Table 2 Performance of Pima stratified by venous and capillary blood collection and presented by reference test used
Author, year | Performance of Pima on capillary blood | Author, year | Performance of Pima on venous blood | ||||||
|---|---|---|---|---|---|---|---|---|---|
Bias/LoA; Sample size (N) | Sensitivity Specificity | Total Misclassification | Failure rate | Bias/LoA Sample size (N) | Sensitivity | Total Misclassification | Failure rate | ||
Specificity | |||||||||
Reference test = FACSCalibur | |||||||||
(van Rooyen, Barnabas et al. 2013) [14] | Mean bias: 16 cells/μl (LoA: −1 to 32; N = 193) | Not reported (NR) | NR | NR | (Rathunde, Kussen et al. 2014) [32] | Bias/LoA NR; N = 107 | At CD4 threshold of 350 cells/μl: Sensitivity 94 % Specificity 93 % | NR | NR |
(Jani, Sitoe et al. 2011) [34] | Accurate absolute counts | NR | At CD4 thresholds of 200 cells/μl: 5.2 % | NR | (Galiwango, Lubyayi et al. 2014) [45] | Pima significantly underestimate CD4 count particularly at higher CD4 count. | At CD4 threshold of 350 cells/μl: Sensitivity 88.6 % specificity 87.5 % | 12.2 % | NR |
Bias: −52.8 cell/μl (LoA: −250.9 to 145.2; N = 135). | |||||||||
Bias was smaller at lower CD4 count (<500: −24.4) than at higher CD4 count (>500: −107.9). | At CD4 thresholds of 350 cells/μl: 17 % | Bias: −34.6 cells/μl (LoA: −219.8 to +150.6; N = 903) | At CD4 threshold of 500 cells/μl: Sensitivity 96.1 % specificity 83.0 % | 9.5 % | |||||
At 350 cut-off: +5.1 cells/μl (LoA: −126.6 to +136.8) vs −51.0 cells/μl (LoA: −245.4 to +143.4) | |||||||||
At 500 cut-off:-10.9 cells/μl (LoA: −147.3 to +125.5) vs −66.3 cells/μl (LoA: −286.6 to +154.0) | |||||||||
(Mtapuri-Zinyowera, Chideme et al. 2010) [46] | Mean bias: 7.6 cells/μl (LoA: −173.8 to +189.0). | At CD4 threshold of 200 cells/μl: Sensitivity: 95.1 % Specificity: 91.6 % | 6.7 % | NR | (Mwau, Adungo et al. 2013) [49] | Bias: −64.8 cells/μl (LoA: −332.5 to +203.0; N = 396) | At CD4 threshold of 350 cells/μl (in those ≥ 5 years old N = 389): Sensitivity: 89.7 % Specificity: 87 % | 11.9 % (47/396) | NR |
Bias was small at both low (<400 cells/μl) and high (>400 cells/μl) | With sub-samples of FACSCalibur results of (100 to 300 cells/μl) | 12.8 % | At CD4 threshold of 200 cells/μl (in those ≥ 5 years old N = 389): Sensitivity: 86.7 % Specificity: 94.1 % | NR | |||||
At CD4 threshold of 350 cells/μl: Sensitivity: 94.7 % Specificity: 87.5 % | 6.7 % | ||||||||
With sub-samples of FACSCalibur results of (200 to 500 cells/μl) | 14.1 % | ||||||||
(Thakar, Mahajan et al. 2012) [31] | Relative bias: −9.1 %; LoA: −46 % to 27 %; N = 175 | NR | NR | NR | (Thakar, Mahajan et al. 2012) [31] | Among patients with CD4 < 350 cells/μl: relative bias: +4 % (N = 121) | At CD4 threshold of 350 cells/μl: Sensitivity: 96 %; Specificity: 91 % | NR | NR |
(Manabe, Wang et al. 2012) [41] | Bias: −66.3 cells/μl (LoA: −83.4 to +49.2; p < 0.001; N = 176) | NR | NR | 17.7 % | (Manabe, Wang et al. 2012) [41] | Bias: −68.5 cells/μl (LoA: −79.6 to −57.4; p < 0.001; N = 206) | NR | NR | 8.1 % |
Bias was smaller at lower CD4 counts (−10.8 cells/μl; LoA: −27.3 to +5.6; p = 0.19 for CD4 range 0–250 cells/μl) and much greater at higher CD4 count (−120.6 cells/μl; LoA: −162.8 to −78.4; p < 0.001 for CD4 > 500 cells/μl) | Bias was smaller at lower CD4 counts: +13.6 cells/μl (LoA: 2.52 to 24.7; p = 0.02 for CD4 range 0–250 cells/μl) and much greater at higher CD4 counts: −121.7 cells/μl (LoA: −147.9 to −95.4; p < 0.001 for CD4 > 500 cells/μl) | ||||||||
(Wade, Daneau et al. 2014) [35] | Relative bias: −0.9 %; (LoA: −57.3 to +55.6); N = 200 | At CD4 threshold of 200 cells/μl: Sensitivity: 100 % Specificity: 94 % | 4 % (16/410) | 4.5 % (9/200) | (Wade, Daneau et al. 2014) [35] | Relative mean bias: −9.4 % (LoA: −54.4 to +35.6) | At CD4 threshold of 200 cells/μl: Sensitivity: 98 % Specificity: 95 % | 3 % (14/440) | 6.5 % (13/200) |
Sub-samples of CD4 ≤ 200: 5 % (−78 to +89); CD4 200–500: 0 % (−49 to +49); CD4 ≥ 500: −8 % (−49 to +34) | At CD4 threshold of 350 cells/μl: Sensitivity: 87 % Specificity: 90 % | 13.4 % (55/410) | Sub-samples of CD4 ≤ 200: 1 % (LoA: −75 to 77); CD4 200–500: −11 % (LoA: −46 to +25); CD4 ≥ 500: −15 % (LoA: −34 to +4) | At CD4 threshold of 350 cells/μl: Sensitivity: 97 % Specificity: 80 % | 9 % (40/440) | ||||
At CD4 threshold of 500 cells/μl: Sensitivity: 97 % Specificity: 82 % | NR | At CD4 threshold of 500 cells/μl: Sensitivity: 99 % Specificity: 78 % | NR | ||||||
(Zeh, Inzaule et al. 2014) [30] | Bias: −44 cells/μl; N = 147 | At CD4 threshold of 350 cells/μl: Sensitivity: 86 % Specificity: 99 % | NR | NR | (Zeh, Inzaule et al. 2014) [30] | Bias: −86 cells/μl; N = 147 | At CD4 threshold of 350 cells/μl: Sensitivity: 94 % Specificity: 95 % | NR | NR |
(Arnett N 2013) [28] | Bias: 0 (PIMA –Microtube) and −20 cell/μl (PIMA –direct); N = 1060 | NR | NR | 8.6 % (Micro-tube) and 10.1 % (direct) | (Arnett N 2013) [28] | Bias: −10 cells/μl | NR | NR | 7.7 % |
Reference test = FACSCount | |||||||||
(Mwau, Adungo et al. 2013) [49] | Mean bias: +8.6 cells/μl (LoA: −235.4 to 252.7; N = 521) | At CD4 threshold of 350 cells/μl: Sensitivity: 79.4 % Specificity: 86.9 % | 16.5 % (86/521) | NR | (Mwau, Adungo et al. 2013) [49] | Mean bias: +7.8 cells/μl (LoA: −168.9 to 184.4; N = 822) | At CD4 threshold of 350 cells/μl (in those ≥ 5 years old N = 813): Sensitivity: 79.4 % Specificity: 83.4 % | NR | NR |
At CD4 of 200 cells/μl (in those ≥ 5 years old N = 813): Sensitivity: 83 % Specificity: 98.2 % | NR | ||||||||
(Thakar, Mahajan et al. 2012) [31] | Among patients with CD4 < 350 cells/μl: Mean relative bias: −5 % (N = 206) | At CD4 threshold of 350 cells/μl: Sensitivity: 92 %; Specificity: 91 % | NR |  | |||||
(Diaw, Daneau et al. 2011) [39] | Of 95 HIV (+) patients, Absolute bias: −39 cells/μl (LoA: −258 to +179) | At CD4 threshold of 200 cells/μl: Sensitivity: 91 % | 5.3 % (5/95); of finger-prick samples | 14 % total; 23 % in one study site | (Diaw, Daneau et al. 2011) [39] | For 100 HIV(+) patients, Absolute bias: −32 cells/μl (LoA: −146 to +84) | At CD4 threshold of 200 cells/μl: Sensitivity: 90 % | 4 % | 4.8 % |
Specificity: 97Â % | Specificity: 98Â % | ||||||||
Sub-samples of CD4 < 200: bias: +15 cells/μl (LoA: −89 to 118); Sub-samples of CD4 > 500: bias: −112 cells/μl (LoA: −429 to 204) | At CD4 threshold of 350 cells/μl: Sensitivity: 91 %; Specificity: 80 % | NR | Sub-samples of CD4 < 200: bias: +9.4 cells/μl (LoA: −76 to 94); Sub-samples of CD4 > 500: bias: −77 cells/μl (LoA: −217 to 63) | At CD4 threshold of 350 cells/μl: | NR | ||||
Sensitivity: 98Â % | |||||||||
Specificity: 79Â % | |||||||||
For 99 HIV(−) controls Absolute bias: −125 cells/μl (LoA: −434 to +184 cells/μl for all ranges of CD4 | |||||||||
|  |  |  |  |  | (Wade, Diaw et al. 2013) [37] | Bias: −30 cells/μl (LoA: −160 to 101; N = 128: 111 HIV+ & 17 HIV-) | At CD4 threshold of 200 cells/μl: | NR | NR |
Sensitivity 95Â % | |||||||||
Specificity 96Â % | |||||||||
Sub-samples of CD4 < 200: Bias: +6.0 cells/μl (LoA: −39 to +51) | At CD4 threshold of 350 cells/μl: | ||||||||
Sensitivity 97Â % | |||||||||
Specificity 90Â % | |||||||||
Sub-samples of CD4 > 500: Bias: −65 cells/μl (LoA: −224 to +93) | At CD4 threshold of 500 cells/μl:: Sensitivity 99 % Specificity 72 % | ||||||||
|  |  |  |  |  | (Malagun, Nano et al. 2014) [33] | Urban clinic: Bias: −46.4 cells/μl (LoA:-199.8 to 107.0); N = 139 | At CD4 threshold of 350 cells/μl: Sensitivity: 99.2 %; specificity: 77.1 % | 10.7 % | Error rate: 5.1 % |
Rural clinic: Bias: −55.8 cells/μl (LoA: −182.9 to 71.2); N = 98 | |||||||||
Reference test = Beckman-Coulter flow cytometry using Pan-leucogating (PLG) method | |||||||||
(Mnyani, McIntyre et al. 2012) [47] | Bias: −20.5 cells/μl (LoA: −175.0 to +133.9; p < 0.001; N = 296) | At CD4 threshold of 350 cells/μl: | 10.8 %; mostly in favor of patient treatment. | NR | (Myer, Daskilewicz et al. 2013) [40] | Bias: −22.7 cells/μl (LoA: −174.6 to 129.2); N = 546. | At CD4 threshold of 350 cells/μl: | 10 % | 61/546 samples required 83 additional test; 4 returned no result due to repeated machine errors |
No significant variability in the level of agreement related to age and gestational age | Sensitivity: 93 % (95 % CI 87–96), Specificity: 86 % (95 % CI 80–91) | Bias increased with increasing gestational age | Sensitivity: 92 % | ||||||
Specificity: 89Â %; | |||||||||
Sensitivity & specificity did not vary significantly across gestational age | |||||||||
(Glencross, Coetzee et al. 2012) [42] | Phase II (Hospital ANC clinic: Bias: −37.9 cells/μl (LoA: −389.1 to 309.8; N = 77 | NR | NR | NR | (Glencross, Coetzee et al. 2012) [42] | Phase II (Hospital ANC clinic: Bias: −19.6 cells/μl (LoA: −149.1 to 110.0; N = 91) | NR | NR | 10.4 % (5/48) & 20.9 % (9/43) for 2 devices |
Phase IIIA Rural/poor resourced clinic: Not applicable (NA) | NA | NA | NA | Substantial, clinically significant difference to predicate: Bias +105.7 cells/μl (LoA −336.1 to 547.5; N = 96) | Among 32 patients with CD4 < 350: 10 patients (31.2 %) would have missed ART (upward misclassification) | 6.8 % (7/103) | |||
Larger bias and wider LoA for samples with CD4 < 350: +131.4 cells/μl (LoA: −275.8 to +538.6; N = 32) as compared to samples with CD4 < 500: +102.3 cells/μl (LoA: −289 to 493.6; N = 52) = > increasing error at CD4 range of less than 350 cells/μl | |||||||||
Phase IIIB well resourced clinic: NA | NA | NA | NA | Results showed considerably less bias and tighter LoA variance, as compared to phase IIIA, irrespective of lancet used: lancet 1 (Sarstedt) bias: +8.9 cells/μl (LoA: −211.1 to 229; N = 87); lancet 2 (Caralet Blue) bias: −11.2 cells/μl (LoA: −147 to 124; N = 52) |  | 9 % (14/153) | |||
(Gous, Scott et al. 2013) [48] | Phase I: Multiple POC testing from multiple finger-sticks: mean bias was −32 cells/μl (N = 98) PIMA overestimate at low CD4 count (<350) and underestimate at high CD4 count (>500 cells/μl) | At CD4 threshold of 350 cells/μl: Sensitivity 86.4 %, Specificity 88.5 % | 12.4 % | 16.3 % |  | NA | NA | NA | NA |
Phase II: Multiple POC testing from single finger-stick: Mean bias - 30 cells/μl (N = 73) | At CD4 threshold of 350 cells/μl: Sensitivity 97.5 %, Specificity 95 % | 4.1 %; | 19.2 % |  | NA | NA | NA | NA | |
(Picken, Williams et al. 2014) [38] | Bias: 23.8 cells/μl (LoA: −166.1 to 213.8; N = 50 | At CD4 threshold of 350 cells/μl: Sensitivity: 88.9 %, specificity: 90.6 % | 10 % | 1.9 % |  |  |  |  |  |
Reference test = Partec Cyflow | |||||||||
(Mwau, Adungo et al. 2013) [49] | Mean bias: −10.0 cells/μl (LoA: −261.4 to 241.4; N = 162) | NR | NR | NR | (Mwau, Adungo et al. 2013) [49] | Mean bias: −24.2 cells/μl (LoA: −277.6 to +229.3; N = 407) | NR | NR | NR |
|  |  |  |  |  | (Thakar, Mahajan et al. 2012) [31] | Among patients with CD4 < 350 cells/μl: mean relative bias +8 % (N = 550) | At CD4 350 threshold: Sensitivity: 91 %; Specificity: 96 % | NR | NR |
Reference test = GUAVA | |||||||||
(Mwau, Adungo et al. 2013) [49] | Mean bias: +23.9 cells/μl (LoA −329.6 to 281.9; N = 176) | NR | NR | NR | (Mwau, Adungo et al. 2013) [49] | Mean bias: −0.3 cells/μl (LoA: −315.0 to 315.6; N = 191) | NR | NR | NR |