Detection of Acute HIV Infections in an Urban HIV Counseling and Testing Population in the United States

Abstract

Summary: 

The southeastern United States has an increasing burden of HIV, particularly among blacks, women, and men who have sex with men. To evaluate HIV nucleic acid amplification testing (NAAT) and antibody-based algorithms in determination of HIV incidence, detection of acute HIV infections, and surveillance of drug-resistant virus transmission in the urban southeastern United States, we conducted a cross-sectional analysis of prospectively collected data from 2202 adults receiving HIV testing and counseling at 3 sites in Atlanta, GA from October 2002 through January 2004. After standard testing with an HIV enzyme immunoassay (EIA) and Western blot confirmation, HIV-positive specimens were tested with 2 standardized assays to detect recent infection. HIV antibody-negative specimens were pooled and screened for HIV using NAAT. Seventy (3.2%) of 2202 subjects were HIV infected. Only 66 were positive on the standard HIV antibody test; 4 were antibody-negative but acutely HIV infected. The overall annual HIV incidence was 1.1% (95% confidence interval [CI]: 0.4 to 1.8) based on the Vironostika-LS assay and 1.3% (95% CI: 0.6 to 2.1) based on the BED Incidence Enzyme Immunoassay (EIA). The prevalence of acute HIV infection was 1.8 per 1000 persons (95% CI: 0.7 to 4.6). The sensitivity of the current testing algorithm using an EIA and Western blot test for detectable infections was only 94.3% (95% CI: 86.2 to 97.8). All 3 of the acutely infected subjects genotyped had drug resistance mutations, and 1 had multiclass resistance. Adding NAAT-based screening to standard HIV antibody testing increased case identification by 6% and uncovered the first evidence of multidrug-resistant HIV transmission in Atlanta. Antibody tests alone are insufficient for public health practice in high-risk urban HIV testing settings.

Recent HIV prevalence data show the increasing burden of HIV in the southeastern United States, with increasing prevalence among blacks, particularly among women and men who have sex with men (MSM).¹ Georgia has the seventh highest rate of AIDS among all states and the seventh highest number of persons living with AIDS.² Similar to other southern states, Georgia's HIV epidemic is increasingly black, female, and heterosexual, with 63% of reported AIDS cases occurring in metropolitan Atlanta.² It has been difficult to monitor HIV incidence, to access patients early in infection, or to identify high-incidence populations in Atlanta accurately, however.³

Two-step HIV antibody testing algorithms using a less-sensitive enzyme immunoassay (EIA) or a capture EIA can detect recent infections, usually within the past 6 months, and have been used to estimate HIV incidence.^4-10 High rates of individual misclassification, potential confounding of results by antiretroviral therapy, and the critical failure to detect acute HIV infections can all limit the usefulness of these antibody-based algorithms for monitoring local epidemiologic trends, however.^11-13 Recently, a screening methodology for identifying acute HIV infection in HIV counseling and testing populations has been described, using nucleic acid amplification testing (NAAT) to screen antibody-negative specimens for HIV RNA.¹⁴ Although the individual benefits of identifying acute HIV infection, such as early initiation of antiretroviral therapy, remain uncertain,^15,16 a statewide trial using NAAT for public HIV testing of 109,250 individuals in North Carolina showed that detection of acute infections had important public health benefits, including increasing HIV case identification (by 4% overall), creating opportunities for secondary prevention of transmission, and providing new detail on local patterns of HIV incidence that led to detection of new risk populations.^17,18 Several studies of acute HIV infection in urban areas in the United States and United Kingdom have found increasing rates of drug-resistant HIV transmission.^19-21 Targeting secondary prevention toward acutely infected persons may help to decrease HIV incidence and prevent the spread of multidrug-resistant HIV. We wished to evaluate and compare the use of HIV NAAT and antibody-based algorithms in determination of HIV incidence, detection of acute HIV infections, and surveillance of transmission of drug-resistant virus in Atlanta, the largest city in the southeastern United States, with a significant HIV epidemic among black and MSM populations.

METHODS

Three HIV testing sites were selected based on testing volume, HIV prevalence, and client characteristics. The 3 sites serve primarily black clients from the following risk groups: MSM, sexually transmitted disease (STD) patients, and noninjection drug abusers. All 3 sites are located in the county with the highest HIV testing volume, the highest HIV prevalence among testing clients, and the highest cumulative number of AIDS cases in Georgia.²² The study was approved by the Institutional Review Boards of Emory University, the University of North Carolina, and the study sites.

After obtaining written informed consent, additional blood was collected from adult clients presenting for HIV testing at the 3 sites from November 2002 through January 2004. HIV antibody testing was performed with the sensitive second-generation Genetic Systems HIV-1 rLAV EIA (Sanofi, Paris, France). Repeatedly EIA-reactive specimens were confirmed with Western blot analysis (BioRad, Hercules, CA). Western blot-positive specimens were then tested using 2 standardized assays designed to have reduced sensitivity during recent infection, the Vironostika-LS EIA (BioMerieux, Durham, NC) and the Calypte HIV-1 BED incidence EIA (Rockville, MD). Vironostika-LS testing was conducted by the Georgia Public Health Laboratory under a US Food and Drug Administration Investigational New Device program that included a quality assurance protocol. A standardized optical density (SOD) cutoff <1.0 was interpreted as recent infection (within the past 170 days).²³ BED EIA testing was performed by the HIV Diagnostics Laboratory at the US Centers for Disease Control and Prevention. An normalized optical density (OD-n) cutoff <0.8 was interpreted as recent infection (within the past 153 days).²⁴

Incidence (I) was calculated as follows:

• I (% per year) = Number of Recent Infections/Number of HIV-1 Negative + Number of Recent Infections × 365/w × 100

where 365/w is a conversion factor that annualizes the HIV-1 incidence estimate based on the window period estimate: 170 (95% confidence interval [CI]: 145 to 200) days according to the Vironostika-LS EIA, and 153 (95% CI: 146 to 165) days according to the BED EIA.^23,24 The 95% CI for the incidence estimate is I ± 1.96I divided by the square root of the number of recent infections.¹¹ Exact methods were used to calculate the 95% CI when no HIV infections were identified. Agreement between the Vironostika-LS EIA and BED EIA assays was assessed using the kappa statistic. Differences in ethnicity/race or gender among recent and chronically infected patients were determined using the Fisher exact test. Difference in age was determined using a Wilcoxon rank sum test. All tests were 2-sided, with a significance level of 0.05.

HIV antibody-negative plasma specimens were mechanically pooled using an automated pooling instrument (Microlab AT Plus2; Hamilton, Bonaduz, Switzerland) and screened for HIV using NAAT according to a modification of a previously described pooling algorithm.¹⁴ Master pools of 48 specimens were screened using the HIV-1 nucleic acid testing (NAT) assay (Gen-Probe, Inc., San Diego, CA), a transcription-mediated amplification assay developed to screen donated blood products for viral infection. The assay has a rated lower limit of detection of HIV RNA from 30 to 60 copies/mL. Pooling 48 specimens reduces the assay sensitivity from a lower limit of 30 to 60 copies/mL to 1440 to 2880 copies/mL.

HIV RNA-positive master pools received further testing, with the Gen-Probe NAT assay on the 8-specimen intermediate pools within each master pool and on all individual specimens in each HIV RNA-positive intermediate pool. Confirmatory NAT was performed on all individual HIV RNA-positive specimens using the Roche Amplicor 1.5 reverse transcriptase polymerase chain reaction (RT-PCR; Roche Diagnostic Systems, Branchburg, NJ). Antibody-negative HIV RNA-positive specimens were retested for HIV antibody using the Genetic Systems HIV-1 and 2 plus O EIA and the Genetic Systems HIV-1 Western blot test. Genotyping was performed on HIV RNA-positive specimens to evaluate cross-contamination and resistance mutations using the Trugene HIV-1 Genotyping Test (Bayer Diagnostics, Tarrytown, NY). Acute HIV infection was defined as being negative or indeterminate for HIV antibody on screening EIA or Western blot analysis and positive for HIV RNA on both screening NAAT assays. Sensitivity of the standard antibody testing algorithm was calculated based on all detectable infections. Specificity of the standard EIA/Western blot algorithm was assumed to be 100%. CIs for acute and overall infection prevalence, sensitivity, and specificity were calculated using the Wilson adjusted method, which is appropriate for extremely low or extremely high observed proportions.²⁵

Costs of testing were estimated by multiplying the numbers of tests used by estimated per-sample testing costs provided by the laboratory. Per-samle cost estimates ($10 per EIA, $60 per WB and $60 per NAAT) reflected all lab costs including labor. All NAATs used to screen 48-specimen master pools, 8-specimen intermediate pools and individual specimens in the HIV RNA-positive intermediate pools were included to provide an overall estimate of the cost of NAAT. Cost of NAAT per patient screened was calculated by dividing the total estimated NAAT cost by the total number of patients screened. Cost per additional case identified by NAAT was calculated by dividing the total estimated NAAT cost by the numbe of new HIV infections identified by NAAT.

The overall cost of EIA/Western blot testing was calculated based on per-test costs, the number of total samples submitted for EIA testing, and the number of EIA positive samples submitted for re-testing by WB. Cost per additional case identified by EIA/Western blot was calculated as total cost divided by number of new HIV infections identified by EIA/Western blot.

Linked demographic and HIV risk data, including self-reported ethnicity/race, were obtained from the Georgia Division of Public Health HIV Counseling and Testing database. Inclusion of repeat testers at each site was minimized by flagging medical charts of prior participants and using the same study technician to enroll participants for the duration of the study. Linked clinical data for participants with acute infection were abstracted from charts at the testing sites.

RESULTS

During the study period, 77% (2202) of eligible HIV testing clients at the 3 sites were enrolled in the study. Because only clients who reported not previously testing HIV-positive were recruited into the study, participants were less likely to have previously tested HIV-positive than nonparticipants and had a lower overall HIV prevalence. Study participants did not differ meaningfully by age, race, or HIV risk behavior from nonparticipants seeking HIV testing at the study sites, however (Table 1). Of the 2212 specimens collected, 8 were insufficient for HIV testing and 2 were from individuals who reported a previous HIV-positive test. Of the remaining 2202 specimens, 66 (3.0%) were found to be HIV antibody-positive (Fig. 1).

FIGURE 1: Flow chart of testing algorithm and results.

TABLE 1: Characteristics of Study Participants Versus Nonparticipants at HIV Testing Sites

Adding Nucleic Acid Amplification Testing

Of 2135 HIV antibody-negative specimens, 8 were not available or had insufficient specimen for NAAT testing. A total of 2127 HIV antibody-negative specimens and 1 specimen indeterminate by Western blot analysis were pooled and screened for HIV RNA. Four specimens were HIV RNA-positive and met the study definition of acute HIV infection (Table 2). Two acutely infected clients receiving confidential testing seroconverted during follow-up. One client was not located. One client with an indeterminate Western blot result was instructed by the testing site to return for repeat testing in a month. Because he tested anonymously, however, tracing of follow-up testing was not possible.

TABLE 2: Clinical and Laboratory Features of True and Possible Acute HIV Infections

HIV viral loads of acute HIV infections were measured for 3 of 4 patients and ranged from 4.3 to 5.4 log₁₀ copies/mL (see Table 2). One subject was a black MSM STD clinic patient who presented with penile discharge and tested positive for gonococcal urethritis. HIV genotype showed resistance to nucleoside reverse transcriptase inhibitor (NRTI) and nonnucleoside reverse transcriptase inhibitor (NNRTI) drug classes. The second subject was a black male STD clinic patient who presented with penile discharge, inguinal lymphadenopathy, and a truncal rash and tested positive for gonococcal urethritis. The HIV genotype showed NNRTI resistance. The third subject was a white MSM who tested anonymously at the community testing center; no clinical findings were recorded, and the specimen collected was insufficient for genotyping. The fourth subject was a black woman with a history of a recent STD who tested at the drug treatment clinic; no clinical findings were recorded. HIV genotype showed NNRTI resistance. Both subjects with penile discharge were given appropriate STD treatment and encouraged to refer their sexual partners. Partner management data were not linked to the index subject, however, and thus were not available for analysis.

Acute HIV infection prevalence was 1.8 per 1000 persons (95% CI: 0.7 to 4.6). Including these infections in the total, the initial global HIV prevalence estimate of 3.0% was revised upward to 3.2% (95% CI: 2.5 to 4.0). In all, 5.7% of all detectable HIV infections were acutely infected and antibody-negative or indeterminate by Western blot analysis. Considering all detectable infections, the sensitivity of the standard second-generation recombinant-peptide EIA used to screen for HIV in this study was estimated as 66 of 70, or 94.3% (95% CI: 86.2 to 97.8). The positive predictive value (70 of 70) and overall specificity (2124 of 2124) of the EIA Western blot NAAT algorithm were 100%.

One hundred four NAAT tests were needed to screen 2212 subjects. NAAT cost $2.82 per patient screened, or $1560 per additional case identified by NAAT. By comparison, EIA and Western blot testing cost $11.79 per patient screened, or $395 per additional case identified.

Testing for Recent (nonacute) Infections

Of the 65 Western blot-positive specimens available for testing, 11 (16%) were classified as recent infections using the Vironostika-LS and 12 (18%) were classified as recent by the BED EIA (see Fig. 1). Based on the Vironostika-LS, the estimated annual HIV incidence was 3.1% at the community testing site (95% CI: 0.5 to 5.7), 0.9% at the county STD clinic (95% CI: 0.2 to 1.6), and 0.0% at the drug treatment program (95% CI: 0.0 to 2.6). The estimated annual HIV incidence using the BED EIA was 4.1% at the community testing site (95% CI: 0.8 to 7.3), 1.0% at the county STD clinic (95% CI: 0.2 to 1.8), and 0.0% at the drug treatment program (95% CI: 0.0 to 2.8) (Table 3).

TABLE 3: HIV Prevalence and Estimated Incidence at Study Sites

The 11 persons classified as recently infected by the Vironostika-LS ranged in age from 19 to 74 years (Table 4). Recently infected persons were younger than chronically infected persons (26 vs. 36 years; P = 0.05) but did not differ significantly by demographic or HIV risk groups. Recently infected persons were primarily male and black, and all came from the 3 risk groups: MSM, sex while using noninjection drugs, or STD diagnosis. Although MSM had the highest measured incidence, 4.1% per year (95% CI: 0.8 to 7.3), incidence rates were not significantly different between demographic or risk groups or study sites.

TABLE 4: Characteristics of HIV Testing Clients by HIV Status

We compared the results of the Vironostika-LS and BED EIA for Western blot-positive specimens. The BED EIA classified 2 cases as recent infections that were classified as chronic using the Vironostika-LS. The Vironostika-LS classified 1 case as recent infection that was classified as chronic using the BED EIA. There was excellent agreement between the 2 methods, and the kappa statistic was 0.85 (95% CI: 0.67 to 1.00).

DISCUSSION

Standard HIV antibody testing in Atlanta missed 6% of HIV-infected persons presenting for HIV testing at public and community-based testing sites because these patients were acutely infected and had not yet seroconverted. These findings are strikingly similar to results from a recent study in a lower HIV prevalence testing population in North Carolina and a study of San Francisco STD clinic patients with a similar overall HIV prevalence of 3.6% and suggest that the addition of pooled NAAT to standard antibody testing algorithms can significantly improve HIV testing performance even in urban settings, where the HIV burden is high and concentrated in specific populations.^17,26

The relative failure of the current standard-of-care HIV testing system in this study is emphasized by several additional observations. First, 2 of the 4 “missed” patients had acute HIV infection and a symptomatic STD, a hazardous situation in which HIV transmission probabilities could be elevated as much as 100-fold greater than average.^27-33 Second, only 1 had symptoms suggestive of acute retroviral syndrome, which went unrecognized by the counselor at the time of testing. Third and most importantly, at least 3 of these highly contagious missed patients were infected with drug-resistant virus and, in 1 case, multiclass-resistant virus. The current HIV testing system is ill-equipped to diagnose persons with acute HIV infection or to address secondary transmission of drug-resistant virus; in our study, 4 patients with acute infection received “HIV-negative” or “indeterminate” results and returned to the community: 2 were lost to follow-up, whereas the other 2 remained in the community for at least 9 months before repeat standard HIV screening diagnosed HIV infection. The current HIV screening system failed to detect these infections at multiple points by initial antibody-based testing, by the standard recommendation to retest, by failure to recognize acute retroviral syndrome symptoms, and by failure to achieve follow-up of high-risk subjects. For these reasons, the continued practice of exclusive antibody testing needs to be urgently reconsidered.

Our results suggest that transmission of multidrug- and multiclass-resistant HIV may be more common than previously suspected among acute HIV infections in Atlanta. The only other report of acute HIV infection in the southeastern United States found decreasing rates of drug-resistant virus, with no drug-resistant mutations identified in cases after 2000.³⁴ In contrast, at least 3 of 4 acutely infected patients identified in our study had multidrug- or multiclass-resistance mutations. Two subjects had NNRTI multidrug resistance, whereas 1 subject had multiclass (NRTI and NNRTI) resistance and was lost to follow-up after standard HIV testing. This difference may be attributable to differences in the populations studied or to a greater likelihood of HIV drug resistance among patients with another STD. In the recent study of San Francisco STD clinic patients that detected 11 acute HIV infections among 3789 persons, none of the acute infections had clinically significant genotypic resistance mutations, however.³⁵ Our results are consistent with earlier observations of apparently increasing transmission of drug-resistant virus in San Francisco and other US cities^19,20 and support the importance of identifying acute infections to investigate emerging patterns of disease transmission prospectively. Moreover, because acutely infected persons are at high risk for secondary spread of infection and for having drug resistance, targeting secondary prevention efforts toward acutely infected persons may be critical to reducing the spread of drug-resistant strains.

Other studies of acute infections in the United States relied on detection and referral of symptomatic cases and found acute infections primarily in white non-Hispanic MSM.^19-21 This study demonstrates that NAAT-based screening can find diverse acutely HIV-infected persons wherever they present for HIV testing without the involvement of specialty physicians. STD clinics, sites serving MSM in urban areas, and probably sites serving high-risk women should consider adopting HIV testing that includes NAAT.

We explored the usefulness of 2-step antibody testing algorithms for describing HIV incidence and the local epidemiology of HIV transmission in our testing population. The Vironostika-LS and BED EIA methods performed on the same samples have excellent agreement when estimating incidence, indicating that methods could be useful for regional- or national-level HIV surveillance programs. HIV incidence estimates are biased by care-seeking and testing behaviors, however.^5,36 The higher incidence at the site serving primarily MSM in this study, for instance, was an expected observation that may be partly attributable to more frequent testing in the MSM population.^5-7 In addition, with small numbers of observations in groups, it was impossible to assess differences in incidence across groups of potential interest within the study sample. Most importantly, the key finding of this study, local transmission of drug-resistant virus by acutely HIV/STD-coinfected patients, was only able to be documented through surveillance for acute infections using NAAT. At the local level, tracking individual-level data on confirmable acute HIV infections can be used to target HIV prevention interventions.

There are several additional limitations of this study. The sampled population does not reflect all high-risk populations in Atlanta but is limited to clients presenting for HIV testing at 3 major public HIV testing sites. In addition, the base population that was sampled is unknown and may have varied during the course of the study. With regard to testing, it is possible that patients identified as acutely infected may have been missed because of laboratory error rather than EIA performance. It is also possible that other more sensitive antibody or antigen/antibody screening tests could have identified some or all of the acute HIV infections detected by NAAT. Even third-generation IgM-sensitive antibody tests cannot reduce the RNA-positive antibody-negative window period to fewer than 7 days,³⁷ however, and it is unclear whether an expensive and complex EIA would be preferred as a screening test to the relatively less expensive and maximally sensitive EIA plus NAAT algorithms demonstrated in this and other studies.^13,16,31 In addition, most state public health laboratories (42 of 49 laboratories) continue to rely on second-generation tests to screen for HIV.³⁸ Perhaps more importantly, a highly sensitive antibody-based testing approach might detect some acute infections but would not recognize them as acute, missing crucial opportunities for contact tracing, secondary prevention, and clinical follow-up. Finally, although women comprised 41% of study participants, we detected only 1 recent and 2 acute infections among women, suggesting that women with acute or recent infection are not accessing HIV testing or that alternative testing sites should be targeted to diagnose incident infections among women.

The results of our work lead to several related conclusions. First, existing strategies for HIV voluntary counseling and testing fail to detect the people most likely to be involved in ongoing HIV transmission in urban areas. This failure limits, and may actually reduce, the effectiveness of HIV testing as a prevention strategy. Second, the frequent transmission of drug-resistant strains makes the acutely infected population even more critical to identify and target for important secondary prevention activities. Third, research on the best clinical and public health management of acute and early HIV infections, where synergistic treatment and prevention may be possible, is clearly needed. Finally, the spread of HIV among women and minorities and a recrudescence among young MSM³⁹ painfully illustrate that our current prevention efforts are not adequate. Revision of current HIV testing recommendations should be urgently considered. Implementation of screening for acute HIV infections as a public health strategy is likely to have associated financial, human resource, and capacity costs that must be weighed when determining how widely to implement this strategy.

ACKNOWLEDGMENTS

The authors thank the study participants and staff at the study sites for their cooperation, Howard Pope of Emory University for enrolling participants, Lee Rueei-Shu and Rom Morales of the Georgia Public Health Laboratories for Vironostika-LS EIA performance, Mark Turner of the University of North Carolina for NAAT performance, Trudy Dobbs and Bharat Parekh of the Centers for Disease Control and Prevention HIV Serology and Diagnostics Laboratory for BED CEIA performance and interpretation, and Judi Duffy of the HIV Statistics Section of the Georgia Department of Human Resources for providing data on HIV testing in Georgia.