Assessing the agreement of biomarker data in the presence of left-censoring

BMC Nephrology

Table 1 Simulation results based on 1000 data sets with sample size of 100 -- Per cent censoring (25%, 25%)

Per cent censoring	True ρ	True ρ _c	Method	Mean ${\hat{ρ}}_{c}$	Relative bias (%)	Empirical SD	Mean SE	The percentage of 95% confidence intervals that include true value of CCC
(25%, 25%)	0.25	0.238	1. Delete the pair	0.1405	−40.97	0.1119	0.1087	85.2
			2. Replace by LOD	0.2072	−12.94	0.1214	0.0856	91.0
			3. Replace by 0.5 × LOD	0.1863	−21.72	0.1630	0.0848	88.2
			4. Replace by c × LOD	0.1951	−18.03	0.1240	0.0851	89.0
			5. ML	0.2346	−1.43	0.0944	0.0939	94.3
	0.50	0.476	1. Delete the pair	0.2966	−37.69	0.1197	0.0999	61.5
			2. Replace by LOD	0.4169	−12.42	0.1569	0.0735	86.4
			3. Replace by 0.5 × LOD	0.3664	−23.03	0.3516	0.0742	78.0
			4. Replace by c × LOD	0.3821	−19.73	0.1541	0.0744	78.3
			5. ML	0.4701	−1.24	0.0807	0.0780	93.6
	0.75	0.714	1. Delete the pair	0.5307	−25.67	0.0886	0.0762	38.6
			2. Replace by LOD	0.6582	−7.82	0.0996	0.0497	85.1
			3. Replace by 0.5 × LOD	0.6140	−14.01	0.1482	0.0519	71.5
			4. Replace by c × LOD	0.6140	−14.01	0.1662	0.0521	72.1
			5. ML	0.7077	−0.88	0.0525	0.0503	94.8

Abbreviations: LLD Lower Limit of Detection, ML Maximum Likelihood, C = Random number from uniform distribution (0, 1), CI confidence interval, Relative bias = $\frac{(Mean {\hat{ρ}}_{c} - True ρ_{c})}{True ρ_{c}} \times 100 %$ , CCC⁵= $ρ_{c} = \frac{2 COV (X, Y)}{VAR (X) + VAR (Y) + {(E (X) - E (Y))}^{2}}$ .

ISSN: 1471-2369