Over the past 5 years, the PHSKC HIV NAAT program has had a unique opportunity to compare the performance of different HIV tests. Not surprisingly, standard antibody testing using a first- or second-generation EIA identified 92% of HIV-infected MSM, but rapid antibody testing identified only 91% of antibody-positive MSM and 80% of all HIV-infected MSM. These numbers may actually overestimate the sensitivity of rapid antibody testing in our population, compared with EIA and NAAT, because only 75% of MSM screened by OraQuick had blood samples obtained for additional HIV testing.
Rapid testing may have advantages over EIAs in some settings because of patient preferences [15
] and because more testers may receive results [16
]. Rapid tests have been considered to be as sensitive as second-generation EIAs for detection of recent infection [18
]; however, our data suggest that this may not be true for OraQuick. Many of the OraQuick-negative, EIA-positive individuals had likely been infected for a couple of months, because some subjects reported symptoms consistent with the acute retroviral syndrome up to 6 weeks prior to testing and because of their Western blot patterns and the window periods associated with these early-generation EIAs [4
]. The exception is patient 13 in , who, to our knowledge, is the second example of a false-negative OraQuick result in late infection [21
]. The finding that persons with negative OraQuick results may actually be antibody positive and have been infected for months or years is relevant not only for HIV testing programs but also for research programs that must accurately estimate the timing of HIV acquisition to understand viral dynamics during acute infection and implications for transmission.
It has been suggested that the Uni-Gold Recombigen HIV Test may have greater sensitivity during early infection, because it uses a larger specimen volume and an “antigen sandwich” detection system similar to that of third-generation EIAs [22
]. Although one study [23
] supports this possibility, our limited data (which were not intended for this purpose) and a study of seroconversion panels [24
] revealed no differences between OraQuick and Uni-Gold. It is theoretically possible that every false-negative OraQuick test result that we observed in EIA-reactive persons resulted from operator error, particularly because each of the 5 cases we retested yielded positive OraQuick results. However, we believe that this latter finding was more likely associated with our use of frozen serum samples for retrospective testing instead of the oral fluid specimens or small volumes of whole blood that were evaluated in real time. We believe that our data show that caution should be used when extrapolating results from retrospective HIV tests of frozen specimens and that real-time comparisons of rapid antibody tests are needed.
Our data reinforce the fact that no HIV test can have a sensitivity of 100% and that the operating characteristics of a test depend not only on the test itself, but also on the likelihood that individuals will seek testing during the interval when testing could produce false-negative results (i.e., the “window period”), which depends both on HIV incidence and the extent to which persons randomly or purposefully seek testing shortly after HIV acquisition. is a simplified schematic that shows how the relative sensitivities of antibody testing and NAAT vary depending on the testing frequency of a population of individuals who acquire HIV infection after their most recent negative test results. As public health programs strive to increase the frequency of testing to reduce the duration of time that people are unaware of their infection status, the proportion of HIV-infected individuals who are misdiagnosed will increase unless sensitive tests are used to mitigate the expected greater number of false-negative antibody test results during acute and early infection.
Figure 2 Effect of frequency of testing among individuals newly infected with human immunodeficiency virus (HIV) on HIV test sensitivity. If the timing of testing is independent of the risk for HIV acquisition, and if HIV antibody and the nucleic acid amplification (more ...)
The failure to diagnose acute HIV infection represents an important public health problem. Persons with primary infection may be up to 10 times more likely to transmit HIV per sex act than are individuals with established infection [3
], and secondary transmission from recently infected persons likely contributes to a significant proportion of overall HIV transmission [25
]. Like other NAAT programs [32
], our program successfully identifies cases of acute infection with costs per case detected that are comparable to case-finding based on more traditional methods [12
]. However, <10% of the estimated 300–400 MSM who acquire HIV infection annually in Seattle receive a diagnosis during acute infection. Greater numbers of persons must be diagnosed during acute infection for NAAT programs to impact HIV incidence. Yet the CDC recommends focused use of NAAT only if clinical providers are astute enough to recognize at-risk persons and symptoms of seroconversion. It is illustrative that, even in settings where the topic of acute infection is frequently discussed, only one-half of PHSKC clinicians considered the diagnosis in MSM with consistent symptoms. Efforts to increase availability of routine NAAT should be promoted, and programs that increase the frequency of HIV testing in high-risk persons and prompt such persons to seek NAAT for symptoms consistent with the acute retroviral syndrome merit evaluation.
Although NAAT remains the gold standard for the diagnosis of acute infection, limitations of NAAT include cost, the requirement for venipuncture and more complex technology, and the time interval between specimen collection and receipt of results. Furthermore, a small percentage of HIV-infected persons may have undetectable HIV RNA levels without antiretroviral therapy. Fourth-generation combination assays detect both anti-HIV antibodies and p24 antigen and could be alternatives to NAAT. The Architect HIV Ag/Ab Combo appears to detect acute infection if HIV RNA levels are above ~30,000 copies/mL [36
], and our data suggest that the Architect HIV Ag/Ab Combo would miss few acute infections that would be detected by NAAT. These results are similar to previous studies that suggested that p24 antigen testing could identify from 79% to >90% of individuals who seek testing during acute infection [37
Our findings may not be generalizable to populations with a lower prevalence and incidence of HIV infection. However, NAAT could also increase case finding in such settings. In North Carolina, where the proportion of antibody-positive (0.5%) and antibody-negative, NAAT-positive (0.02%) testers was nearly 10-fold lower than in our program, antibody testing detected only 96% of HIV-infected testers [32
]. The insensitivity of antibody testing during acute infection also has relevance to general health care settings outside of public HIV testing programs, because it has been estimated that 0.7% of all general health care visits for rash and 0.5% of visits for fever may represent acute HIV infection [40
], and persons who seek attention for seroconversion symptoms frequently receive misdiagnoses [41
]. Data are needed on the sensitivity and cost-effectiveness of different testing strategies in settings with varying prevalences and incidences of HIV infection.
Unfortunately, our data do not allow us to compare the sensitivity of OraQuick on oral fluid and finger-stick blood specimens, because these tests were performed in subpopulations of MSM who had different rates of acute infection. Therefore, it would be impossible to interpret whether differences were due to a greater likelihood of testing in the period following HIV acquisition or actual differences in test performance. We did identify 2 MSM who had concurrent OraQuick-negative results with oral fluid specimens and Ora-Quick-reactive whole blood specimens—findings consistent with the lower sensitivity of oral fluid tests reported in the package insert. Prospective studies are needed to compare the sensitivity of OraQuick performed with oral fluid specimens versus finger-stick blood specimens.
In summary, although expanded antibody testing may increase the number of individuals with established infection who are aware of their HIV status, there may be detrimental public health effects in some settings if less-sensitive tests are widely implemented, if NAAT is not performed, and if highly infectious individuals are falsely reassured by recent negative antibody test results. In populations with high levels of HIV transmission and frequent testing, we believe that pooled HIV NAAT should be the standard of care and should be integrated with antibody tests to act as a back-up for false-negative antibody test results. In light of these data, in settings where NAAT cannot be made available, it behooves programs to utilize the most sensitive antibody tests, which may or may not be rapid tests. We look forward to a time when the existence of future technologies such as rapid, point-of-care fourth-generation assays or even point-of-care NAAT could make these discussions moot.