International
Journal of STD & AIDS
Royal Society of Medicine, October 2002
EDITORIAL
REVIEW
HIV
infections in sub-Sahara Africa not explained by sexual
or vertical transmission
by
David
Gisselquist, PhD, independent consultant; Richard
Rothenberg, MD, MPH, Department of Family and Preventive
Medicine, Emory University School of Medicine, Atlanta,
Georgia, USA; John Potterat, BA, independent consultant;
Ernest Drucker, PhD, Dept of Epidemiology
and Social Medicine, Montefiore Medical Center/Albert
Einstein College of Medicine, NYC, USA
Summary
An
expanding body of evidence challenges the conventional hypothesis
that sexual transmission is responsible for more than 90%
of adult HIV infections in Africa. Differences in epidemic
trajectories across Africa do not correspond to differences
in sexual behavior. Studies among African couples find low
rates of heterosexual transmission, as in developed countries.
Many studies report HIV infections in African adults
with no sexual exposure to HIV and in children with HIV-negative
mothers. Unexplained high rates of HIV incidence have been
observed in African women during antenatal and postpartum
periods. Many studies show 20%-40% of HIV infections
in African adults associated with injections (though direction
of causation is unknown). These and other findings that
challenge the conventional hypothesis point to the possibility
that HIV transmission through unsafe medical care may be
an important factor in Africa's HIV epidemic. More research
is warranted to clarify risks for HIV transmission through
health care.
Introduction
Within
two years after the first AIDS cases were described in homosexual
men in Los Angeles in 1981, AIDS was diagnosed in Haitians(1)
and among Africans in Europe,(2) Zaire(3) (now Democratic
Republic of Congo [DRC]), Rwanda,(4) and Zambia(5). Unlike
AIDS in the US and Europe, which seemed concentrated among
injection drug users (IDUs), men-who-have-sex-with-men (MSM),
and hemophiliacs, AIDS in Haitians and Africans occurred
about equally in women and men, and was found among the
well-to-do, including those who could afford to go to Europe
for medical care.
Experts
at a World Health Organization (WHO) meeting on AIDS in
November 1983 puzzled over possible channels for HIV transmission
among Africans and Haitians.(6) While noting that spouses
of AIDS patients were at risk, experts were undecided about
heterosexual promiscuity, concluding that "whether
persons with multiple heterosexual sex partners are at greater
risk of acquiring AIDS is unknown " Meeting participants
considered that "injections with unsterile needles
and syringes may play a role " WHO's 1983 recommendations
focused on sterilization of medical equipment, blood safety,
and MSMs.
During
1983-88, researchers in Africa found high rates of HIV prevalence
among female commercial sex workers (CSWs) and patients
at sexually transmitted disease (STD) clinics.(7-9) By the
end of the 1980s, a consensus emerged among AIDS experts
dealing with Africa that over 90% of adult HIV infections
in sub- Sahara Africa were acquired through heterosexual
contact and less than 2% through unsafe injections.(10-13)
Unfortunately, this consensus was achieved without research
to address confound between sexual and medical exposures.
As Packard, Epstein, Minkin, and others have noted, CSWs
and STD patients have relatively high levels of medical
exposures that may be channels for transmission of blood
borne pathogens.(14, 15) Further, the consensus ignored
evidence from 1980s research suggesting non-trivial levels
of HIV transmission to African children and adults through
unsafe injections and other medical care.(16-19)
Observations
on heterosexual transmission
During
the past decade, researchers have struggled to fit emerging
facts about Africa's evolving HIV epidemic into the consensus
view that heterosexual transmission accounts for nearly
all adult infections and that iatrogenic transmission is
minimal. Many facts do not fit well.
Divergent
epidemic trajectories.
Differences
in sexual behavior across countries do not explain differences
in epidemic trajectories. In some countries and regions
with high HIV prevalence during the second half of the 1980s,
such as DRC, Uganda, and Kagera in Tanzania, the epidemic
has been stable or declining during the 1990s. In others,
such as South Africa and Botswana, the epidemic reportedly
doubled in less than two years among the low risk population
(viz, antenatal women) during the early 1990s. A series
of sexual behavior surveys in 12 African countries during
1989-93 shows no apparent correlation between the percent
of adults in a country reporting non- regular sexual partners
in the last year and HIV prevalence.(20) A more recent
study of sexual behavior and HIV prevalence in four African
cities reports that partner change, contacts with sex workers,
and concurrent partnerships were no more common in the two
high prevalence cities studied than in the two low prevalence
cities.(21, 22)
Unexplained
high implicit rates of heterosexual transmission in Africa.
The
assumption that historic and continuing high rates of epidemic
increases among African adults are almost exclusively due
to sexual transmission requires much higher rates of heterosexual
transmission in Africa than in the developed world. However,
a recent study of HIV incidence in serodiscordant couples
in Africa (only 1.2% reported consistent condom use) estimated
a rate of transmission per coital act of only 0.0011,(23)
comparable to rates of 0.0003-0.0015 from similar studies
in the US and Europe.(24, 25, 26) .
Epidemiologists
who design computer models to support heterosexual transmission's
role in fueling Africa's HIV epidemic characteristically
choose and/or adjust assumptions about sexual behavior,
rates of heterosexual transmission, and/or other parameters
to allow the model to reproduce observed prevalence.(35-38)
These assumptions are often distant from empiric observations
from African studies. While such models show that
it is possible to imagine patterns of heterosexual transmission
that can "explain" the epidemic, they do not show
that imagined patterns are realistic.
In
one model, for example, Anderson and colleagues assumed
a mean rate of annual partner change of 3.4.(35) In contrast,
surveys in 12 African countries show unweighted averages
of 74% of men and 91% of women aged 15-49 years with no
non-regular sex partners in the past year, and only 3.7%
of men and 0.7% of women with more than four non-regular
partners.(20) At about the
same time, a survey in Denmark found that 19% of adults
aged 18-59 years reported more than one sex partner in the
past year;(39) a survey in France found that 17% of men
and 7.9% of women aged 18-44 years reported more than one
sex partner in the past year;(40) and a survey in the UK
found that 17% of men and 8.4% of women aged 16-44 years
reported more than one sex partner in the past year.(41)
Studies of sexual behavior do not show as much partner change
in Africa as modelers have assumed, nor do they show differences
in heterosexual behavior between Africa and Europe that
could explain major differences in epidemic growth.
Model-builders
often use the transmission co-factor effect imputed to STDs
to generate desired rates of heterosexual propagation. For
example, Korenromp and colleagues(37) assumed that genital
ulcers from syphilis or chancroid in either partner enhance
HIV transmission by a factor of 100, implying an HIV transmission risk from a single coital act
of 30% from men to women and 8% from women to men. These
rates are at odds with most empiric studies which indicate that STDs enhance HIV transmission
2-5 fold. Assumed
high co-factor effects for STDs are based on findings from
two prospective studies in Kenya that enrolled cohorts of
CSWs and male clients visiting STD clinics (42). In such studies, the cohorts are not
representative, and STD treatments - through injections
- may be a source of risk. Although information about
the structure of sexual networks in Africa is sparse, many
studies of incidence show over two-thirds of HIV occurring
in African adults reporting only one sexual partner in the
reporting period27,(43-45). The network geometry of such a pattern would be predominantly
dendritic, with open ends formed by infected spouses with
no outside partners. Rapid STD/HIV transmission has been
associated with cyclic (with closed loops providing network
cohesion) rather than dendritic networks (46,47).
Adult
HIV without sexual exposure to HIV
During
the last 14 years, a number of studies have reported adults
contracting HIV without sexual exposures to HIV. A study
in Zimbabwe in the 1990s found 2.1% HIV prevalence among
933 women with no sexual experience.(48) In a 1988 study
of discordant couples in Rwanda, 15 of 25 HIV-positive women
with HIV- negative partners reported only one lifetime sex
partner.(49) reported only one lifetime sex partner49.
In a 1990 study of teenagers in Uganda, 6.9% of women with
no sex partners in the last five years were HIV positive
compared to 23% for those with one or more partners;
for men, 1% with no partners in the last five years were
HIV-positive compared
2.5% of those reporting partners
(50). Among young adults 15± 24 years old
in Tanzania, a 1995 study found HIV prevalence of 5.6% among
men and 3.6% among women who did not report any lifetime
sexual activity vs 4.8% and 12% for men and women reporting one or more sexual
partners (51). In a 1999 study in South
Africa, 6.8% of women and 1.2% of men 14-24 years old who
reported never having sex were HIV positive; however, a
validation study found some under-reporting of sexual activity.(52).....
When
HIV prevalence or incidence is found in adults and adolescents
with no reported sexual exposures to HIV, it may be assumed
that a share of the HIV in those who are sexually exposed
comes from non-sexual transmission as well.....
whatever exposures are responsible for
non-sexual transmission - injections or vaginal examinations,
for example - may well increase with age and sexual activity.
Observations
suggesting medical transmission
HIV-positive
children with HIV-negative mothers
A
study in Kinshasha in 1985 found 39% (17 of 44) of HIV-positive
inpatient and outpatient children 1-24 months old to have
HIV-negative mothers; only five of 16 had been transfused.(17).
A study
in Rwanda in 1984- 86 found that 20% (15 of 76) of children
1- 48 months old with AIDS or AIDS related complex had HIV-negative
mothers…. In a later report from Rwanda, 7.3% (54 of 704)
of mothers of children with AIDS were HIV-negative; transfusions
were identified as the risk factor for 22 (40%) of the 54
children (54).
Of 26 children less than 15 years old
admitted to the Uganda Cancer Institute with Kaposi’s sarcoma
during 1989± 94 for which the mother was tested for HIV,
19% (five of 26) had HIV-negative mothers55. A study in Burkina Faso in 1989± 90
found 23% (11 of 48) of HIV-positive children to have HIV-negative
mothers; six of 11 had been transfused, and the others reported
multiple injections (56).
In a 1994 report from Cote d’Ivoire, De Cock and colleagues
report that 21% (three of 14) of children with HIV-1 had
mothers without HIV-1, and one of two with HIV-2 had a mother
without HIV-2 (57)….. These and other findings suggest that
a significant proportion of paediatric HIV in Africa - as
much as a fifth (20%) or more in many studies - has been
acquired through health care rather than through vertical
transmission from mothers.
Shortfalls
in accounting for incidence during antenatal and postpartum
periods.
Studies
from seven African countries over the last 15 years show
rates of HIV incidence during antenatal and/ or postpartum
periods exceeding what could be expected solely from sexual
transmission (Table 1).(43, 45, 60-68)…..
Overall,
four studies in Malawi, Zimbabwe, South Africa, and Kenya
show unexplained HIV-incidence ranging from 5-19 per 100
PYs (person-years) during antenatal and postpartum periods
(see Table 1). These rates of unexplained incidence among
African women are comparable to rates of maternal mortality
from puerperal fever of 6% to 16% observed by Semmelweis
during 1841-46 in the First Clinic at the University of
Vienna's obstetric department.(73) ...
Variation
of unexplained incidence from country-to-country and over
time most notably within the Malawi study suggests that
something more than simply heterosexual transmission is
involved. Excess
HIV incidence associated with pregnancy - whatever its cause
- helps to explain the spread of HIV among low-risk populations
in many African countries. In
Malawi, for example, antenatal and postpartum women seroconverted
at the rate of 21.3 and 12.8 per 100 PYs in 1990 and 1991,
so that within one year, prevalence among women who were
HIV-negative at first antenatal visit was well over half
of observed prevalence from sentinel surveys of 22% and
26% in 1990 and 1991.(60) ... In other words, whatever happens
during one or two pregnancies and postpartum periods whether
iatrogenic or sexual or something else may largely account
for observed high levels of HIV among low risk women in
at least some African communities.
|
Table
1. HIV incidence in antenatal and post-partum periods
vs expected
incidence from heterosexual transmission
|
|
Location
and years of study
|
Number
of HIV negative women
|
Incidence
observed during ANC and PP (per
100 PYs)
|
Maximum
expected incidence from sexual
transmission* (per 100 PYs)
|
Incidence
not explained by sexual transmission
(per 100 PY
|
|
Blantyre,
Malawi, 1990± 95
(43, 60)
|
>1,000
|
21 during
ANC and PP in 1990
12 during ANC and PP in 1991
8.0 during ANC 1990± 93
|
2.2-
3.3
|
19 during
ANC and PP in 1990
10 during ANC and PP in 1991
5 during ANC 1990± 93
|
|
Harare,
Zimbabwe, 1990± 94
(61, 62)
|
372
|
17 during
ANC**
>=13 during
0± 6 months PP
|
2.4-3.0
|
14 during
ANC
>=10 during
0± 6 months PP
|
|
Durban,
South Africa, 1993
(63)
|
178
|
9.0 during
ANC**
|
0.7
|
8 during ANC
|
|
Nairobi,
Kenya, 1986± 91
(64, 65)
|
353
|
6.2 during
0± 6 months PP
|
0.27-
1.6
|
5 during
0± 6 months PP
|
|
Kigali,
Rwanda, 1988± 90
(66, 67)
|
216
|
7.2 during
0± 6 months PP
4.2 during
7± 18 months PP
|
3.0-
3.2
|
4 during
0± 6 months PP
1 during 7± 18 months PP
|
|
Rakai,
Uganda, 1994± 96
(45)
|
1,305
|
3.2 during
pregnancy
|
1.4^
|
1.8 during
pregnancy
|
|
Lusaka,
Zambia, 1987± 88
(68)
|
634
|
3.0 during
first year PP
|
1.2
|
1.8 during
first year P
|
|
*See
text
**Estimated
from reported seroconversions from first antenatal
visit to delivery, assuming three months in antenatal
care
^Observed
incidence in women during all other intervals
ANC =
Antenatal care; PP = post-partum
period; PYs = person
years
|
HIV
infections associated with induced abortions and assisted
delivery.
In
addition to these prospective studies of pregnant and postpartum
women, some other studies also suggest that health care
for pregnant women may be a risk factor for HIV. In Congo,
among 1,770 women at an antenatal clinic in 1987-88, 17
of 282 with a history of induced abortions were HIV-positive
vs. 54 of 1,488 without for a crude population attributable
fraction (PAF) of HIV associated with induced abortions
of 10%; complications from abortions were a common cause
of hospitalization, which was also associated with HIV infection.(74)
... Finally, a study of 5690 pregnant women
in Rwanda in 1989 - 91 reports an odds ratio of 2.7 (95%
confidence interval 1.9- 3.6) for last delivery with assistance
by medical personnel vs unassisted
Delivery (76).
From reported data (221 HIV-positive out of 1872 with last
delivery assisted by medical personnel vs
130 HIV positive out of 2422
with assistance by traditional birth assistant or no one),
the crude PAF for last delivery assisted by medical personnel
can be calculated as being 34%.
Studies
associating African HIV infections with injections.
At
least 15 large studies (with more than 500 subjects or 50
cases in a case-control study) of risk factors for HIV prevalence
or incidence in a general population sample (i.e., not CSWs
or patients seeking treatment for an STD or other illness)
in Africa have reported sufficient data to calculate crude
PAFs associated with one or more vs. no injections over
some period ranging from 4 months to lifetime (see Table
2).(16, 19, 77-89) Of the 20 PAFs calculated from these
15 studies (with PAFs for two samples in five studies),
only four are below 22%, and the unweighted average is 29%.
...
Several
investigators(19, 85, 90) noted that some of the association
may be due to people seeking treatment for HIV/AIDS symptoms
or STDs, but the assertion is not adequately supported by
research. ... In a parallel survey among 150 health workers,
prevalence for those with STDs and injections for STDs (47%)
was almost double prevalence for those with STDs only (24%).(90)
Discussion
The
recognition that significant shares of HIV in African adults
and children cannot be explained on the basis of current
knowledge about sexual and vertical transmission leaves
open several transmission hypotheses. There may, for example,
be co-factors for sexual transmission not yet identified
that are particularly influential during pregnancy or for
young women. However, an accumulating body of evidence from
Africa and other countries suggests that iatrogenic transmission
may explain many if not most of the observations previously
held to be anomalous and detailed in this review.
|
Table 2. Crude population attributable
fractions (PAFs) associated with medical injections
in studies of HIV Prevalence and Incidence in
Africa
|
|
Location, year of study
|
Type of data
|
Type of cohort
|
RR for>0 vs 0injections
|
p with
>0 injections*
(%)
|
PAF
|
|
DRC, 1984 (16)
|
P
|
Adults
|
1.82
|
81
|
40
|
|
DRC, 1984± 86 (19)
|
I
|
Adults
|
1.54^
|
73^
|
28^
|
|
Uganda, 1987 (77)
|
P
|
Adults
|
1.68
|
66.0
|
31
|
|
Zimbabwe, 1987{ (78)
|
P
|
Men
|
3.61**
|
95.0**
|
71**
|
|
Tanzania, 1987 (81)
|
P
|
Rural adults
|
2.6
|
79.2
|
56
|
|
|
|
Urban adults
|
3.0
|
89.7
|
64
|
|
Uganda, 1989 (79)
|
P
|
Men
|
1.67
|
41.8
|
22
|
|
|
|
Women
|
1.75
|
57.2
|
30
|
|
Uganda, 1989± 90 (80)
|
I
|
Adults
|
1.12
|
75.7
|
8
|
|
Malawi, 1989± 90 (88)
|
P
|
Women
|
0.91
|
64
|
-6
|
|
Kenya, 1989± 90 (89)
|
P
|
Women
|
0.73
|
96.6
|
-3.5
|
|
|
|
|
|
|
|
|
Tanzania, 1990± 91 (82)
|
P
|
Men
|
1.98
|
42.9
|
30
|
|
|
|
Women
|
1.66
|
|
