Skip to main content
. 2021 Oct 13;2(12):e704–e714. doi: 10.1016/S2666-5247(21)00214-7

Table 3.

Pooled sensitivity and specificity using Reitsma's random-effects model, pooled diagnostic OR, AUC, and heterogeneity for primary subgroup analysis

Number of studies (sample size) Pooled sensitivity (95% CI) Pooled specificity (95% CI) Pooled diagnostic OR (95% CI) AUC I22)
LAMP
Overall 31 (3453) 83·3% (76·9–88·2) 96·3% (93·8–97·8) 188·37 (84·10–421·91) 0·963 0% (3·96)
Specimen
Nasopharyngeal swabs 10 (1004) 87·7% (78·3–93·4) 92·4% (85·6–96·2) 107·46 (29·47–391·83) 0·955 19% (3·43)
Oropharyngeal or throat swabs 11 (1046) 87·6% (77·9–93·4) 91·3% (84·9–95·1) 99·10 (30·17–325·49) 0·952 21% (3·09)
Saliva 5 (773) 86·2% (75·5–92·6) 89·4% (83·0–93·5) 68·15 (22·631–205·25) 0·939 0·3% (2·72)
Sputum 7 (513) 81·7% (76·3–86·1) 98·6% (96·1–99·5) 397·60 (127·59–1238·96) 0·949 0% (0·0)
RNA extraction
With RNA extraction 17 (2445) 88·0% (82·1–92·2) 95·1% (90·9–97·4) 288·67 (88·63–940·17) 0·961 4·14% (4·72)
Without RNA extraction 14 (1008) 73·6% (62·4–82·5) 97·0% (94·1–98·5) 106·50 (37·60–301·69) 0·962 0% (2·13)
Primer–probe set
ORF1ab primer 6 (1008) 86·2% (60·8–96·2) 98·1% (96·4–99·3) 362·39 (33·76–3890·00) 0·984 8% (7·14)
N primer 10 (1136) 84·4% (72·9–91·6) 97·5% (93·9–99·0) 260·98 (115·11–591·70) 0·975 0% (0·0)
E primer 2 (146)* 81·6% (71·0–89·5) 100% (94·9–100) 300·35 (38·53–2341·37) 0·901 0% (0·0)
RNA extraction method
Magnetic beads 5 (408) 81·3% (53·7–94·3) 97·2% (89·3–99·3) 180·22 (56·55–574·39) 0·973 0% (0·0)
Silica spin column 19 (1634) 83·7% (77·8–88·2) 95·1% (91·1–97·4) 198·02 (71·74–546·61) 0·943 0% (3·4)
dPCR
Overall 15 (783) 94·1% (88·9–96·9) 78·5% (57·4–90·8) 77·47 (24·68–243·19) 0·946 0·5% (2·79)
Specimen
Pharyngeal swabs 7 (459) 95·0% (87·5–98·1) 90·1% (60·3–98·2) 237·354 (39·35–1431·75) 0·958 7% (3·38)
Saliva 3 (152)* 89·7% (75·8–97·1) 77·0% (68·1–84·4) 18·76 (3·46–101·62) 0·865 0% (1·16)
Sputum 2 (98)* 100% (93·4–100) 88·6% (75·4–96·2) 283·29 (32·304–2484·26) 0·979 0% (0·0)
Primer–probe set
ORF1ab primer 11 (544) 97·3% (93·8–98·8) 80·9% (53·7–93·9) 158·584 (39·77–632·42) 0·964 0·8% (2·46)
N primer 15 (783) 94·1% (88·9–96·9) 78·5% (57·4–90·8) 77·47 (24·68–243·19) 0·946 0·5% (2·79)
RNA extraction method
Silica spin column 8 (536) 92·3% (83·0–96·7) 81·6% (52·9–94·6) 76·52 (15·35–381·44) 0·937 10% (3·6)
Automatic 4 (173) 96·5% (86·7–99·2) 91·0% (75·2–97·1) 310·37 (59·85–1609·41) 0·972 0% (0·0)
qPCR
Overall 25 (3667) 92·7% (88·3–95·6) 92·9% (87·2–96·2) 231·27 (86·77–616·44) 0·967 0% (4·60)
Specimen
Pharyngeal swabs 13 (2250) 96·1% (91·5–98·3) 94·7% (88·1–97·8) 638·96 (215·32–1896·09) 0·982 0% (2·07)
Saliva 10 (1175) 86·3% (78·6–91·6) 90·9% (79·3–96·3) 88·02 (24·89–311·28) 0·921 0% (3·07)
RNA extraction
With RNA extraction 19 (3230) 93·7% (88·4–96·6) 92·4% (85·3–96·3) 262·47 (81·57–844·57) 0·971 0% (5·63)
Without RNA extraction 6 (437) 89·7% (81·9–94·3) 94·3% (79·2–98·6) 157·61 (28·96–857·70) 0·940 16% (2·53)
Primer–probe set
ORF1ab primer 5 (1369) 97·3% (87·4–99·5) 88·8% (85·7–91·2) 527·47 (44·33–6276·02) 0·892 0% (6·16)
N primer 12 (899) 92·9% (87·0–96·3) 91·3% (79·9–96·5) 192·23 (83·16–444·36) 0·961 4% (0·53)
E primer 11 (2442) 92·3% (83·3–96·6) 94·1% (85·5–97·7) 208·13 (40·84–1060·61) 0·971 8% (6·3)
RNA extraction method
Magnetic beads 7 (1170) 92·9% (86·8–96·3) 97·0% (88·4–99·3) 462·91 (83·03–2580·88) 0·968 25% (3·76)
Silica spin column 10 (1805) 93·4% (82·9–97·6) 84·0% (69·4–92·4) 96·50 (21·89–425·37) 0·948 6% (4·52)
Automatic 2 (224)* 96·7% (88·5–99·6) 91·5% (86·1–95·3) 350·48 (42·61–2882·13) 0·961 0% (0·0)

Data are n (N); sensitivity, specificity, or OR (95% CI); AUC; or I22). Summary of subgroup analyses for LAMP, dPCR, and qPCR assays based on the number of true false-positive and false-negative cases reported in the 71 experiments. All dPCR assays required RNA extraction. OR=odds ratio. AUC=area under the curve. LAMP=loop-mediated isothermal amplification. ORF1ab=open reading frame 1 ab. N=nucleocapsid. E=envelope. dPCR=digital PCR. qPCR=quantitative PCR.

*

A univariate random-effects model was used to estimate sensitivities and specificities for subgroups with less than four included studies, as bivariate models do not converge when the sample size is small.