Skip to main content

Table 2 Description of unexplained discrepancies (N = 89)

From: Evidence of unexplained discrepancies between planned and conducted statistical analyses: a review of randomised trials

Unexplained changesN (%)
Analysis population
 Changed set of patients included by specifying additional exclusions12 (13%)
Analysis model
 Changed model13 (15%)
 Changed method of selecting analysis model2 (2%)
Covariates
 Changed from unadjusted to adjusted analysis7 (8%)
 Changed from adjusted to unadjusted analysis4 (4%)
 Changed set of covariates included in analysis5 (6%)
Missing data
 Changed from complete case to multiple imputation1 (1%)
 Changed imputation strategy2 (2%)
Unexplained additions
Analysis population
 Not mentioned in the original analysis plan8 (9%)
 Incomplete detail given in the original analysis plan7 (8%)
 Allowed analyst to subjectively choose analysis population based on the trial dataset1 (1%)
Analysis model
 Not mentioned in the original analysis plan5 (6%)
 Incomplete detail given in the original analysis plan5 (6%)
 Allowed analyst to subjectively choose analysis model based on the trial dataset7 (8%)
Covariates
 Not mentioned in the original analysis plan2 (2%)
 Incomplete detail given in the original analysis plan2 (2%)
 Allowed analyst to subjectively choose covariates based on the trial dataset3 (3%)
Missing data
 Not mentioned in the original analysis plan9 (10%)
 Incomplete detail given in the original analysis plan3 (3%)
 Allowed analyst to subjectively choose missing data approach based on the trial dataset1 (1%)