Skip to main content

Table 2 Details of adjusted analysis

From: Reporting on covariate adjustment in randomised controlled trials before and after revision of the 2001 CONSORT statement: a literature review

Year of publication 2000
(n = 84)*
2006
(n = 113)*
% difference
(95% CI)
P-value
Performed adjusted analysis on primary outcome 50 (90.9%) 93 (93.9%) 30.0 (-5.9 to 12.0) 0.5
Reason for adjustment     0.4
   Imbalance in covariates 9 (10.7%) 12 (10.6%)   
   Prognostic covariates 6 (7.1%) 15 (13.3%)   
   Both 0 3 (2.6%)   
   Other reasons 3 (3.6%) 4 (3.5%)   
   Not mentioned 66 (78.6%) 79 (69.9%)   
Choice of covariates     0.5
   All pre-specified 5 (5.9%) 8 (7.1%)   
   All suggested by data 12 (14.3%) 20 (17.7%)   
   Combination of pre-specified and post hoc 0 3 (2.6%)   
   Not mentioned 67 (79.8%) 82 (72.6%)   
Number of covariates adjusted for §     0.02 ||
   1 39 (46.4%) 36 (31.8%)   
   2 23 (27.4%) 33 (29.2%)   
   3-5 14 (16.7%) 25 (22.1%)   
   6-9 2 (2.4%) 12 (10.6%)   
   Not mentioned 6 (7.1%) 7 (6.2%)   
Covariate used for adjustment     
   Outcome assessed at baseline 33/62 (53.2%) 55/81 (67.9%) 14.7 (-1.4 to 30.7) 0.07
   Centre/Country 31/49 (63.3%) 25/61 (41.0%) -22.3 (-40.6 to -39.9) 0.02
   Assessed after randomisation 9/78 (11.8%) 9/107 (8.4%) -3.1 (-12.0 to 5.7) 0.5
All stratification factors were adjusted for 11/25 (44.0%) 20/46 (43.5%) -0.5 (-24.7 to 23.6) 1.0
Explicitly specified nature of analysis     
   Primary analysis 2 5   
   Secondary/sensitivity analysis 5 5   
Type of outcomes     0.8
   Binary 13 (15.5%) 19 (16.8%)   
   Continuous 65 (77.4%) 81 (71.7%)   
   Ordinal 1 (1.2%) 3 (2.6%)   
   Time-to-event 5 (5.9%) 10 (8.9%)   
Adjusted analysis method used was mentioned for specific outcome in the Method section 62/78 (79.5%) 88/109 (80.7%) 1.2 (-10.4 to 12.9) 0.8
  1. *One adjusted analysis selected per study only
  2. Number of studies that have specified primary outcomes: Year 2000 = 55 and Year 266 = 99
  3. Year 2000: Clinical relevance (n = 1), significant at 3 weeks after randomisation (n = 1), exploring role of baseline variables (n = 1); Year 2006: Mediated treatment effect on outcome (n = 1), related to compliance/adherence of treatment (n = 2), effect of outcome decline over time (n = 1)
  4. §Year 2000: 6 studies did not report number of covariates; Year 2006: 6 studies did not report number of covariates and 1 stated at least 2 covariates
  5. || Mann-Whitney test
  6. Did not have statistical method section: Year 2000 (n = 6), Year 2006 (n = 5)