Background: It is unclear as to the extent to which psychological interventions focusing specifically on depression and anxiety are helpful for people with physical health conditions, with respect to mood and condition management. Aims: To evaluate the effectiveness of a modified evidence-based psychological intervention focusing on depression and anxiety for people with type 2 diabetes mellitus (T2DM), compared with a control intervention. Method: Clients (n = 140) who experienced mild to moderate depression and/or anxiety and had a diagnosis of T2DM were allocated to either diabetes specific treatment condition (n = 52) or standard intervention (control condition, n = 63), which were run in parallel. Each condition received a group intervention offering evidence-based psychological interventions for people with depression and anxiety. Those running the diabetes specific treatment group received additional training and supervision on working with people with T2DM from a clinical health psychologist and a general practitioner. The diabetes specific treatment intervention helped patients to link mood with management of T2DM. Results: Both conditions demonstrated improvements in primary outcomes of mood and secondary outcome of adjustment [95% confidence interval (CI) between 0.25 and 5.06; p < 0.05 in all cases]. The diabetes specific treatment condition also demonstrated improvements in secondary outcomes of self-report management of T2DM for diet, checking blood and checking feet, compared with the control condition (95% CIs between 0.04 and 2.05; p < 0.05 in all cases) and in glycaemic control (95% CI: 0.67 to 8.22). The findings also suggested a non-significant reduction in NHS resources in the diabetes specific treatment condition. These changes appeared to be maintained in the diabetes specific treatment condition. Conclusions: It is concluded that a modified intervention, with input from specialist services, may offer additional benefits in terms of improved diabetic self-management and tighter glycaemic control.