The electronic searches retrieved a total of 4,777 records; two additional reports were identified. In addition, eight reports were included in the 2006 systematic review. Duplicate records were removed, and the titles and abstracts of the remaining 2,896 records were screened yielding 632 potentially relevant studies. Full text screening identified 53 reports of 50 evaluations (Figure 1). Two of the three comparison groups include evaluations where more information is known about endorsement of CONSORT as a more direct intervention. Subsequently, we report the findings of these two comparison groups, ‘CONSORT endorsing journals compared to non-endorsing journals’ and ‘CONSORT endorsing journals before and after’. Please refer to the Cochrane review for results of the third comparison group, cross-sectional samples of RCTs before and after the publication of the CONSORT Statement.
CONSORT-endorsing journals compared to non-endorsing journals
Twenty-nine (of 53) evaluations were eligible for this comparison group. Across the 27 outcomes the number of studies per meta-analysis varied (median (IQR1, IQR3), 6 [5, 8]). Adequate reporting of the method of ‘Allocation Concealment’ and the description of flow of participants through the trial, ‘Participant Flow’, had the largest number of included studies (n=16), evaluating adequacy of reporting in 2,396 and 2,140 RCTs, respectively.
Of 27 outcomes evaluated, 25 had effect estimates indicating a relatively higher proportion of completely reported RCTs published in CONSORT-endorsing journals compared to non-endorsing journals. Of these, five were statistically significant at the 1% level (Figure 2); adequate details of method of ‘Allocation Concealment’ RR = 1.81 (1.25, 2.61) (16 evaluations, 2,396 RCTs, I2= 75%). This suggests that 81% more RCTs published in CONSORT-endorsing journals described allocation concealment more completely compared to those published in non-endorsing journals.
Other outcomes which resulted in statistically significant effects, favouring the completeness of reporting in CONSORT-endorsing journals were providing an adequate scientific explanation and rationale detailed in the ‘Introduction’ of the trial, RR = 1.07 (1.01, 1.14) (five evaluations, 513 RCTs, I2 = 0%), adequate description of how ‘Sample Size’ was determined RR = 1.61 (1.13, 2.29) (11 evaluations, 1,843 RCTs, I2 = 76%), and adequate description of the method used for ‘Sequence Generation’ RR = 1.59 (1.38, 1.84) (14 evaluations, 2,231 RCTs, I2 = 24%). Seven included studies evaluating 560 RCTs contributed to a statistically significant pooled effect in favour of CONSORT for ‘Total Sum Score’: SMD = 0.68 (0.38, 0.98) (I2 = 0%). This indicates that when adequate reporting was summarised across all CONSORT items, RCTs published in CONSORT-endorsing journals were more completely reported than RCTs published in non-endorsing journals.
Precise details of ‘Interventions’, CONSORT checklist item four, were equally reported in endorsing and non-endorsing journals, RR=1.0 (0.95, 1.05) (six evaluations, 638 RCTs, I2 = 0%), and eligibility criteria for trial ‘Participants’ was reported in six evaluations assessing a total of 683 RCTs, and resulted in an effect estimate less than 1.0, RR= 0.95 (0.56, 1.62). This suggests that the relative completeness of reporting for this item is slightly less in CONSORT-endorsing journals compared to non-endorsing journals.
Eight studies were not strictly compliant with the definition of a CONSORT-endorsing journal used in this review. Sensitivity analyses showed that only one outcome (of 27), although only minimally different, differed when evaluations that did not directly meet our definition of endorsement were excluded. Completeness of reporting of the scientific rational and background in the ‘Introduction’, was adjusted from RR = 1.07 (1.01, 1.14) to 1.05 (0.87, 1.27).
CONSORT-endorsing journals before and after CONSORT endorsement
Eleven (of 53) evaluations assessed journals that endorse the CONSORT Statement and presented RCT completeness of reporting of at least one CONSORT item before and after the journal’s date of endorsement of CONSORT. The number of RCTs assessed per outcome had a median (IQR) of 532 (512, 919). The number of reported CONSORT checklist items varied over evaluations, with a median (IQR) of 3 (2, 5). Adequate reporting of the method of ‘Sequence generation’ and the flow of participants through the trial, ‘Participant Flow’, were both reported in eight evaluations. For 15 of 27 outcomes data were reported in fewer than five evaluations. The results across all outcomes in this comparison are presented (Figure 3).