5. Introduction Major depressive disorder (MDD) is a devastating psychiatric disorder with significant morbidity and mortality from a host of conditions including cardiovascular disease (CVD) (Kemp and Quintana, 203). Unfortunately, patients must remain on their prescribed medication for at least four weeks without knowing whether their chosen antidepressant will be effective. This uncertainty prolongs patient suffering, increases societal burden, and imposes a huge economic cost through reduced productivity. Sometimes patients must try a variety of treatment options before symptoms are controlled, delaying the correct treatment for several months and increasing the risk of suicide. The Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study demonstrated that over four successive treatment steps cumulative remission rate is only 67 per cent (Rush et al., 2006). Disorder chronicity exaggerated by a lack of appropriate follow-up and care increases risk of CVD over the long-term (Rudisch and Nemeroff, 2003), highlighting an urgent need for adopting a per-sonalized medicine approach to MDD treatment. A recent study on over 500 000 participants (Scherrer et al., 202) who were free of cardiovascular and cerebrovascular disease at baseline reported that patients with treatment-resistant depression were .7 (95% CI .05–2.79) times more likely to die over an average follow-up period of 39 months, while insufficiently treated patients were 3.04 (95% CI 2.2–4.35) likely. A major impediment to research has been a focus on the heterogeneous diagnoses of MDD, as defined by current classifications (Diagnostic Statistical Manual, DSM, and the International Classification of Diseases, ICD), leading to inclusion of individual patients into studies with completely different symptoms. Unfortunately, the most recent version of the DSM, DSM-5, continues to base diagnoses on the presence of symptoms rather than specific features of the disorder such as underlying neural circuitry and associated behaviours. Several points regarding our review should be noted. First, we have previously reviewed the literature on predicting treatment response in depression (Kemp et al., 2008) and for brevity, do not reiterate key findings here. Instead, we comment on the progress that has been made over the last five years. Secondly, it is important to distinguish between markers of treatment response (i.e. a marker of current state) from predictors of treatment response, which refers to indicators of a future state. while many studies have focused on the impact of available treatments, these studies do not help clinicians wanting to determine 54 CHAPTER 5 Predictors of treatment response in MDD which treatment to prescribe to a particular patient. Thirdly, it is also important to note the distinction between treatment response, which is often used to refer to a 50 per cent reduction in symptoms on a primary outcome measure such as the Hamilton Rating Depression Scale (HRDS), versus treatment remission, which refers to a complete amelio-ration of symptoms. while studies have typically sought to predict whether or not a patient will respond to a particular treatment, more recent studies have placed a stonger emphasis on sustained remission (e.g.McGrath et al., 203). It is important to focus on remission rather than a reduction in symptoms because the clinically important residual symptoms are associated with ongoing functional disability and disorder recurrence. 5.2 Clinical information alone is insufficient for adequate prediction