Randomized controlled trials always report the dose assessed and usually include a measure of adherence. By comparison, observational studies assessing medication safety often fail to report the dose used and rarely report any measure of adherence to therapy. This limits the ability to control for differences in doses used when undertaking meta-analyses. Non-adherence with therapy is common in the practice setting and varies across countries and settings. Inter-country differences in the registration of medicines may also result in different product strengths being available in different countries. These two factors combined means that observational studies undertaken for the same medicine in different settings may be assessing the same medicine but in circumstances where quite different dosages are used. Given that many adverse drug effects are dose dependent, differences in dosages used could be a factor explaining differences in risk estimates observed across studies. We argue that dose intensity, which can be defined as a product of the dose prescribed and adherence to the dose prescribed over the course of treatment, should be routinely reported in observational studies of medication safety. We illustrate the issue with the example of dabigatran. The randomized controlled trial evidence underpinning dabigatran's marketing authorization resulted in uncertainty about the appropriate dose for efficacy versus safety. As a result, different dosages of dabigatran were registered in the USA and Europe. The USA registered the 150- and 75-mg dabigatran products, while the 150- and 110-mg dabigatran products were registered in Europe. Among five observational studies subsequently undertaken to resolve the safety question concerning dabigatran and risk of bleeding, only one stratified results by dose. None of the US studies stratified results by the 75-mg dabigatran dose, despite this dose not being assessed in the original trial. None of the five studies reported adherence measures, despite three separate observational studies finding between 25 and 40 % of patients were non-adherent to dabigatran. The STROBE and RECORD statements should consider adding the requirement for reporting measures of dose intensity and its component products to improve observational study reports.