Since the first AIDS cases were identified more than 30 years ago [1], significant advances have been made towards diagnosing HIV infection, developing anti-retroviral therapy (ART) regimens, and implementing transmission prevention measures. Despite this ever-evolving progress, an HIV vaccine remains elusive and an estimated 34 million people globally are living with HIV infection [1]. HIV surveillance methods have been instrumental in monitoring the status of the epidemic; however, surveillance has primarily focused on the prevalence of HIV in the population. Estimates of HIV incidence are crucial for understanding the dynamics of the epidemic and assessing the efficacy of prevention measures [2]; however, recent acquisition of HIV infection is difficult to extrapolate from standard diagnostic test results.

In 1998, a ground-breaking study by Janssen et al. described the development of a detuned or less-sensitive serologic assay for distinguishing recent from long-term HIV-1 infection, which allowed for incidence estimation from cross-sectional patient samples. Several serology-based laboratory tests for recent infection (TRI) have been developed that measure a specific biomarker, primarily HIV-1-specific antibody [3], [4], avidity [5]–[9], or both [10], [11], that evolves in a predictable pattern from early to late infection. To date, two TRIs have been commercialized for HIV-1 surveillance purposes, the BED-CEIA (Sedia Biosciences Corp., Portland, OR; Calypte Biomedical Corp., Portland, OR) and HIV-1 Limiting Antigen (LAg)-Avidity EIA (Sedia Biosciences Corp.; Maxim Biomedical, Inc., Rockville, MD). The BED assay has been used to calculate HIV-1 incidence estimates in the United States and worldwide. [12]–[15].

Given that the majority of current TRI approaches are serology-based, the most commonly used sample type for incidence testing is plasma and/or serum. Although HIV-1-specific antibodies are stable in plasma and serum, sample collection is somewhat limited due to processing and storage requirements. Dried blood spots (DBS) have been used extensively for HIV testing and surveillance, as they can be prepared from a finger stick, do not require centrifugation, and can be shipped at room temperature. HIV biomarkers remain stable on the DBS filter paper; with minimal risk of infectivity once the sample is fully dried [16]. Applications for DBS are numerous, including HIV-1-specific antibody testing, genotyping, viral nucleic acid amplification [17], [18], and drug resistance testing [19]. Furthermore, the use of DBS as a sample source for HIV-1 incidence estimation has been reported [20]–[22]. Protocol adaptations for DBS are currently supplied by the manufacturer for use with both the BED (Sedia Biosciences Corp.; Calypte Biomedical Corp.) and LAg assays (Maxim Biomedical Inc.). Collection of whole blood on DBS filter paper increases the applicability of laboratory-based TRIs, given that sample collection is suitable for multiple settings, including resource limited sites where the epidemic is often concentrated.

To improve upon the accuracy of TRIs for estimating incidence, it has been demonstrated that multiassay algorithms (MAAs), combining one or more TRIs with clinical data, yield improved incidence estimates and reduced false-recent rates as compared to each individual test [23]–[26]. Recently, we described the development of an in-house HIV-specific multiplex assay, based on the Bio-Plex format, which measures HIV-specific antibody levels and avidity to multiple analytes [10]. Improved incidence estimates have been demonstrated using multi-analyte algorithms based on three or more assay measures obtained from the multiplex format [27]. Thus far, all assay development for the HIV-1-specific Bio-Plex assay has been performed with plasma samples. In the current study, we evaluate the use of DBS as an additional sample source for determining recent infection with the multiplex assay. We compared assay performance with matched plasma and DBS samples from HIV-1 infected individuals obtained from diverse sources.

In previous studies, we described the development and performance characteristics of an in-house HIV-1-specific Bio-Plex assay for determination of recent HIV-1 infection with plasma specimens [10], [27], [30]. Here, we evaluate the use of DBS eluates as an additional sample source for the customized Bio-Plex assay. The detection of HIV-1 antibodies from DBS using the Luminex technology (Luminex Corporation, Austin, TX) has been demonstrated for diagnostic purposes [31]; however, tests for recent infection require unique considerations given that the time of infection must be estimated with a fair degree of accuracy. Furthermore, our assay approach incorporates measurement of both total HIV-1-specific antibody binding and avidity; therefore, it is crucial to demonstrate that neither measure is significantly altered by sample type. Overall, comparable performance was demonstrated between matched plasma and DBS samples for both direct antibody binding and avidity measures. The consistency between sample types suggests that plasma and DBS can be used interchangeably with the Bio-Plex format, without negatively impacting the misclassification rate of the assay.

Given that most assays are subject to some degree of innate variation, it is important to characterize the baseline intra- and inter-assay variation prior to assessing potential differences in sample types. Specimen replicates yielded highly reproducible assay measures, irrespective of sample type. These findings were consistent with a previous study demonstrating performance of our in-house HIV-1 Bio-Plex assay versus a prototype kit, using plasma specimens [30]. Although samples were plated in duplicate for the purposes of the assay evaluation, the high degree of reproducibility supports the use of a single well for specimen testing, which improves the cost effectiveness of the assay. For all samples tested, the assay measures were remarkably similar between plasma and DBS for most analytes and any differences between the sample types could not be discerned from the innate variation associated with assay.

Specimens from three separate cohorts were evaluated in this study to assess potential inconsistencies due to site-specific sample collection or storage methods. Not surprisingly, the avidity measures were the most consistent between the cohorts. As observed in previous studies, avidity measures tend to be less affected by slight alterations in the assay or sample, given that avidity index is calculated relative to a control included within the same assay plate [10], [27], [30]. The correlation in normalized values between plasma and DBS tended to vary slightly between the cohorts. The SIPP cohort yielded highly consistent results for all analytes, with r values of 0.99 or 1. This may be the result of well-controlled sample and data collection, along with the fact that the specimens included in the current study were collected from a single site and freeze-thaw cycles were minimal for the plasma specimens. The NHBS cohort also exhibited high correlation between sample types; however, a lower r value of 0.79 for gp160-n was skewed by a single outlier. Unfortunately, limited sample volume for this particular study participant prevented repeat testing. For this particular cohort, site-prepared DBS were also available for some of the study participants. We did not observe any differences in reactivity between the site and CDC-prepared samples, which demonstrates that samples can be collected at field sites, shipped, and stored for later testing. The Cameroon specimens were the oldest of the three cohorts, as they were collected and stored 10–15 years ago and the DBS had been freeze-thawed multiple times for previous testing. Additionally, the Cameroon cohort exhibited a large degree of HIV-1 subtype diversity. Despite the age of the specimens and subtype diversity, the correlation between sample types was high for all analytes. It is important to note that any differences in reactivity between sample types might be exaggerated given the small sample numbers included in current study.

One of the most important considerations for a TRI is whether the recent/long-term classification is impacted by the use of DBS, thereby affecting incidence estimates. Since “true” recent/long-term status is not known for all of the samples included in this study, we compared the number of recent infections as determined by the assay values for plasma versus DBS. This evaluation was done strictly for demonstrative purposes, as finalized cutoffs have yet to be determined for the assay. Although demonstrated with a relatively small sample set, the number of specimens classified as recent were similar (differences ≤2) between plasma and DBS for all analytes except gp160-n. The number of specimens classified as recent differed between the analytes, given that each analyte has a distinct mean duration of recency (MDR) and, therefore, the determination of recent infection is analyte-specific [27]. Differences in classification between sample types were associated with specimens that elicited assay values close to the analyte cutoff. It is likely that misclassifications due to values close to the cutoff will decrease when the analyte measures are used in an algorithm to determine recent infection. The largest difference in recent classifications was observed for gp160-n, which is not surprising given the narrow range of reactivity between recent and long-term specimen. Given the high MFI values elicited from the gp160 bead set, titration of the gp160 protein on the beads can be performed to improve the measurable distinction between recent and long-term specimens (data not shown). Although each analyte was evaluated separately in the current study, an algorithm of analyte measures employing a common cutoff and MDR will be used to determine recent infection, as described previously [27]. The optimal algorithms for the current assay will be addressed in a separate study.

Given that most TRI formats are laboratory-based, the use of DBS as an additional sample source is highly valuable since samples can be collected in resource-limited settings and shipped for later testing. We demonstrate stable reactivity of antibodies from DBS, even after specimens have been stored for several years. The modification to utilize DBS for the HIV-1 Bio-Plex assay described here expands its utility as a TRI for estimating HIV incidence.