*2.4. Potential Confounders*

From the questionnaire data at SAP2, we extracted additional potential confounders of the association between RTN at SAP2 and change in HbA1c between SAP2 and SAP3. These included age (continuous), sex (male/female), formal education ( ≤9 (primary level)/10–13 (secondary, apprenticeship level)/>13 years (tertiary level)), neighborhood socio-economic index (SEI) (continuous; incorporating education and income of household head, median rent and crowding of households within the neighborhood) [31], and green areas within 2 km residential buffer, available from the European Environment Agency hectare resolution dataset (continuous; CORINE CLC-2006 Version 13). We also considered smoking status (never/former/current), exposure to passive smoke (yes/no), consumption of alcohol (including liquor, wines, beers and spirits; ≤1/>1 glass/day), and body mass index (BMI), defined as weight and height-squared ratio (continuous). Towards sensitivity analyses, we extracted participants' responses to the question "do you often feel that you have not slept enough after you wake up in the morning?" from which we derived insufficient sleep (yes/no), as well as their responses to the Short Form 36-item mental health questionnaire (continuous) [32]. In SAP2, average annual residential exposure to nitrogen dioxide (NO2), a marker of road traffic-related air pollution and potential confounder of RTN [33], was assigned to participants' residences based on a hybrid model (adjusted R<sup>2</sup> = 0.83) derived by fitting NO2 sampler measurements using land-use and traffic variables, and predictions from a Gaussian dispersion model as previously described [34].
