Table 2.
Hierarchical summary ROC model | No. of studiesb | Pooled no. of ER‐positive lesions | Sensitivity (95% confidence region) | Variation in sensitivity attributed to study heterogeneity (test of homogeneity) | Pooled no. of ER‐negative lesions | Specificity (95% confidence region) | Variation in specificity attributed to study heterogeneity (test of homogeneity) | Model area under the ROC curve |
---|---|---|---|---|---|---|---|---|
Metastatic lesions, IHC (primary analysis) | 4 | 69 | 0.78 (0.65–0.88) | I2 = 46% | 44 | 0.98 (0.65–1) | I2 = 0% | 0.98 |
Breast lesions, IHC (secondary analysis 1) | 3 | 60 | 0.86 (0.73–0.94) | I2 = 32% | 18 | 0.76 (0.52–0.90) | I2 = 0% | 0.87 |
Combined, IHC (secondary analysis 2) | 7 | 143 | 0.83 (0.72–0.90) | I2 = 53% | 64 | 0.83 (0.64–0.93) | I2 = 35% | 0.89 |
Combined, all reference standards (secondary analysis 3) | 11 | 211 | 0.81 (0.73–0.87) | I2 = 52% | 116 | 0.86 (0.68–0.94) | I2 = 78% | 0.89 |
A continuity correction of 0.1 was applied to all cells; random effects not fitted when sample size inadequate.
Excludes studies with no ER‐negative lesions (“Indeterminate” specificity in Table 1, although only one study is excluded from secondary analyses 2 and 3 for this reason—see Table 1 footnotes).
Abbreviations: ER, estrogen receptor; IHC, immunohistochemistry; ROC, receiver operating characteristic.