We took several steps to improve the estimation of alcohol use as a risk factor. First, on the exposure side, we added 26 survey series, which contributed 12 195 datapoints in our models. Second, we developed and implemented a method that adjusts total consumption for tourism and unrecorded consumption for each location-year. Third, we calculated the TMREL. We chose TMREL as being the exposure that minimises an individual's risk of suffering burden from any given cause related to alcohol (appendix 1 p 22 for more detail). Fourth, we performed a systematic review of all cohort and case-control studies reporting a RR, hazard ratio, or odds ratio for any risk-outcome pairs studied in GBD 2016 and then modelled a dose-response relationship using DisMod ordinary differential equations (ODE).18 Fifth, we estimated injury PAFs from cohort studies and adjusted them to account for victims.
