbanisation, ecological zone, district,
level of education and wealth. The country’s population is
essentially Christian, with various churches operating. Here
again, contrary to expectations, HIV seroprevalence was
significantly higher among circumcised men than among
uncircumcised men, aged 15–59 (22.8% versus 15.2%
prevalence, respectively), and this was true in relation to
any socio-economic characteristics considered in the DHS
final report. Therefore, the difference in prevalence between
the two groups could not be explained by any single
socio-economic confounder, with the exception of urbanisation
(urban areas had higher HIV prevalence and lower
proportions of circumcised men). In a multivariate analysis
of the risk factors of HIV seroprevalence, male circumcision
was no longer found to be a risk factor (OR = 1.00,
95% CI = 0.79–1.27), after controlling for age, age-square,
place of residence, ecological zone, level of education,
wealth and religion.
Other countries
Other countries with small proportions of uncircumcised
men provide additional evidence, although they are more
difficult to interpret because of the high selectivity of groups
with different circumcision practices.
In Burkina Faso (2003 DHS), where 90% of men are
circumcised, the difference of HIV seroprevalence between
circumcised and uncircumcised groups was not significant
(p = 0.142). Furthermore, the capital city Ouagadougou had
among regions in that country the highest level of circumcision
(97%), but also the highest male HIV seroprevalence
(3.9%). In the eastern province of Burkina Faso,
where almost every male is circumcised (99.1%), male
HIV seroprevalence was also above the national average
(3.1%). Two ethnic groups that live in the southern part of
the country had a low level of circumcision, the Gurunsi
and the Lobi: they had higher male HIV seroprevalence
than average (4.2%), but not significantly different from that
of two other neighbouring ethnic groups also living in the
south, the Dioula and the Bissa (3.8% prevalence), who are
almost all circumcised (97% circumcised).
In Cameroon (2004 DHS), where 92% of men are circumcised,
the tiny groups of uncircumcised men had lower HIV
seroprevalence than others (1.1% versus 4.1% prevalence,
respectively), the difference being significant (p = 0.010).
In Ethiopia (2005 DHS), the prevalence of HIV among
men was low (1%), and most men were circumcised (92%),
so that any differential analysis was virtually impossible.
Table 2: HIV prevalence in the male adult population according to circumcision status; data are from demographic and health surveys (DHS)
conducted in 2003–2006 (see separate references listing); RR = risk ratio; 95% CI = 95% confidence interval; NS = no significant difference
in prevalence between groups (RR <> 1, p > 0.05); + denotes higher prevalence for circumcised groups (RR > 1, p < 0.05); – denotes lower
prevalence for circumcised groups (RR < 1, p < 0.05)
% HIV-positive Comparison of the two groups
Percent Circumcised Not 95% CI
circumcised circumcised RR Min. Max. Signif.
Burkina Faso 89.7 1.8 2.9 0.62 0.328 1.174 NS
Cameroon 91.8 4.1 1.1 3.73 1.369 10.150 +
Cote d’Ivoire 96 2.8 3.8 0.74 0.32 1.71 NS
Ethiopia 92.3 0.9 1.1 0.82 0.309 2.165 NS
Ghana 95.3 1.6 1.4 1.14 0.356 3.673 NS
Kenya 83.4 3 12.6 0.24 0.171 0.332 –
Lesotho 48.6 22.8 15.2 1.50 1.236 1.820 +
Malawi 20.7 13.2 9.5 1.39 1.050 1.839 +
Niger 99.5 1 0 NS
Rwanda 11.1 3.5 2.1 1.67 0.975 2.848 NS
Tanzania 69.7 6.5 5.6 1.16 0.905 1.489 NS
Uganda 24.9 3.8 5.6 0.68 0.531 0.868 –
Zimbabwe 10.5 16.6 14.2 1.17 0.95 1.44 NS
Total 0.99 0.94 1.05 NS
Downloaded By: [Australian National University Library] At: 23:00 29 January 2011
African Journal of AIDS Research 2008, 7(1): 1–8 5
In Ghana (2003 DHS), where 95% of men are circumcised,
there was no difference in male HIV seroprevalence
between the two groups (1.6% versus 1.4% prevalence
for circumcised and uncircumcised groups, respectively;
p = 0.823), and the numbers were too small for pursuing the
analysis by ethnicity.
In Malawi (2004 DHS), a minority of men are circumcised
(21%). Plotted according to ethnicity, the correlation
between circumcision and male HIV seroprevalence
was positive (r = 0.404), meaning that circumcised groups
tended to be more HIV-infected than the uncircumcised
groups. The group practicing almost universal circumcision
(the Yao) was more infected than the national average
(12.4% prevalence), as was the case for the Lomwe (30%
circumcised; 18% HIV-positive), whereas groups hardly
practicing circumcision (the Chewa, Tumbuka and Nkonde)
had lower values of male HIV prevalence (5.7%, 5.1% and
4.9%, respectively). The Ngoni were a counter-example,
practicing a low level of circumcision (4.2%) and having
higher male HIV seroprevalence (14.5%). These correlations
could not be explained by interactions with any known
socio-economic factor, although, notably, higher education
was negatively related with circumcision and positively
related with HIV prevalence, probably for independent
reasons. In a multivariate analysis of the risk factors of
HIV seroprevalence, male circumcision was no longer
found to be a risk factor (OR = 1.38, 95% CI = 0.97–1.96),
after controlling for age, age-square, place of residence,
level of education, wealth and religion. The ethnic correlations
appeared primarily the product of a regional pattern:
males in southern Malawi were more often circumcised
(33.1%) and more often HIV-infected (15.1% prevalence);
in northern Malawi, males were less often circumcised
(5%) and less often HIV-infected (5.4% prevalence). These
findings show the complexity of the relationship between
circumcision and HIV prevalence. Numerous reasons
may be evoked for understanding the Malawi pattern: the
south is more urbanised, poorer, and tends to send more
migrant workers abroad. A recent study (based on small
numbers and not employing proper sampling) indicates that
southern populations in Malawi are more prone to partnermobility,
to late marriage and to divorce than are northern
populations (Obare, 2007). In any case, and whatever the
ultimate determinants in Malawi, male circumcision did not
protect the southern populations from acquiring HIV in high
In Niger (2006 DHS), where virtually all men are circumcised,
there was no case of HIV infection reported among
the tiny minority of males who were not circumcised.
In Rwanda (2005 DHS), the prevalence of circumcision
and that of HIV were both too low to permit a differential
analysis. Note, however, that Rwanda has kept low levels
of HIV prevalence despite low levels of circumcision (11%),
and despite the fact that the country is located near the
epicentre of the HIV epidemic.
In Kenya (2003 DHS), data on individual risk and population
effects were consistent with expectations. The only
sizeable minority not practicing routine circumcision, the
Luo (about 12% of the population) who live near Lake
Victoria, had higher male HIV seroprevalence than other
groups (17.5% versus 2.8% prevalence; p < 0.001), and
even among the Luo the few circumcised men were less
often HIV-infected than were uncircumcised men, although
the difference was not significant (about 10% versus 20%
HIV prevalence among circumcised versus uncircumcised
Luo men, respectively; p = 0.091).
Similarly, in Uganda (2005 AIS), where only a minority
of men are circumcised, the correlation among the ethnic
groups between prevalence of circumcision and HIV was
negative, but remained small (r = –0.203). The pattern
according to ethnicity was complex. The only group widely
practicing circumcision (80% of Bagisu/Sabiny males
circumcised) had lower than average HIV seroprevalence
(3.5%). Two other groups also practicing circumcision,
although at a much lower level (the Baganda and Basoga,
31% and 35% of males circumcised, respectively), had
on the contrary higher than average HIV seroprevalence
(5.8% and 5.6%, respectively). The highest seroprevalence
(12.8%) was found in a group practicing circumcision to a
certain extent (22% of Batoro males circumcised), whereas
the lowest seroprevalence (1.1%) was found in a group (the
Karimojong) who was not practicing circumcision. Furthermore,
circumcision was more prevalent in urban areas
than in rural areas of Uganda (36% urban versus 23%
rural males circumcised); HIV infection was more prevalent
in urban areas as well (6.7% urban versus 4.7% rural HIV
prevalence), a pattern opposite from that expected. So,
even if the correlation among ethnic groups was somewhat
negative in Uganda, the pattern was by no means straightforward.
The demographic evidence indicates that the relationship
between male circumcision and HIV seroprevalence is
complex, and that both positive and negative relations can
be found for a variety of reasons. No evidence of an overall
protective effect of male circumcision was found for the
countries studied, and if there is an effect at the individual
level it is buried in a mix of many other intervening factors.
One might argue that the ‘natural experiment’ in South
Africa is not identical to systematic voluntary circumcision.
However, if circumcision were to be recommended, it is
unlikely to become universal, so that the natural experiment
is probably close to a real-world situation. Surveys about
males’ intention to be circumcised have shown that about
two-thirds of men expressed an interest in the practice
(Westercamp & Bailey, 2007), so that the natural situation
of groups practicing circumcision is probably close to what
could happen if programmes of voluntary circumcision were
successful in the country.
The situation in Tanzania also seems close to an experimental
situation. One could argue that the two groups —
circumcised and uncircumcised males — are not exposed to
the virus the same way. However, since circumcised groups
typically live farther away from Lake Victoria, they may be
less exposed or later exposed to HIV (which reinforces
the main argument). One exception so far, in Tanzania,
is the Zanzibar Islands, which have low HIV prevalence.
The populations on these islands are highly conservative
Muslims, and they are fairly isolated from the continent’s
mainland, so their exposure to HIV is likely different from
Downloaded By: [Australian National University Library] At: 23:00 29 January 2011
6 Garenne
populations in the rest of the country.
The most puzzling finding from the study concerns
Lesotho, where HIV seroprevalence was found consistently
higher among circumcised men, despite otherwise high
homogeneity of the country’s population. The case of this
country, perhaps typical of southern African countries with
high levels of HIV seroprevalence, clearly shows that little
may be expected from voluntary circumcision programmes
in advanced generalised epidemics.
Population-effects over long periods of time are not
necessarily identical with measures of individual risk for
shorter periods. Indeed, under intense and repeated
exposure, a 50% lower annual risk does not translate into
a large reduction at the population level. A good example
is provided by contraception. A 50% efficacious contraceptive
method, such as periodic abstinence (the rhythm
method), does not protect against pregnancy in the long
run: certainly, it has an effect on the length of birth intervals
and mean-age at birth stratified by birth order, but not on a
woman’s lifetime probability of getting pregnant.
Similarly, comparisons with vaccines are worth mentioning
and many contrasting situations have been documented
in this field. An example of equivalence between individualeffects
and population-effects is provided by the measles
vaccine: its efficacy (about 96% in clinical trials) translates
readily into population-effect. Mass vaccination stops
measles epidemics for several years, and has a strong
impact on incidence of the disease. An example of low
population-effect, with a 50% efficacy in clinical trials, is
provided by the cholera vaccine. This vaccine protects only
for a short period of time in about half the cases, does not
have much effect on cholera epidemics, and is currently not
even recommended for mass usage, even though its effect
on herd immunity clearly appears when proper multivariate
analysis is conducted (as was often the case for male
circumcision in this study) (Ali, Emch, Von Seidlein, Yunus,
Scoak, Rao et al., 2005). A third example of populationeffect
that is higher than that expected from clinical trials is
given by the vaccine against Bordetella pertussis (whooping
cough). In clinical trials this vaccine has shown low efficacy
to protect against infection, though it has a much higher
efficacy for preventing severe disease and death. Despite
its low efficacy in clinical trials, the whooping cough vaccine
has a major population-effect, and can stop an epidemic in
a way similar to the measles vaccine. The British health
authorities who stopped vaccinations against whooping
cough in the 1970s quickly learned that this led to a return
of large epidemics, so that they had to reinstall the vaccination
policy (Pollard, 1980). These three examples show
that there is not always a one-to-one relationship between
findings from clinical trials and the dynamics of communicable
diseases in large populations. The case of HIV is further
complicated by sexual transmission modes, and hence the
possibility of different dynamics among men and women.
To return to the original argument proposed by Fink (1987),
epidemics of syphilis were at least as severe in the United
States, where most men were circumcised, as in Western
Europe, where more often they were not circumcised.
Likewise, heterosexual transmission of HIV never became
widespread in Europe, as it is the case in the United States.
In his comprehensive review on the topic, Van Howe
(1999, p. 61) concluded that “the medical literature does
not support the theory that circumcision prevents STDs.”
Similarly, the Cochrane Library review on the effect of male
circumcision on HIV prevalence found “insufficient evidence
to support an effect of male circumcision on HIV acquisition
in heterosexual men” (Siegfried, Muller, Volmink, Deeks,
Egger, Low et al., 2003).
Some authors have questioned the quality of data on
circumcision reported in Demographic and Health Surveys
and the quality of the circumcision itself when done by
traditional practitioners. However, the reported effect of
male circumcision in clinical trials is basically the same as in
observational studies, and also in general populations after
controlling for confounding factors, which reinforces the
value of the demographic information.
Decisions of whether or not to circumcise for nontherapeutic
reasons remains controversial in situations of
high HIV incidence. On one hand, any intervention that can
reduce HIV-infection risk seems desirable; on the other
hand, it does not seem likely that circumcising a large
proportion of the male population will have a large populationeffect.
Furthermore, universal circumcision would require
massive involvement of health personnel, already in short
supply in many African countries. Even in South Africa, a
fairly developed country, physicians and nurses are scarce
and their numbers are not large enough to manage antiretroviral
therapy for all who need it. The 2006 World Health
Report (WHO, 2006) puts strong emphasis on health
personnel, their importance for developing countries, and
their shortage throughout sub-Saharan Africa. In contrast,
alternative strategies do exist for controlling the spread of
HIV, such as the well-known A-B-C strategy (abstain, be
faithful, use condoms); these may be less costly and are
known to be efficient at the population level. For instance,
the dynamics of the HIV epidemic in Thailand were changed
with massive use of condoms; in Uganda a major change
was obtained primarily by changing sexual behaviour;
even in South Africa, the incidence of the disease, given
the size of the HIV-infected population, has decreased
since 2001, most likely as the result of increased condom
use (documented in HSRC surveys) and some changes in
sexual behaviour.
Given what is known at the individual level, one would
have expected HIV incidence or prevalence in circumcised
groups of men to be consistently about 20% lower than in
uncircumcised groups. This was not the case according
to the results of this study. Until this discrepancy between
demographic evidence and expectations from epidemiological
evidence is resolved, is it wise to recommend mass
circumcision? Also, the implementation of mass circumcision
raises serious ethical concerns, well documented
elsewhere (Clark, 2006).
Once more, the dynamics of generalised HIV epidemics
in Africa appear more complex than originally thought.
Male circumcision appears only as a minor factor amid
many others contributing to the spread of HIV, such as the
complex web of social, cultural and economic interactions
surrounding sexual behaviour, especially among adolescents
and young adults. These complex relationships merit
further field studies — as well as investigations through
mathematical modelling.
Downloaded By: [Australian National University Library] At: 23:00 29 January 2011
African Journal of AIDS Research 2008, 7(1): 1–8 7
Acknowledgements — The author warmly thanks all those who
shared ideas, comments and data for this study, in particular Paul
Pronyk, Slim Abdool Karim, Rob Dorrington, Pauline Leclerc, Muriel
Vray and Arnaud Fontanet.
The author — Michel Garenne (PhD, demography) is Director of
Research at the French Institute for Research and Development
(IRD) and is currently working at the Pasteur Institute, Emerging
Diseases Unit, in Paris. He is also honorary Associate Professor at
the University of the Witwatersrand, Johannesburg. He directed the
Niakhar Demographic Surveillance System in Senegal in the 1980s
and has collaborated with the Agincourt Health and Demographic
Surveillance System in South Africa since 1992. He is author of
numerous publications on population and health issues in Africa,
and has taught demography at several universities in Europe
(Paris, Clermont-Ferrand, Heidelberg, Antwerp) and in the United
States (Harvard).
Demographic and Health Surveys (DHS)
Burkina Faso — Institut National de la Statistique et de
la Démographie (INSD) & ORC Macro (2004) Enquête
Démographique et de Santé du Burkina Faso 2003. Calverton,
Maryland, INSD and ORC Macro.
Cameroon — Institut National de la Statistique (INS) & ORC Macro
(2004) Enquête Démographique et de Santé du Cameroun 2004.
Calverton, Maryland, INS and ORC Macro.
Cote d’Ivoire — Institut National de la Statistique (INS), Ministère
de la Lutte contre le Sida [Côte d’Ivoire] & ORC Macro (2006)
Enquête sur les Indicateurs du Sida, Côte d’Ivoire 2005. Calverton,
Maryland, INS and ORC Macro.
Ethiopia — Central Statistical Agency (CSA) [Ethiopia] & ORC Macro
(2006) Ethiopia Demographic and Health Survey 2005. Addis
Ababa, Ethiopia and Calverton, Maryland, CSA and ORC Macro.
Ghana — Ghana Statistical Service (GSS), Noguchi Memorial
Institute for Medical Research (NMIMR) & ORC Macro (2004)
Ghana Demographic and Health Survey 2003. Calverton,
Maryland, GSS, NMIMR and ORC Macro.
Kenya — Central Bureau of Statistics (CBS) [Kenya], Ministry of
Health (MoH) [Kenya] & ORC Macro (2004) Kenya Demographic
and Health Survey 2003. Calverton, Maryland, CBS, MoH and
ORC Macro.
Lesotho — Ministry of Health and Social Welfare (MoHSW)
[Lesotho], Bureau of Statistics (BOS) [Lesotho] & ORC Macro
(2005) Lesotho Demographic and Health Survey 2004. Calverton,
Maryland, MoH, BOS and ORC Macro.
Malawi — National Statistical Office (NSO) [Malawi] & ORC Macro
(2005) Malawi Demographic and Health Survey 2004. Calverton,
Maryland, NSO and ORC Macro.
Niger — Institut National de la Statistique (INS) & Macro International
Inc. (2007) Enquête Démographique et de Santé et à
Indicateurs Multiples du Niger 2006. Calverton, Maryland, INS
and Macro International Inc.
Rwanda — Institut National de la Statistique du Rwanda (INSR)
& ORC Macro (2006) Rwanda Demographic and Health Survey
2005. Calverton, Maryland, INSR and ORC Macro.
Tanzania — Tanzania Commission for AIDS (TACAIDS), National
Bureau of Statistics (NBS) & ORC Macro (2005) Tanzania HIV/
AIDS Indicator Survey 2003–04. Calverton, Maryland, TACAIDS,
NBS and ORC Macro.
Uganda — Ministry of Health (MoH) [Uganda] & ORC Macro (2006)
Uganda HIV/AIDS Sero-Behavioural Survey 2004–2005. Calverton,
Maryland, MoH and ORC Macro.
Zimbabwe — Central Statistical Office (CSO) [Zimbabwe] & Macro
International Inc. (2007) Zimbabwe Demographic and Health
Survey 2005–06. Calverton, Maryland, CSO and Macro International
Ali, M., Emch, M., Von Seidlein, L., Yunus, M., Scoak, D.A., Rao,
M., Hilmgren, J. & Clemens, J.D. (2005) Herd immunity conferred
by killed oral cholera vaccines in Bangladesh: a re-analysis. The
Lancet 366, pp. 44–49.
Auvert, B., Taljaard, D., Lagarde, E., Sobngwi-Tambekou, J., Sitta,
R. & Puren, A. (2005) Randomized, controlled intervention trial of
male circumcision for reduction of HIV-infection risk: the ANRS
1265 trial. PLoS Medicine 2(11), pp. 1–111.
Bailey, R.C., Moses, S., Parker, C.B., Agot, K., MacLean, I.,
Krieger, J.N., Willams, C.F.M., Campbell, R.T. & Nidnya-Achola,
J.O. (2007) Male circumcision for HIV prevention in young men
in Kisumu, Kenya: a randomised controlled trial. The Lancet 369,
pp. 643–656.
Bongaarts, J., Reining, P., Way, P. & Conant, F. (1989) The
relationship between male circumcision and HIV infection in
African populations. AIDS 3(6), pp. 373–377.
Clark, P.A. (2006) To circumcise or not to circumcise? Health
Progress 87(5), pp. 1–9.
Fink, A. (1986) A possible explanation for heterosexual male
infection with AIDS. New England Journal of Medicine 315(8), p.
Fink, A. (1987) Circumcision and heterosexual transmission of
HIV infection to men. New England Journal of Medicine 316, pp.
Gray, R.H., Kigozi, G., Serwadda, D., Makumbi, F., Watya, S.,
Nalugoda, F., Kiwanuka, N., Moulton, L.H., Chaudhary, M.A.,
Chen, M.Z., Sewankambo, N.K., Wabwire-Mangen, F., Bacon,
M.C., Williams, C.F.M., Opendi, P., Reynolds, S.J., Laeyendecker,
O., Quinn, T.C. & Waver, M.J. (2007) Male circumcision
for HIV prevention in men in Rakai, Uganda: a randomised
controlled trial. The Lancet 369, pp. 657–666.
Gray, R.H., Li, X., Wawer, M.J., Gange, S.J., Serwadda, D.,
Sewankambo, N.K., Moore, R., Wabwire-Mangen, F., Lutalo, T.,
Quinn, T.C. & the Rakai Project Group (2003) Stochastic simulation
of the impact of antiretroviral therapy and HIV vaccines on
HIV transmission; Rakai, Uganda. AIDS 17(13), pp. 1941–1951.
Human Sciences Research Council (HSRC) (2002) Nelson
Mandela/HSRC Study of HIV/AIDS. South African National HIV
Prevalence, Behavioural Risks and Mass Media, Household
Survey 2002. Cape Town, South Africa, HSRC Publishers.
Human Sciences Research Council (HSRC) (2005) South African
National HIV Prevalence, HIV Incidence, Behaviour and
Communication Survey, 2005. Cape Town, South Africa, HSRC
Moses, S., Bradley, J.E., Nagelkerke, N.J., Ronald, A.R., Ndinya-
Achola, J.O. & Plummer, F.A. (1990) Geographical patterns of
male circumcision practices in Africa. International Journal of
Epidemiology 19(3), pp. 693–697.
Moses, S., Plummer, F.A., Bradley, J.E., Ndinya-Achola, J.O.,
Nagelkerke, N. & Ronald, A.R. (1994) The association between
lack of male circumcision and risk for HIV infection. Sexually
Transmitted Diseases 21(4), pp. 201–210.
Nagelkerke, N.J., Moses, S., De Vlas, S.J. & Bailey, R.C. (2007)
Modelling the public health impact of male circumcision for HIV
prevention in high prevalence areas in Africa. BMC Infectious
Diseases 7, p. 16. Available online: http://www.pubmedcentral.
Obare, F. (2007) An assessment of the HIV risk factors and
variations in HIV prevalence in rural Malawi. Paper presented
at the Fifth African Population Conference, Arusha, Tanzania,
10–14 December 2007.
Orroth, K.K., Freeman, E.E., Bakker, R., Buvé, A., Glynn, J.R.,
Boily, M.C., White, R.G., Habbema, J.D.F. & Hayes, R.J. (2007)
Understanding the differences between contrasting HIV epidemics
in east and west Africa: results from a simulation model of the
Downloaded By: [Australian National University Library] At: 23:00 29 January 2011
8 Garenne
Four Cities Study. Sexually Transmitted Infections 83 (supplement
1), pp. i5–i16.
Pollard, R. (1980) Relation between vaccination and notification
rates for whooping cough in England and Wales. The Lancet
1(8179), pp. 1180–1182.
RADAR (Rural AIDS and Development Action Research
Programme) (2002) Social interventions for HIV/AIDS: intervention
with micro-finance for AIDS and gender equity. IMAGE
Study, Evaluation Monograph No 1.
Siegfried, N., Muller, M., Volmink, J., Deeks, J., Egger, M., Low,
N., Weiss, H., Walker, S. & Williamson, P. (2003) Male circumcision
for prevention of heterosexual acquisition of HIV in men
(Cochrane Review). In: The Cochrane Library, Issue 3, 2003.
Oxford, Update Software.
Van Howe, R.S. (1999) Does circumcision influence sexually
transmitted diseases? A literature review. BJU International
83(supplement 1), pp. 52–62.
Weiss, H.A., Quigley, M.A. & Hayes, R.K. (2000) Male circumcision
and risk of HIV infection in sub-Saharan Africa: a systematic
review and meta-analysis. AIDS 174, pp. 2361–2370.
Westercamp, N. & Bailey, R.C. (2007) Acceptability of male circumcision
for prevention of HIV/AIDS in sub-Saharan Africa: a
review. AIDS and Behavior 11(3), pp. 341–355.
Williams, B.G., Lloyd-Smith, J.O., Gouws, E., Hankins, C., Getz,
W.M., Hargrove, J., De Zoysa, I., Dye, C. & Auvert, B. (2006)
The potential impact of male circumcision on HIV in sub-Saharan
Africa. PLoS Medicine 3(7), p. e262.
World Health Organization (WHO) (2006) The World Health Report
2006 — Working Together for Health. Geneva, WHO.

No comments:

Post a Comment