Table 2.

The variables which have been tested as predictors of TRS in the twelve studies included in this review

	Chan et al. (2014)	Demjaha et al. (2017)	Kim et al. (2017)	Lally et al. (2016)	Meltzer et al. (1997)	Sorensen et al. (2014)	Üçok et al. (2016)	Wimberley et al. (2016b)
Younger age of onset
Alcohol misuse during follow-up period
Antipsychotic polypharmacy during follow-up period
C-reactive protein
Comorbid diagnosis of personality disorder
Comorbid diagnosis of suicide attempts
Schizophrenia diagnosis
Paranoid schizophrenia diagnosis
Duration of first episode
Longer DUP
Early parental loss
Lower education qualification
Fewer years in education
Employment status
Black ethnicity
Family history of schizophrenia
Worse functioning
Worse premorbid functioning
Male
Living arrangements
Living arrangements during follow-up period
Marital status
Mode of onset
Paternal age
Polygenic risk score for schizophrenia
Relapse despite adherence
Relapse in first 6 months
More relapses in the first three years
Born in Autumn/winter
Substance misuse
Substance misuse during follow-up period
More symptoms of psychosis
Inpatient
Urbanicity at birth
Living in a more rural area at diagnosis
Violent offence
	Chan et al. (2014)	Demjaha et al. (2017)	Kim et al. (2017)	Lally et al. (2016)	Meltzer et al. (1997)	Sorensen et al. (2014)	Üçok et al. (2016)	Wimberley et al. (2016b)

NB: grey squares, variables not significantly associated with TRS; dark grey squares, variables significantly associated with TRS; all analyses using the Wimberley et al., Danish dataset were grouped under Wimberley et al. (2016b)