Methods

Pre-Registration

The research questions, hypotheses, and planned analyses were pre-registered prior to data analyses: https://osf.io/946xu (adolescent samples) and https://osf.io/p79dm (student samples). Deviations are reported in Supplementary Material I.

Participants

This study was conducted following the principles of the Declaration of Helsinki. For all samples, participants provided informed consent before inclusion, and ethical approval was obtained. A short description of the included samples can be found below (for more information, see Supplementary Material II).

Sample 1: Kandinsky Longitudinal Study (KLS)

KLS is an ongoing project on Dutch adolescents’ socio-emotional well-being with annual waves of data collection of students in grades 7–10 (https://osf.io/8ztk5/). For this study, we included 828 adolescents from three cohorts: 7th-grade students in 2012 (cohort 1, n = 284), 2013 (cohort 2, n = 259), and 2014 (cohort 3, n = 286). At baseline, all students were in 7th grade (first year of secondary school in the Dutch educational system). Data from 196 adolescents were removed because they initiated smoking before baseline (n = 34), had missing data at baseline (n = 162), or did not have smoking information at follow-up (n = 32). Thus, the analytic sample consisted of 600 adolescents (cohort 1: n = 242, cohort 2: n = 177, cohort 3: n = 181). At W1, the sample included 52.7% females and 47.3% males. The mean age at W1 was 12.59 years.

Sample 2: G(F)OOD Together project (GFT)

GFT is a 6-wave longitudinal study on Dutch adolescents’ and their parents’ health behaviours (van den Broek et al., 2020, 2022); https://osf.io/bysgq/). At baseline (W1), students were in 7th or 8th grade. W1 and 3 took place before the COVID-19 pandemic (fall 2017, n = 667; spring 2019, n = 674), and W4 to W6 during the COVID-19 pandemic (i.e. fall 2020, n = 306; spring 2021, n = 142; fall 2021, n = 129). As loneliness with peers was not assessed in W2, this wave (and corresponding data) will not be used in the current study. At W4, adolescents were asked to retrospectively report about the first COVID-19 lockdown (W4 retro, spring 2020, n = 306). Data from 232 adolescents were removed because they initiated smoking before W1 (n = 28), had missing data at baseline (n = 131), or did not have smoking information at follow-up (n = 73). The analytic sample consisted of 551 adolescents. At W1, the sample included 54.3% females and 45.7% males. The mean age at W1 was 12.81 years.

Sample 3: Family & Health study (F&H)

F&H investigated socialization processes underlying health behaviours (Harakeh et al., 2005). In this study, 428 Dutch families were included with at least two children aged 13–16 years. Six waves of data were collected over 5 years. Baseline (W1) was collected between November 2002 and April 2003. The number of families at the yearly follow-up waves was 416 (W2), 404 (W3), 356 (W4), 326 (W5), and 313 (W6). In this project, data from the youngest sibling was selected. Data from 167 adolescents were removed because they initiated smoking before W1 (n = 153), had missing data at baseline (n = 8), or did not have smoking information at follow-up (n = 6). The analytic sample consists of 261 adolescents. At W1, the sample included 46.7% females and 53.3% males. The mean age at W1 was 13.80 years.

Sample 4: Healthy Student Life project (HSL)

HSL (van Hooijdonk et al., 2023; https://osf.io/e4d5b/) is a longitudinal online survey study that investigates the well-being, lifestyle, and individual and contextual factors of Dutch university students. So far, three waves of survey data have been collected (W1 fall 2021, n = 4.902; W2 spring 2022, n = 4.431; W3 spring 2023, n = 5.214). At each wave, all students enrolled at Radboud University (± 25.000) were invited to participate. In addition, at W3, students who graduated after W2 and gave permission for a follow-up were invited (W3 alumni, _n_invited = 484). After the exclusion of participants above 30 years old, with no longitudinal data (participation at W1 and at least W2 or W3 was required) or who reported inconsistently on smoking, the analytic sample consisted of n = 2.636 (W1), n = 2.168 (W2), and n = 1.666 (W3) participants, with complete data on both smoking and loneliness. At W1, the sample included 75.2% females, 23.3% males, and 1.5% gender other. The mean age at W1 was 22 years.

Sample 5: al-RISCO lifestyle project (al-RISCO)

Al-RISCO (Rubio et al., 2024; https://osf.io/urhdt/) is a two-year longitudinal survey study (early 2021–May 2023) to examine the cognitive, social, and emotional factors of Dutch university students’ lifestyles who reported heavy (episodic) drinking before the COVID-19 outbreak. Out of 1176 students screened (fall 2020), 634 students were classified as heavy drinkers and were invited to participate. The sample includes students of applied and research universities who were in the last year of their bachelor studies or any year of their master studies at the time of screening. For the current study, three waves of data were selected: W1 (spring 2021, n = 439), W2 (fall 2021, n = 415), and W3 (fall 2022, n = 366). After the exclusion of participants who reported inconsistently on smoking, the analytic sample consisted of n = 404 (W1), n = 377 (W2), and n = 326 (W3) participants, with complete data on both smoking and loneliness. At W1, the sample included 65.6% females, 33.9% males, and 0.5% gender other. The mean age at W1 was 22.55 years.