IPDfromKM: reconstruct individual patient data from published Kaplan-Meier survival curves

Table 3 Hypothesis testing using log-rank test with true IPD versus with reconstructed IPD based on data points extracted using different software

	Results based		Results based on reconstructed IPD
	on true IPD		IPDfromKM		DigitizeIt		ScanIt
	HR(SE)	p-value	HR(SE)	p-value	HR(SE)	p-value	HR(SE)	p-value
Simulated curves:
1	0.47(0.12)	<0.001	0.48(0.13)	<0.001	0.47(0.12)	<0.001	0.47(0.12)	<0.001
2	0.40(0.18)	<0.001	0.40(0.18)	<0.001	0.40(0.18)	<0.001	0.40(0.18)	<0.001
3	0.70(0.12)	0.002	0.70(0.12)	0.003	0.71(0.12)	0.003	0.70(0.12)	0.002
4	1.06(0.16)	0.735	1.06(0.16)	0.727	1.05(0.16)	0.761	1.05(0.16)	0.780
5	0.76(0.13)	0.028	0.77(0.13)	0.044	0.77(0.13)	0.039	0.77(0.13)	0.037
6	0.48(0.16)	<0.001	0.48(0.16)	<0.001	0.48(0.16)	<0.001	0.48(0.16)	<0.001
Real trial example:
1	0.49(0.22)	0.068	0.48(0.41)	0.064	0.46(0.38)	0.041	0.45(0.41)	0.041
2	0.54(0.14)	0.014	0.54(0.26)	0.017	0.56(0.26)	0.020	0.53(0.26)	0.013
3	0.59(0.11)	0.005	0.58(0.19)	0.005	0.58(0.19)	0.005	0.59(0.19)	0.007
4	1.04(0.29)	0.871	1.06(0.26)	0.830	0.99(0.25)	0.959	1.03(0.26)	0.917
5	0.73(0.12)	0.040	0.72(0.15)	0.031	0.70(0.14)	0.019	0.72(0.15)	0.036

ISSN: 1471-2288