Article Text

other Versions


Interventions to prevent self-harm: what does the evidence say?
  1. Kate E Saunders1,2,
  2. Katharine A Smith1,2,3
  1. 1Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford, UK;
  2. 2Oxford Health NHS Foundation Trust, Warneford Hospital, Oxford, UK;
  3. 3NIHR Oxford Cognitive Health Clinical Research Facility, Warneford Hospital, Oxford, UK
  1. Correspondence to Dr Kate E Saunders, Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 7JX, UK; kate.saunders{at}

Statistics from


Self-harm (SH) refers to intentional self-poisoning or self-injury, irrespective of motive or the extent of suicidal intent.1 Thus, it includes acts intended to result in suicide (sometimes referred to as ‘attempted suicide’), those without suicidal intent (eg, to communicate distress or reduce unpleasant feelings) and those with a mixed (or unclear) motivation. It is a major public health concern and a common cause of assessment, treatment and admission to general hospitals.2 However, the majority of those who SH do not present to hospital, and community prevalence is estimated to be as high as 20% in those aged 15 years.3 SH presentations increase significantly from the age of 12 years and decline from the mid-20s onwards.4 Repetition of SH is common, with up to 25% of individuals who present to hospital following SH returning to the same hospital following further SH within a year. Repetition is also frequent (54.8%) in adolescents who do not present to clinical services.5 At least 80% of individuals presenting to hospital following SH are suffering from at least one mental disorder6 (including major depression, bipolar disorder, anxiety and substance misuse, often in combination with personality disorders). However, the aetiology is complex; SH is often associated with acute life events, often against a background of longer term social and personal difficulties (such as relationship problems, financial difficulties or social isolation). A history of SH is associated with a significantly increased risk of subsequent suicide.7 One to three per cent of those who present to hospital after SH will die by suicide in the following year.7 Of those who die by suicide, over 50% have a history of SH and 15% have presented to hospital with SH in the preceding year.8 SH is not only a problem in terms of healthcare. It is also associated with poorer educational outcomes as well as significant health and social care costs.9 ,10 Improving the treatment of SH has been a focus of a number of national and international guidelines,1 ,11 although the paucity of evidence for effective treatments has been highlighted.12 The aetiology of SH is complex, encompassing not only mental health problems, but also psychological and social influences. Treatment options to be appraised should include pharmacological and psychological treatments, and studies of the role of social and societal change. As SH is a problem that often emerges during the teenage years, it is important to encompass studies in children and adolescents as well as adults.


To identify relevant evidence for the treatment and clinical management (whether short or long term) of SH, we searched PubMed, PsycINFO, Google Scholar and the Cochrane Library for systematic reviews and meta-analyses of randomised controlled trials (RCTs) published between 1980 and May 2016. No language constraints were applied and the following key words were used: self-harm, suicide attempt, non-suicidal self-injury, treatment. In case of multiple publications on the same topic, only the most recent or most comprehensive article was considered. The reference lists of reports identified were used to find additional publications.


Three Cochrane reviews were identified13–15 which focused on pharmacological treatments, psychological treatments and interventions for children and adolescents. These Cochrane reviews are related. They were completed by the same group of authors (encompassing experts in suicide research in the UK, Ireland, Australia and Belgium), and the three together update a single Cochrane review originally published in 1999. The update was divided into three reviews to allow space for the assessment of secondary outcomes where possible.

Pharmacological treatments for SH

Given the high prevalence of depressive illness and depressive symptoms in people who SH,6 antidepressants would seem to be possible candidates for prevention of recurrence. Mood stabilisers may also be a possibility, as there is evidence of a specific antisuicidal effect of lithium when used to treat people with affective disorders.16 Antipsychotic medication, particularly in low doses, might also be considered, especially in those who frequently repeat SH, and or those with a diagnosis of borderline personality disorder.

The Cochrane review included seven RCTs with a total of 546 patients.14 Interestingly, the reviewers did not find any new trials when they searched for this update, compared with their original search for the 1999 Cochrane review.12 It is not clear why this might be, but the difficulties in conducting RCTs of in this group in general are particularly pronounced when considering drug treatment. All studies were assessed using the risk of bias tool17 and were reported as being of low or very low quality, and no data on adverse events were reported. Information on psychiatric diagnosis was reported only in some trials and additional comorbidity in only one trial. Thus, it was not possible to make any association between the effects of different types of pharmacological treatment on repetition of SH according to diagnostic group.

  • Meta-analysis revealed no significant treatment effects on repetition of SH for antidepressants (OR 0.76, 95% CI 0.42 to 0.36), lithium (0.99, 0.33 to 2.95), low-dose fluphenazine (1.51, 0.50 to 4.58) or natural products (1.33, 0.38 to 4.62). The antipsychotic flupenthixol was found to significantly reduce SH in a single trial (0.09, 0.02 to 0.50) but numbers were small (n=30). These findings are in contrast to the observational data, particularly for lithium. The Cochrane review's analysis (of 167 participants) found no beneficial effect for lithium on repetition of SH, or on the secondary outcomes of depression score, hopelessness, suicidal ideation or suicide. Although with such a relatively small number of participants, it is very unlikely that there was sufficient power to show an effect on a relatively rare event such as suicide, the other data are somewhat surprising, given the evidence supporting the role of lithium in the reduction of suicidal behaviour. A systematic review and meta-analysis of 48 RCTs of lithium in 6674 patients with mood disorders found that lithium was more effective than placebo at reducing suicides but had little discernible effect on SH.16 However, the results of a number of observational studies suggest that lithium may reduce SH18 ,19 when compared with other anticonvulsants. In the specific population of patients with bipolar disorder, a recent large (n=14 396) population-based electronic health records study20 showed that rates of SH and unintentional injury were lower in the group treated with lithium compared with other mood stabilisers (valproate, olanzapine, quetiapine). It has been argued that the findings may be confounded by the secondary benefits of being on lithium (such as repeated blood tests, more clinic attendances). However, in a large naturalistic longitudinal study of non-fatal SH in individuals with bipolar disorder which replicated the protective effects of lithium, no difference was observed in the number of physician contacts in patients on lithium compared with those on other medications.

  • Other potential treatments such as ketamine and buprenorphine were not included in the review. There is an emerging evidence to support a rapid antidepressant effect of ketamine, a glutamate N-methyl-d-aspartate receptor antagonist, in the treatment of unipolar and bipolar depression.21 ,22 In depressed patients in the emergency department, a rapid reduction in suicidal ideation was observed following an intravenous infusion of ketamine with the reduction being maintained for up to 10 days.23 Suicidal cognition was eradicated in patients with depression given three times weekly infusions. Recent evidence suggests that these effects are mediated by the reduction in non-suicide-related depressive symptoms.24 Open-label studies investigating suicidal ideation and anhedonia suggest that both are reduced after acute intravenous ketamine.25 Further studies are needed to assess whether this effect is related to the antidepressant effect or is independent of it. In a multisite randomised, double-blind, placebo-controlled trial, ultra low-dose sublingual buprenorphine was associated with a significant reduction in Beck Suicidal Ideation Scores at 2 and 4 weeks in severely suicidal patients without substance abuse.26

  • Electroconvulsive therapy (ECT) is widely acknowledged as an effective treatment for severe depressive episodes and is cited as a treatment for suicidal behaviour by the American Psychiatric Association.26 In some observational studies, ECT appears to significantly reduce the frequency of suicide attempts in depressed patients.27 ,28 While none of these approaches are included in the Cochrane review as no RCTs are yet available, initial findings suggest that they may be a helpful approach in the treatment of SH.

Psychological treatments

This Cochrane review identified 55 RCTs for psychological interventions for SH.13 The most common treatment modality was cognitive behavioural therapy (CBT)-based psychological therapy (18 trials). The majority of these trials explored the effects of a 1:1 intervention delivered in fewer than 10 sessions. The remaining 37 trials evaluated a range of other psychological interventions including dialectical behavioural therapy (DBT), mentalisation-based therapy (MBT), group-based psychotherapy and remote contact interventions (postcards, emergency cards, telephone contact). A significant treatment effect for CBT-based therapy compared with treatment as usual was observed with respect to repetition of SH (OR 0.70, 95% CI 0.55 to 0.88) although no reduction in the frequency of SH was found. While just 6% fewer people repeated SH following CBT, improvements in mood, hopelessness and suicidal thoughts were also observed. Most studies were small in nature and there was considerable variability between trial outcomes. The effects of the other therapeutic approaches remain unclear as they were mostly evaluated in small single studies. The evaluation of psychological treatments within an RCT is more challenging than that of pharmacological treatments as patients and therapists are aware of the treatment modality that they are receiving such that true blinding cannot be achieved. There is also the issue of the comparison arm of the trial. In the review, all ‘active’ treatments were compared with ‘treatment as usual’. While this addresses to some extent the non-specific effects of any type of psychological treatment (face-to-face contact, feeling that another individual is listening, etc), treatment as usual varies between centres in each country and between countries. This means that combining the studies together in a meta-analysis, while increasing the power, is not strictly comparing like with like.

In addition, the studies of psychological treatments did not routinely document the ‘fidelity’ to the treatments. In other words, we do not know for certain how many of the offered sessions (either active treatment or treatment as usual) the participant actually attended. In addition, we cannot be certain how closely therapists adhered to the manualised therapy. So, by drawing a parallel with pharmacological studies, we do not know the ‘dose’ of active or placebo treatment that the participants received. While the findings for brief CBT-like therapy are encouraging, the trials did not identify for whom this intervention would be most effective. This is important—while the data suggest the best evidence that we have is that CBT should be offered to those who SH, we do not know whether it should be offered to all or a subgroup. Given the finite resources in psychological services, identifying the ‘active’ elements of the CBT intervention is a key question. Only studies comparing the full CBT intervention with the same intervention but missing a defined key element could answer this question.

Implementing this strategy into clinical care also presents other problems. Only a small percentage present to services following SH,29 so hospital-based interventions will inevitably miss a significant number of the target population. In addition, psychiatric services are focused on providing treatment resources to those with a diagnosed mental illness, and many patients with SH will not meet standardised criteria. The most persuasive argument for implementing such a service would be an economic one. SH carries a significant direct and indirect cost in healthcare (in addition to the costs to the individual and their family).10 Future studies of CBT would be enhanced by including assessments of healthcare costs as an integral part of the study. While CBT is the most widely available modality of psychological treatment in the UK following the launch of the IAPT (Improving Access to Psychological Therapies) scheme,30 those who SH or are expressing suicidal ideation are often deemed too high risk to be eligible and are referred to specialist services where waiting times are often much longer. The single trials of other psychotherapies that were associated with favourable outcomes, for example, DBT and MBT, were all conducted in patients diagnosed with borderline personality disorder. The use of a specified diagnostic group means that there is clearer and more consistent framework for psychological intervention and understanding of SH as well as a more homogeneous patient group.

Treatments for children and young people

In their third Cochrane review Hawton et al15 identified only 11 trials of interventions for SH in children and adolescents. This was a surprisingly low number of trials, given that under 18s are a key target age group for SH interventions and suicidality is a relevant clinical issue in child and adolescent psychiatry.31 ,32 Most interventions were limited to single trials and the quality of the evidence was mostly graded as low. No trials of pharmacological interventions were identified. Neither group-based therapy nor DBT (adapted for use in adolescents) was found to be associated with a reduction in the proportion of participants engaging in SH compared with treatment as usual (OR 0.72, 0.12 to 4.40 and 1.72, 0.56 to 5.24 for DBT and group therapy, respectively), although a reduction in the frequency of SH over time was observed following DBT. There were also significantly greater reductions in depression, hopelessness and suicidal ideation in this group. MBT was associated with fewer adolescents scoring above the cut-off for repetition of SH based on the Risk-Taking and Self-Harm Inventory 12 months postintervention, although this study was limited to individuals who had multiple episodes of SH or emerging personality problem.

School-based interventions were not included in the review as they generally target individuals irrespective of whether they have self-harmed or not. In a recent multicentre cluster RCT, a Youth Aware of Mental Health Programme intervention significantly reduced the incidence of suicide attempts (OR 0.45, 0.24 to 0.85) and suicidal ideation at 12-month follow-up (0.50, 0.27 to 0.92) compared with the control group. The reported reduction was more than 50% compared with the control group. Similar reductions were not associated with a manualised gatekeeper programme or screening for high-risk individuals by professionals.33 However, there are concerns about large-scale implementation of interventions in schools. While it is logical to focus on schoolchildren at an early age before SH is likely to start, careful assessment of risk needs to be undertaken. It is possible that interventions may harm as well as benefit. For the child and adolescent population, the issue of ‘contagion’ is particularly important (Hawton 2012).34


All three Cochrane reviews summarise relatively small numbers of studies which are of poor or moderate quality. Meta-analysis, where it was feasible, revealed small changes in SH but the majority of approaches evaluated involved single studies. The Cochrane methodology exclusively explores RCT data, and in the area of SH, this significantly limits their interpretation because of the many methodological challenges (see box 1). Observational data may be just as informative and perhaps more generalisable to clinical practice. The best example if this is the overwhelming observational evidence supporting the antisuicidal effects of lithium for which no RCT data exist. Testing interventions for the prevention or reduction of SH are limited by a number of methodological considerations. The ethics of randomising individuals who report suicidal ideation often means that trials can never be truly representative. The question of who to target an intervention towards is also an area of debate. The majority of studies focus on hospital presenters or those known to mental health services, whereas the majority of SH (particularly in younger people) occurs in individuals where this is not the case.29 Other studies chose to focus on those who repeatedly harm themselves or limited recruitment to individuals who had taken overdoses. More effective interventions may be those that target individuals prior to the emergence of SH or promote mental well-being in the population as a whole. The majority of completed suicides occur in individuals who are not known to mental health services at the time of their death such that there is an imperative for population intervention as opposed to focusing on high-risk groups.

Box 1

Methodological challenges in designing studies to assess the efficacy of interventions for self-harm

Methodological challenges in assessing the efficacy of interventions for self-harm:

Heterogeneity of a study population recruited on the basis of a behaviour;

Which population to study;

Definitions of self-harm;

Study context;

Timing of intervention;

Time to follow-up;

Ethics of randomising suicidal patients;

Finding an appropriate control intervention (particularly in psychological treatment trials).

SH describes a range of different behaviours and is associated with nearly all mental disorders.6 The reasons for SH vary widely as to the methods and lethality. While a number of models for the emergence of SH and suicidality have been proposed, all highlight the complex nature of the behaviour and none provide a clear single focus for intervention. In view of the heterogeneous nature of the SH as well as the absence of a mechanistically valid treatment target, it is not surprising that treatment effects are small at best. Data from the treatment of specific disorders, for example, mood disorders indicate that significant reductions in SH can be achieved by targeting the disorder rather than the SH per se.16 The settings in which individuals who SH are most commonly encountered are the emergency department, general practice or in education. All of the treatment trials included in the three Cochrane reviews were limited to trials conducted in inpatient or outpatient settings which further limits the generalisability of the findings.

The three Cochrane reviews identified highlight the challenges of testing interventions for a SH that are generalisable to routine clinical practice. They summarise the available randomised controlled evidence available, and therefore provide useful suggestions for effective therapies and inform areas for further research. However, the reliance on RCT data, while methodologically robust, leads to a limited summary of the available evidence and overlooks a number of important interventions for the reduction of SH. Given the many ethical and methodological challenges inherent in RCTs for SH, observational data are an important complementary source of evidence which is generalisable to clinical and non-clinical settings. In conclusion, there is limited RCT evidence for interventions for SH, whereas observational studies in specific diagnostic group highlight a number of effective treatments, in particular lithium.


The authors acknowledge support from the NIHR Oxford Cognitive Health Clinical Research Facility.



  • Disclaimer The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.