Data specific to Karnataka were used to parameterize the model and to reflect the demographic characteristics of the general and ANC populations [14–16]. The behavioural parameter values were based on surveys among low and high-risk populations in Karnataka [14,17–20,32]. The model was not fitted to any specific district but to reflect the overall range of HIV prevalences observed among different ANC populations, which seem to have reached equilibrium in 2003–2004 as suggested in Fig. 1a [2–4]. Given observed sexual behaviour patterns, the following assumptions were made: women become sexually active at a younger age than men; women tend to mix with men 5–10 years older; and the client population is larger but less sexually active than the FSW population [17–20]. As younger women necessarily acquire one new partner when they first become sexually active, the rate of partner acquisition of 15–19-year-old women was assumed always to be lower than one new partner per year, and to decline with age [18,19] (Table 1). The levels of sexual activity of older women were thus always lower than those of younger women. Fertility rates were higher among nulliparous sexually active women , and could set to be lower among seropositive women . Attendance at ANC is marginally higher at the first than at following pregnancies .
Our null hypothesis was that it is unlikely that even a very intensive targeted intervention could lead to a 30% and 50% decline in ANC HIV prevalence within a 3 and 4-year time period. Biological parameters that we thought may maximize the impact of this intervention among the ANC population were thus sometimes chosen, such as 15–19-year-old women being at increased risk of HIV infection  and HIV infectivity being elevated in the HIV acute phase (RR1 = 15–20) . Data on the exact coverage of the intervention and changes in behaviour in high-risk groups remains limited at present but should become available in future [6,32,33]. For the intervention itself, we thus made some very optimistic assumptions to test our hypothesis, under various concentrated epidemic conditions. Although Moses et al.  reported that 80% of FSW may have been reached by the intervention only in 2005, we assumed that all FSW (fcond′ = 100%) were reached instantaneously at the start of the modelled intervention rather than gradually over time. In addition, we also assumed that the fraction of FSW–client sex acts (factc) protected by condoms increased instantaneously from an assumed baseline level prevailing before the start of the intervention to a level exceeding 80% at the start of the modelled intervention. Data on the baseline level of condom use before the Avahan intervention is very uncertain. Although it is believed to be low , unpublished data (M. Alary) from behavioural surveys among FSW in three Karnataka districts, who recalled ‘always’ using condoms with new/occasional clients, suggest that it could have been between 15% and 45% in 2002. The baseline frequency of condom use thus varied from 25% to 40% in the sensitivity analysis (Fig. 1c). In reality, the intervention started during 2003, coverage increased and may have been at scale by mid-2005. Therefore, in each simulation, we randomly chose to start the intervention between June 2003 and 2004 and started measuring the impact of the simulated intervention from 2004 for 3 (until 2007) and 4 years (until 2008).
Initially, 147 000 different parameter sets were selected by Latin hypercube sampling over prespecified uniform parameter ranges (Table 1) . As previously mentioned, the number of partners of low-risk women was constrained to decrease with age (m′11 > m′12 > m′13 > m′14). Among these, 2258 sets satisfied the fit criteria of reflecting the observed range of HIV prevalences by risk and age groups in the different Karnataka districts, and were included in the analysis. The fit criteria required the overall 2004 female HIV prevalence to be between 0.5% and 4.0%, the relative change in the female HIV epidemic to be less than ±2% between 2003 and 2004, the 2004 HIV prevalence in FSW and clients by age to be between 5% and 45%, and 1.5% and 20%, respectively, and the 2004 client age-specific HIV prevalence to be lower than the FSW prevalence [3,4,17–20]. The 2258 parameter sets were run with and without the intervention to measure the intervention impact.
The characteristics of the different concentrated epidemics simulated are presented in Tables 2 and 3. Figures 1, 2 and Table 3a summarize the projections of the model fits for the impact of a very effective targeted FSW intervention on different HIV outcomes over 3 and 4 years for different simulated concentrated epidemics. Our results [median (10th, 90th percentiles)] suggest that a very effective intervention may result in an approximately 39% (29%, 47%) and 47% (36%, 55%) intervention decline in HIV incidence after 3 and 4 years among general population low-risk 15–19-year-old nulliparous women (Fig. 1b, Table 3b–c). A corresponding decline in overall general population HIV prevalence is more modest at approximately 15% (11%, 19%), compared with approximately 21% (14%, 27%) in the overall ANC population over 3 years (Fig. 1b). The HIV prevalence among 15–19-year-old ANC attendees (including high-risk women and all parity levels) was more sensitive, with an approximately 27% (19%, 35%) 3-year intervention decline. The 3-year intervention decline in ANC HIV prevalence among low-risk 15–19-year-old women at first pregnancy is lower, whereas the 3-year intervention decline in HIV prevalence among clients and FSW for this intervention is large: 33% (28%, 38%) and 44% (39%, 50%), respectively (Table 3b, Fig. 1b). These intervention impacts were obtained after an absolute increase in condom use post-intervention of 58% (51%, 63%) (Fig. 1c, Table 3b). The 4-year intervention declines are presented in Table 3c.
The sensitivity analysis results shown in Fig. 2a demonstrate how the magnitude of intervention decline depends on the magnitude of the absolute increase in condom use. Varying the latter from 65% to 45% reduced the intervention decline among FSW from approximately 50% to approximately 40%. Varying the intervention decline in client HIV prevalence from 40% to 25% resulted in a reduction in the 3-year intervention decline in HIV prevalence among 15–19-year-old ANC attenders from approximately 30% to approximately 20% (Fig. 2b). In addition, the intervention impact depends on epidemic size, with the 3-year intervention decline in overall ANC HIV prevalence being lower at higher overall general population female prevalence in 2004, partly as a result of the fraction of infections among low-risk women in 2004 caused by clients (2004 client population attributable fraction; PAF) being smaller for larger simulated epidemics (Fig. 2c). Indeed, the 3-year intervention decline in overall ANC HIV prevalence was 15% (11%, 21%), compared with 24% (19%, 29%) when the 2004 client PAF was less than 60% or more than 80%, respectively (Table 2). Figure 2d compares the intervention decline among the 15–19-year-old ANC population after 3 and 4 years of intense intervention, in function of the 2004 client PAF among 15–19-year-old women. The decline in HIV prevalence among 15–19-year-old ANC attenders remain below 55% even after 4 years of intervention and a large 2004 PAF. For epidemics when the client PAF among 15–19-year-olds was above 78%, the intervention decline was 35% (26%, 44%) after 4 years compared with 27% (20%, 35%) after 3 years.
Finally, the predicted time to observe a 30% decline in overall ANC prevalence, and in prevalence among 15–19-year-old women, was approximately 5.0 (4.0, 6.4) and 3.6 (2.8, 4.8) years, respectively, whereas the predicted time to observe a 55% decline was approximately twice as long for both outcomes (Table 3d).
Our results do not fully support the assertion that the observed decrease in ANC HIV prevalence was caused by the Avahan FSW targeted intervention. Under the very favourable conditions modelled, perfect coverage and an instantaneous, absolute, increase in condom use between FSW and clients as high as approximately 58%, and assuming that the majority of HIV infections among low-risk women are caused by male clients, the median decline in HIV prevalence among the overall and 15–19-year-old ANC population 3 years after implementation of the intervention was 21% and 27%, and 4 years after implementation was 27% and 35%. These estimates are lower than the observed decline in HIV prevalence between 2004 and 2007 among overall (52% decline) and 15–24-year-olds (59% decline) among female ANC attenders from sentinel surveillance sites in the Avahan intervention districts in Karnataka . In contrast, the decrease in FSW HIV prevalence projected for the modelled intervention is 44% over 3 years, much larger than the small decrease in FSW HIV prevalence (from 25.6% to 24.3% between 2004 and 2006) observed in the only follow-up survey with available data so far in Mysore, Karnataka . The current collection of further serial cross-sectional data among FSW in Karnataka will provide additional useful data to compare against our model projections, with large decreases in FSW HIV prevalence first being required to support the assertion that the observed decrease in ANC HIV prevalence is caused by the Avahan FSW interventions [6,33].
The predicted changes over time applying to fairly stable pre-intervention epidemics has seemed to have been occurring in Karnataka . Given the same intervention impact (i.e. the same increase in FSW–client condom use), however, the decline over time (and the time to produce a 30% decline) would appear faster/slower in declining/rising epidemics because of the natural dynamics of the epidemics, which would explain some of the changes over time. These predictions of intervention impact are applicable to other states in India with concentrated epidemics and similar sexual behaviour and fertility patterns. The intervention impact is, however, expected to be lower in populations in which women become sexually active some years before giving birth, and in larger generalized HIV epidemics [9,12].
It is important to note that our results are not an estimate of the impact of the Avahan intervention. They suggest that the Avahan intervention has the potential to be very effective, but that it is probably too early for it to produce the magnitude of decline in overall ANC HIV prevalence observed since 2003 (Fig. 1a).
Considerable uncertainty remains around current estimates of condom use rates by FSW with clients in the different districts before, and following, the start of the intervention, and it is thus not known whether the range of increase modelled [median 58%, 10th, 90th percentiles (51%, 63%)] could actually have occurred. Data from two successive rounds of integrated behavioural and biological assessment (IBBA) surveys among FSW in Mysore in 2004 and 2006 suggest that the biggest increases in condom use rates were with the last new client, from 65% to 90% [17,35]. The situation in northern Karnataka districts, however, where the HIV epidemic has been more severe, may be different. Condoms may have been less widely available to FSW before the start of the programme, particularly in Shimoga, Davangere, rural Belgaum, Bijapur, and Bagalkot (S. Moses, J. Blanchard, personal communication). In IBBA surveys carried out in the districts of Belgaum, Bellary, urban Bangalore and Shimoga, in 2005/2006, consistent condom use with new clients was 93%, 74%, 78% and 53%, respectively . Data on prevailing levels of condom use before the intervention are scarce. In three independent special behavioural surveys conducted in Bellary, Belgaum and Bangalore in 2006, 70%, 75% and 64% of FSW reported always using condoms with new clients; whereas based on their report of condom use in the past the proportion consistently using condoms 24%, 54% and 47%, respectively, reported that they had started to use condoms consistently in 2003 (M. Alary, personal communication). This contrasts with the observed decline in ANC HIV prevalence, which is larger and more consistent in Belgaum than Bellary (Fig. 3).
Our ability to detect the impact of the intervention among the ANC population depends greatly on the extent to which FSW are represented among the ANC population. In five ANC sites (Bangalore, Belgaum, Bellary, Mysore, Shimoga) less than 0.7% of women reported two or more lifetime partners, whereas the size estimate of the FSW population is approximately 0.5–2% across districts, but these estimates may be biased [17,18]. In addition, observed trends in HIV prevalence among the ANC population vary considerably across the different districts of Karnataka (Fig. 3). The largest declines were observed among high-prevalence districts (i.e. 2003 HIV prevalence >1%), be they Avahan or non-Avahan. There are, however, more Avahan than non-Avahan high-prevalence districts. The majority of the continuously rising epidemics are among the lowest prevalence districts (<0.75%), which are disproportionately non-Avahan. A similar relationship has also been observed for India as a whole, where ANC HIV prevalence has declined in the southern states, which have a higher prevalence than the northern states, where prevalence has not declined [2,4,36–38]. Therefore, it is possible that a large portion of the observed decrease in ANC HIV prevalence in Avahan districts could be caused by the non-random selection of high-prevalence districts for Avahan, with the variations in observed ANC HIV trends across the different districts possibly reflecting differences in HIV epidemic stages or dynamics, variations in increases in condom use among FSW before 2003, or differential biases unrelated to the targeted intervention [7–12].
Documenting earlier prevention efforts among FSW, particularly in non-Avahan districts, would provide additional empirical information, which would be useful to assess the extent to which core group interventions are effective at reducing HIV rates in the general population in this context. In Karnataka, some intervention activities may have commenced before 2003 in urban Belgaum, Bijapur, Bagalkot, Dharwar, Raichur, Koppal and Tumkur (B.M. Ramesh, personal communication); districts that include both Avahan and non Avahan sites [39,40]. In addition, a better characterization of the ANC population would also help address the issue of changes in selection biases over time.
In conclusion, this study indicates that, in concentrated epidemics and under specific conditions, a significant decrease in ANC prevalence can be achieved as a result of highly intensive and effective targeted interventions. Based on the available empirical data and our modelling results, it seems highly unlikely that the observed decline in HIV prevalence among the ANC population in Karnataka could be solely the result of the Avahan intervention over such a short period of time. The modelling results indicate that in order to achieve a 30% decline in overall ANC HIV prevalence in a 3-year period through targeted FSW interventions, a very rapid and substantial increase in condom use among high-risk groups, as well as a very rapid reduction in FSW and client HIV prevalence of at least 35–40% must be demonstrated to have occurred. As part of the monitoring and evaluation of the Avahan intervention, future rounds of IBBA surveys among high-risk groups will provide relevant behavioural and biological information to validate and interpret trends among the ANC population in southern India [5,6,41]. In the meantime, more efforts should be made to examine complementary explanations for these declines before we can attribute a fraction of the decline to the intervention and sound conclusions and policy recommendations concerning intervention impact can be drawn [9,41].
M.C.B., M.P., P.V., S.M., J.B. and M.A. contributed to the different aspects of the study design. M.C.B. developed and coded the mathematical models. S.M., J.B., R.B. provided the data and parameter estimates. M.C.B., M.P., R.B., S.I., K.D., B.M.R. and E.D. performed the simulations, modelling or statistical analyses. M.C.B., P.V., M.A., S.M., J.B., C.L. interpreted the results, M.C.B. wrote the first version of the manuscript. All authors contributed to the following versions of the manuscript. M.C.B. and P.V. finalised the manuscript.
The authors would like to thank Gina Dallabetta, Padma Chandrasekaran and Geoff Garnett for very useful discussions and suggestions.
Sponsorship: Support for this research was provided by the Bill and Melinda Gates Foundation through Avahan, its India AIDS initiative.
The views expressed herein are those of the authors and do not necessarily reflect the official policy or position of the Bill and Melinda Gates Foundation and Avahan.
Conflicts of interest: James F. Blanchard receives funding from Canada Research Chairs, Health Canada.
All other authors declare no conflict of interest.
1. Chandrasekaran P, Dallabetta G, Loo V, Rao S, Gayle H, Alexander A. Containing HIV/AIDS in India: the unfinished agenda. Lancet Infect Dis 2006; 8:508–521.
3. Moses S, Ramesh BM, Nagelkerke NJD, 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.
5. Avahan – the India AIDS initiative. The business of HIV prevention at scale
. New Delhi, India: Bill & Melinda Gates Foundation; 2008.
6. Boily MC, Lowndes CM, Vickerman P, Kumaranayake L, Blanchard J, Moses S, et al
. Evaluating large-scale HIV prevention interventions: study design for an integrated mathematical modelling approach. Sex Transm Infect 2007; 83:582–589.
7. Zaba B, Boerma T, White R. Monitoring the AIDS epidemic using HIV prevalence data among young women attending antenatal clinics: prospects and problems. AIDS 2000; 14:1633–1645.
8. Glynn JR, Buvé A, Carael M, Zaba B. Adjustment of antenatal clinic HIV surveillance data for HIV-associated differences in fertility. AIDS 1999; 13:1598–1599.
9. Garnett GP, Gregson S, Stanecki KA. Criteria for detecting and understanding changes in the risk of HIV infection at a national level in generalised epidemics. Sex Transm Infect 2006; 82(Suppl. 1):i48–i51.
10. Gupte N, Sastry J, Brookmeyer R, Phadke MA, Bhosale RA, Bollinger RC. Declining HIV infection rates among recently married primigravid women in Pune, India. J Acquir Immune Defic Syndr 2007; 45:570–573.
11. Walker PT, Hallett TB, White PJ, Garnett GP. Interpreting declines in HIV prevalence: impact of spatial aggregation and migration on expected declines in prevalence. Sex Transm Infect 2008; 84(Suppl. 2):ii42–ii48.
12. Boily MC, Lowndes C, Alary M. The impact of HIV epidemic phases on the effectiveness of core group interventions: insights from mathematical models. Sex Transm Infect 2002; 78(Suppl. 1):i78–i90.
13. Glynn JR, Carael M, Auvert B, Kahindo M, Chege J, Musonda R, et al
. Study Group on the Heterogeneity of HIV Epidemics in African Cities. Why do young women have a much higher prevalence of HIV than young men? A study in Kisumu, Kenya and Ndola, Zambia. AIDS 2001; 15(Suppl. 4):S51–S60.
15. Sekher TV, Raju KNM, Sivakumar MN. Fertility transition in South India; Karnataka component of the South India fertility project (SIFP)
. Population Research Centre Institute for Social and Economic Change, Bangalore, 560 072; January 2000. Available at: http://www.demographie.net/sifp/phase1.htm
. Accessed: 2 December 2008.
17. A workshop organised by Karnataka State AIDS Prevention Society, Karnataka Health Promotion Trust. Policy, practice, research, dissemination and dialogue: using evidence for improving HIV prevention and care in Karnataka. Findings from Karnataka
. Bangalore, India: University of Manitoba; 26 February 2008.
18. Munro HL, Banandur P, 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.
19. Rajaram S, Sangameshwar BM, Jayachandran AA, Bradley J, Alary M, Ramesh BM, Washington R. HIV and STIs in Belgaum district, Karnataka, India. A general population survey
. CHARME Working paper No. 1; June 2008. 178 pages. (available on request).
21. Lewis JJC, Ronsman C, Ezeh A, Gregson S. The population impact of HIV on fertility in sub-Saharan Africa. AIDS 2004; 18(Suppl. 2):S35–S43.
22. Boily MC, Baggaley RB, Wang L, Masse RB, White RG, Hayes RJ, Alary M. Risk of HIV-1 infection per sexual contact: a systematic review and meta-analysis of observational studies. Lancet Infectious Diseases
24. Powers KA, Poole C, Pettifor AE, Cohen MS. Rethinking the heterosexual infectivity of HIV-1: a systematic review and meta-analysis. Lancet Infect Dis 2008 4 August E-pub ahead of print.
25. Hira SK, Shroff HJ, Lanjewar DN, Dholkia YN, Bhatia VP, Dupont HL. The natural history of human immunodeficiency virus infection among adults in Mumbai. Natl Med J India 2003; 16:126–131.
26. Collaborative Group on AIDS Incubation, HIV Survival. Time from HIV-1 seroconversion to AIDS and death before widespread use of highly active antiretroviral therapy: a collaborative re-analysis. Collaborative Group on AIDS Incubation and HIV Survival including the CASCADE EU Concerted Action. Concerted Action on Seroconversion to AIDS and Death in Europe. Lancet 2000; 355:1131–1137.
27. San Andres Rebollo FJ, Rubio Garcia R, Castilla Catalan J, Pulido Ortega F, Palao G, de Pedro Andres I, et al
. Mortality and survival in a cohort of 1,115 HIV-infected patients (1989–97). An Med Interna 2004; 21:523–532.
28. Williams JR, Foss AM, Vickerman P, Watts C, Ramesh BM, Reza-Paul S, et al
. What is the achievable effectiveness of the India AIDS initiative intervention among female sex workers under target coverage? Model projections from southern India. Sex Transm Infect 2006; 82:372–380.
29. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li C, Wabwire-Mangen F, et al., for the Rakai Project Study Group. Viral load and heterosexual transmission of immunodeficiency virus type 1. N Engl J Med 2000; 342:921–929.
30. Rottingen JA, Cameron DW, Garnett GP. A systematic review of the epidemiologic interactions between classic sexually transmitted diseases and HIV: how much really is known? Sex Transm Dis 2001; 28:579–597.
31. Foss AM, Hossain M, Vickerman PT, Watts CH. A systematic review of published evidence on intervention impact on condom use in sub-Saharan Africa and Asia. Sex Transm Infect 2007; 83:510–516.
32. Use it or lose it: how Avahan used data to shape its HIV prevention efforts in India
. New Delhi, India: Bill & Melinda Gates Foundation; 2008.
33. Chandrasekaran P, Dallabetta G, Loo V, Mills S, Saidel T, Adhikary R, et al
, for the Avahan Evaluation Partners. Evaluation design for large-scale HIV prevention programmes: the case of Avahan, the India AIDS initiative. AIDS 2008; 22(Suppl. 5):S1–S15.
34. Iman RL, Helton JC, Campbell JE. An approach to sensitivity analysis of computer models. Part 1. Introduction, input variable selection and preliminary variable assessment. J Quality Technol 1981; 13:174–183.
35. Reza-Paul S, Beattie T, Syed HR, Venukumar KT, Venugopal MS, Fathima M, 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.
36. Arora P, Kumar R, Bhattacharya M, Nagelkerke NJ, Jha P. Trends in HIV incidence in India from 2000 to 2007. Lancet 2008; 372:289–290.
37. Arora P, Cyriac A, Jha P. India's HIV-1 epidemic
. Can Med Assoc J
38. Kumar R, Jha P, Arora P, Mony P, Bhatia P, Millson P, et al
, International Studies of HIV/AIDS (ISHA) Investigators. Trends in HIV-1 in young adults in south India from 2000 to 2004: a prevalence study. Lancet 2006; 367:1164–1172.
39. India-Canada Collaborative HIV/AIDS Project (ICHAP). Community-based HIV prevalence study in ICHAP demonstration project area (additional tables)
. Bangalore, March 2003 March. Available at: http://www.ichapindia.org/
. Accessed: November 2008.
41. Garnett G, Gracia-Calleja JM, Rehle T, Gregson S. Behavioural data as an adjunct to HIV surveillance data. Sex Transm Infect 2006; 82(Suppl. 1):i57–i62.
© 2008 Lippincott Williams & Wilkins, Inc.