Sexually transmitted HIV infection may stem from 2 groups of people: HIV+ persons unaware they are infected and HIV+ persons aware of their infection status. The Centers for Disease Control and Prevention (CDC) estimates that 850,000 to 900,000 persons are living with HIV in the United States and that approximately 650,000 are aware of their status and 250,000 are not.1,2 Little is known about the relative difference between these 2 groups in exposing uninfected sex partners to HIV. A better understanding of this relative difference can help to guide public health strategies and allocation of resources to fight the HIV/AIDS epidemic in the United States.
To reduce the risk of HIV transmission, the CDC recommends a variety of interventions in medical care and community settings for those who are aware of their status. Linkage to medical care and antiretroviral therapy, which may reduce viral load and transmission risk,3-6 and ongoing behavioral interventions are important approaches for those who know they are living with HIV. To reach those who are unaware of their infection, the CDC recommends making HIV counseling and testing more accessible and acceptable for people at high risk for HIV infection.7 Of course, both strategic pathways are important, and prevention efforts can be informed by even roughly estimating the differences in sexual risk behaviors between awareness groups. We performed a meta-analysis of findings that compared the prevalence of unprotected anal and vaginal intercourse (UAV) between persons who were aware and unaware of their HIV infection.
Literature Search and Study Selection
Electronic databases, including MEDLINE, PubMed, PsycINFO, AIDSLINE, and SocioFile from January 1987 through January 2004, were searched to identify relevant published articles and conference abstracts. We crossed multiple search terms (ie, keywords and medical subject headings) reflecting 2 categories: (1) HIV status (HIV, AIDS, antibody, positive, counseling, testing, and aware) and (2) sexual behavior or sexually transmitted diseases (sexual behavior, sex, unsafe, unprotected, condom, anal, vaginal, intercourse, risk, high risk, exposure, transmission, sexually transmitted disease, STD, sexually transmitted infection, STI, gonorrhea, and syphilis). Additional material was identified from reference lists of pertinent articles. Studies were included in the meta-analysis if they met all the following criteria:
- Data from the United States
- Compared a group of HIV+ aware persons with an independent group of HIV+ unaware persons (ie, between-group comparison) or measured seroconverting individuals before and after notification of HIV+ status (ie, within-subject comparison)
- Measured any of the following sexual behaviors during a specified recall period:
- Unprotected insertive or receptive anal intercourse or unprotected vaginal intercourse
- Consistency of condom use during sexual intercourse
- Other sexual risk measures that combined components of the previous categories
- Data were reported or available to calculate the prevalence of UAV (ie, 1 or more instances during the recall period)
A total of 620 abstracts were screened. Although many studies identified in the search examined the prevalence of sexual behavior or the effects of HIV testing and counseling, few of the screened sources provided the specific types of comparisons (criterion 2) needed for this meta-analysis. Four studies8-11 published in peer-reviewed journals and 4 multisite data sets met inclusion criteria. The 4 data sets were (1) the Multicenter AIDS Cohort Study (MACS),12 (2) the HIV Epidemiology Research Study (HERS),13 (3) the Supplement to HIV/AIDS Surveillance (SHAS phase 1, 1995-2000),14 and (4) SHAS phase 2 (2000-2003).15 The 2 SHAS data sets were independent (no dual participation). We obtained the data necessary for our meta-analysis directly from the investigators. The Women's Interagency HIV Study (WIHS) was also eligible for inclusion but not used because it had too few seroconversions for a meaningful analysis.
Using a standardized spreadsheet, information on the following variables was abstracted: author(s), location and/or setting of data collection, study period, sample characteristics (eg, gender), sample size, type of comparison (between-group vs. within-subject), data collection methods (self-administered vs. interviewer-administered questionnaire), sexual behavior measured, recall period, and findings.
The data abstraction was guided by the following rules:
- If a study had independent samples of participants (eg, men, women) and reported data on each sample, we calculated an effect size (ES) for each independent sample.8,14,15
- To ensure independence of data in the meta-analysis, only 1 sexual behavior measure per study was used in calculating the overall ES. If separate data were available for unprotected insertive and receptive anal intercourse, we included only insertive activity because it has a higher risk of transmitting HIV when performed by an HIV+ person.16 One study8 used a composite index (UAV or oral sex).
Calculation of Effect Size
Effect sizes were estimated with a prevalence ratio (PR). For between-group comparisons, the HIV+ aware groups' prevalence of UAV was divided by the HIV+ unaware groups' prevalence of UAV, resulting in the PR. For within-subject comparisons, PR was calculated by dividing the prevalence of UAV at the postnotification period by the prevalence at the prenotification period. For each type of comparison, 1−PR was then calculated to reflect the percentage reduction in the prevalence of UAV in aware groups relative to unaware groups (or prenotification vs. postnotification period).
Standard meta-analytic methods were used.17,18 A random-effects model for aggregating individual ESs was used because it provides a more conservative estimate (than a fixed-effects model) of the variance and generates more accurate inferences about a population of studies beyond those included in the review.19 We first used the natural logarithm to obtain the log PR (lnPR) and calculated its corresponding weight (ie, inverse variance) for each independent sample. For within-subject comparisons, the variance of lnPR was adjusted for nonindependence of pre- and postnotification sexual behavior before aggregation (Appendix). To calculate the overall ES, we multiplied each lnPR by its weight, summed the weighted lnPR across samples, and then divided by the sum of the weights. In presenting the results, the aggregated lnPR was converted back to the PR by exponential function. 1−PR was then derived, along with a 95% confidence interval (CI).
Sensitivity analyses examined the effect of outliers by comparing the aggregated ES with estimates obtained after iterations using k−1 findings (k = number of independent samples). In other words, we removed a finding and calculated the aggregated ES. We then replaced that finding, removed another, and repeated the process. Stratified analyses were also conducted to examine whether ESs differed by type of comparison (between-group vs. within-subject) and gender of participants (men vs. women).
Adjustment for HIV Serostatus of Sex Partners
The meta-analysis was conducted with and without applying an adjustment factor to primary studies that assessed self-reported sexual behavior with any partner.8-10,12,14 Data on the prevalence of UAV with any partner may bias the meta-analysis, because some people who know they are HIV+ only have unprotected sex with partners perceived to be HIV+.20-22 The adjustment factor focused the analysis on behavior with partners at risk for infection.
The adjustment was applied to the prevalence data of HIV+ aware participants (or prevalence during the postnotification period) before conducting the meta-analysis. The adjustment was not applied to the HIV+ unaware group (or prenotification period) under the reasonable assumption that unawareness leaves risky behavior patterns unaltered and that persons who are unaware of their infection are not likely to be having unprotected sex only with HIV+ partners. Accordingly, we assume that some of their sexual activity is with persons at risk for infection.
The empirically derived adjustment factor was calculated with data from 6 US studies conducted from 1995 to 2001 (J. L. Richardson, S. Stoyanoff, J. Milam, et al, unpublished data, 1999)22-26 different from those in the meta-analysis. The participants in the “adjustment studies” were all aware of their HIV+ status and were recruited at a variety of locations (eg, HIV clinics, AIDS service organizations, bars, clubs). The adjustment studies contained data on self-reported UAV with any partners and with partners reported to be HIV+. The behavioral window varied (eg, past 3 months, last 2 sexual encounters). Half of the studies used self-administered surveys, and the other half used face-to-face interviews.
For each adjustment study, we first calculated the percentage of participants who engaged in UAV with any partner (factor “A”). Next, participants who engaged in those risk behaviors only with partners reported to be HIV+ were removed from the numerator. The percentage was then recalculated using all participants in the denominator to estimate the prevalence of UAV with partners at risk for infection (factor “B”). Dividing this “at-risk prevalence” (B) by the “any partner prevalence” (A) and then subtracting from 1 (ie, 1−B/A) yields the percentage reduction in the prevalence of UAV with at-risk partners relative to any partner. Calculations were performed for 3 groups to identify appropriate adjustment indices for the samples in the meta-analysis: men who have sex with men (MSM; 5 samples), men in general (4 samples that pooled MSM and heterosexual men), and women (3 samples).
The adjustment index was calculated as the unweighted average of the reductions across studies within each group. The unweighted average minimizes the influence of sample size and methodologic factors that may affect behavioral assessment. The unweighted average reduction was 43% (range: 27.3%-52.5%) for MSM, 47.5% (range: 42.0%-51.8%) for men in general, and 45.1% (range: 38.9%-56.0%) for women. These indices were applied only to studies that assessed UAV with any partner. For example, HIV+ aware MSM's prevalence of UAV with any partner was reduced 43% to derive an estimated prevalence of UAV with at-risk partners. The adjustment factor was not applied to studies that assessed behavior with at-risk partners (study by Colfax et al,11 SHAS phase 2 data set), nor was it applied to “casual partners” (HERS data set), because many of those casual partners were probably at risk. Sensitivity analyses examined whether there were any changes in the meta-analytic results when the adjustment index was calculated as a weighted average (ie, applying an inverse variance weight to each adjustment study before averaging).
Eleven independent findings from 1988 through 2003 were included in the meta-analysis. Six were between-group comparisons, and 5 were within-subject comparisons. Table 1 presents the characteristics of the studies and the unadjusted findings. Every study showed that the prevalence of UAV was lower in the HIV+ aware group (or postnotification period) compared with the HIV+ unaware group (or prenotification period).
Two sets of results of the random-effects models are presented in Table 2: findings based on the unadjusted data and findings based on data adjusted to focus on at-risk sex partners. The combined ES for all 11 findings in the unadjusted model indicates that the prevalence of UAV was an average of 53% (95% CI: 45%-60%) lower in HIV+ aware persons relative to HIV+ unaware persons. The reduction was 68% (95% CI: 59%-76%) in the adjusted model (k = 11). Each of these reductions differed significantly from 0 (Z > 25, P < 0.001). In both models, the reduction was significantly larger in the between-group findings (k = 6) compared with the within-subject findings (k = 5) (unadjusted model: χ(1) 2 = 38.71, P < 0.0001; adjusted model: χ(1) 2 = 8.32, P < 0.01). The ESs of the men and women did not differ significantly in either model (unadjusted: χ(1) 2 = 0.18, P > 0.50; adjusted: χ(1) 2 = 2.47, P > 0.10).
Sensitivity analyses indicated that no individual finding appreciably affected the overall (k = 11) or between-group (k = 6) ES in the two models. The ESs were affected less than ±3% (eg, absolute change from 68% to 70% reduction in the prevalence of UAV) with any single finding removed. For the unadjusted and adjusted models, the within-group ES (k = 5) changed no more than ±5% and the men's ES (k = 7) changed no more than ±4%. There was slightly more instability in the women's aggregated findings. The ES changed no more than ±4% in the women's unadjusted model but changed up to 8% in the adjusted model. For example, the ES was reduced nearly 8% in the adjusted model when the SHAS phase 1 finding was removed and was increased 4% in the adjusted model when the SHAS phase 2 finding was removed. These individual findings (especially the SHAS phase 1 finding) were heavily weighted in the aggregated analysis because of large sample sizes. In summary, the overall models (k = 11) showed a high level of stability. The stability decreased only slightly when analyses were performed separately for between-group and within-subject comparisons. The stability of the men's data exceeded the stability of the women's data in the adjusted model.
Finally, we examined the effect of using a weighted (as opposed to an unweighted) average of the adjustment studies that focused the analysis on at-risk sex partners. The meta-analytic findings were virtually identical when using these 2 methods (absolute difference of 1% in the estimated ESs).
Our meta-analysis shows that the prevalence of high-risk sexual behavior is markedly lower in HIV+ persons aware of their seropositive status than in HIV+ persons unaware of their status. The findings of studies conducted in the United States from 1988 through 2003 were highly consistent despite methodologic differences among the investigations. A highly similar picture emerges from studies that examined STD acquisition rates between HIV-infected aware and unaware persons.27-30 These studies were not included in the meta-analysis because of the qualitatively different nature of the outcome variable and the fact that the partner serostatus adjustment could not validly be applied to those studies. Of persons presenting at an STD clinic in New Orleans (1989-1991) with a first-time diagnosis of gonorrhea, new gonorrhea diagnoses were 50% lower during 2 to 3 years of follow-up in persons aware relative to unaware that they were HIV+.27 In an STD clinic in Miami in 1988 through 1989, the percentage with a newly diagnosed STD declined 11% in the 6 months after they learned they had HIV infection compared with the 6 months before their HIV+ diagnosis.28 STD studies conducted in other countries show even stronger findings. Of HIV+ persons presenting at a genitourinary clinic in London in 1994, the prevalence of diagnosed STDs was 73% lower in the year after HIV diagnosis than in the year before.29 Finally, of HIV+ women presenting at prenatal and pediatric clinics in Rwanda (1988-1989), the prevalence of clinically diagnosed gonorrhea was 54% lower in the year after than the year before HIV diagnosis.30
In our analysis, the between-group and within-subject comparisons demonstrated significant reductions in self-reported UAV, although the reduction was higher in the between-group design. One explanation is that a group of HIV+ aware persons may differ from a group of HIV+ unaware persons on other variables associated with unsafe sex, potentially accentuating the difference between the groups. For example, on average, people who are unaware they are infected with HIV are likely to be younger than people who are aware they are HIV+.31 The prevalence of unsafe sex has been found to be higher among younger than older MSM. 31,32
Another issue concerns whether behavior change in HIV+ aware persons is maintained over time. Additional analysis of the SHAS phase 2 data set showed that the prevalence of UAV in the most recent encounter with an at-risk partner was remarkably similar among subgroups of HIV+ aware men who differed in length of time they knew they were HIV+ (range: 1-24 months [17%], 25-48 months [16%], 49-72 months [17%], 73-96 months [14%], and >96 months [13%]). The prevalence was 39% in HIV+ unaware men. A highly similar pattern was seen in HIV+ aware women in SHAS phase 2 and in men and women in SHAS phase 1. One other study not included in the meta-analysis because it assessed sexual behavior of HIV+ aware persons but not HIV+ unaware persons did not confirm this pattern. The prevalence of UAV was higher in persons aware of their infection for 5 or more years compared with those diagnosed more recently.33 These mixed results make it difficult to reach conclusions at this time about the stability of behavior change after being diagnosed HIV+. This issue merits attention in future research.
Our meta-analytic findings must be viewed within the context of the methodologic limitations of the primary studies. First, the studies used self-reported sexual behavior. Self-reports are open to socially desirable responding, and some HIV+ aware persons may underreport unprotected sex with at-risk partners.34 US studies of HIV serodiscordant homosexual35 and heterosexual36 couples demonstrate high levels of agreement (approximately 90%) between couple members in self-reports of condom use, however. We were not able to gauge the level of partner agreement in the studies analyzed here. Second, we were not able to examine the number of sex partners placed at risk by HIV+ aware and HIV+ unaware persons, because those data were not available in the literature. The difference between these groups in the prevalence of UAV may not necessarily reflect the magnitude of difference in the number of sex partners placed at risk. Our analysis, however, provides a starting point for a more refined model when additional data become available. Third, the difference in the prevalence of UAV between HIV+ aware and HIV+ unaware persons may not reflect differences in HIV transmission rates between groups. Actual transmission depends on a host of biologic factors. For example, transmission risk is increased when an individual who is the source of exposure is in the primary HIV infection stage,37 has a high viral load,3,4 or has an STD.38,39 These factors would elevate transmission risk more from HIV+ unaware than HIV+ aware persons, because unaware persons as a group are more likely to be in the primary infection stage, to a have higher viral load because they are not in medical care, and to have an STD.27-30
Our findings reinforce the need for a multidimensional approach to HIV prevention.7 Resources and efforts are needed to make HIV testing opportunities more accessible (eg, rapid tests) and to reduce barriers to testing so that infected persons learn their status. Public health campaigns targeting young MSM, especially young MSM of color, are urgently needed, because many of these men are unaware they are infected.40 Promising but underused methods for reaching such persons include offering HIV testing routinely in all health care settings in high HIV prevalence areas7; offering testing at venues that attract high-risk persons; adding HIV testing capacity to all effective educational outreach and risk reduction interventions41; and gaining the cooperation of current HIV+ aware persons to reach members of their sexual and social networks for HIV testing, counseling, and care if needed.42
Clearly, HIV counseling and testing alone are not enough to control the HIV epidemic. Behavioral interventions for people aware they are infected and for those at high risk for HIV are needed. Those interventions may reduce sexual risk behavior by as much as 30% to 40%.43-45 For those aware that they are HIV+, the challenge is to find settings and approaches for delivering prevention programs to this population over time. The HIV clinic is an ideal setting for offering prevention messages and counseling to HIV+ persons and for integrating prevention with routine medical care.46 Such counseling from HIV providers has been shown to be efficacious in reducing unprotected intercourse in HIV+ patients.44 Other promising interventions have been delivered by HIV+ peers in community settings.47 Assisting HIV+ people to establish social networks that encourage risk reduction and provide social support for seeking medical care and adhering to treatment regimens has also shown promise in demonstration projects.48 Together, these approaches may contribute to more rapid control of the HIV epidemic in the United States and elsewhere.
The authors thank Paul Denning, Michael Campsmith, and Glenn Nakamura for providing data from the SHAS; the Executive Committee of the HERS for permission to use HERS data; Merle Hamburger for assistance in constructing the HERS seroconversion data set; the Executive Committee of the MACS for permission to use MACS data; Mary Elizabeth Gore and Lisa Jacobson for assistance in constructing the MACS variables used in the analysis; Ramses Sadek for statistical consultation; and Ida Onorato for helpful comments on several substantive issues.
1. Fleming PL, Byers RH, Sweeney PA, et al. HIV prevalence in the United States, 2000. Presented at: 9th Conference on Retroviruses and Opportunistic Infections; 2002; Seattle.
2. Centers for Disease Control and Prevention. Advancing HIV prevention: new strategies for a changing epidemic-United States, 2003. MMWR Morb Mortal Wkly Rep
3. Quinn TC, Wawer MJ, Sewankambo D, et al. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. N Engl J Med
4. Pedraza MA, del Romero J, Roldan F, et al. Heterosexual transmission of HIV-1 associated with high plasma viral load levels and a positive viral isolation in the infected partner. J Acquir Immune Defic Syndr
5. Vernazza PL, Gilliam BL, Flepp M, et al. Effect of antiviral treatment on the shedding of HIV-1 in semen. AIDS
6. Vernazza PL, Troiani L, Flepp MJ, et al. Potent antiretroviral treatment of HIV-infection results in suppression of the seminal shedding of HIV. AIDS
7. Janssen RS, Holtgrave DR, Valdiserri RO, et al. The Serostatus Approach to Fighting the HIV Epidemic: prevention strategies for infected individuals. Am J Public Health
8. Cleary PD, Devanter NV, Rogers TF, et al. Behavior changes after notification of HIV infection. Am J Public Health
9. McCusker J, Stoddard AM, Mayer KH, et al. Effects of HIV antibody test knowledge on subsequent sexual behaviors in a cohort of homosexually active men. Am J Public Health
10. Valleroy LA, MacKellar DA, Karon JM, et al. HIV prevalence and associated risks in young men who have sex with men. JAMA
11. Colfax GN, Buchbinder SP, Cornelisse PGA, et al. Sexual risk behaviors and implications for secondary HIV transmission
during and after HIV seroconversion. AIDS
12. Saah AJ, Hoover DR, Weng S, et al. Association of HLA profiles with early plasma viral load, CD4+ cell count and rate of progression to AIDS following acute HIV-1 infection. Multicenter AIDS Cohort Study. AIDS
13. Smith DK, Warren DL, Vlahov D, et al. Design and baseline participant characteristics of the Human Immunodeficiency Virus Epidemiology Research (HER) Study: a prospective cohort study of human immunodeficiency virus infection in US women. Am J Epidemiol
14. Buehler JW, Diaz T, Hersh BS, et al. The Supplement to HIV/AIDS
Surveillance Project: an approach for monitoring HIV risk behaviors. Public Health Rep
15. Centers for Disease Control and Prevention. Late versus early testing of HIV-16 sites, United States, 2000-2003. MMWR Morb Mortal Wkly Rep
16. Vittinghoff E, Douglas J, Judson F, et al. Per-contact risk of human immunodeficiency virus transmission between male sexual partners. Am J Epidemiol
17. Cooper H, Hedges LV. The Handbook of Research Synthesis
. New York: Russell Sage Foundation; 1994.
18. Lipsey MW, Wilson DB. Practical Meta-Analysis
. Thousand Oaks: Sage; 2001.
19. Hedges LV, Vevea JL. Fixed and random effects models in meta-analysis
. Psychol Methods
20. Marks G, Cantero PJ, Simoni JM. Is acculturation associated with sexual risk behaviours? An investigation of HIV-positive Latino men and women. AIDS Care
21. Posner SF, Marks G. Prevalence of high-risk sex among HIV-positive gay and bisexual men: a longitudinal analysis. Am J Prev Med
22. Kalichman SC. Psychological and social correlates of high-risk sexual behaviour among men and women living with HIV/AIDS
. AIDS Care
23. Wolitski RJ. Dataset: Seropositive Urban Men's Intervention Study (SUMIT). Atlanta: Centers for Disease Control and Prevention; 2000.
24. Marks G. Dataset: HIV attributions and unsafe sex in seropositive men. Los Angeles, CA: University of Southern California; 1996.
25. Marks G. Dataset: drug use and unsafe sex in HIV+ men and women. Los Angeles, CA: University of Southern California; 1995.
26. Mansergh G, Colfax GN, Marks G, et al. Dataset: HIV prevention messages and sexual risk behavior in men who have sex with men. San Francisco: San Francisco Public Health Department and Centers for Disease Control and Prevention; 2001.
27. Chamot E, Couglin SS, Farley TA, et al. Gonorrhea incidence and HIV testing and counseling among adolescents and young adults seen at a clinic for sexually transmitted diseases. AIDS
28. Otten MW, Zaidi AA, Wroten JE, et al. Changes in sexually transmitted disease rates after HIV testing and posttest counseling, Miami, 1988 to 1989. Am J Public Health
29. George N, Green J, Murphy S. Sexually transmitted disease rates before and after HIV testing. Int J STD AIDS
30. Allen S, Serufilira A, Bogaerts J, et al. Confidential HIV testing and condom promotion in Africa. JAMA
31. Mansergh G, Marks G. Age and risk of HIV infection in men who have sex with men. AIDS
32. Crepaz N, Marks G, Mansergh G, et al. Age-related risk for HIV infection in men who have sex with men: examination of behavioral, relationship, and serostatus variables. AIDS Educ Prev
33. McGowan JP, Shah SS, Ganea CE, et al. Risk behavior for transmission of human immunodeficiency virus (HIV) among HIV-seropositive individuals in an urban setting. Clin Infect Dis
34. Catania JA, Gibson DR, Chitwood DD, et al. Methodological problems in AIDS behavioral research: influences on measurement error and participation bias in studies on sexual behavior. Psychol Bull
35. Seage GR III, Mayer KH, Horsburgh CR, et al. Corroboration of sexual histories among male homosexual couples. Am J Epidemiol
36. Padian NS. Sexual histories of heterosexual couples with one HIV-infected partner. Am J Public Health
37. Koopman JS, Jacquez JA, Welch GW, et al. The role of early HIV infection in the spread of HIV through populations. J Acquir Immune Defic Syndr Hum Retrovirol
38. Balzquez MV, Madueno JA, Jurado R, et al. Human herpesvirus-6 and the course of human immunodeficiency virus infection. J Acquir Immune Defic Syndr Hum Retrovirol
39. Fleming DT, Wasserheit JN. From epidemiological synergy to public health policy and practice: the contribution of other sexually transmitted diseases to sexual transmission of HIV infection. Sex Transm Infect
40. MacKellar DA, Valleroy LA, Secura GM, et al. Unrecognized HIV infection, risk behaviors, and misperceptions of risk among young men who have sex with men-6 United States cities, 1994-2000 [abstract MoPeC3427]. Presented at: 14th International AIDS Conference; 2002; Barcelona.
41. Academy for Educational Development. Best Practices for Prevention With Positives
. Conference Proceedings
. Washington, DC; 2001.
42. Jordan WC, Tolbert L, Smith R. Partner notification and focused intervention as a means of identifying HIV-positive patients. J Natl Med Assoc
43. Centers for Disease Control and Prevention. HIV/AIDS
Prevention Research Synthesis Project. Compendium of HIV Prevention Interventions with Evidence of Effectiveness
. Atlanta: Centers for Disease Control and Prevention; 1999.
44. Richardson JL, Milam J, McCutchan A, et al. Effect of brief provider safer-sex counseling of HIV-1 seropositive patients: a multi-clinic assessment. AIDS
45. Kalichman SC, Rompa D, Cage M, et al. Effectiveness of an intervention to reduce HIV transmission
risks in HIV-positive people. Am J Prev Med
46. Marks G, Burris S, Peterman TA. Reducing sexual transmission of HIV from those who know they are infected: the need for personal and collective responsibility. AIDS
47. Collins C, Morin S, Shriver M, et al. Designing Primary Prevention for People Living with HIV
. San Francisco: University of California at San Francisco; 2000.
48. Senterfitt JW, Dooley W, Abdul-Quader AS. Developing comprehensive prevention services for persons living with HIV/AIDS
: the Prevention for HIV-Infected Persons Project (PHIPP) [abstract 635, Track D]. Presented at: National HIV Prevention Conference; 2001; Atlanta.
APPENDIX: ADJUSTMENT FOR NONINDEPENDENCE OF RESPONSE IN THE WITHIN-SUBJECT STUDIES
We had access to individual level data from a within-subject study of MSM (MACS) and a within-subject study of women (HERS) and calculated the covariance terms for those investigations. The other within-subject studies did not provide individual level data to calculate the before and after correlations, so we imputed values based on the MACS (r = 0.49) and HERS (r = 0.20). The MACS value was used for the subgroup of men in the studies by Cleary et al8 and Colfax et al.11 The HERS value was used for the subgroup of women in the study by Cleary et al.8
We performed a sensitivity analysis to examine how much the aggregated PRs and CIs changed when these correlation values were increased or decreased by 0.20 correlation units of the original imputed values (ie, the MACS value was changed by ±0.20, the HERS value was changed by ±0.20). This ±0.20 represents a reasonable boundary for capturing the correlation between before and after behavior in the population. The aggregated PR changed half of 1% or less in each direction, and the width of the CIs increased or decreased only 4 units (percentage points) from the original. Thus, the imputation did not unduly affect the results.
Keywords:© 2005 Lippincott Williams & Wilkins, Inc.
HIV-positive persons; HIV/AIDS; high-risk sexual behavior; HIV transmission; meta-analysis