Introduction

Globally, an estimated 45% of people living with HIV are aware of their HIV status[1], far below the 90% target set out recently by WHO and UNAIDS[1]. The greatest numbers of people living with undiagnosed HIV are in sub-Saharan Africa [1], where fragile health and laboratory systems have hindered efforts to scale up HIV testing coverage [2]. In low-income, high HIV burden countries with limited technological infrastructure, rapid diagnostic tests (RDTs) are widely used in diagnostic algorithms [1].

RDT accuracy may be undermined by poor operator practice [3, 4], and by storage at ambient temperatures above manufacturer guidelines [5]. Poor accuracy of RDTs and subsequent misclassification can result in a) if falsely diagnosed HIV-positive, unnecessary worry and initiation on antiretroviral therapy; or b) if falsely diagnosed HIV-negative, missed opportunities for linkage to HIV care and prevention services[6, 7].

The scale-up of HIV self-testing (HIVST) using oral fluid testing will pose additional challenges to maintaining accuracy in HIV testing programmes. Self-testing is defined by an individual performing and interpreting their own HIV test result [8]. Oral fluid RDTs intended for self-testing may be stored in uncontrolled settings (e.g. people’s houses) for prolonged periods before use and are less amenable to formal quality assurance (QA) programmes [9]. We therefore set out to evaluate two aspects of RDT kit stability: the effect of prolonged field exposure to high temperature on the accuracy of whole blood and oral fluid RDTs; and the stability of oral RDT results with delayed visual re-reading.

Methods

Results

Characteristics of study participants

Of 404 individuals screened for eligibility, a total of 378 (93.6%) were recruited into the study (Fig 1). Median age was 31 years (IQR 25–40), and 198 (52.4%) participants were male. The majority (242, 64.0%) of participants had previously tested for HIV, with comparable proportions of prior testing among men (121/238, 50.8%) and women (117/238, 49.2%; p = 0.096). A total of 108 (28.6%) HIV-positive individuals were already on ART.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow of participant recruitment and HIV testing.

https://doi.org/10.1371/journal.pone.0158107.g001

Sensitivity and specificity of index tests

Results from 5/378 (1.3%) participants were excluded from the computation of sensitivity and specificity (3 excluded were due to discordant results between Determine and Uni-Gold and 2 due to missing result from ≥ 1 reference standard test). All RDTs had a sensitivity > 98.0% except pre-incubated OraQuick with sensitivity 97.2% (95% CI: 93.6–99.6) Table 1. Specificity was higher than 99.5% for all RDTs. There were seven false negative results in total across all the pre-incubated RDTs compared with three false negatives when RDTs were optimally stored.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sensitivity and specificity compared to a reference standard (parallel testing).

https://doi.org/10.1371/journal.pone.0158107.t001

Reliability of pre-incubated RDTs

There was excellent agreement between pre-incubated and optimally-stored RDTs with Kappa of 1.00 for Determine and Uni-Gold (Table 2). Similar results were observed for pre-incubated OraQuick, with Kappa of 0.97 (95% CI: 0.95; 1.00).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Reliability comparison of optimally stored and pre-incubated tests.

https://doi.org/10.1371/journal.pone.0158107.t002

Discussion

The main findings of this study were that the contribution of suboptimal storage to misclassification of HIV results is likely to be relatively minor, and that re-reading of used oral fluid tests (OFT) may be a feasible and valid option for implementing quality assurance (QA) of OraQuick test kits as these are scaled up for use by less qualified personnel, including during HIV self-testing [9].

Pre-incubating HIV RDTs at 37°C for 28 days did not substantially affect their diagnostic accuracy, and, in all but two cases (both known to be taking ART), initial visual read results on OraQuick test kits were stable over the one-year follow up period in this study. The results from our study demonstrate the stability of OraQuick, Determine and Uni-Gold when pre-incubated at 37°C for 28 days, which is an important consideration in resource poor settings where poor storage conditions are common [5].

The scale up of HTC in Africa was not initially accompanied by systematic QA or retraining, but the need to build systematic proficiency testing into routine HIV testing programmes has become strikingly apparent with reports of poor operator practice and high rates of misclassification in many African national programmes [10, 11]. Lessons from this experience need to be applied to HIVST and other types of highly decentralised community-based HIV testing services.

Our results, showing robustness of the kits to poor storage and late read, are consistent with current understanding that identifies poor operator practice as the major determinant of misclassification [10, 11], in part reflecting “task-shifting” of HTC to cadres such as lay counsellors. The introduction of very simple oral fluid test kits will enable further task-shifting including as part of HIV self-testing (HIVST) [9]. Systems for providing external QA for oral fluid tests and HVST are not yet well established. These data provide reassurance that countries currently revising QA systems to accommodate oral fluid tests and HIVST could consider including systems based on visual re-reading. While outside of manufacturer’s recommendations, our results suggest that for OraQuick at least, this provides a simple and practical approach to monitoring operator-dependant misclassifications if used kits can be retrieved along with the users’ interpretation of results. In this context, we have previously reported a high willingness to return HIVST kits to the distributor, together with a self-completed questionnaire that includes results interpretation [9].

More work is needed to develop fully comprehensive external quality assurance (EQA) systems for oral fluid RDTs. False-positive results have been described for OFTs, particularly in the context of pregnancy and concomitant infections [12, 13]. With respect to sensitivity, subsets where oral-based RDTs should be used with caution because of reports of lower than expected sensitivity [14] are patients at high risk of acute HIV infection [15] and those established on ART. In this study, we report one potential false-positive from 369 optimally and 375 sub-optimally stored OFT kits, and a suggestion that poor storage conditions may reduce sensitivity but with no obvious effects on specificity. We also report two participants already on ART who had an observed change in the result of their oral test kits when re-read over time. Otherwise, there was minimal change in the interpretation of used OraQuick RDT results when read up to one year after testing in this relatively small sample.

The implementation of a multistep approach to HIV testing has been recommended [1], and it will be vital to have rigorous QA mechanisms in place in an effort to reduce error amongst testers in both home-based and laboratory settings. Re-evaluation of used test kits could contribute to internal and external quality assurance programs [1, 16].

There were three notable limitations in this study. First of all, our reference standard was not carried out separately from the index test rendering blinding of index tests impossible. Secondly, we did not exclude patients already on ART, despite the known issues of low sensitivity of RDTs in this important subgroup. Lastly, we did not measure the impact of humidity a long with temperature in this study.

In summary, this study has shown that prolonged exposure of RDTs to temperatures above manufacturer’s recommendation did not substantially affect the sensitivity and specificity of commonly used RDTs including OraQuick OFT. Furthermore, visual interpretation of OraQuick results was highly stable for 12 months after testing, supporting the use of re-reading of used test kits as part of programmatic quality assurance.

Acknowledgments

Staff at Ndirande Health Centre, HitTBhard project nurses who performed the study procedures including late Fellie Guba. Mr Frank Chikapa who performed second reading of the used OraQuick test kits and study participants.