Table 2.
HIVST concordance, reasons for disagreement, and errors in performance among studies (n=25)
| HIVST concordance* | Reasons for disagreement† | HIV positivity | Type of observation | Errors in performance | Invalid results (invalid result/tests performed) | Reasons for invalid result | |
|---|---|---|---|---|---|---|---|
| Directly assisted studies | |||||||
| Prazuck et al (2016)‡37 | 97·1% (142/147) | Non-reactive as reactive 2·7% (4/147), invalid as reactive 2·7% (4/147) or non-reactive 2·7% (4/147), non-reactive as invalid 1·4% (2/147) | .. | Observed | .. | 1% (2/264) | .. |
| Majam et al (2016)40 | 88% (53/60) | Non-reactive as reactive 1·7% (1/60), non-reactive as invalid 1·7% (1/60), reactive as non-reactive 1·7% (1/60), invalid 6·7% (4/60) as reactive or non-reactive | .. | Observed | 20 participants made mistakes; common errors were with blood collection and transferring and use of buffer | .. | .. |
| MacGowan et al (2014; oral fluid arm)29 | 95% (21/22) | Reactive as non-reactive 4·5% (1/22) | 22·7% (5/22) | Observed | 13·6% (3/22) participants made mistakes, common errors were spilling buffer and incorrect time to read the results | 4·5% (1/22) | .. |
| MacGowan et al (2014; blood-based arm)29 | 95% (20/21) | One HIV-positive participant with an invalid result interpreted his result as reactive 4·8% (1/21) | 19% (4/21) | 23·8% (5/21) participants made mistakes; common errors were incorrectly pushing the device into test holder and incorrect timing to read the results; one participant broke the device | 9·5% (2/21) | Operational error | |
| Choko et al (2015)§6 | 99·4%, 98·9%–99·7% (1639/1649) | Reactive as non-reactive 0·5% (9/1649), non-reactive as reactive 0·06% (1/1649) | 8·6% (141/1649) | Observed | .. | .. | .. |
| Choko et al (2011)¶20 | 99·2%, 97–100% (256/258)‖ | One HIV-positive participant with a faint reactive result interpreted his result as uncertain, one HIV-positive participant had an invalid result | 16·9% (48/283) | Non-observed | Common errors were touching collection pad, incorrect or incomplete swabbing, removing kit from developer too early, buffer spills, reading incorrectly, and fumbling vial or cap when opening developer fluid | 0·4% (1/260) | .. |
| Marley et al (2014)30 | 93·9% (215/229), κ 0·551, p=0·012 | Reactive as invalid 3·1% (7/229), non-reactive as reactive 1·3% (3/229), invalid as non-reactive 1·3% (3/229), non-reactive as invalid 0·4% (1/229) | 5·6% (13/229) | Observed | Common errors were unpreparedness before start 42% (94/229), inability to swab correctly 10% (23/229), buffer 15·9% (36/229), testing and reading test results 7·5% (17/229) | 3·5% (8/229) | Six participants used test paper incorrectly |
| Martínez Pérez et al (2016)36 | 99·4% (2184/2198), κ 0·99** | Reactive as non-reactive 0·2% (4/2181) | 15·3% (337/2198) | Observed | Two participants had to repeat the self-test, they accidentally spilled buffer vial; excluding known people living with HIV, 0·18% (4/2181) interpreted their tests as negative whereas the HCW interpreted the result as positive | 0·5% (11/2198) | .. |
| Sarkar et al (2016)38 | 98%, κ 0·566, p<0·001 | Invalid as non-reactive 0·5% (1/202), non-reactive as invalid 0·9% (2/202) | 0·9% (2/202) | Observed | .. | 0·9% (2/202) | .. |
| Pant Pai (2013)34 | 98·8% (248/251) | Reactive as non-reactive 1·2% (3/251), two of which had a faint reactive line | 3·6% (9/251) | Non-observed | Errors were in conducting and interpreting results | .. | .. |
| Pant Pai et al (2014)35 | 100% (145/145) | No difference between self-tester and HCW interpretation | 0 | Non-observed | .. | .. | .. |
| Asiimwe et al (2014; observed arm)18 | 99·2% (122/123) | Non-reactive as invalid 0·8% (1/123) | 10·6% (13/123) | Observed | 19·5% (24/123) participants made mistakes; common errors were incorrect swabbing of gums, touching the collection pad and buffer spills | 0·8% (1/123) | .. |
| Asiimwe et al (2014; unobserved arm)18 | 94·3% (116/123) | .. | 16·3% (20/123) | Non-observed | No errors were reported | 0·8% (1/117)†† | |
| Unassisted studies | |||||||
| Lee et al (2007)27 | κ 0·277, p<0·001 | Invalid as non-reactive 50·1% (176/350), invalid as reactive 4·6% (16/350) and reactive as non-reactive 0·3% (1/350) | 25% (88/350) | Observed | .. | 56·3% (197/350) | 85% failed to perform all steps correctly |
| Gras et al (2014)25 | 100% | No difference between self-tester and HCW interpretation | 100% (40/40) | Observed | Common errors were insufficient blood, wrong lancet utilisation, and mixing of samples | 5·7% (2/35) | .. |
| Dong et al (2014)22 | 98·7% (230/233) | Reactive as non-reactive 0·5% (1/195), invalid as non-reactive 0·5% (1/195), non-reactive as invalid 0·5% (1/195) | 18·9% (44/233) | Observed (video recorded) | .. | 0·4% (1/233) | .. |
| Chavez et al (2016; oral fluid arm)‡‡19 | 98% (500/511) | Non-reactive as reactive 1·4%, non-reactive as invalid 0·8% | 2% (11/622) | Non-observed | .. | .. | .. |
| Chavez et al (2016; blood-based arm)‡‡19 | 99% (506/511) | Non-reactive as reactive 0·6%, non-reactive as invalid 0·4% | 1% (7/565) | 4·6% (26/565) | Operational error | ||
| Gaydos et al (2011)§§23 | 99·6%, 0·41–1·00 (476/478) weighted κ 0·75 | Reactive as non-reactive 0·2% (1/478) | 0·8% (4/478) | Observed | Difficulties were interpreting results, reading result chart, reading or following instructions, swabbing or pricking properly, or both, and opening the kit | 0·2% (1/478) | Insufficient blood |
| Spielberg et al (2003; oral fluid arm)39 | 95% | .. | 100% (340/340) | Non-observed | Difficulties performing test decreased through changes made to instructions and labelling from 4·3% to 4% | 4·1% (14/340) | Failure to put the test device in the vial with developer solution |
| Spielberg et a; (2003; blood-based arm)39 | 97% | 100% (340/340) | Difficulties performing test decreased through changes made to instructions and labelling from 14% to 9% | 7·9% (27/340) | |||
| Gaydos et al (2013)24 | 100% | No difference between self-tester and HCW interpretation | 0·2% (1/467) | Observed | .. | .. | .. |
| Kurth et al (2016)26 | κ 0·92 (0·84–0·99) | Non-reactive as invalid 12·5% (30/239), reactive as non-reactive 1·2% (3/239), non-reactive as reactive 0·4% (1/239) | 14·6% (35/239) | Observed (video recorded) | Common errors were difficulty opening bottle, incorrect or incomplete swab of gums, and incorrect time to read the results; some individuals could have made multiple errors | 15·1% (36/239) | All individuals recognised something went wrong with their test |
| Li et al (2016)28 | 95% (190/200) | Non-reactive as invalid 2·5% (5/200), reactive as non-reactive 1·5% (3/200), non-reactive as reactive 0·5% (1/200) | 27·5% (55/200) | Observed | Common errors were incorrect or incomplete swab of gums, incorrect time to read the results, touching the collection pad, and buffer spills | 3% (6/200) | .. |
| Nour et al (2012)33 | 100% | No difference between self-tester and HCW interpretation | 1·6% (4/249) | Observed | .. | .. | .. |
| Mavedzenge et al (2015; urban arm)¶¶31 | 93% (160/172) | Non-reactive as invalid 2% (3/172) | 9% (16/172) | Observed (video recorded) | Common errors were confusion with desiccant, buffer spills, dipping test device in developer before collecting sample, incorrect sampling, and incorrect time to read the results. | 2·9% (5/172) | Participants with invalid results typically did not follow instructions |
| Mavedzenge et al (2015; rural arm)31 | 90% (56/62) | Non-reactive as reactive 4·8% (3/62) | 8% (5/62) | 3·2% (2/62) | |||
| Ng et al (2012)32 | κ 0·97, 0·95–0·99 | Reactive as non-reactive 2·6% (5/983), reactive as invalid 0·5% (1/983), non-reactive as invalid 0·3% (2/983) and non-reactive as reactive 0·1% (1/983) | 19·3% (192/994) | Observed | Common errors were incorrect or incomplete swab of gums, touching the collection pad during removal from packaging, or buffer spills | 0·3% (3/983) | .. |
| FDA phase 2b (2012; observed arm)‖‖4 | 93% (942/1013) | Reactive as non-reactive 0·9% (10/1013), non-reactive as reactive 0·1% (1/1013) | 2·1% (120/5662) | Observed | Common errors were interpreting results (11/986), dipping device in developer prior to swabbing gums (11/986), buffer spills (4/986), incorrect swabbing (5/986), and could not find developer (2/986) | 3·3% (33/986) | Operational errors |
| FDA phase 3 (2012; unobserved arm)***4 | 99·8% (5490/5499) | Reactive as non-reactive 0·1% (8/5499), non-reactive as reactive 0·01% (1/5499) | 51% (526/1031) | Non-observed | Not understanding where to place the test stick after sample collection (1/4999) | 0·6% (31/4999) | .. |
| Directly assisted and unassisted studies | |||||||
| de la Fuente et al (2012; directly assisted arm)21 | 85·4% (445/521) | Invalid as reactive 2·8%, non-reactive as reactive 2·7%, non-reactive as invalid 2%, reactive as invalid 1·9%, invalid as non-reactive 1·5% and reactive as non-reactive 1·1% | .. | Observed | .. | 0·9% (2/208) | Most difficult step was obtaining blood and depositing it in the correct place |
| de la Fuente et al (2012; directly assisted arm)21 | 85·4% (445/521) | Invalid as reactive 2·8%, non-reactive as reactive 2·7%, non-reactive as invalid 2·1%, reactive as invalid 1·9%, invalid as non-reactive 1·5% and reactive as non-reactive 1·1% | .. | Observed | .. | 0·9% (2/208) | Most difficult step was obtaining blood and depositing it in the correct place |
Data are % (n/N); %, 95% CI, (n/N), κ, p value; or κ (95%CI). HIVST=HIV self-test. HCW=health-care worker. FDA=Food and Drug Administration.
*
Reported as percentage of agreement or κ.
†
Reason for disagreement assumes the self-tester perspective compared with the HCW.
‡
The study was divided into two substudies: 264 participants performed the self-test, and 147 participants interpreted contrived pictures.
§
Four participants were on antiretrovirals.
¶
260 of 283 participants self-tested.
‖
Two participants had no confirmatory results.
**
17 known people living with HIV were not considered to calculate the κ.
††
Six participants had no results.
‡‡
515 participants had all three results (both self-tests and dried blood home collection [dried blood spot]), 622 reported the oral fluid-based result, 565 reported the blood-based result, and 548 had the dried blood spot cards processed.
§§
Disaggregated results by type of specimen were not available. One participant was on antiretrovirals with undetectable viral load.
¶¶
One participant in the urban arm was on antiretrovirals.
‖‖
1013 of 1031 participants completed the study.
***
18 positives and 482 negatives were excluded from the accuracy analysis.