The study population has been previously described  and comprises three national Norwegian health studies conducted between 1974 and 2003 by the Norwegian National Health Screening Service. Selection of participants was based on year of birth and residence (municipality or county). The response rate in the three studies varied from 56 to 88% . Briefly, the three surveys used a similar protocol and study design, but there were some modifications made during different time periods, mainly due to questionnaires regarding smoking, physical activity and other lifestyle factors. Altogether 595,675 participants remained in the analytical cohort after exclusion of 40,091 participants due to emigration or death before the start of follow-up, missing information on vital status, measures of smoking exposure, education, or missing of any of the covariates included in the analyses.
The present study was approved by the Regional Committee for Medical Research Ethics South-East, Norway, and the National Data Inspectorate.
The questionnaires elicited information on current and former daily smoking, smoking duration in years (continuous), and average number (continuous) of cigarettes smoked per day.
Among the 373,283 ever smokers in our sample, the proportion of missing values was 5% (n = 18,886) for smoking duration, number of cigarettes per day, and pack-years (i.e., number of cigarettes smoked per day, divided by 20, multiplied by the smoking duration in years).
We categorized current smokers according to smoking duration in years (1–19, 20–29, ≥30), number of cigarettes smoked per day (1–10, 11–20, > 20), and pack-years (1–9, 10–19, ≥20).
We classified participants by level of education into three categories: < 10, 10–12, and ≥ 13 years by using the most recent information regarding duration of education obtained from Statistics Norway. We classified for physical activity in three: [sedentary (reading, watching television, and sedentary activity), moderate (walking, bicycling, and/or similar activities ≥4 h per week), and heavy (light sports or heavy gardening ≥4 h per week, heavy exercise, or daily competitive sports)] categories. We calculated BMI as weight in kg divided by height in m2 and classified in three and classified in three (< 18.5 kg/m2, 18.5–24.9 kg/m2, ≥25.0 kg/m2) categories. All variables were obtained at study enrollment. As questions on alcohol consumption were only included from 1994 onwards, information on alcohol consumption was missing in 73% of the participants in the analytical cohort.
Follow-up and endpoints
The data were linked to the Cancer Registry of Norway, the Norwegian Cause of Death Registry, and the Central Population Register by the national, unique 11-digit personal identification number. Lung cancer mortality was classified according to the eight, ninth and tenth revisions of The International Classification of Diseases (ICD-8, ICD-9, ICD-10). Follow-up ended at the time of death from primary lung cancer, death from any other causes, emigration, or the end of follow-up (December 31, 2013), whichever occurred first.
All deaths connected to primary incident carcinomas of the trachea, bronchus, and lung (ICD-8 code 162 or corresponding codes from ICD-9 and ICD-10) were included as endpoint, i.e. death from lung cancer.
We calculated the age-standardized (European Standard Population) overall lung cancer mortality rate by smoking status, and categories of education .
We used Cox proportional hazards model with attained age between cohort entry and exit as the underlying time scale to estimate the multivariable-adjusted hazard ratios (HRs) with 95% confidence intervals (CI), for the associations between different measures of smoking exposure and lung cancer mortality. We used stratified Cox models by cohort study and birth cohort (≤ 1950 and > 1950) to overcome any probable heterogeneity for these variables. A priori we considered alcohol, physical activity, BMI and education as possible confounders. We tested for interaction between smoking status and sex, and between smoking and education, and decided to stratify by sex and by education. We decided to adjust on BMI and physical activity, but did not include alcohol as a covariate because of a lot of missing data. We estimated dose-response associations among current smokers for the following variables measured continuoulsy: smoking duration in 10 years, number of 10 cigarettes smoked per day, number of 10-pack-years, and lung cancer mortality overall. Never smokers were not inluded in these analyses.
Subsequently, we tested for linear trend for smoking exposure (smoking duration, cigarettes smoked per day and pack-years) among current smokers based on the median value in each category, using the lowest category of each measure of smoking exposure as reference.
We used the Wald test to assess heterogeneity by sex and by education for the associations between different measures of smoking exposure and lung cancer mortality. We tested and found that the criteria for the proportional hazard assumption were met using Schoenfeld residuals (data not shown).
Subsequently, we performed the same analyses after excluding individuals who died from lung cancer within < 2 years of follow-up, and we also performed the same analyses after excluding participants with prevalent cancer.
We conducted all analyses using STATA version 14.0 (Stata Corp.). We considered two-sided p-values of < 0.05 as statistically significant.