Loneliness and desire for thinness as pathways to girls' higher internalizing problems in the Tokyo Teen Cohort

The study provides comparable numbers for the sex difference in internalizing problems at age 16: depressive symptoms 15.9% in girls versus 8.2% in boys, and suicidal ideation 29.8% versus 18.1%, in a sample of 3,171 adolescents (46.9% girls) assessed at ages 12, 14 and 16. Earlier evidence on sex differences in adolescent mental health came largely from cross-cohort or cross-national pooling (the reference list cites Salk et al.'s meta-analysis of representative samples and de Looze et al.'s study of the widening gender gap across 43 countries, 2002–2022); what is added here is prevalence figures and a follow-up structure within one population cohort at three fixed ages. A prevalence description from an observational cohort: more than three thousand participants across three waves, with outcomes from self- and parent-report instruments (the reference list names SMFQ, SDQ and CBCL among the tools).

Candidate pathways were identified through coproduction with young people; loneliness and desire for thinness measured at age 14 were estimated to carry part of the sex effect on outcomes at age 16, at 34.7% of the total effect for depressive symptoms and 46.3% for suicidal ideation. The work moves from asking whether the difference exists or is widening to asking which social intermediates carry it, and the candidate factors came from young people's own perspectives rather than being set by researchers alone. The mediation estimates use the mediational g-formula for time-varying exposure and repeated-measured multiple mediators; the reported proportions are model estimates within a causal mediation inference on observational data.

The analysis pipeline is openly traceable: the parametric mediational g-formula was implemented with custom code in R version 4.5.1 and the main analysis code is hosted on GitHub, while the data cannot be shared openly because of TTC usage restrictions and proprietary variables, and de-identified data are available upon request subject to approval by the TTC steering committee. It brings causal mediation tools for time-varying mediators to the question of sex differences in adolescent mental health while giving both a code location and a data-request route; the Supplementary Information lists Methods 1–6, Fig. 1 and Tables 1–26. The article states the code repository, the data availability conditions and the supplementary material list, which makes the method and the conditions for replication locatable.

The readers this speaks to are researchers on adolescent mental health, designers of school and community mental health programmes, and public health or policy audiences interested in sex disparities; the setting to which it applies is an urban adolescent population with an observation window from age 12 to 16 and outcomes reported as internalizing problems at 16. Its value lies in unpacking the overall difference — girls reporting more internalizing problems — into concrete, measurable social intermediates, loneliness and desire for thinness, which can anchor intervention pilots, indicator selection and further measurement. It also demonstrates a research workflow in which young people help generate hypotheses, and this workflow can be borrowed by other cohorts to test local pathways. Whether the magnitudes travel depends on the distribution of loneliness and weight concerns in a target population and on local cultural expectations of girls.

Readers citing these proportions should keep in mind that mediation analysis rests on identification assumptions for observational data, so 34.7% and 46.3% are estimates under this cohort and model specification. The loaded record consists mainly of the abstract, author information, data and code availability statements, references and the supplementary material list; the specific values behind Fig. 1–3 and Supplementary Tables 1–26 are not included, so the covariate set, pathway coefficients and sensitivity details cannot be checked item by item here — that is what further reading would fill in. Data access is restricted under TTC licensing and some variables are proprietary, so external replication depends on request and approval; whether single-cohort, single-city results hold at a similar magnitude in other social and cultural settings is an open question for cross-cohort comparison.