Fig. 3

Feasibility and growing use of cautious headlines and main claims in observational research (error bars are 95% CIs). Feasibility is indicated by the increase in spontaneous use in pre-intervention (draft) press releases since the baseline period (2014/15). Final press releases showed small further increases in cautious wording following suggestions in the trial. For each bar, n reports the total number of press releases in that analysis group (i.e. the denominator of the proportion that the bar displays)