Online Resource 5
Adjusting Overall Survival Estimates After Treatment Switching: A Case Study in Metastatic Castration-Resistant Prostate Cancer
Konstantina Skaltsa1, Cristina Ivanescu2, Shevani Naidoo3, De Phung4, Stefan Holmstrom5, Nicholas R. Latimer6
1Quintiles Advisory Services, C/Sardenya, 537 – 539 08024, Barcelona, Spain; 2Quintiles, Siriusdreef 10 Beukenhorst Zuid 2132 WT, Hoofddorp, Netherlands; 3Astellas Medical Affairs, Global Health Economic Outcomes Research (HEOR), 2000 Hillswood Dr, Chertsey, Surrey KT16 0PS, UK; 4Astellas Pharma Global Development, Sylviusweg 62, 2333 BE, Leiden, Netherlands; 5Astellas Medical Affairs, Global HEOR, Sylviusweg 62, 2333 BE, Leiden, Netherlands; 6School of Health and Related Research, University of Sheffield, Regent Court, 30 Regent Street, Sheffield, S1 4DA, UK
Correspondence: Shevani Naidoo, Astellas Medical Affairs, Global HEOR, 2000 Hillswood Dr, Chertsey, Surrey KT16 0PS, UK; email:
Table IPCW: Pooled logistic regression analysis on probability of remaining uncensored by informative censoring defined as switching, conditioned on baseline and time-dependent covariates, used to calculate the stabilised weight denominator. Full and restricted models (September 2013 data cutoff)
/ Enzalutamide / Placebo /Parameter / Parameter estimate (SE) / OR
[95% CI] / Parameter estimate (SE) / OR
[95% CI] /
Full model
Intercept / 4.7983 (1.4588) / 5.3623 (1.1804)
Age (continuous) / 0.0007 (0.0129) / 1.001 [0.976; 1.026] / -0.0182 (0.0104) / 0.982 [0.962; 1.002]
Time since diagnosis (<5 years vs. ≥5 years) / 0.6492 (0.2078) / 1.914 [1.274; 2.876] / 0.1295 (0.1668) / 1.138 [0.821; 1.578]
Number of bone metastases at screening (≤5 bone mets vs. >5 bone mets) / -0.0456 (0.2103) / 0.955 [0.633; 1.443] / -0.4227 (0.2105) / 0.655 [0.434; 0.990]
Presence of visceral disease at baseline (yes vs. no) / -0.4105 (0.2804) / 0.663 [0.383; 1.149] / 0.3559 (0.2597) / 1.427 [0.858; 2.374]
Type of disease progression at study entry (PSA progression only vs. no disease progression) / 0.0574 (0.7951) / 1.059 [0.223; 5.031] / -0.2705 (0.5394) / 0.763 [0.265; 2.196]
Type of disease progression at study entry (radiographic progression with or without PSA vs. no disease progression) / 0.0868 (0.7938) / 1.091 [0.230; 5.168] / -0.4240 (0.5390) / 0.654 [0.228; 1.882]
Study day (linear term) / -0.0372 (0.0050) / 0.964 [0.954; 0.973] / -0.0577 (0.0063) / 0.944 [0.932; 0.956]
Grade 3/4/5 AE at treatment discontinuation
(yes vs. no) / 0.3008 (0.3494) / 1.351 [0.681; 2.679] / -0.2507 (0.5394) / 0.778 [0.374; 1.620]
History of grade 3/4/5 AE (yes vs. no) / -0.1987 (0.2913) / 0.820 [0.463; 1.451] / 0.6746 (0.3110) / 1.963 [1.067; 3.612]
Corticosteroid use at treatment discontinuation (yes vs. no) / 0.3189 (0.2868) / 1.376 [0.784; 2.413] / 0.1461 (0.2465) / 1.157 [0.714; 1.876]
ECOG PS prior to treatment discontinuation
(0 vs. ≥1) / -0.3578 (0.2198) / 0.699 [ 0.455; 1.076] / -0.5186 (0.1851) / 0.595 [ 0.414; 0.856]
LDH level prior to treatment discontinuation
(≤240 IU/mL vs. >240 IU/mL) / -0.6795 (0.3104) / 0.507 [0.276; 0.931] / -0.2617 (0.2478) / 0.770 [0.474; 1.251]
PSA level prior to treatment discontinuation (continuous) / 0.0006 (0.0006) / 1.001 [0.999; 1.002] / 0.0002 (0.0002) / 1.000 [1.000; 1.001]
Baseline utility index (continuous) / -1.9352 (0.9251) / 0.144 [0.024; 0.885] / -0.7467 (0.6969) / 0.474 [0.121; 1.857]
Utility index prior to treatment discontinuation (continuous) / -0.7152 (0.6004) / 0.489 [0.151; 1.587] / 0.5021 (0.5643) / 1.652 [0.547; 4.993]
Baseline FACT-P total score (continuous) / 0.0053 (0.0078) / 1.005 [0.990; 1.021] / -0.0054 (0.0068) / 0.995 [0.981; 1.008]
FACT-P total score prior to treatment discontinuation (continuous) / 0.0019 (0.0070) / 1.002 [0.988; 1.016] / -0.0048 (0.0054) / 0.995 [0.985; 1.006]
Time to progression (continuous) / 0.8123 (0.1256) / 2.253 [1.761; 2.882] / 1.5212 (0.1838) / 4.578 [3.193; 6.563]
Progression (yes vs. no) / 0.3488 (0.2038) / 1.417 [0.951; 2.113] / 0.4958 (0.1670) / 1.642 [1.184; 2.278]
Time since treatment discontinuation (linear term) / 0.0332 (0.0056) / 1.034 [1.023; 1.045] / 0.0653 (0.0062) / 1.067 [1.055; 1.081]
Restricted model
Intercept / 5.0910 (0.7799) / 3.6183 (0.6586)
Time since diagnosis (<5 years vs. ≥5 years) / 0.6343 (0.1950) / 1.886 [ 1.287; 2.764]
Study day (linear term) / -0.0361 (0.0049) / 0.965 [ 0.955; 0.974] / -0.0566 (0.0063) / 0.945 [ 0.933; 0.957]
ECOG PS prior to treatment discontinuation (0 vs. ≥1) / -0.4406 (0.2004) / 0.644 [ 0.435; 0.953] / -0.4938 (0.1723) / 0.610 [ 0.435; 0.856]
LDH level prior to treatment discontinuation (≤240 IU/mL vs. >240 IU/mL) / -0.8017 (0.2989) / 0.449 [ 0.250; 0.806]
Baseline utility index (continuous) / -1.7065 (0.7379) / 0.181 [ 0.043; 0.771]
Time to treatment discontinuation (continuous) / 0.8127 (0.1241) / 2.254 [ 1.767; 2.875] / 1.4861 (0.1817) / 4.420 [ 3.096; 6.311]
Time since treatment discontinuation (linear term) / 0.0360 (0.0043) / 1.037 [ 1.028; 1.045] / 0.0641 (0.0062) / 1.066 [ 1.053; 1.079]
Number of bone metastases at screening (≤5 bone mets vs. > 5 bone mets) / -0.5293 (0.1995) / 0.589 [ 0.398; 0.871]
History of grade 3/4/5 AE (yes vs. no) / 0.5568 (0.2054) / 1.745 [ 1.167; 2.610]
Baseline FACT-P total score (continuous) / -0.0103 (0.0049) / 0.990 [ 0.980; 0.999]
Progression (yes vs. no) / 0.4519 (0.1630) / 1.571 [ 1.142; 2.163]
AE adverse event, CI confidence interval, ECOG PS Eastern Cooperative Oncology Group Performance Status, FACT-P Functional Assessment of Cancer Therapy-Prostate, IPCW inverse probability of censoring weights, LDH lactate dehydrogenase, mets metastases, OR odds ratio, OS overall survival, PSA prostate-specific antigen, SE standard error