Evaluation of HIV testing algorithms in Ethiopia: the role of the tie-breaker algorithm and weakly reacting test lines in contributing to a high rate of false positive HIV diagnoses

Table 8 Summary of reports in the literature of performance of RDTs and algorithms with weak positives included and excluded

Paper	Algorithm	Weak positives	N	HIV prevalence%	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)
Gray, Uganda	D-S-U	Included	639	14.6	97.8	94.1	74.0	99.6
Gray, Uganda	D-S-U	Excluded	602	14.4	97.7	99.6	97.7	99.6
Klarkowski, DRC	D-U	Included	2728	7.5	100	99.0	89.5	100
	D-U	Included	2728	7.5	(98.2-100)	(98.6-99.4)	(84.8-93.2)	(99.9-100)
	D-U	Excluded	2711	7.6	100	99.7	96.7	100
	D-U	Excluded	2711	7.6	(98.2-100)	(99.4-99.9)	(93.3-98.7)	(99.9-100)
Galiwango, Uganda	D-S-U*	Included	2520	35.2	97.4	99.6	99.3	99.8
	D-S-U*	Included	2520	35.2	(96.1–98.4)	(99.2–99.9)	99.3	99.8
	D-S-U*	Excluded		35.2	97.4	99.8	99.7	99.8
	D-S-U*	Excluded		35.2	(96.1–98.4)	(99.5–100)	99.7	99.8
	S-D-U**	Included	2520	35.2	99.7	96.9	94.6	99.8
	S-D-U**	Included	2520	35.2	(99.0–99.9)	(96.0–97.7)	94.6	99.8
	S-D-U**	Excluded		35.2	99.7	99.7	99.4	99.8
	S-D-U**	Excluded		35.2	(99.0–99.9)	(99.3–99.9)	99.4	99.8
	D-S-U**	Included	2520	35.2	97.4	99.7	99.4	98.6
	D-S-U**	Included	2520	35.2	(96.1–98.4)	(99.3–99.9)	99.4	98.6
	D-S-U**	Excluded		35.2	97.3	99.9	99.8	98.6
	D-S-U**	Excluded		35.2	(96.0–98.3)	(99.6–100)	99.8	98.6
Kagulire, Uganda	D	Included	150	NR	NR	85.2	67.3	NR
	D	Included	150	NR	NR	85.2	(52.9-79.7)	NR
	D	Excluded	138	NR	NR	95.1	NR	NR
Kroidl, Tanzania	D (whole blood)	Included	1696	1.5	100	96.8	32.9	100
	D (whole blood)	Included	1696	1.5	(86.7-100)	(95.9-97.6)	(22.8-44.4)	(99.8-100)
	D (whole blood)	Excluded	1646	1.5	100	NR	86.2	NR
	D (plasma)	Included	12916	9.2	100	97.9	82.6	100
	D (plasma)	Included	12916	9.2	(99.7-100)	(97.6-98.1)	(80.5-84.5)	(99.8-100)
	D(plasma)	Excluded	12692	9.2	100	NR	96.4	NR

S = STAT-PAK, U = Unigold, D = Determine.
DRC = Democratic Republic of Congo.
NR = Not reported.
*Parallel tiebreaker: both Determine and STAT-PAK are done initially, with discordant results resolved by Unigold.
**Serial tiebreaker: first RDT used as a screen; if negative, no further testing. If positive, second test is employed. Discordant results resolved by Unigold.

ISSN: 1471-2334