Table 2.

Operating characteristics for CD4 monitoring technologies with different measurement precision and bias

CD4 threshold (cells/μl)	Bias	Relative change in SD of short-term variability (δ)	Median ART start time (months)	Mean over 61 months per patient		% starting ART
				Pre-treatment CD4 measurements	Months on ART
						late1	very late2
200	0%	0.5	>60	9.7	8.0	58.6	1.1
	0%	0.75	>60	9.5	9.3	47.3	2.2
	0%	1	>60	9.2	11.1	38.0	2.6
	0%	1.25	60	8.8	13.6	31.0	3.2
	0%	1.5	54	8.5	15.9	25.7	3.3
	−10%	1	60	8.9	13.3	25.7	1.3
	+10%	1	>60	9.5	9.3	50.7	5.4
350	0%	0.5	48	8.0	18.7	54.3	0.5
	0%	0.75	42	7.6	21.2	43.2	1.0
	0%	1	36	7.1	24.0	35.3	1.6
	0%	1.25	30	6.6	27.3	29.3	1.9
	0%	1.5	30	6.2	29.8	24.9	2.2
	−10%	1	30	6.4	28.2	21.5	0.5
	+10%	1	42	7.7	20.5	50.1	3.2
500	0%	0.5	24	5.8	31.9	47.9	1.2
	0%	0.75	24	5.4	34.6	39.8	2.4
	0%	1	18	4.9	37.3	34.4	3.6
	0%	1.25	18	4.5	39.9	30.5	4.3
	0%	1.5	12	4.2	41.7	27.9	4.8
	−10%	1	12	4.1	42.1	19.6	1.5
	+10%	1	24	5.7	32.8	49.8	7.8

¹The percentage starting ART late is calculated as the number who start ART at any time after their latent CD4 count is first below the threshold divided by the number who have a latent CD4 count below the threshold at any time during the 60 months simulated.

²The very late thresholds for ART initiation thresholds of 500, 350 and 200 cells/μl are defined as 350, 200 and 100 cells/μl, respectively. The percentage starting ART very late is then calculated as the number who start ART at any time after their latent CD4 count is first below the very late threshold divided by the number who have a latent CD4 count below the very late threshold at any time during the 60 months simulated.