Skip to main content

Table 5 Performance measures of various methods for handling 45% missingness in recurrences for individuals treated with artemether-lumefantrine

From: Dealing with indeterminate outcomes in antimalarial drug efficacy trials: a comparison between complete case analysis, multiple imputation and inverse probability weighting

Full data Kaplan-Meier estimate of day 28 cure (SE) = 0.960 (0.1559)

 

Complete case analysis

MI

IPW

IPW-E

Mechanism 1

 Bias

0.0159 (0.0165)

-0.0026 (0.0079)

0.0002 (0.0075)

0.0039 (0.0097)

 Relative bias

1.65 %

-0.27 %

0.02 %

0.41 %

 Model based SE

0.2152 (0.0007)

0.2001 (0.0008)

0.2097 (0.0009)

0.2536 (0.0014)

 Empirical SE

0.2032 (0.0045)

0.1854 (0.0041)

0.2015 (0.0045)

0.2582 (0.0058)

 Coverage

13.4 % (1.1)

94.4 % (0.7)

95.4 % (0.7)

85.1 % (1.1)

 RMSE a

0.5732 (0.0079)

0.1914 (0.0014)

0.2028 (0.0023)

0.2918 (0.0044)

Mechanism 2a (Weak scenario)

 Bias

0.0163 (0.0173)

-0.0022 (0.0084)

0.0001 (0.0077)

0.0039 (0.0104)

 Relative bias

1.70 %

-0.23 %

0.01 %

0.41 %

 Model based SE

0.2176 (0.0008)

0.2041 (0.0008)

0.2123 (0.0009)

0.2594 (0.0017)

 Empirical SE

0.2053 (0.0046)

0.1901 (0.0044)

0.2028 (0.0046)

0.268 (0.0062)

 Coverage

12.3 % (1.0)

95.5 % (0.7)

95.7 % (0.6)

83.7 % (1.2)

 RMSE a

0.5907 (0.0085)

0.1938 (0.0018)

0.2039 (0.0024)

0.3019 (0.0051)

Mechanism 2b (Strong scenario)

 Bias

0.0167 (0.0169)

-0.0026 (0.008)

0.0001 (0.0076)

0.004 (0.0102)

 Relative bias

1.74 %

-0.27 %

0.01 %

0.41 %

 Model based SE

0.2200 (0.0008)

0.2060 (0.0008)

0.2151 (0.0009)

0.2659 (0.0016)

 Empirical SE

0.2060 (0.0046)

0.1978 (0.0043)

0.2051 (0.0045)

0.2772 (0.006)

 Coverage

10.1 % (1.0)

94.5 % (0.7)

95.7 % (0.6)

83.1 % (1.2)

 RMSE a

0.6085 (0.0082)

0.2031 (0.0016)

0.2063 (0.0023)

0.3119 (0.0046)

  1. MI Multiple Imputation, IPW Inverse Probability Weighting, IPW-E Inverse Probability Weighting with recurrence status excluded; SE= Standard Error; RMSE= Root Mean Squared Error; K-M = Kaplan-Meier estimates
  2. a=Monte Carlo error for the RMSE presented on mean squared error scale
  3. Monte Carlo Standard Errors shown in parentheses