Skip to main content
. 2017 Jan 13;24(4):697–708. doi: 10.1093/jamia/ocw168

Table 1.

Clinical characteristics of the included subjects and precautionary drug exposure

Algorithm CLEAR ClDE pairs MetaLAB MLlDE pairs MetaLAB MLLRS MetaNurse MNNRS
Gold standard for ADRs DE pairsa RS-ADR
Type of ADR signals Predefined laboratory abnormalities MedDRA PTs
No. of target drugs 10 101
No. of ADR signals 47 (from 40 tests) 117 (from 48 tests) 757 (from 1357 SNSs)
No. of drug-ADR pairs 470 11 817 76 457
No. of positive pairsb 221 2210 34 857
No. of negative pairsc 249 9607 41 600
AUC, not integratedd,e 0.55 ± 0.06 0.61 ± 0.18 0.69 ± 0.11 0.62 ± 0.07
AUC, SOC-integratedf 0.84 ± 0.13 0.84 ± 0.09
No. of patients 68 769 88 038 215 088 220 954
No. of exposure casesg 90 804 127 171 1 028 724 1 187 037
Age (years)h 51.3 ± 18.9 50.7 ± 19.9 46.5 ± 23.0 46.22 ± 23.1
Female, n (%)i 38 290 (55.67) 47 470 (53.91) 108 410 (50.40) 110 864 (50.17)
Disease severity, n (%)j
 1 22 755 (33.08) 29 706 (33.74) 106 997 (49.74) 111 410 (50.42)
 2 29 225 (42.49) 37 611 (42.72) 82 306 (38.26) 83 422 (37.75)
 3 10 485 (15.24) 12 619 (14.33) 19 238 (8.94) 19 824 (8.97)
 4 3306 (4.80) 4167 (4.73) 4130 (1.92) 4025 (1.82)
 5 2882 (4.19) 3747 (4.25) 2304 (0.01) 2174 (0.09)
 6 116 (0.16) 188 (0.21) 113 (0.0005) 99 (0.0004)

aPredefined DE pairs for 10 drugs and 47 laboratory abnormalities reported by Yoon et al.13

bPositive pairs were established by expert review, and the remaining pairs were considered to be cNegative DE pairs.

dAUCs were computed by considering all the drugs and ADRs as a single dataset.

eCLEAR significantly outperformed MetaCLEAR (DeLong’s test for 2 ROC curves, P = .0137).

fAUCs were computed for each MedDRA SOC by stratifying ADRs and then integrating.

gNumbers of patient exposures to target drugs were summed by separately counting the exposures of each patient to different target drugs.

hAge differed significantly between CLEAR and MetaLAB (P = .01) and between MetaLAB and MetaNurse (P = 4.68 × 10–7) in Student t test.

iGender did not differ significantly between the comparison groups (P = .89 and .93).

jDisease severity did not differ significantly between CLEAR and MetaLAB (P = .029), but it did differ significantly between MetaLAB and MetaNurse (P = 3.09 × 10–6) in Fisher’s exact test.