The Scottish Government

« Previous | Contents | Next »

Listen

CHAPTER FOUR: EVIDENCE FROM THE HIGHEST QUALITY STUDIES

4.1 The scoping review provided an overview of outcomes for the primary literature largely independent of any consideration of study quality. To evaluate the best evidence currently available and also to identify aspects of current research design which could be improved on in future, we also evaluated outcomes on the basis of the quality of design and implementation shown by individual studies. Details of the approach taken are set out in Annex H. To summarise, we used a total of 15 aspects of study design and implementation to evaluate each study individually and then, in order to compare 'like with like' we selected the strongest studies from within the two categories of quantitative and qualitative design and, within each of these broad categories, selected the highest quality studies evaluating, respectively, interventions for completed suicide, for attempted suicide, for self-harm and for suicidal ideation. The 'best' studies in each category were defined as those studies achieving a total 'quality score' equal to or exceeding the median quality score for that category (that is for either quantitative or qualitative studies addressing a particular aspect of suicidal behaviour or ideation). It is important to recognise that there are significant caveats surrounding the now quite common use of summative quality scores to evaluate research studies. However, they can provide a useful rule of thumb in evaluating the likely robustness of identified outcomes and it is in this light we present findings from our quality evaluation of the available studies.

Suicide

4.2 Five of the highest quality studies identified (Owens et al 2004, Meltzer et al 2003, Milstein et al 1986, Tondo et al 1998, Zenere & Lazarus 1997) addressed completed suicide as an outcome. Studies following a quantitative methodology identified no significant improvements in outcomes following the use of clozapine or olanzapine (Meltzer et al 2003) or ECT (Milstein et al 1986) in psychiatric populations. One study (Tondo et al 1998) of lithium treatment for bipolar disorder identified a higher rate of suicidal acts prior to lithium treatment and, for those discontinuing lithium treatment, also in the first of 5 years following treatment discontinuation. This study failed, however, to specifically differentiate completed suicide from 'suicidal acts' as a whole. Positive outcomes from this study are to an extent supported also by the outcomes from a broader range of lower quality studies. However, some evidence of possible increases in suicide following discontinuation of treatment with lithium and the subsequent reduction in incidents of suicide only back to baseline following 2 or more years of discontinuing lithium in this study argue for caution in over-interpreting the possible benefits of this intervention.

4.3 One final quantitative study (Zenere & Lazarus 1997) reported positive outcomes following the introduction of a school crisis management programme. However, the overall number of suicides reported in this study were so low (7 at initiation of the programme down to 5 in the fifth year of the programme) that in practice random variation may equally well have accounted for the outcomes observed. The only high quality qualitative study addressing completed suicide (Owens et al 2004) followed a retrospective psychological autopsy design. This study evaluated whether or not detection and treatment of mental ill health by GPs had any association with the rate of completed suicide. The authors concluded that detection and treatment of mental ill health by GPs had been adequate and could therefore not be held as accountable for suicide. This study relates specifically to 'detection and treatment' of those who chose to present to a GP and of these nearly one quarter (24%) of patients failed to have their mental illness detected or treated. It therefore remains open to question whether for this sub-group of patients or for the broader group of patients with similar problems who fail to present themselves to a GP mental illness is or is not subsequently associated with suicide.

4.4 In summary, evidence from high quality studies of the effectiveness of available interventions to reduce completed suicide is very limited. This somewhat pessimistic outcome should be set against the broader range of studies which, whilst having less robust methodological approaches, nevertheless provide some suggestions for promising avenues to pursue. For example, studies addressing the restriction of access to means and ongoing contact with suicidal people following discharge from hospital. It is clear that a major research initiative is required in this field if practitioners are to be given the opportunity to pursue evidence-based intervention. In the interim, it could be of value to cautiously pursue the more promising approaches identified by the broader range of literature, bearing in mind the lack of high quality studies reporting unequivocal outcomes for interventions to prevent completed suicide.

Attempted suicide

4.5 Five of the highest quality studies (Kuipers & Lancaster 2000, Perseius et al 2003, Brown et al 2005, Meltzer et al 2003, Zenere & Lazarus 1997) directly addressed outcomes for attempted suicide, a further study of lithium treatment (Tondo et al 1998), referred to above, addressed attempted suicide in combination with completed suicide under the umbrella term 'suicidal acts'. All of these studies reported positive outcomes. In respect of lithium treatment, the caveats set out above also apply to outcomes for attempted suicide. Considering the other studies, three were quantitative studies (Brown et al 2005, Meltzer et al 2003, Zenere & Lazarus 1997) and two were qualitative studies following a content analysis/grounded theory methodology (Kuipers & Lancaster 2000, Perseius et al 2003). Of the quantitative studies, 2 RCTs (Brown et al 2005, Meltzer et al 2003) provide statistical support for their outcomes.

4.6 The first of the RCTs reported significantly greater reductions in attempted suicide following treatment with CBT in comparison with treatment as usual ( TAU) for people attending A&E as a consequence of self-harm (a repetition rate of 24% versus 42%). The second RCT reported significantly greater reductions in attempted suicide following treatment with clozapine versus treatment with olanzapine for people with schizophrenia (a repetition rate of 7% versus 11%). It is unfortunate that the latter study did not include a placebo or non-pharmaceutical TAU arm. Without either of these comparisons in place, it is not possible to evaluate whether clozapine, in addition to outperforming olanzapine, is also able to achieve better outcomes than interventions with fewer side-effects. Both olanzapine and clozapine have known side-effects, but the main side-effect associated with clozapine (agranulocytosis) is particularly severe and also results in increased treatment costs via the need for ongoing monitoring of patients during the course of treatment. It is also worth noting in this context, that a broader range of additional, albeit lower quality, studies are more equivocal in their support for clozapine.

4.7 The one quantitative study (Zenere & Lazarus 1997) which failed to provide statistical analysis of outcomes nevertheless provided figures demonstrating a substantive decline in attempted suicide over the course of a school crisis intervention programme. The study is also referred to above in relation to completed suicide. With regards to attempted suicide, however, outcomes are more convincing. The reported incidence of attempted suicide fell from 243 at baseline to a figure of 95 during the fourth and fifth years of the study. Whilst these figures look very promising, it should still be noted that the number of individuals to whom the programme was delivered was extremely large (330,000) and that the authors did not control for random variation or for natural trends downwards. It is important therefore to avoid over-interpreting these results. Finally, two qualitative studies addressed outcomes for attempted suicide. The first study (Kuipers & Lancaster 2000) evaluated informal social support for brain injured patients. Themes identified using content analysis of interview scripts from structured interviews with patients and their carers identified two consistent mechanisms for successful intervention in suicide attempts. The first, restriction of access to means, had successfully resolved prior suicide attempts, but was cited by only a minority of patients. The second, cited by the majority of participants (total N =14) was informal social support by family, friends and clinicians.

4.9 Whilst qualitative research, in particular research with such limited numbers of participants, is not ideally suited to providing unequivocal support for the success or otherwise of interventions, provision of social support (including ongoing contact) is also consistently identified in the broader range of literature as a successful intervention. Studies reporting successful outcomes from this type of intervention include larger scale quantitative studies with adequate statistical evaluation. Restriction of access to means is similarly cited by a broader range of studies, primarily in the context of firearms control but also with reference to other national and individual level initiatives as having successful outcomes. Broader support in the literature for interventions identified as promising by at least one well conducted and more in-depth study provides a degree of confidence in assuming that these are useful avenues to explore further. However, it should again be borne in mind that the number of studies addressing any given intervention is in absolute terms quite small.

4.10 The final qualitative study evaluating interventions for attempted suicide (Perseius et al 2003) addressed the use of DBT for people with borderline personality disorder. The study provides only limited outcome details and again includes only a very small number of participants. Nevertheless, support for the intervention shows a measure of consistency, with themes derived from the transcripts of all 10 female patients suggesting that DBT was regarded as having 'saved their lives' by reducing the frequency of suicide attempts. This limited but comparatively robust evidence is also supplemented by a broader range of support from both quantitative and qualitative studies within the full range of studies identified by the review. Although positive outcomes are not as consistently reported as is the case for social support and the restriction of access to means, DBT does appear to be a promising intervention to evaluate further for some client groups, in particular for people with personality disorders.

4.11 In summary, the outcomes of the highest quality studies, supplemented by additional evidence from the broader range of studies outlined earlier, present a more promising picture for interventions to prevent attempted suicide than for interventions to prevent completed suicide. There is both a greater consistency in support for particular interventions and more substantive evidence to suggest which interventions can be effective in preventing attempted suicide than is the case for completed suicide. This having been said, the most promising interventions for attempted suicide have primarily been evaluated in the context of mental ill health and further research is required to confirm that the same interventions could have a similar impact in other populations, including the general population. Particularly promising interventions which have the potential for widespread implementation in a range of populations are the provision of informal social support and the restriction of access to means. Further studies exploring the latter approach in a broader range of contexts would be helpful. More specific clinical approaches which also appear promising are DBT (primarily evaluated in the context of borderline personality disorder) and, although this intervention receives less consistent support in the literature as a whole, CBT.

Self-harm

4.12 Although under-represented in respect of intervention studies as a whole, the number of higher quality studies addressing self-harm as an outcome is similar to that for other modes of suicidal behaviour. Six of the highest quality studies evaluated interventions for self-harm. Of these, two addressed self-harm defined broadly as any form of self-harm. One addressed self-mutilation, one self-injury and two self-poisoning. It is notable both here and across the full-range of studies identified that self-cutting as a specific form of self-harm is rarely addressed. The three quantitative studies addressing self-harm focussed on quite diverse interventions. The first (Bennewith et al 2002) evaluated a general practice based intervention whereby GPs were given management guidelines for good practice in respect of self-harm and subsequently pro-actively offered clients with self-harming behaviour the opportunity for a consultation. This study was a particularly well conducted RCT, with a large sample size (N=1,932) but failed to find any significant differences between the intervention and non-intervention groups on any of the three outcome measures evaluated (repeat episodes of self-harm, the number of repeat episodes and time to first repetition). This rather disappointing outcome is unfortunately supported by the broader range of General Practice-based training and other initiatives evaluated in the wider literature.

4.13 A further RCT (Carter et al 2005ps) evaluating ongoing contact, via postcards sent to people following discharge from hospital for self-poisoning, provided slightly more optimistic but still limited positive outcomes. No significant differences were found in the absolute likelihood of further admissions. However, the intervention group - who received 8 supportive postcards enquiring about their well-being over a 12 month period - did show a substantive and significant reduction in the total number of episodes recorded. One hundred and ninety two episodes were recorded for the control group versus 101 for the intervention group. For a very minimalist intervention, this is a quite substantial outcome in clinical terms. Further evidence from this study demonstrated that the impact primarily related to improvements for women rather than men, suggesting that the intervention may benefit from targeted rather than general implementation. The final quantitative study (Kapur et al 2004) addressing self-harm was a retrospective cohort study evaluating emergency department management strategies for people attending with self-poisoning. Following adjustment for baseline differences, receiving a psychosocial assessment was not found to be associated with reduced repetition rates. However, being referred for specialist follow-up did reduce rates of subsequent repetition. Again this was a particularly well conducted study with a large sample size (N=658).

4.14 Three of the six studies evaluating interventions for self-harm were qualitative studies. One of these has already been discussed in relation to suicide attempts (Perseius et al 2003) and outcomes in the context of self-harm were as for suicide attempts, with DBT showing some promise in respect of patients with borderline personality disorder. The other two studies (Bloxham et al 1993, Cowdery et al 1990) are case studies in effect traversing the borderline between quantitative and qualitative approaches. Both addressed the use of behaviour therapy to reduce self-harm using single case studies and both reported positive outcomes. In the first study, a 35 year old woman who had consistently self-injured over a lengthy period of time ceased to self-injure by the 26 ^th week of an inpatient admission. Treatment during the admission focussed on behaviour therapy incorporating a combined token-economy and 'time-out' strategy. In the second study, a nine-year old boy substantially decreased the frequency with which he self-mutilated over the course of 50 therapy sessions, using differential reinforcement of other (non-self-harming) behaviour ( DRO). Neither client was reported as having any specific mental health diagnosis.

4.15 It is worth noting, that whilst behaviour therapy shows consistently positive outcomes both in the small number of studies included in the current review and more generally in relation to studies in the wider public health and mental health literature, it appears to be an intervention which has more recently 'gone out of fashion'. Interventions adding a cognitive component to behaviour therapy, including DBT and CBT, which are shown to be promising approaches in the current context appear to have displaced behaviour therapy per se as an intervention of choice ⁷. Interestingly, however, we were unable to find any studies providing evidence that the addition of a cognitive element improved outcomes and/or that it is the cognitive element specifically which successfully addresses the behaviour. It is therefore unclear that this evolution from a simpler to a more resource intensive and complex intervention is itself evidence-based.

4.16 In summary, the evidence evaluating particular interventions for self-harm is more limited overall than is the case either for other suicidal behaviours or for suicidal ideation. Whilst the proportion of all studies which are of high quality is greater for self-harm than for the other outcomes evaluated, the messages for future intervention are also more equivocal and, in comparison with outcomes for attempted suicide, less positive overall. As the evidence base currently stands, there is, as with attempted suicide, some evidence that DBT may be of value, although it is important that in future research outcomes relating to the cognitive components of both this therapy and of CBT are distinguished from outcomes attributable solely to the behavioural component. There is again also some support for the efficacy of ongoing contact, although in the context of self-harm outcomes for this form of intervention are slightly less convincing than is the case for attempted suicide. There is currently no support for the efficacy of GP-based contact and training initiatives or for psychosocial assessment carried out in the context of hospital presentation. There is some limited evidence in the latter context that referral for specialist support may be of value. It is important that future research address the relative paucity of studies focussed specifically on interventions for self-harm.

Outcomes for suicidal ideation

4.17 In line with the broader range of studies considered earlier, suicidal ideation was also addressed by a greater proportion of the highest quality studies than suicidal behaviour. Three qualitative studies and 8 quantitative studies evaluated interventions to reduce suicidal ideation. One of the quantitative studies (Brown et al 2005), also discussed in relation to outcomes for attempted suicide, presented evidence from a randomised controlled trial of CBT versus TAU in an A&E setting. Although a substantive reduction in attempted suicide was reported for the CBT group, this study found no significant differences in outcomes for suicidal ideation at any assessment point for the intervention group treated with CBT versus the TAU group. A further study (Zenere & Lazarus 1997) has been discussed in relation to outcomes for both suicide and attempted suicide. In respect of suicidal ideation this study, presenting outcomes from a school-based crisis intervention programme, also provided little evidence of more than a temporary decline in suicidal ideation, with the prevalence of suicidal ideation at the end of the study back to baseline figures.

4.18 The six quantitative studies which focussed exclusively on suicidal ideation all reported positive outcomes, with greater or lesser support from the statistical analyses presented. All six studies used scale-based measures of ideation. Five studies using the Hamilton Rating Scale for Depression ( HAMD 1960) reported significant decreases in suicidal ideation from baseline for depressed patients treated with moclobemide (Gagiano et al 1995), fluvoxamine (Gonella et al 1990, Kasper et al 1995) and sertraline (Lapierre 1991a and b) but not for comparator groups receiving imipramine (compared with fluvoxamine, Kasper et al 1995) or amitriptyline (compared with sertraline, Gonella et al 1990). One study (King et al 2003) using a scale developed during the study but based on the MINI International Neuropsychiatric Interview (Sheehan et al 1998) reported significant reductions in suicidal ideation from the beginning to the end of telephone counselling sessions conducted in the context of a community based helpline.

4.19 The one study supporting the use of moclobemide is hampered by the fact that its main goal was to compare different dosages of moclobemide (all of which were reported as reducing suicidal ideation to the same degree) without any attempt to compare this intervention to either placebo or active comparators. One of the fluvoxamine studies (Gonella et al 1990) failed to provide adequate statistics to support narrative outcomes and also failed to employ a placebo comparator. However, outcomes are similarly positive for the second study (Kasper et al 1995) which, although again giving sparse details of statistical analyses, did match the active treatment against placebo. Of the two studies evaluating treatment with sertraline, one again failed to provide an adequate account of the statistical analyses carried out and also failed to provide a placebo comparator (Lapierre1991b), the other (Lapierre1991a) employed a placebo but failed to differentiate outcomes specific to suicidal ideation from outcomes relating to total HAMD scores in statistical analyses. Whilst these pharmaceutical studies are of high quality in respect of design and implementation therefore, the focus and presentation of their analyses leave something to be desired.

4.20 The one non-pharmaceutical study in this group of quantitative studies (King et al 2003) provides more concrete support for a reduction in suicidal ideation. However, this study has a very limited follow-up, restricting, for pragmatic reasons, the evaluation of outcomes to the course of a single telephone conversation, albeit for a large number (N=1010) of individuals. Nevertheless, consistently positive outcomes were reported with quite substantive mean differences in suicidal ideation from beginning to end of call (t=12.6 p<0.005) and similarly a substantive mean decrease in suicidal urgency (t=-8.4 p<0.0005). Comparable, although smaller, differences were reported for the subset of scale items relating to 'imminent' thoughts of suicide and to raters' perceptions of changes in suicidal ideation.

4.21 Three qualitative studies focussed on the evaluation of interventions for suicidal ideation. Two of these studies (Kuipers & Lancaster 2000, Perseius et al 2003) have been discussed earlier in relation to interventions for attempted suicide. Both reported similarly positive outcomes for suicidal ideation in respect of the effectiveness of informal social support and DBT respectively. The third study (Mishara et al 1997) reported more equivocal outcomes. This study involved non-participant observation of telephone intervention styles used by helpline staff. The study compared directive versus non-directive 'Rogerian' styles of communication. Overall, there were no significant differences between the two styles in respect of changes in suicidal ideation from the beginning to end of calls, as evaluated by the Suicide Urgency Scale (Morisette 1984). However, when outcomes were evaluated on the basis of whether or not a caller was regarded as 'chronic' versus 'non-chronic' in respect of the frequency of their calls to the helpline, the authors reported that Rogerian (non-directive) telephone styles improved outcomes for non-chronic callers. The Rogerian style of communication also significantly increased the likelihood of establishing a 'no suicide' contract with callers.

4.22 In summary, outcomes for the higher quality studies evaluating interventions for reducing suicidal ideation broadly match those observed for the wider range of lower quality studies. The two sets of studies also match one another in a focus on pharmaceutical intervention. The prevalence of pharmaceutical intervention in this context may be accounted for by the ease of adding scale-based measures of suicidal ideation to trials with a main focus on depression. Additional studies specifically focussed on suicidal ideation and studies exploring suicidal ideation outside the context of mental health problems would be helpful. Outcomes from the available pharmaceutical studies are also hampered by a lack of placebo control and poor reporting of statistical outcomes. The evidence as it stands suggests that there is some, very limited, support for the use of moclobemide and rather stronger, but not unequivocal, support for the use of fluvoxamine and sertraline in reducing suicidal ideation. There is currently no support from high quality studies or from the broader range of studies to support the use of imipramine or amitriptyline. All of these outcomes relate to interventions for people with depression only. In respect of non-pharmaceutical interventions, there is currently no support from higher quality studies and only equivocal support from the range of other studies available for the efficacy of CBT or school-based intervention programmes in reducing suicidal ideation. There is some limited evidence of the efficacy of telephone-based support over very short follow-up periods (the length of the call) with again some evidence that non-interventionist styles of communication may be beneficial with first-time callers.

Specific issues in research design and implementation

4.23 In evaluating the quality of the available primary research material, we were able to identify particular problems with current study design and implementation and also to identify where consistent differences in quality existed between different approaches to research. A detailed overview of these findings, which can be used to inform the commissioning of future research, is given in Annex H. Here we present a brief summary of the main points:

• Whilst there is substantial room for improvement in the quality of study designs, the literature relating to interventions for suicidal behaviour and suicidal ideation compares favourably with other public health literature in terms of overall quality
• The design and implementation of qualitative studies is substantially poorer than that of quantitative studies, with very few qualitative studies making any attempt to follow an identified methodology
• There are few substantive differences in the quality of studies addressing different aspects of suicidal behaviour or ideation, although overall the quality of studies addressing completed or attempted suicide is slightly higher than that of studies addressing either self-harm or suicidal ideation
• Both pragmatic and ethical constraints on the conduct of research were commonly cited to account for acknowledged failings in either study design or implementation.

4.24 A number of aspects of study design and implementation seem to pose particular problems for this literature. Specifically, a substantial proportion of studies were adversely affected by:

• high drop-out rates (losing one third or more of participants to follow-up)
• failure to randomise (in particular in studies addressing completed suicide)
• failure to 'blind' investigators to the allocation of participants
• a lack of attention to the adequacy of implementation of interventions
• failure to control for the impact of other (unevaluated) ongoing interventions
• failure to control for the baseline frequency of the behaviour used as an outcome measure

« Previous | Contents | Next »