SINCE THE LATE 1980S, HETEROSEXUAL transmission has been considered the predominant route of transmission of HIV-1 in sub-Saharan Africa. However, recently, it has been suggested that unsafe medical injections may account for the majority of transmission in this region.1,2 This has sparked much debate and several studies have presented results that are not consistent with the iatrogenic transmission hypothesis.3,4
Proponents for the theory that iatrogenic transmission is the primary mode of transmission argue that unsafe medical injections are common in sub-Saharan Africa, that HIV-1 transmission efficiency for injections is 2.3% or greater, that there is a strong association between history of injections and HIV infection, and that there is only a weak association between sexual behavior and HIV-1 infection in adults in sub-Saharan Africa.1,2 It has been argued that this weak association may be a product of unsafe injections in adults attending sexually transmitted infection (STI) clinics.5 A study in Rwanda found higher HIV prevalence in people with injections for sexually transmitted diseases (STDs) in the last 2 years than in people who reported STDs but had not had any injections for STDs.6
The occurrence of HIV in people reporting no sexual experience7 and in children2 has been cited to strengthen the argument for nosocomial transmission. However, it has also been argued that the age–sex distribution of HIV-1 in sub-Saharan Africa cannot be explained by injection patterns but that it is similar to the age–sex distribution of other STIs3 and HIV-2. HIV-1 epidemics in sub-Saharan Africa are frequently characterized by a higher incidence in young people and a 5- to 10-year age delay in prevalence between men and women with women experiencing higher prevalence of HIV at younger ages than men. Prevalence of HIV in the 5- to 14-year age group is low and is much lower than that observed in adults. The age–sex pattern and the discrepancy between children and adults is not seen in the number of injections received.4 Schmid and colleagues observe that “if injections were a major mode of transmission, a much smaller discrepancy between HIV-1 prevalence in children and adults would be expected.”3
Those who ascribe a limited fraction of HIV transmission to an iatrogenic route argue that unsafe injections are not sufficiently common to have played a dominant role in transmission and that transmission efficiency from unsafe injections has been overestimated at 2.3%.3 The average number of injections administered for health reasons in sub-Saharan Africa has been estimated to be between 2.0 and 2.3 per person per year.8 The proportion of these procedures in which needles are reused has been estimated to be between 13% and 23%.8 Using these data, Hauri and colleagues9 calculated that the proportion of HIV infections as a result of unsafe injections in sub-Saharan Africa was 2.5% (confidence intervals, 1.9–3.1%).
It has also been argued that analyses of the association between injections and HIV-1 do not adequately take into account reverse causation and confounding, and that analyses of the relationship between sexual behavior and HIV-1 infection are oversimplified and do not consider issues related to measurement of exposure.3 Studies have shown associations between HIV and an increased number of sexual partners,10 and Kiwanuka and colleagues4 observed that their data suggested sexual transmission was the dominant mode of transmission. Several studies have found no association between injections and HIV in Rakai, Uganda,4 in Kinshasa, Democratic Republic of the Congo,11 and in rural Zimbabwe.12 One study in Rwanda found no association after adjustment for other variables.13
Studies of the risks of individuals acquiring infection provide a limited understanding of the population-level factors shaping the spread of infection through a community. The distribution of cases over time is a function of the contact patterns of individuals. To understand how different contact patterns change an epidemic, we can use mathematical models to predict the spread of infection and compare observed epidemics with those expected. In this article, we explore the epidemics produced by different transmission routes and determine whether it is possible to predict the size and pattern of the observed HIV epidemics through the two transmission routes and distinguish between them.
An age-structured deterministic compartmental model of HIV transmission was developed to compare the HIV epidemic generated by different routes of transmission. The model is defined in a set of partial differential equations that represent the relationships among biological, behavioral, and demographic variables. A full description of the model can be found in the appendix. Briefly, the model divides the population into susceptible individuals and HIV-infected individuals with respect to age, sex, and time. It allows the transmission of HIV through either of two routes: sexual contact or unsafe injections. We define unsafe injections as injections with reused needles. Only one route of transmission was used at a time. The duration of infection with HIV was assumed to be 10 years consistent with HIV in a nontreatment setting,14 and a Weibul distribution was approximated using stages of infection. For simplicity, the probability of transmission (by either route) was assumed to be constant over the duration of infection. Demographic parameters, age-specific fertility and mortality rates, for the model were taken from a study of HIV and fertility in Zimbabwe (see Appendix).15
In all simulations of iatrogenic transmission of HIV, we assumed that 90% of the adult population (aged over 13 years) received unsafe medical injections. The first scenario was of homogenous receipt of unsafe injections with the same number of unsafe medical injections experienced by all the injection-receiving population. It was assumed that individuals received between one and five unsafe injections per year and that none of the needles used were effectively cleaned. This latter assumption was made to explore the upper boundaries of any possible epidemic. The model assumes that all needles are reused, creating a link between two people who use the needle in sequence. This allows for transmission from the “first” recipient to the “next” but not further as a result of the reduction in virus associated with the potential infectious injection. This precludes more than one individual becoming infected from a single needle. The infection status of the person who was last injected with the needle determines whether the needle is contaminated. This assumption is reasonable given the ease with which HIV is flushed from a syringe.16
In addition, two scenarios with different levels of heterogeneity in receipt of unsafe injections were examined. Lopman and colleagues found that two groups of individuals experienced more injections in rural Zimbabwe: pregnant women and female STD patients.12 The adjusted risk ratios of receiving injections for these two groups were 1.30 and 1.39, respectively.12 These two population groups comprise approximately 20% of the adult population. To capture these characteristics, the model was parameterized with 20% of the adult (injection-receiving) population receiving 1.5 times the number of unsafe injections received by the remaining 80%. The range of average number of unsafe injections received per year (1–5) was the same as that used for the homogeneous scenario. In the second heterogeneous scenario, we investigated higher values of these parameters to identify circumstances within which faster epidemics would occur. We assumed that 10% of the injection-receiving population receiving five times the number of unsafe injections of the rest of the injection-receiving population. Like with the two previous scenarios, the same range of average number of unsafe injections received per year was explored. For both heterogeneous scenarios, a mixing parameter between injection levels of 0.5 was used. This parameter describes the similarity of the current recipient of the needle to the previous recipient. A value of 0.0 represents assortative mixing in which the current recipient is exposed to the same average number of unsafe injections per year as the previous recipient, and a value of 1.0 represents random mixing in which the population size of the activity classes determines whether the two recipients of the needle are similar or different in terms of the annual number of unsafe injections they are exposed to.
A range of probabilities for iatrogenic transmission of HIV on receipt of an infected needle was investigated (0.01–0.5). This range extends well beyond the upper bound of estimates for transmissibility of HIV from contaminated needles reported in a systematic review (0.000–0.058)17 and also those estimates used in recent analyses of the importance of unsafe injections to the transmission of HIV: 0.019 to 0.069,2 0.023,1 0.005,18 and 0.003.3
To examine HIV transmission through sexual contact, the model was parameterized with four sexual activity classes. These classes correspond to reported sexual behavior in Zimbabwe15 in terms of size, relative partner change rates, and the number of sex acts per partnership. Mixing between sexual activity classes was set to 0.3, also to correspond to that reported in Zimbabwe.15 Individuals are assigned to an activity class when they reach adulthood. For simplicity, they remain in that activity class throughout their lives. In addition, behavior was kept the same for men and women. We used average yearly rates of sexual partner change between one and four, which is consistent with studies in sub-Saharan Africa.19 The ratio of partner change rates between activity classes was consistent with the data reported in a study of HIV and fertility in Zimbabwe,15 and average annual partner change rates were varied between 1.5 and four. A range of probabilities for sexual transmission of HIV per unprotected sex act were investigated (0.002–0.03). A systematic review reported transmission probabilities between 0.003 to 0.2 for men to women and 0.003 to 0.082 for women to men.17
In all model simulations, HIV was introduced into the model after the demographic processes had stabilized. Simulations explored all combinations of partner change rates or injection rates per year and transmission probabilities. HIV prevalence in adults 22 years after the introduction of HIV to the population was examined for each combination of parameters. This can be compared with observed prevalence in sub-Saharan Africa in 2003 based on HIV first being observed in Africa in 1981. Adult prevalence was examined because the observed age structure of the epidemic shows prevalence is much higher in adults than in children. UNAIDS estimates that the adult HIV prevalence in sub-Saharan Africa at the end of 2003 was 7.5% (or between 6.9% and 8.3%) and was as high as 38.8% in Swaziland and 37.3% in Botswana.20 In Zimbabwe (where the demographic data used in these simulations are from), HIV prevalence at the end of 2003 was estimated to be 24.6% (between 21.7% and 27.8%).
Increasing the per-needle transmission probability or average number of unsafe medical injections increases the prevalence of HIV in the adult population (aged 15–49 years) reached 22 years after the introduction of HIV (Fig. 1A). However, for transmission probabilities under 0.14, with five or fewer average unsafe injections per year, no epidemic is achieved after 22 years. Once an epidemic has occurred, the size of the epidemic after 22 years with each combination of parameters increases rapidly to over 90% of the population. If an epidemic is possible, there is a small parameter range in which the spread of infection does not quickly become unreasonably great, i.e., over 90%. With an average of two unsafe injections a year,8 iatrogenic transmission probabilities of 0.39 and 0.43 are required to reach prevalence of 7.5% and 25%, respectively (Table 1).
The time course of iatrogenically transmitted epidemics is shown in Figure 2A. Oscillations in prevalence are a result of the adult population becoming saturated with HIV leading to declines in incidence and brief replenishment of susceptible numbers. Prevalence then decreases until aging of the population introduces more people vulnerable to HIV infection. The population rapidly declines as a result of the HIV epidemic and the majority of adults are infected very quickly. Peak adult prevalence is over 85% in most of the epidemics (Fig. 2A). In the smaller epidemics, the peak is closer to 80% or the peak is not reached in the 100 years of the simulation.
Results for the first heterogeneous scenario (scenario 2) were not significantly different to results from the simulations with homogeneous receipt of unsafe injections (results not shown). However, in scenario 3, with greater heterogeneity in receipt of unsafe injections, the transmission probabilities and average number of injections required to generate an epidemic of HIV are much less (Fig. 1B). With an average of five unsafe injections per year, a transmission probability of 0.06 (within the upper bounds suggested by Gisselquist and colleagues)2 adult prevalence of HIV after 22 years is projected to be 4.3%. This prevalence is comparable to HIV prevalence in 2003 in the Democratic Republic of Congo. An average of five unsafe injections per year equates to an average of 3.6 unsafe injections per year in 90% of the injection-receiving population and an average of 17.9 unsafe injections per year in the remaining 10% and is considerably more unsafe injections per year than the total number of annual injections reported by Hutin and colleagues.8 With an average number of injections of two per year,8 assuming all are unsafe, the transmission probability required for an epidemic of 7.5% after 22 years is 0.16 (Table 1).
When infection is able to spread widely, peak adult prevalence is similar for the case of heterogeneous (Fig. 2B) and homogeneous injection patterns (Fig. 2A). However, in regions of parameter space where the infection spreads less widely, heterogeneous injection patterns lead to faster and smaller epidemics when compared with those generated by homogeneous injection patterns. This is particularly true in the epidemic generated with an average of two unsafe injections per year and transmission probability of 0.15 which, in the case of homogeneous receipt of injections, does not reach peak prevalence within the 100 years shown on the graph. The main difference between the two scenarios is the speed of growth: when heterogeneity is introduced, HIV moves through the population more quickly.
The prevalence of HIV in the adult population (aged 15–49 years) 22 years after the introduction of HIV for each combination of average partner change rate and sexual transmission probabilities is shown in Figure 1C. Like with the iatrogenically transmitted epidemics, prevalence increases with increasing partner change rate and transmission probability. However, in this case, the increase is more gradual. The peak epidemic, which is only reached with the highest partner change rates, is also lower at around 75% to 80% of the population. The combinations of partner change rates and sexual transmission probabilities required to achieve adult prevalence of 7.5% or 25% after 22 years are shown in Table 1. With an average annual partner change rate of 2.5, the prevalence of HIV in adults reaches 25% in 22 years if the transmission probability of HIV is 0.0085 and 7.5% in 22 years if the transmission probability is 0.0065. As Figure 2C illustrates, the time course of the epidemic changes with different combinations of partner change rates and transmission probabilities. This change is more marked than in iatrogenic transmitted epidemics. Sexual transmission generates slower, less widespread epidemics than the iatrogenic route.
Our results can be interpreted either quantitatively or qualitatively. Quantitatively, we explore the parameter values that would be associated with the observed HIV epidemics as a result of iatrogenic or sexual transmission. Qualitatively, we explore the pattern of the epidemic. The key to generating a pattern broadly similar to observed epidemics is the inclusion of heterogeneity in behavior. The impact of the variation in risk behaviors is well known and has been realized for sexual behavior, less so for receipt of injections. However, the explicit representation of both iatrogenic and sexual behavior heterogeneities used here allows a clear illustration of the different epidemic patterns.
The sexual transmission model presented here reproduces observed variation in country epidemics if the partner change rate is varied within a plausible range. HIV epidemics in sub-Saharan Africa vary considerably between countries. For example, adult prevalence in Gambia at the end of 2003 was estimated to be 1.2%; in Uganda, it was estimated to be 4.1%; in Kenya, it was 6.7%; in Mozambique, it was 12.2%; in South Africa, it was 21.5%; and in Botswana, it was 37.3%.20 If a sexual transmission probability of 0.009 is assumed, our simulations indicate that small variations in average annual partner change rate between 1.5 and 2.5 will result in adult prevalence between 0.6% and 31.6% (see Fig. 1C).
A transmission probability of between 0.019 to 0.069 on receipt of an infected needle has been suggested.2 With the highest of these transmission probabilities, our results indicate that the average number of unsafe needles received per person per year would need to be much greater than five in the homogeneous scenario and around five or just over for the heterogeneous scenario. Gisselquist and colleagues2 noted that the proportion of injections that were unsafe was greater than 50% in the majority of studies. It has been estimated that on average, each individual receives two injections each year in sub-Saharan Africa, and only 17% to 19% are reused,8 giving an average of less than 0.4 unsafe injections a year or one a year using the estimate from Gisselquist and colleagues. These numbers of unsafe injections are not sufficient to create the sizes of epidemics seen in sub-Saharan Africa. However, it is important to note that there is very little information available on the number of unsafe injections received by different populations, and it is clear that more data on the number of unsafe injections, the groups who are likely to receive the same needle and, importantly, the heterogeneity in receipt of unsafe injections is required.
In all scenarios of iatrogenic transmission, with lower annual rates of unsafe injections, middling prevalence is more likely. Very high rates of unsafe injections (over 80% of all injections) have been observed in Pakistan,21 where HIV has been reported since the mid-1980s, but very high prevalence of HIV is not observed (adult prevalence in 2003 and 2001 was 0.1%).20 This corroborates the low transmission probabilities reported by Baggaley and colleagues.17 Therefore, it is difficult to justify the high iatrogenic transmission probabilities required to produce the HIV epidemics seen in sub-Saharan Africa. However, if HIV were to become established in Pakistan, the very high rates of unsafe injections would be a cause for concern.
Despite the apparently clear difference between the plausibility of the two routes of transmission, model assumptions may have served to exaggerate some of the differences. In simulations of sexual transmission, four sexual activity classes were used to capture heterogeneity in sexual behavior. Heterogeneity in receipt of unsafe injections was not explored to the same degree. Analysis of simulations using two groups of injection-receiving adults parameterized using data from Lopman and colleagues12 showed no difference in epidemic sizes or trajectories from simulations with all the population receiving the same number of unsafe injections. When more heterogeneity is included in the model (like in scenario 3), more plausible numbers of unsafe injections and transmission probabilities are required to generate observed epidemics. As already noted, heterogeneity in contact patterns is of key importance and further information on possible groups of the population who are likely to receive more injections is needed.
The iatrogenic transmission model contains some assumptions that will underestimate the predicted epidemic, including no transmission in children (to increase chances of representing adult HIV prevalence) and no onward transmission after one reuse. Alternatively, the assumption that all injections are risky (i.e., all needles unclean) will overestimate predicted spread. The sexual transmission model assumes that individuals remain in the same sexual activity class throughout their adult (aged 15–49 years) lives, overestimating spread compared with a model in which high-risk sexual behavior is transitory. Transitory risk behavior would result in smaller, more peaked epidemics, more similar to those observed. It should also be observed that rather than measuring prevalence 22 years after the introduction of HIV into the population, an alternative would have been to measure prevalence 22 years after a threshold prevalence had been reached. HIV was spreading silently in Africa well before it was “discovered” in 1982–1983, possibly some 50 to 100 years prior. However, the change in results in terms of which combinations of parameters produce epidemics is likely to be small as most of the simulations, which do not produce epidemics after 22 years do not go on to produce epidemics at all.
We suggest that the transmission probabilities, the average number of unsafe injections, and degree of iatrogenic heterogeneity required to generate the observed HIV epidemics in sub-Saharan Africa through transmission by unsafe medical injections are unlikely. Conversely, for heterosexual transmission to have generated the observed epidemics, the transmission parameters fall well within previously published ranges. Therefore, heterosexual transmission seems a more likely main route of HIV-1 transmission in the region. However, more studies are required to accurately quantify levels of heterogeneity in medical injections in the region. Although it is possible that both iatrogenic and sexual transmission are responsible for the spread of HIV in the region, we have treated them as mutually exclusive. The current lack of data on levels of heterogeneity for medical injections precludes a more systematic approach in which a combined model could be used to determine relative likely levels of transmission.
Mathematical Model Description
The description of the model that follows includes both routes of transmission for ease of description; however, during the simulations, each route of transmission was removed in turn. The number of people in the susceptible category is denoted Skli, Ikli for the infected category and Nkli for the total number of people. The subscript k represents gender, with k′ representing the opposite gender, and the subscripts l and i denoting sexual activity class and receipt of unsafe injections (if i >0), respectively. The model is defined by the following set of partial differential equations. The boundary conditions for age zero are:
where r is the sex ratio of births, f is the age-specific fertility rate, and v is the probability of vertical transmission. The initial population distribution was derived at steady state in the absence of HIV from a starting population along with age- and sex-specific fertility and mortality rates. The following partial differential equations define the model for children aged zero to 12 inclusive:
where μ(a) is the crude mortality rate for age a and ω(a) is the age-specific AIDS-related mortality rate. The following partial differential equations define the model for adults aged 13 to 99 inclusive:
where γk(a) is the age-specific rate of becoming sexually active for gender k with φkl the proportion of people of gender k being recruited to sexual activity class l, σk(a) is the age-specific rate of starting to receive medical injections for gender k with ψki the proportion of people of gender k being recruited to injection receiving class i, and age a is the crude mortality rate for age a, and ω(a) is the age-specific AIDS-related mortality rate.
The force of infection of HIV through sexual contact is given by:
Equation (Uncited)Image Tools
where ρkll′im(a,a′) is the probability that an individual of gender k, sexually activity class l, injection receiving class i, and age a has sexual contact with an individual of the opposite gender in sexual activity class l′, injection receiving class m and age a′, the sexual mixing matrix, ckli(a) is the rate of acquisition of sexual partners per year by an individual of gender k, sexual activity class l, injection receiving class i, and age a, β is the transmission probability of HIV per sex act and naa′ll′ is the number of unprotected sex acts in a partnership between an individual of gender k, age a, and sexual activity class l and an individual of the opposite gender, age a′, and sexual activity class l′.
The force of infection of HIV through injection contact is given by:
Equation (Uncited)Image Tools
where ρ′im(a,a′) is the probability that an individual of age a, in injection receiving class i receives a needle previously received by someone of age a′, in injection receiving class m, the needle receiving mixing matrix, kki(a) is the age-specific rate of receiving needles for an individual of gender k and injection-receiving class i, α is the transmission probability of HIV through an infected needle, and v is the proportion of needles that have been effectively cleaned.
The mixing matrices were defined following the approach of Garnett and Anderson,19 and for ease of calculation, they were constructed for discrete age groups q and q′. The sexual mixing matrix can be defined as:
Equation (Uncited)Image Tools
where ε1 is the degree of assortative sexual mixing by age (0 = random, 1 = assortative), ε2 is the degree of assortative mixing within sexual activity class, ε3 is the degree of assortative sexual mixing within injection-receiving class, and δim is the identity matrix, which equals 1 when i = m and 0 otherwise.
Rate of sexual partner change must be balanced between genders, age, and activity classes; therefore, if the discrepancy:
Equation (Uncited)Image Tools
then c is updated such that:
Equation (Uncited)Image Tools
and the partner change rate in reverse direction is:
Equation (Uncited)Image Tools
where ϕ is the degree to which men or women alter their behavior. We assume that they alter their behavior equally (ie, ϕ = 0.5).
The injection-receiving mixing matrix can be defined in a similar manner as:
Equation (Uncited)Image Tools
where ε4 is the degree of assortative needle-sharing by age and ε4 is the degree of assortative needle-sharing by injection-receiving class.
The demographic parameters used in simulations of both routes of transmission were obtained from Garnett and Gregson15 for Zimbabwe (see Table 1A). Cited Here...
1.Gisselquist DP. Estimating HIV-1 transmission efficiency through unsafe medical injections. Int J STD AIDS 2002; 13:152–159.
2.Gisselquist D, Rothenberg R, Potterat J, et al. HIV infections in sub-Saharan Africa not explained by sexual or vertical transmission. Int J STD AIDS 2002; 13:657–666.
3.Schmid GP, Buve A, Mugyenyi P, et al. Transmission of HIV-1 infection in sub-Saharan Africa and effect of elimination of unsafe injections. Lancet 2004; 363:482–488.
4.Kiwanuka N, Gray RH, Serwadda D, et al. The incidence of HIV-1 associated with injections and transfusions in a prospective cohort, Rakai, Uganda. AIDS 2004; 18:342–344.
5.Gisselquist D, Potterat JJ. Confound it: Latent lessons from the Mwanza trial of STD treatment to reduce HIV transmission. Int J STD AIDS 2003; 14:179–184.
6.Van de Perre P, Carael M, Nzaramba D, et al. Risk factors for HIV seropositivity in selected urban-based Rwandese adults. AIDS 1987; 1:207–211.
7.Zaba BW, Carpenter LM, Boerma JT, et al. Adjusting ante-natal clinic data for improved estimates of HIV prevalence among women in sub-Saharan Africa. AIDS 2000; 14:2741–2750.
8.Hutin YJ, Hauri AM, Armstrong GL. Use of injections in healthcare settings worldwide, 2000: Literature review and regional estimates. BMJ 2003; 327:1075.
9.Hauri AM, Armstrong GL, Hutin YJ. The global burden of disease attributable to contaminated injections given in health care settings. Int J STD AIDS 2004; 15:7–16.
10.Auvert B, Ballard R, Campbell C, et al. HIV infection among youth in a South African mining town is associated with herpes simplex virus-2 seropositivity and sexual behaviour. AIDS 2001; 15:885–898.
11.N′Galy B, Ryder RW, Bila K, et al. Human immunodeficiency virus infection among employees in an African hospital. N Engl J Med 1988; 319:1123–1127.
12.Lopman BA, Garnett GP, Mason PR, et al. Individual level injection history: A lack of association with HIV incidence in rural Zimbabwe. Public Library of Science: Medicine 2005; 2:e37.
13.Bulterys M, Chao A, Habimana P, et al. Incident HIV-1 infection in a cohort of young women in Butare, Rwanda. AIDS 1994; 8:1585–1591.
14.UNAIDS Reference Group on Estimates Modelling and Projections. Improved methods and assumptions for estimation of the HIV/AIDS epidemic and its impact: Recommendations of the UNAIDS Reference Group on Estimates, Modelling and Projections. AIDS 2002; 16:W1–14.
15.Garnett GP, Gregson S. Monitoring the course of the HIV-1 epidemic: The influence of patterns of fertility on HIV-1 prevalence estimates. Mathematical Population Studies 2000; 8:251–277.
16.Abdala N, Gleghorn AA, Carney JM, et al. Can HIV-1-contaminated syringes be disinfected? Implications for transmission among injection drug users. J Acquir Immun Defic Syndr 2001; 28:487–494.
17.Baggaley R, Boily M-C, White R, et al. Systematic review of HIV-1 transmission probabilities in absence of antiretroviral therapy. In: UNAIDS Reference Group on Estimates Modelling and Projection. Geneva, 2004.
18.Kane A, Lloyd J, Zaffran M, et al. Transmission of hepatitis B, hepatitis C and human immunodeficiency viruses through unsafe injections in the developing world: Model-based regional estimates. Bull World Health Organ 1999; 77:801–807.
19.Garnett GP, Anderson RM. Factors controlling the spread of HIV in heterosexual communities in developing countries: Patterns of mixing between different age and sexual activity classes. Philos Trans R Soc Lond B Biol Sci 1993; 342:137–159.
20.UNAIDS. 2004 Report on the Global HIV/AIDS Epidemic: 4th Global Report, 2004.
21.Simonsen L, Kane A, Lloyd J, et al. Unsafe injections in the developing world and transmission of bloodborne pathogens: A review. Bull World Health Organ 1999; 77:789–800.