Table 4.

Univariable and multivariable negative binomial regression model testing the association between any/monthly/weekly disordered eating behaviours at age 14 years and autistic social traits at age 16 years. Model based on participants with complete exposure data and imputed confounder and outcome data

	SCDC scores at age 16 years
	Any univariable model RRa	Any multivariable model RRa
Any disordered eating at age 14 years
Any vs. none	1.29 (1.13, 1.49) p = .0002	1.07 (0.95, 1.22), p = .28
Frequency of disordered eating at age 14 years
Monthly vs. none	1.20 (1.01, 1.41), p = .04	1.04 (0.85, 1.21), p = .66
Weekly vs. none	1.47 (1.17, 1.84), p = .0008	1.14 (0.92, 1.40), p = .23

Multivariable model adjusted for child’s sex and BMI, depressive symptoms and autistic traits at age 14 years (baseline); maternal: education, age, prenatal BMI, history of eating disorders and depressive symptoms.

Abbreviations: RR, relative risk; SCDC, Social and Communication Disorders Checklist.

^a Relative risk is derived from exponentiating the coefficients of negative binomial regressions, modelling the logs of SCDC scores in adolescents with disordered eating (i.e. any, monthly or weekly) compared to those with no disordered eating, holding confounder variables in the model constant.