AIMS: To (1) examine several factors associated with trends in heavy episodic drinking (HED) in Finland, Norway and Sweden, (2) investigate similarities in these associations across the countries and (3) analyse the contribution of these factors to the trend in HED and the differences across the countries.

DESIGN AND SETTING: Observational study using five waves of the European School Survey Project on Alcohol and Other Drugs (ESPAD) from Finland, Norway and Sweden between 1999 and 2015.

PARTICIPANTS: A total of 18,128 male and 19,121 female 15- to 16-year-old students.

MEASUREMENTS: Monthly HED, perceived access to alcohol, truancy, parental control, leisure time activities and daily smoking. The Cochran-Armitage test was used to examine linear time trends in HED. Logit regression models using the Karlson-Holm-Breen (KHB) method were fitted for each country separately, including all the independent variables together with time and adjusted for family status, parental education and gender.

FINDINGS: In Finland, Norway and Sweden, perceived access to alcohol, truancy and daily smoking decreased significantly between 1999 and 2015 while risk perceptions, parental control and participation in sports increased in the same period. The confounding percentage of all the independent variables related to the trend in HED was 48.8%, 68.9% and 36.7% for Finland, Norway and Sweden, respectively. Decline in daily smoking (p<.001) and perceived access to alcohol (p<.001) were positively and increase in parental control (p<.001) negatively associated with the decline in HED in all three countries. Changes in truancy, going out with friends, and engaging in sports and other hobbies had little or no impact on the decline in HED or displayed no consistent results across the countries.

CONCLUSIONS: The decline in adolescent heavy episodic drinking in Finland, Norway and Sweden between 1999 and 2015 appears to be associated with a decline in adolescent daily smoking and perceived access to alcohol and an increase in parental control.