An estimated one quarter of the 1 039 000–1 185 000 persons in the United States infected with HIV do not know they are infected . Failure either to seek HIV testing or to receive test results hampers early diagnosis and thus overall HIV prevention and treatment efforts in the United States. New initiatives from the Centers for Disease Control and Prevention (CDC), such as the Serostatus Approach to Fighting the Epidemic  and Advancing HIV Prevention , emphasize the need to increase HIV testing and knowledge of HIV status to link HIV-infected persons to treatment.
The conventional method for HIV counseling and testing (HIV-CT) involves pre-test counseling, phlebotomy, submission of a serum specimen to a central laboratory for screening with enzyme-linked immunoassay (EIA), and confirmation of repeatedly reactive EIA results with Western blot or immunofluorescence assay. This approach requires that a person returns to the testing site, typically 1–2 weeks later, for test results and post-test counseling; however, this frequently does not occur. Up to half of persons tested at publicly funded clinics do not return to receive their HIV test results and some may be less likely to accept testing if a return visit is required [4,5].
Alternative testing methods have the potential to mitigate some of these problems for HIV testing because they are less invasive or otherwise eliminate the need for clinic visits. Less-invasive testing methods include urine and oral fluid testing. Providing results of conventional HIV tests by telephone can also eliminate the need for a return visit . A home collection kit for HIV testing allows individuals to mail a dried blood spot collected from a finger prick to a testing laboratory and then receive results by telephone .
Rapid HIV screening tests provide immediate results and eliminate the need for a return visit by persons who test negative. Persons with reactive rapid tests receive preliminary positive results but must undergo confirmatory testing. Although rapid tests have been available for more than 10 years, they have been seldom used in the United States because the 1989 guidelines from the US Public Health Service (PHS) required confirmatory testing before reporting any positive HIV test results to minimize the chance of reporting false-positive results . In 1998, this recommendation was changed to encourage provision of preliminary positive test results before confirmatory results are available to help to increase the number of persons who learn their HIV test results . The revised PHS recommendation was based on a decision model that compared receipt of test results with conventional HIV-CT and that with a commercially available rapid test [Single Use Diagnostic System for HIV-1 (SUDS), Abbott-Murex, Norcross, Georgia, USA]. The model used the number of tests performed at publicly funded testing sites in 1995 to calculate the number of persons who would learn their HIV status, and the number who would receive false-positive results, under each strategy .
Although evaluations of alternative HIV-CT methods have emerged in the past several years, there has been no comprehensive assessment of the relative effectiveness of receipt of HIV test results with the different approaches. Two published meta-analyses of HIV-CT interventions focused on the reduction in HIV risk behaviors but did not evaluate testing acceptance or receipt of HIV test results [10,11]. The present study is a meta-analysis to assess the effectiveness of alternative HIV-CT methods compared with conventional serum testing on receipt of HIV test results. The study also analyzed the effects on rates of HIV testing, receipt of confirmatory results and linkage to medical care.
The HIV-CT literature was searched systematically from March 1990 to May 2005 to identify articles for review, including the following databases: Medline, Embase, and AIDSline. Abstracts were also reviewed from the International AIDS Conferences (2000, 2002, and 2004) and the CDC's National HIV Prevention Conferences (2003 and 2005); references in retrieved articles were examined and experts in the field were contacted. For studies that did not contain enough information for effect size computation, authors were contacted for additional information. Studies that involved mandatory HIV-CT and those conducted in perinatal or occupational exposure settings were excluded. The following inclusion criteria were used: (i) primary studies of voluntary HIV-CT interventions conducted in the United States that reported receipt of HIV test results and had a control or comparison group; (ii) treatment groups that included rapid, oral fluid, urine or home testing as an alternative to serum EIA testing, or another alternative that eliminated a return visit for test results; and (iii) published English language articles, abstracts, and unpublished studies that provided sufficient detail for effect size computation.
Using a data abstraction form developed for this study, two independent reviewers systematically coded outcome variables, which included study design, population description, treatment and control groups, sample size, study setting, and additional items identified during the review. Disagreements were resolved by discussion and consensus. If an article reported data separately for different populations or study settings, they were considered as separate, independent studies.
Receipt of HIV test results and acceptance of testing were used as outcome measures. For rapid testing studies, receipt of preliminary positive rapid test results were considered as receipt of test results. The effect sizes were calculated for the individual studies and the meta-analysis examined the overall effect; subgroup analyses identified effect modifiers and sources of heterogeneity (between-study differences not attributable to sampling error variance alone). The effect size measure was the risk ratio (RR) . Values > 1 indicated that clients were more likely to receive their test results or to undergo HIV testing with the alternative method than with conventional HIV-CT. If the event rate contained a value of zero, 0.5 was added to each cell so that the RR could be calculated .
The DerSimonian and Laird random-effects model was used, indicated when heterogeneity was suspected, to conduct the meta-analysis . The Q statistic was used to test for homogeneity and the tau (τ2) statistic to measure the magnitude of between-study heterogeneity. A Begg's funnel plot was used to evaluate publication bias, the tendency for studies with non-significant results not to be published . STATA 8.0 (StataCorp LP; College Station, Texas, USA) was used for data analysis.
The initial search generated 3151 citations. Most were not HIV-CT interventions (e.g., assessments of HIV test performance) or did not include the appropriate outcome measures. Six studies were excluded because they had insufficient data to compute effect sizes. Thirteen articles or abstracts [16–28] met the inclusion criteria with a combined sample size of 21 096 (Table 1). These 13 articles included 15 different studies, two of which had two treatment arms, so 17 effect sizes (k = 17) were included in the meta-analysis. These studies encompassed interventions conducted in sexually transmitted disease (STD) clinics, emergency departments, HIV testing, and outreach venues (needle exchanges, bath houses, mobile health vans). All studies were conducted in urban areas of the United States and enrolled persons with risk behaviors for HIV: men who have sex with men, injection drug users, homeless youth, and high-risk heterosexuals. Several research designs were represented, including eight randomized controlled trials and nine non-randomized cohort designs with concurrent (n = 4) or historical (n = 5) comparison groups (Table 1).
Twelve studies assessed the effectiveness of rapid testing [16–24,26]; two assessed oral fluid EIA testing (non-rapid) ; two assessed telephone post-test counseling [25,27], and one study evaluated home collection tests . Eleven of the rapid testing studies used the SUDS rapid test, no longer commercially available, which required phlebotomy for a serum specimen and 30–120 min to perform. One study  used the Oraquick HIV-1 Rapid Antibody Test (Orasure Technologies, Bethlehem, Pennsylvania, USA), which can be performed with whole blood specimen and generates results in approximately 20 min. All but one study had a control group of conventional serum EIA. The remaining study compared delivery of test results by telephone and in person, both the treatment and control groups undergoing oral fluid testing . Three studies that compared rapid, oral fluid, and conventional CT used telephone post-test counseling in both the treatment and control groups because it was the standard of care in the study sites: a STD clinic (study A), a bath house (study B), and a needle exchange (study C) . The two outreach studies (B and C) compared rapid and oral fluid testing with conventional serum EIA. To ensure that the control group was not over-represented in the pooled effect size, the pooled meta-analysis was also run with the study weights of studies B and C adjusted downward by 50%. Another study had two different sites, an anonymous testing clinic (study A) and a STD clinic (study B) .
Meta-analysis of the effect of HIV testing methods on receipt of HIV test results
In the pooled analysis that included all the alternative testing methods (n = 21 096), clients were more likely to receive their HIV test results with alternative methods compared with conventional HIV-CT [RR, 1.61; 95% confidence interval (CI), 1.36–1.90; P < 0.0001) (Table 2). There was also evidence of heterogeneity (Q = 1429; P < 0.0001) among the studies. Adjustments to the study weights (studies B and C of Spielberg et al.) did not change the pooled effect size. The forest plot (Fig. 1) provides a visual representation of the individual and pooled effect sizes ordered from least to most effective, and the study weights. Effect sizes for receipt of results ranged from a value of RR of 0.98 for oral fluid testing to 4.20 for rapid testing. The effect sizes were much larger for rapid testing (RR, 1.80; 95% CI, 1.46–2.22; P < 0.0001) than for non-rapid methods (Fig. 1 and Table 2). The next most effective method was the use of the telephone to deliver test results (RR, 1.38; 95% CI, 1.24–1.47; P < 0.0001) followed by home collection testing (RR, 1.28; 95% CI, 1.07–1.53; P < 0.01). Receipt of test results were similar for oral fluid testing and conventional HIV-CT. Examination of a Begg's funnel plot showed no evidence of publication bias (not shown).
Subgroup analysis of rapid testing studies
Rapid testing was the most evaluated (n = 17 520; 12 studies) and most effective alternative. Rapid testing significantly increased receipt of test results in all settings evaluated, with the largest effect sizes in emergency departments (RR, 2.19; 95% CI, 1.29–3.70; P < 0.01) followed by STD clinics (RR, 1.82; 95% CI, 1.30–2.54; P < 0.01) and outreach settings (RR, 1.54; 95% CI, 1.21–1.96; P < 0.01) (Table 2).
The increase in receipt of results was greater for HIV-negative clients (RR, 2.00; 95% CI, 1.19–3.36; P < 0.05) than for HIV-positive clients (RR, 1.19; 95% CI, 1.08–1.31; P < 0.0001). There was no heterogeneity in the HIV-positive subgroup analyses, indicating that the effectiveness of rapid testing for HIV-positive persons was similar across studies.
Study design moderated the effect size in the subgroup analysis of rapid testing. The point estimates of the effect sizes were larger for the non-randomized trials (RR, 2.04; 95% CI, 1.39–2.99; P < 0.0001), which were heterogeneous, than for the randomized controlled trials (RR, 1.43; 95% CI, 1.39–1.47; P < 0.0001), which were homogeneous. The studies with concurrent comparison groups showed less of an effect (RR, 1.53; 95% CI, 1.35–1.73; P < 0.0001) than the studies with historical comparison groups (RR, 2.51; 95% CI, 1.21–5.19; P < 0.05).
Additional outcomes: test acceptance, linkage to care, confirmatory and false-positive rapid test results
One randomized controlled trial evaluated both test acceptance and receipt of HIV test results to determine the net effect of the alternatives on knowledge of HIV status . Persons at the needle exchange (RR, 2.31; 95% CI, 1.68–3.19) and bath houses (RR, 1.40; 95% CI, 1.10–1.77) were more likely to accept oral fluid than conventional serum testing. Acceptance was also greater with rapid testing than conventional HIV-CT at the needle exchange (RR, 1.31; 95% CI, 1.02–1.68) and bath houses (RR, 1.55; 95% CI, 1.09–2.19), but not at the STD clinic (RR, 1.05.95–1.15). At the needle exchange, a higher percentage of persons accepted oral fluid testing than serum rapid testing, and because of this higher acceptance, more persons learned their HIV status with oral fluid testing even though the percentage who received their test results with rapid testing (82%) was higher than with oral fluid testing (61%).
Three studies, all conducted in emergency departments, reported outcomes for linkage into medical care (e.g., first clinic visit) for HIV-infected clients. Because entry into care was not defined similarly across studies, this was not evaluated using a meta-analytic framework. However, two of the three studies found that clients were more likely to enter into care when tested with rapid testing than with conventional testing [16,18–20]. Six rapid testing studies documented receipt of confirmatory results, with an average of 75% (range, 0–100) receiving confirmatory results [16–18,20,21,22]. However, one outreach study conducted in a mobile van required those positive by the rapid test to return to an STD clinic to receive their test results; none of them did . Receipt of confirmatory HIV test results after a preliminary positive rapid test result averaged 90% in the other five studies. Additionally, 11 (0.6%) false-positive rapid test results occurred among 1839 persons tested in the two studies that reported this outcome [18,21].
This meta-analysis clearly demonstrates that alternative methods for HIV-CT, particularly rapid testing, lead to substantial increases in receipt of test results. The most compelling finding is the magnitude of the effect of rapid testing. Compared with conventional HIV-CT, clients were 1.5–2.2 times more likely to receive results with rapid testing. The pooled RR values were large and statistically significant even when non-randomized trials and studies with historical comparison groups (which showed larger effect sizes but more heterogeneity) were excluded. This meta-analysis confirms that the anticipated benefit from revising the PHS recommendation to permit delivery of preliminary positive rapid test results has been realized. Rapid testing substantially increases receipt of both HIV-positive and HIV-negative test results, and the rate of false-positive test results (< 1%) was low.
Although rapid testing appeared to be more effective at increasing receipt of results among persons who tested negative than those who tested positive, this finding may be confounded because of the extensive outreach efforts that are usually made to communicate HIV-positive (but not HIV-negative) test results to persons who fail to return after conventional testing. In the one study that examined this differential , Kassler and colleagues  documented that 79% of the patients at an STD clinic received their conventional HIV-positive test results but that only 45% returned on their own; 34% received results only after outreach, and 21% never learned their HIV-positive results despite efforts to contact them. By contrast, 94% of the patients at an STD clinic who received a preliminary positive rapid test result returned on their own to receive their confirmatory test results and only 3% required outreach. This suggests that rapid testing is particularly likely to increase receipt of HIV-positive test results, with less additional effort required to contact persons who fail to return for their test results.
The alternatives to conventional HIV-CT that reduced the need for a return visit were most effective at increasing receipt of results. After rapid testing, telephone post-test counseling and home testing were the next most effective methods. Although acceptance of testing was greater when oral fluid (instead of blood) was collected for EIA, receipt of oral fluid test results was similar to that with other methods that require a return visit for test results. Consistent with this demonstrated effectiveness and the CDC's recommendation to make HIV testing a routine part of medical care , healthcare providers should provide HIV test results in the same manner used for other diagnostic and screening tests (e.g., by telephone or by ancillary personnel) and not require a return visit to receive test results.
There are as yet no data on changes in risk behavior after receipt of a preliminary positive HIV test result. It is clear that persons substantially reduce HIV risk behavior after learning that they are HIV positive. A recent meta-analysis found a 68% reduction in high-risk behavior among HIV-infected persons aware of their status compared with HIV-infected persons who were unaware of their status . When individuals are given a preliminary positive rapid test result, they are counseled to avoid activities that might transmit HIV until additional tests confirm whether they are HIV infected. Hopefully, persons who receive such information are more likely to adopt safer behaviors during the interim period between testing and receipt of confirmed results than are conventionally tested persons who receive no preliminary information about their HIV status.
Only two studies documented the incidence of false-positive rapid test results, and neither described any adverse consequences. In future studies, it will be important to document both the frequency of false-positive results and how these affect patients tested with rapid tests.
This analysis likely underestimates the effect sizes that might be expected with newer rapid HIV tests because most of these studies used the SUDS rapid test, which required phlebotomy and processing of a serum specimen in a laboratory. Since 2003, the US Food and Drug Administration has approved four additional rapid HIV tests, some of which use finger prick whole blood or oral fluid specimens and can be performed in 10–20 min. Some of these newer, easy-to-use rapid tests have been waived under the Clinical Laboratory Improvement Amendments of 1988 and are suitable for use in non-clinical settings without a traditional laboratory. Our analytic framework did not permit us to quantify the additional effectiveness that might result from testing at an increased number of diverse venues where conventional testing is impractical.
It would have been most desirable to evaluate a number of outcomes – testing acceptance, receipt of test results, and ultimately, linkage to care and preventive services – to assess the net effect of various alternatives to HIV-CT on important health outcomes, but the studies in this meta-analysis did not consistently report such data. It will be important for future studies to report sufficient data to evaluate these important outcomes.
Our study is subject to several limitations. The pooled estimate of all alternative methods for HIV-CT may be difficult to interpret because it includes several conceptually different variations on testing. However, the subgroup analysis provides separate estimates of the different constructs of alternative tests (speed of results, sample collection, and the need to return for results). There were also some differences in how the authors measured the outcome variable, receipt of HIV test results, and differences in efforts to encourage HIV-positive and HIV-negative clients to return for oral fluid and conventional test results. Many studies reported that HIV-infected clients who did not return were called by telephone, and one study provided an economic incentive for returning for HIV test results .
This meta-analysis demonstrates how study design can affect treatment effects. In our subgroup analyses, effect sizes were larger in studies with non-randomized than with randomized designs, and in studies that used historical rather than concurrent, control groups. This finding is consistent with a previous analysis, which also found that studies with historical controls were more susceptible to bias than prospective designs . Accordingly, the effect size for studies with concurrent controls (RR, 1.53) may most closely approximate the true treatment effect.
Heterogeneity was evident and significant in the pooled analyses and most of the subgroup analyses, indicating that the relationship between HIV-CT and receipt of HIV test result is complex and may be determined by multiple factors. Because the studies differed in setting, HIV risk group, and geographic location, the variability may actually represent the true treatment effects under different circumstances.
Despite these limitations, we find clear evidence that individuals are substantially more likely to receive their HIV test results with rapid testing and with other methods that eliminate the need for a return visit to the testing center.
1. Glynn M, Rhodes P. Estimated HIV prevalence in the United States at the end of 2003
. National HIV Prevention Conference
. Atlanta, June, 2005 [abstract T1-B1101].
2. Janssen RS, Holtgrave DR, Valdiserri RO, Shepherd M, Gayle HD, de Cock KM. The serostatus approach to fighting the HIV
epidemic: prevention strategies for infected individuals. Am J Public Health 2001; 91:1019–1024.
3. Centers for Disease Control and Prevention. Advancing HIV Prevention: new strategies for a changing epidemic, 2003
. MMWR Morb Mortal Wkly Rep
4. Sullivan PS, Lansky A, Drake A. Failure to return for HIV
test results among persons at high risk for HIV
infection. Results from a multistate interview project. J Acquir Immune Defic Syndr 2004; 35:511–518.
5. Valdiserri RO, Moore M, Gerber AR, Campbell CH, Dillon BA, West GR. A study of clients returning for counseling after HIV
testing: implications for improving rates of return. Public Health Reports 1993; 108:12–18.
6. Lyons MS, Lindsell CJ, Ledyard HK, Frame PT, Trott AT. Emergency department HIV
testing and counseling: an ongoing experience in a low-prevalence area. Ann Emerg Med 2005; 46:22–28.
7. Branson B. Home sample collection tests for HIV
infection. JAMA 1999; 280:1699–1701.
8. Centers for Disease Control and Prevention. Interpretation and use of the Western blot assay for serodiagnosis of human immunodeficiency virus type 1 infection
. MMWR Morb Mortal Wkly Rep
1989; 38(Suppl 7)
9. Centers for Disease Control and Prevention. Update: HIV counseling and testing using rapid tests – United States, 1995
. MMWR Morb Mortal Wkly Rep
10. Weinhardt LA, Carey MP, Johnson BT, Bickman NL. Effects of HIV
counseling and testing on sexual risk behavior: a meta-analytic review of published research, 1985–1997. Am J Public Health 1999; 89:1397–1405.
11. Higgins DL, Galavotti C, O'Reilly KR, Schnell DJ, Moore M, Rugg DL, et al
. Evidence for the effects of HIV
antibody counseling and testing on risk behaviors. JAMA 1991; 266:2419–2429.
12. Zang J, Yu K. What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998; 280:1690–1691.
13. Cooper H, Hedges LV (eds) (1994). The Handbook of Research Synthesis.
New York: Russell Sage Foundation; 1994.
14. DerSimonian R, Laird N. Meta-analysis
in clinical trials. Control Clin Trials 1986; 7:177–188.
15. Egger MG, Smith GD, Schneider M, Minder CE. Bias in meta-analysis
detected by a simple, graphical test. Br Med J 1997; 315:629–634.
16. del Rio C, Paredes CF, Duffus W, Cesarz K, Green S, Hicks G, et al
. Routinely recommended HIV
testing at an urban urgent care clinic: Atlanta, Georgia. MMWR Morb Mortal Wkly Rep 2001; 50:538–541.
17. Kassler WJ, Dillon BA, Haley C, Jones WK, Goldman A. On-site, rapid HIV
testing with same-day results and counseling. AIDS 1997; 11:1045–1051.
18. Kelen GD, Shahan JB, Quinn TC for the Project Educate Work Group. Emergency department-based HIV
screening and counseling: experience with rapid and standard serologic testing. Ann Emerg Med 1999; 33:147–155.
19. Kendrick SR, Kroc KA, Couture E, Weinstein RA. Comparison of point-of-care rapid HIV
testing in three clinical venues. AIDS 2004; 18:2208–2210.
20. Kroc K, Kendrick S, Withum D, Couture E, Miller S, Zagorski B, et al.Rapid HIV testing in an emergency department.National STD Prevention Conference
. San Diego, March 2002 [abstract C6F].
21. Kendrick SR, Kroc KA, Withum D, Rydman RJ, Branson BM, Weinstein RA. Outcomes of offering rapid point-of-care testing in a sexually transmitted disease clinic. J Acquir Immune Defic Syndr 2005; 38:142–146.
22. Liang TS, Erbelding E, Jacob CA, Wicker H, Christmyer C, Brunson S, et al
. Rapid HIV
testing of clients of a mobile STD/HIV
clinic. AIDS Patient Care and STD 2005; 19:253–257.
23. Marmour M, Titus S, Hagaerty R, Harrison C. Rapid versus standard HIV testing of persons at high risk of HIV infection.National HIV Prevention Conference
. Atlanta, August–September 1999 [abstract 692].
24. Metcalf CA, Douglas JM, Malotte K, the RESPECT-2 Study Group. Relative efficacy of prevention counseling with rapid and standard HIV
testing: a randomized controlled trial (RESPECT-2). Sex Transm Dis 2005; 32:130–138.
25. Schulter WW, Judson F, Barone AE, McGill WL, Marine WM, Douglas JM. Usefulness of human immunodeficiency virus post-test counseling by telephone for low-risk clients of an urban sexually transmitted diseases clinic. Sex Transm Dis 1996; 23:190–197.
26. Speilberg F, Branson BM, Goldbaum GM, Lockhart D, Kurth A, Rossini A, et al
. Choosing HIV
counseling and testing strategies for outreach settings: a randomized trial. J Acquir Immune Defic Syndr 2005; 38:348–355.
27. Tsu RC, Burm ML, Gilhooly JA, Sells CW. Telephone vs. face-to-face notification of HIV
results in high-risk youth. J Adolesc Health 2002; 30:154–160.
28. Wickrema R, Scotti R, Navaline H, Mark H, Fiore D, Davis-Vogel A, et al.Randomized trial comparing return rates for HIV test results using Homedash collection test kits and Clinicdash based testing.XIV International Conference on AIDS
. Barcelona, July 2002 [abstract ThPeF8057].
29. Marks G, Crepaz N, Senterfitt JW, Janssen RS. Meta-analysis
of high-risk sexual behavior in persons aware and unaware they are infected with HIV
in the United States. Implications for HIV
prevention programs. J Acquir Immune Defic Syndr 2005; 39:446–453.
30. Ioannidis JP, Haidich AB, Pappa M, Pantazis N, Kokori SI, Tektonidou MG, et al
. Comparison of evidence of treatment effects in randomized and nonrandomized studies. JAMA 2001; 286:821–830. Note: The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.