Purpose The return to work (RTW) of people with mood and anxiety disorders is a heterogeneous process. We aimed to identify prototypical trajectories of RTW over a two-year period in people on sick leave with mood and anxiety disorders, and investigate if socio-demographic or clinical factors predicted trajectory membership. Methods We used data from the randomized IPS-MA trial (n = 283), evaluating a supported employment intervention for participants with recently diagnosed mood or anxiety disorders. Information on "weeks in employment in the past 6 months" was measured after 1/2, 1, 1 ½ and 2 years, using data from a nationwide Danish register (DREAM). Latent growth mixture modelling analysis was carried out to identify trajectories of RTW and logistic regression analyses were used to estimate predictors for trajectory membership. Results Four trajectory classes of RTW were identified; non-RTW [70% (196/283)] (practically no return to work); delayed-RTW [19% (56/283)] (6 months delay before full RTW); rapid-unstable-RTW [7% (19/283)] (members rapidly returned to work, but only worked half the time); and the smallest class, rapid-RTW [4% (12/283)] (members rapidly reached full employment, but later experienced a decrease in weeks of employment). Self-reported disability score according to the SDS, not living with a partner, and readiness to change on the CQ scale were found to be significantly associated with RTW. Conclusion The trajectories identified support that many do not benefit from vocational rehabilitation, or experience difficulties sustaining employment; enhanced support of this patient group is still warranted.Trial registration: http://www.clinicaltrials.gov (Identifier: NCT01721824).