Some meta-analyses of randomized placebo-controlled trials on antidepressants conclude that there might be an increased risk for suicidal behaviour, especially in children and adolescents but also in adults. Placebo-controlled trials exclude patients with serious suicidality and might therefore underestimate the risk of respective adverse events. The change of suicidal ideation and the prevalence of suicides and non-fatal suicide attempts were therefore analysed in a large naturalistic prospective multicentre study of depressed in-patients. Additionally, specific risk factors for new emergence of suicidal ideation were investigated. The naturalistic prospective study was performed in 12 psychiatric hospitals of the German research network on depression and suicidality (seven psychiatric university hospitals and five district hospitals) in Germany. All patients (n=1014) were hospitalized and had to meet DSM-IV criteria for major depressive disorder. Six events were defined for the purposes of statistical analysis: ‘emergence’, ‘extended emergence’, ‘improvement’ and ‘worsening of suicidal ideation’, ‘suicide attempts’ and ‘suicides’. Logistic regression analysis and classification and regression trees (CART) analyses were conducted to determine specific risk factors for new emergence of suicidal ideation. The mean HAMD total score decreased from 24.8 at baseline to 10 after 10 wk. An effect on suicidality was evident by week 2 in the sense of a decrease of the mean HAMD item-3 score. Emergence, worsening and improvement of suicidal ideation occurred in 3.2%, 14.74% and 90.79% of patients, respectively. A total of 10 suicide attempts and two suicides were reported. The rate of suicides (13.44/1000 patient-years) was rather low and comparable to the rate observed in randomized controlled antidepressant trials. Five risk factors for emergence of suicidal ideation were determined with two independent statistical methods: age (with higher risk at age <45 yr), treatment resistance, number of hospitalizations, presence of akathisia and comorbid personality disorder. Age <45 yr as one of five risk factors for the emergence of suicidal ideation is in line with the meta-analysis performed for the recent US Food & Drug Administration (FDA) memorandum; although the naturalistic study design does not permit definite conclusions to be made about certain compounds. The rate of suicides was comparable to that seen in randomized controlled trials, as were the rates of emergence and worsening of suicidal ideation, but more improvement was found. Thus, in-patient treatment in a psychiatric care setting, including daily assessments of suicidality by trained psychiatrists adhering to the rules of good clinical practice (e.g. use of specific co-medications, supportive psychotherapy and continuous medical attendance by nursing staff) might be beneficial.