Bivariate random-effects meta-analysis and the estimation of between-study correlation

Table 1 The telomerase data taken from the bladder cancer review of Glas et al. [13]

Study	Number of patients with bladder cancer	Number of patients with bladder cancer and a positive test result	Sensitivity	Logit-sensitivity Y _i1	Standard error of logit-sensitivity S _i1	Number of patients without bladder cancer	Number of patients without bladder cancer and a negative test result	Specificity	Logit-specificity Y _i2	Standard error of logit-specificity S _i2
1	33	25	0.758	1.139	0.406	26	25	0.962	3.219	1.020
2	21	17	0.810	1.447	0.556	14	11	0.786	1.299	0.651
3	104	88	0.846	1.705	0.272	47	31	0.660	0.661	0.308
4	26	16	0.615	0.470	0.403	83	80	0.964	3.283	0.588
5	57	40	0.702	0.856	0.290	138	137	0.993	4.920	1.004
6	47	38	0.809	1.440	0.371	30	24	0.800	1.386	0.456
7*	43	23.5	0.547	0.187	0.306	13	12.5	0.962	3.219	1.442
8	33	27	0.818	1.504	0.451	20	18	0.900	2.197	0.745
9	17	14	0.824	1.540	0.636	32	29	0.906	2.269	0.606
10	44	37	0.841	1.665	0.412	29	7	0.241	-1.145	0.434

* 0.5 was added to each cell of the 2 by 2 table for this study, as a continuity correction was needed given there were zero false negatives

ISSN: 1471-2288