Skip to main content

Table 2 Differences in achievement of remission at 6, 12 and 24 months in the ARCTIC trial (Norway; 2010–2015) compared to the NOR-VEAC observational study (Norway; 2010–2018) according to different approaches to missing dataa

From: Using observational study data as an external control group for a clinical trial: an empirical comparison of methods to account for longitudinal missing data

Approaches to missing data

Estimate

SE

Odds ratio

95% CI

p-value

6 months

 CCb

0.756

0.288

2.13

1.21–3.75

0.009

 Strict censoring + IPCWc

0.423

0.223

1.53

0.99–2.37

0.058

 MI + censoring + IPCWd

0.401

0.222

1.49

0.97–2.31

0.072

 MI + IPCWe

0.401

0.220

1.49

0.97–2.30

0.069

 MI for allf

0.401

0.225

1.49

0.96–2.32

0.074

12 months

 CCb

1.093

0.294

2.98

1.68–5.31

< 0.001

 Strict censoring + IPCWc

0.879

0.259

2.41

1.45–4.00

< 0.001

 MI + censoring + IPCWd

0.797

0.250

2.22

1.36–3.62

0.001

 MI + IPCWe

0.768

0.243

2.16

1.34–3.47

0.002

 MI for allf

0.687

0.248

1.99

1.22–3.23

0.006

24 months

 CCb

0.429

0.288

1.54

0.87–2.70

0.136

 Strict censoring + IPCWc

0.407

0.284

1.50

0.86–2.62

0.151

 MI + censoring + IPCWd

0.448

0.279

1.57

0.91–2.70

0.108

 MI + IPCWe

0.401

0.268

1.49

0.88–2.52

0.134

 MI for allf

0.367

0.253

1.44

0.88–2.38

0.147

  1. SE Standard error, CI Confidence interval, CC Complete case analyses, IPCW Inverse probability of censoring weighting, MI Multiple imputation
  2. aFor all approaches, inverse probability of treatment weighting using the propensity score was used to balance the two cohorts on relevant baseline covariates
  3. bAnalyses in a subset of patients with complete follow-up data for the main outcome
  4. cCensoring of subjects with missing outcome data at a visit or intermittent missing visits, whichever occurred first. IPCW used to account for missing data due to drop-out (naturally occurring or created due to censoring)
  5. dMI used to account for missing outcome data at visits. Censoring of subjects with intermittent missing visits. IPCW used to account for missing data due to drop-out (naturally occurring or created due to censoring)
  6. eMI used to account for missing outcome data at a visit and intermittent missing visits. IPCW used to account for naturally occurring drop-out
  7. fMI used to account for all missing data