Secondary Logo

Journal Logo

Epidemiology and Social

To what extent is the HIV epidemic in southern India driven by commercial sex? A modelling analysis

Vickerman, Petera,b; Foss, Anna Ma; Pickles, Michaela,c; Deering, Kathleend; Verma, Supriyaf; Eric Demers, e; Lowndes, Catherine Me,g; Moses, Stephend; Alary, Michele; Boily, Marie-Claudec

Author Information
doi: 10.1097/QAD.0b013e32833e8663
  • Free

Abstract

Introduction

UNAIDS (2008) estimates 2.4 million people are living with HIV in India. Although the HIV prevalence in India amongst adults was low (0.28%) in 2006 [1], in the southern states of Andhra Pradesh, Karnataka and Maharashtra, the HIV prevalence was up to three-fold higher.

The ‘Avahan Initiative’ aims to control HIV spread in the higher prevalence states of India by concentrating prevention activities among high-risk groups [2,3]. To determine whether this strategy is appropriate, it is important to determine the extent to which HIV transmission is driven by commercial sex.

As part of evaluating Avahan, behavioural and biological data have been collected from female sex workers (FSWs), their clients and the wider population in numerous districts. These data are collected through serial cross-sectional surveys among FSWs and clients, termed the Integrated Biological and Behavioural Assessments (IBBAs); in-depth special behavioural surveys (SBSs) among FSWs; general population biobehavioural surveys (GPSs); and general population polling booth surveys (PBSs) designed to reduce social desirability bias [3–7].

Previous epidemiological studies looking at self-reported risk behaviours amongst HIV-infected sexually transmitted infection (STI)/HIV clinic clients have suggested that the HIV epidemic in Pune and Chennai is focused within FSWs, their clients and women who report one lifetime partner [8–11]. This suggests the HIV epidemic in southern India could be driven by commercial sex, with low-risk women possibly being infected by main partners who are clients of FSWs. However, the accuracy of these studies was limited by the selective nature of the populations surveyed. Additionally, they did not consider whether these transmission routes accounted for all prevalent infections in that setting or how the population attributable fraction (PAF) due to different partnership types changes over an HIV epidemic.

The present study considers these questions through using mathematical modelling to explore the extent to which the HIV and herpes simplex virus-2 (HSV-2) epidemics in two districts of southern India are driven by female commercial sex. The districts considered both have in-depth IBBA data from FSWs and clients, and GPS/PBS data from the general population: Belgaum, a higher prevalence district in northern Karnataka; and Mysore, a lower prevalence district in southern Karnataka. In both settings, few men (<17%) or women (<1%) report more than one lifetime sexual partner [6], supporting the hypothesis that HIV/HSV-2 transmission amongst individuals not involved in commercial sex may be low. Using FSW and client IBBA data, this hypothesis is tested by estimating whether most prevalent HIV/HSV-2 infections observed in the general population can be accounted for by HIV/HSV-2 infections resulting from commercial sex between FSWs and clients, and their noncommercial bridging infections. The model estimates the PAF of different partnership types to these HIV/HSV-2 epidemics over their duration and the impact of targeting a generic intervention to these different partnerships.

Methods

Model description

A deterministic, compartmental model was used to simulate the transmission of HIV and STI amongst FSWs, their clients and the general population in Mysore and Belgaum. The model stratifies the population by sexual activity and HIV/STI status. Sexually active individuals are stratified by whether they have commercial partnerships or not. FSWs are also stratified by whether they have above or below the mean number of commercial clients per month. Individuals of different sexual activity form main, casual and commercial sexual partnerships at defined rates, coital frequencies and levels of condom use. The frequency of commercial sex for clients is determined by the overall frequency of commercial sex reported by FSWs divided by client population size. For simplicity, the transmission of HIV/STIs amongst MSM was not modelled (see discussion). Figure 1 shows a schematic of the risk groups included in the model and the partnerships modelled.

Fig. 1
Fig. 1:
Schematic showing behavioural subgroups included in model and different partnership types modelled. FSW, female sex worker; HSV-2, herpes simplex virus-2; STI, sexually transmitted infection.

Because STIs facilitate HIV transmission [12], the model incorporates the transmission of syphilis, HSV-2 and a generic curable STI (representing Neisseria gonorrhoeae and Chlamydia trachomatis). The model formulations for these STIs are described in the supplementary material, http://links.lww.com/QAD/A86. When individuals are HIV-infected, they enter a short primary infection stage, then progress to the long asymptomatic phase and lastly a symptomatic phase before leaving the population due to AIDS-related morbidity and mortality. The effect of antiretroviral therapy (ART) was not modelled because HIV clinic data from Karnataka suggest coverage was low before 2008.

Individuals enter the simulated population when they become sexually active and have a specified sexual activity. FSWs and clients stop selling and buying sex at a certain rate, enter the noncommercial risk groups (Fig. 1) and are replaced by an equal number becoming FSWs/clients, but spread proportionally over all infection classes. Individuals leave the population due to natural death or AIDS-related morbidity/mortality.

The model uses a system of ordinary differential equations to simulate how HIV/STIs spread over time. The force of infection for all HIV/STIs combines the risk of infection for each individual's partnerships with the mixing probabilities for forming different partnerships – random for commercial and casual partnerships, but ranging from assortative to random for main partnerships. The model incorporates the effect of HSV-2, syphilis and N. gonorrhoeae/C. trachomatis on increasing HIV infectivity and susceptibility, the effect of HIV on increasing the rate of HSV-2 symptomatic recurrences and HSV-2 infectivity, and the higher HIV infectivity associated with primary and symptomatic HIV infection [13]. Multiple cofactors are additive. The effect of circumcision was not included because of its low prevalence (∼10% [6]).

More details are in the supplementary material, http://links.lww.com/QAD/A86.

Model parameterization

The model was parameterized using behavioural data from FSWs/clients in Mysore and Belgaum (round 1 IBBA and SBS), and data from general population surveys (GPS/PBS) undertaken in these districts. Some data from these surveys are published [4–7]. Most parameters were directly obtained from these surveys, but some were derived indirectly (Table 1), such as client population size and temporal changes in condom use between FSWs and clients as explained below [15].

Table 1
Table 1:
Behavioural model input parameter ranges used to obtain the Mysore and Belgaum model fits.

Specifically, data for FSWs/clients on the frequency of commercial sex, proportion with main/casual partners, frequency of sex and consistency of condom use with different partners generally came from the FSW/client IBBA surveys. This was complemented with data from the GPS because the Mysore client IBBA was not fully available, and the FSW SBS for the frequency of sex in noncommercial partners. The IBBA data suggest FSWs in Belgaum have more clients (mean 57 clients per month in Belgaum and 30 in Mysore [7]) and stay in sex work longer. Most FSWs (52–68%) and clients (66–68%) in both settings also report having a ‘main’ noncommercial partner but few (<20%) report having additional noncommercial partners (defined as ‘casual’ partners) in the last year. IBBA and SBS data suggest there is a lower coital frequency associated with these ‘casual’ partners (3–4 per month) than with ‘main’ partners (7–11 per month) and greater condom use (>40% in last sex act compared to <20% for ‘main’ partners).

The GPS was the preferred data source for the sexual behaviour of low-risk individuals, but was supplemented with data from the FSW/client IBBAs and PBS for risk behaviours that were likely to be underreported in the GPS (e.g. proportion of general population currently having casual partners and frequency of casual partners). GPS data suggest that approximately 60% of men and approximately 80% of women in Mysore/Belgaum are sexually active, but with a low reported rate of sexual partner change, and the PBS suggests very few have casual partners (<10% of men and <3% of women).

Data on the proportion of women who are currently FSWs came from the anonymous PBS and suggested more women are FSWs in Mysore (0.7%) than Belgaum (0.5%; Table 1). The PBS also produced an estimate for the proportion of men who paid for sex in the past 12 months (3.5% in Mysore; 4.8% in Belgaum) and was used to give a rough lower bound (7) for the factor difference between the number of FSWs and clients. An upper bound for this factor (19 for Mysore and 35 for Belgaum) was obtained by dividing the mean frequency of clients per month in each setting as reported by FSWs (Table 1) by the lower bound frequency that clients see FSWs per month in the client IBBAs (1.6). These two estimates produced a range for the factor difference between the client and FSW population size.

A range for the consistency of condom use between FSWs and clients at the time of the round 1 FSW IBBA was derived from the FSW/client IBBA, GPS and PBS. To estimate condom use in other years, data from the round 2 FSW IBBA on when FSWs started using condoms consistently were used to reconstruct the yearly increase in consistent condom use before and after the Avahan intervention in Belgaum (2004) and Mysore (2003) [15]. Consistent with condom distribution data for the same period [16], these data suggested consistent condom use between FSWs and clients increased from 15 to 65% in Mysore (2001–2007) and from 30 to 85% in Belgaum (2001–2008) [15]. The same yearly increase was assumed to apply to condom use in last sex act, as used in the model. The supplementary material and supplementary Figure 3 gives further details, http://links.lww.com/QAD/A86.

Other HIV/STI biological parameters were obtained from the scientific literature (see supplementary material, http://links.lww.com/QAD/A86). Uncertainty bounds were estimated for most parameters.

Model fitting

The epidemiological data used to fit and cross-validate the model are shown in Table 2, with fitting data shown in bold. The round 1 IBBA data suggest the HIV prevalence amongst FSWs at that time was 26% in Mysore and 34% in Belgaum, and approximately 5% amongst clients. Over subsequent years, the rounds 2/3 FSW IBBA data suggest HIV prevalence decreased amongst FSWs in Mysore and possibly in Belgaum. Lastly, GPS data suggest the general population HIV prevalence in Mysore was 0.8% in 2006 [6] and 1.3% in Belgaum in 2007.

Table 2
Table 2:
Epidemiological data used to fit, validate and initialize the model for Mysore and Belgaum.

A Bayesian fitting algorithm [14,18,19] was used to identify multiple model fits to available HIV/STI prevalence data from the round 1 FSW IBBA surveys (October 2005 for Belgaum, August 2004 for Mysore) and HIV prevalence data from the round 1 client IBBA surveys (October 2007 for Belgaum, October 2008 for Mysore). In effect, 20 000 000 parameter sets were sampled [20] from the parameter uncertainty bounds in Table 1 and supplementary Table 1, http://links.lww.com/QAD/A86 to identify model simulations within the 95% confidence intervals (CIs) of this HIV/STI data (model fit). Importantly, all model simulations assumed HIV was seeded in 1987, consistent with the first HIV prevalence estimates amongst FSWs in India [17]; and that condom use increased over this period. This fitting algorithm enabled our modelling results to incorporate the uncertainty present in the data used to parameterize the model, and epidemiological data used to fit the model.

Instead of fitting the model to the client STI data (round 1 IBBA) or HIV prevalence data from subsequent FSW IBBA surveys (July 2008 for Belgaum, December 2006 and April 2009 for Mysore), these data were used to cross-validate the model's accuracy. The model was also not fitted to general population HIV/HSV-2 prevalence data from the GPS or antenatal clinic (ANC) surveys (2002–2008) as the model predictions needed to be independent of these data (see analysis plan). More details of the fitting process are in the supplementary material, http://links.lww.com/QAD/A86.

Overall, 3728 model simulations fit the Mysore IBBA FSW HIV/HSV-2/STI and client HIV prevalence data, and 2205 for Belgaum.

Analysis plan

With and without including HIV/HSV-2 transmission between low-risk individuals, all model fits projected the general population HIV/HSV-2 prevalence trends. The exclusion of HIV/HSV-2 transmission between low-risk individuals was initiated at the start of a model simulation. These model projections were compared to empirical prevalence estimates from the GPS for Mysore (June 2006) and Belgaum (June 2007) to establish whether the model could account for all prevalent HIV/HSV-2 infections without including any transmission between low-risk individuals. Because sexual partner change reported by women was low (GPS), the model projections that included transmission between low-risk individuals assumed that low-risk women have the same sexual activity as reported by men in the GPS (Table 1). The model projections were also compared against ANC HIV prevalence estimates.

These model fits were also used to estimate the annual proportion of incident HIV/HSV-2 infections resulting from either commercial sex partnerships, main or casual partnerships of FSWs or clients, and sexual partnerships between low-risk individuals (Fig. 1). For each year of the HIV epidemic in each setting, this PAF was defined as the relative decrease in the number of incident HIV/HSV-2 infections when all HIV and HSV-2 transmission risk due to a specific partnership type was set to zero for that year. Additional projections assumed low-risk men and women have higher sexual activity on the assumption that men may also be underreporting sexual activity (Table 1).

Lastly, to explore the relative impact of intervention strategies targeted to different partnership types, the model estimated the relative decrease in the overall HIV infections over 5 years (from 2009) resulting from an additional generic intervention resulting in a 20% reduction in the HIV force of infection within either commercial or noncommercial partnerships of FSWs or clients.

Results

Female sex worker and client projections

The model's projections in Fig. 2a/b suggest the HIV prevalence amongst FSWs/clients in Mysore and Belgaum increased from a low level in 1987, peaking in the late 1990s in Belgaum and Mysore, and then decreased due to increasing condom use within commercial sex since 2000. As expected, the model projections agree with round 1 IBBA FSW/client HIV prevalence data used to fit the model and also with rounds 2/3 IBBA FSW HIV prevalence data not used in the fitting process. In addition, the model's FSW/client STI prevalence projections generally agree well with round 1 IBBA data (supplementary Figure 1, http://links.lww.com/QAD/A86). The supplementary material discusses the prior/posterior model parameter ranges.

Fig. 2
Fig. 2:
A comparison of the model fit projections (median, 2.5, 25, 75 and 97.5% percentiles) and data estimates (with 95% confidence bounds) for the female sex worker, client and female general HIV prevalence. Data estimates with grey shading were used in fitting, whereas all others were not. Comparisons with female general HIV prevalence include data estimates (with 95% confidence bounds) from the antenatal clinic (ANC) surveys, which were weighted down because model projections include those women who are not sexually active. (a) Female sex worker (FSW) and client data and model projections for Mysore. (b) FSW and client data and model projections for Belgaum. (c) Female general population data and model projections for Mysore. (d) Female general population data and model projections for Belgaum.

General population projections

With all HIV transmission routes included, Fig. 2c/d shows the model's HIV prevalence projections for the female general population show similar trends to their respective FSW/client projections. Due to increasing condom use between FSWs and clients, projections suggest the female HIV prevalence decreased from 2003 to 2009, by 39% (95%CI 17–52%) and 24% (95%CI −13–43%) in Mysore and Belgaum, respectively. Similar trends occur in the male general population (not shown).

Without fitting the model to HIV prevalence data from the general population, the model corresponds well with ANC and GPS HIV prevalence estimates (Fig. 2c/d). Indeed, 52 and 32% of the Mysore and Belgaum model projections are within the 95% CIs of the GPS HIV prevalence estimates. These HIV prevalence projections change little (<1% relative change) if the model excludes HIV transmission due to low-risk partnerships (supplementary Figure 5, http://links.lww.com/QAD/A86). The model's projections also agree well with GPS HSV-2 prevalence estimates (supplementary Figure 5, http://links.lww.com/QAD/A86), with low-risk partnerships again having little effect.

Variability in the model's general population HIV prevalence projections is primarily due to uncertainty in the FSW's client population size (supplementary Figure 6, http://links.lww.com/QAD/A86).

Population attributable fraction due to different partnership types

Projections suggest most incident male HIV/HSV-2 infections (PAF > 90%) in either setting have and continue to be due to commercial sex (Fig. 3 and supplementary Figure 7, http://links.lww.com/QAD/A86), with the remainder mostly due to main noncommercial partnerships of FSWs (PAF < 10%). Most incident female HIV/HSV-2 infections (PAF = 80–90%) are due to main noncommercial partnerships of clients, with the remainder mainly due to commercial sex (PAF = 10–20%). However, earlier in the epidemic, a greater proportion of female HIV infections were due to commercial sex (PAF = 40–50%), but this changed as clients became infected and condom use within commercial sex increased.

Fig. 3
Fig. 3:
Proportion of yearly incident HIV infections in men and women that result from different types of sexual partnership in Mysore and Belgaum (defined as population attributable fraction in figure). Projections show median with 2.5 and 97.5% percentiles. Low-risk women are assumed to have the same sexual activity as reported by men in the general population biobehavioural survey (GPS). (a) Mysore incident infections in men. (b) Mysore incident infections in women. (c) Belgaum incident infections in men. (d) Belgaum incident infections in women. FSW, female sex worker;

The bridging infections from FSWs/clients to their noncommercial casual partners have little effect on HIV/HSV-2 transmission (PAF < 1%) because the frequency of casual partners is low (Table 1). Similarly, low-risk sexual partnerships only contribute approximately 1.5% of all male/female incident HIV infections if their sexual activity is as reported in the male GPS (1.05 main partners per year), and less than 7% if their sexual activity were approximately 50% higher than reported levels (Table 1).

Intervention impact projections

The simple intervention projections suggest that a 20% reduction in transmission risk achieves greatest impact, in terms of a 16% (95%CI 5–22%) reduction in HIV infections in Mysore over 5 years and 13% (95%CI 8–20%) reduction in Belgaum, when targeted to commercial sex partnerships (Supplementary Figure 8, http://links.lww.com/QAD/A86). Targeting noncommercial partnerships of clients achieves similar impact in year 1 but 35–60% less by year 5. Targeting noncommercial partnerships of FSWs achieves little. Interestingly, the yearly impact of targeting commercial sex partnerships nearly doubles over 5 years due to increasing infections being averted in other partnerships, which are indirectly due to the chain of transmissions initially originating from commercial sex. Despite this, the best immediate strategy for reducing infections amongst noncommercial partnerships of clients is still to target these partnerships directly. However, this is likely to be less cost-effective than targeting FSWs because over 10 times more individuals need to be reached if clients are targeted, and a much greater increase in condom use (5–7% increase in condom use within commercial sex versus 21% increase amongst the main noncommercial partners of clients) and distribution is needed to achieve the same 20% decrease in transmission risk.

Discussion

This analysis suggests the HIV epidemic in Mysore and Belgaum could be largely driven by commercial sex, and that intervention strategies targeting commercial sex are likely to achieve greatest impact. This suggests the Indian HIV epidemic is still concentrated amongst high-risk groups and their sexual partners and gives credence to Avahan's strategy of concentrating prevention activities (condom distribution/promotion and STI treatment) amongst FSWs and clients. In partial accordance with time-trend HIV prevalence data from Karnataka [2,21], the model projects the HIV epidemic in these districts is declining due to the hypothesized increase in condom use within commercial sex since 2000, as suggested by two studies [15,16]. In addition, the low initiation rate of new main/casual sexual partners in these districts [6] seems to have prevented HIV/HSV-2 propagating beyond the noncommercial sexual partners of FSWs/clients.

Most incident HIV infections among women in both districts (80–85%) originate from their husbands/main partners, who have acquired HIV from FSWs. This is the main source of bridging infections to the broader population and agrees with studies showing that marriage is an important risk for HIV infection among women [8]. Model projections suggest additional impact could be achieved from increasing the protection to wives/main partners of clients that is not easily achieved through just targeting commercial sex. Traditionally, prevention programmes in India have focused primarily on FSWs. By turning their attention towards clients as well as FSWs, Avahan has re-addressed this balance. However, encouraging clients to use condoms with their main partners will be a challenge [22], and so other possible new technologies such as microbicides or preexposure prophylaxis could become important when they become available.

The assumed client population size is a key factor driving our results. Indeed, client population size largely determines the general population HIV/HSV-2 prevalence in both settings, with the likely higher proportion of men visiting FSWs in Belgaum possibly being the main determinant for their higher general population HIV prevalence compared to Mysore. This highlights the importance of obtaining better size estimates of client populations to improve our understanding of concentrated HIV epidemics and the accuracy of model projections.

One model limitation is that it did not include MSM. Limited existing epidemiological data suggest up to 12% of male HIV infections in these settings [6,23] and elsewhere in southern India [9] could be due to MSM. Although its inclusion could have increased the proportion of infections due to commercial sex in men and the proportion due to bridging infections in women, it is likely that most HIV infections will still have been due to female sex work and bridging infections from their clients. Despite the model's relative complexity, it did not include some risk behaviour subtleties that can affect HIV/STI transmission. These include such things as FSWs having differing levels of condom use, some FSWs having regular commercial clients and not explicitly modelling long-term partnerships. The modelling was also limited by uncertainty in several parameters, such as the number of FSWs/clients and HIV/STI parameters. To reduce this uncertainty, Bayesian fitting methods were used [18,19] that extensively sampled over parameter uncertainty ranges to obtain multiple model fits to FSW/client HIV/STI prevalence estimates. These model fits give an indication of the wide range of epidemics that can agree with available HIV/STI prevalence data (Fig. 2). In addition, the model fits were compared against HIV/STI prevalence data from FSWs, clients and the general population that was not used in the fitting process. The close correspondence of the model fits to this cross-validation data suggested the model accurately represents the HIV/STI epidemics in these settings.

In conclusion, this analysis suggests that virtually all HIV/HSV-2 infections in the general population of these districts can be accounted for by infections occurring between FSWs and clients, and their noncommercial partners. Clients are the main source of bridging infections to the general population and so marriage is the main risk factor for women in these districts. This raises important implications for targeting HIV prevention interventions in India, which are likely to be relevant to other settings with concentrated HIV epidemics. Prevention efforts to reduce HIV transmission within commercial sex must continue and be strengthened, but strategies should also be considered to reduce transmission from clients to their noncommercial partners.

Acknowledgements

The present research was funded by the Bill & Melinda Gates Foundation. The views expressed are those of the authors and cannot be taken to reflect the official opinions of the London School of Hygiene and Tropical Medicine and do not necessarily reflect the official policy or position of the Bill & Melinda Gates Foundation.

P.V., M.A. and M.C.B. devised the analysis plan. P.V. adapted the model and undertook all model analyses. M.C.B., C.M.L., M.P., S.M. and A.F. all commented on the model results. P.V. wrote the first draft of the manuscript and all authors read and commented on the manuscript. K.D., S.V. and E.D. undertook all data analyses for parameterizing the model. M.A., C.M.L. and S.M. played major roles in collecting the data used by the model.

References

1. International Institute for Population Sciences. National Family Health Survey 3 (http://www.nfhsindia.org/nfhs3.html). Mumbai; 2006.
2. Boily MC, Pickles M, Vickerman P, Buzdugan R, Isac S, Deering KN, et al. Using mathematical modelling to investigate the plausibility of attributing observed antenatal clinic declines to a female sex worker intervention in Karnataka state, India. AIDS 2008; 22(Suppl 5):S149–S164.
3. Chandrasekaran P, Dallabetta G, Loo V, Mills S, Saidel T, Adhikary R, et al. Evaluation design for large-scale HIV prevention programmes: the case of Avahan, the India AIDS initiative. AIDS 2008; 22(Suppl 5):S1–S15.
4. Ramesh BM, Beattie TS, Shajy I, Washington R, Jagannathan L, Reza-Paul S, et al. Changes in risk behaviours and prevalence of sexually transmitted infections following HIV preventive interventions among female sex workers in five districts in Karnataka state, south India. Sex Transm Infect 2010; 86(Suppl 1):i17–i24.
5. Reza-Paul S, Beattie T, Syed HU, Venukumar KT, Venugopal MS, Fathima MP, et al. Declines in risk behaviour and sexually transmitted infection prevalence following a community-led HIV preventive intervention among female sex workers in Mysore, India. AIDS 2008; 22(Suppl 5):S91–S100.
6. Munro HL, Pradeep BS, Jayachandran AA, Lowndes CM, Mahapatra B, Ramesh BM, et al. Prevalence and determinants of HIV and sexually transmitted infections in a general population-based sample in Mysore district, Karnataka state, southern India. AIDS 2008; 22(Suppl 5):S117–S125.
7. Ramesh BM, Moses S, Washington R, Isac S, Mohapatra B, Mahagaonkar SB, et al. Determinants of HIV prevalence among female sex workers in four south Indian states: analysis of cross-sectional surveys in twenty-three districts. AIDS 2008; 22(Suppl 5):S35–S44.
8. Gangakhedkar RR, Bentley ME, Divekar AD, Gadkari D, Mehendale SM, Shepherd ME. Spread of HIV infection in married monogamous women in India. JAMA 1997; 278:2090–2092.
9. Mehendale SM, Gupte N, Paranjape RS, Brahme RG, Kohli R, Joglekar N, et al. Declining HIV incidence among patients attending sexually transmitted infection clinics in Pune, India. J Acquir Immune Defic Syndr 2007; 45:564–569.
10. Mehta SH, Gupta A, Sahay S, Godbole SV, Joshi SN, Reynolds SJ, et al. High HIV prevalence among a high-risk subgroup of women attending sexually transmitted infection clinics in Pune, India. J Acquir Immune Defic Syndr 2006; 41:75–80.
11. Newmann S, Sarin P, Kumarasamy N, Amalraj E, Rogers M, Madhivanan P, et al. Marriage, monogamy and HIV: a profile of HIV-infected women in south India. Int J STD AIDS 2000; 11:250–253.
12. Rottingen J, Cameron DW, Garnett GP. A systematic review of the epidemiologic interactions between classic sexually transmitted diseases and HIV. Sex Transm Dis 2001; 28:579–595.
13. Wawer MJ, Gray RH, Sewankambo NK, Serwadda D, Li X, Laeyendecker O, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis 2005; 191:1403–1409.
14. Pickles M, Foss A, Vickerman P, Deering K, Verma S, Demers E, et al. An interim analysis of HIV among high-risk groups in southern India on changes in condom use due to the India AIDS initiative intervention. STI 2010; 86(Suppl 1):i33–i43.
15. Lowndes C, Alary M, Verma S, Demers E, Bradley JE, Jayachandran AA, et al. Assessment of intervention impact in the absence of baseline data: ‘reconstruction’ of condom use time trends using retrospective analysis of survey data. STI 2010; 86:i49–i55.
16. Arora P, Cyriac A, Jha P. India's HIV-1 epidemic. CMAJ 2004; 171:1337–1338.
17. Hallett TB, Gregson S, Mugurungi O, Gonese E, Garnett GP. Assessing evidence for behaviour change affecting the course of HIV epidemics: a new mathematical modelling approach and application to Zimbabwe. Epidemics 2009; 1:108–117.
18. Vickerman P, Platt L, Hawkes S. Modelling the transmission of HIV and HCV amongst injecting drug users in Rawalpindi, a low HCV prevalence setting in Pakistan. Sex Transm Infect 2009; 85:ii23–ii30.
19. Blower SM, Dowlatabadi H. Sensitivity and uncertainty analysis of complex models of disease transmission: an HIV model, as an example. Int Stat Rev 1994; 62:229–243.
20. Moses S, Ramesh BM, Nagelkerke NJ, Khera A, Isac S, Bhattacharjee P, et al. Impact of an intensive HIV prevention programme for female sex workers on HIV prevalence among antenatal clinic attenders in Karnataka state, south India: an ecological analysis. AIDS 2008; 22(Suppl 5):S101–S108.
21. Bradley JE, Moses S, Blanchard J, Rajaram S, Ramesh BM, Alary M. Validating reported condom use among female sex workers in southern India through examination of condom availability data. STI 2010; 86:i44–i48.
22. Foss AM, Watts CH, Vickerman P, Heise L. Condoms and prevention of HIV. BMJ 2004; 329:185–186.
23. Alary M, Jayachandran AA, Lowndes CM, Bradley J, Demers E, Adhikary R, et al. Ecological analysis of the association between high-risk population parameters and HIV prevalence among pregnant women enrolled in sentinel surveillance in four southern India states. Sex Transm Infect 2010; 86(Suppl 1):i10–i16.
Keywords:

HIV; India; modelling

Supplemental Digital Content

© 2010 Lippincott Williams & Wilkins, Inc.