‘ADHD and suicidal spectrum behaviors’ Professor Samuele Cortese

Marketing Manager for ACAMH

Posted on

Professor Sam Cortese on ‘ADHD and suicidal spectrum behaviors’. In this lecture, Professor Samuele Cortese discusses the recent paper published in Neuroscience & Biobehavioral Reviews, Volume 103, August 2019. ‘Association between suicidal spectrum behaviors and Attention-Deficit/Hyperactivity Disorder: A systematic review and meta-analysis’ Mathilde Septier, Coline Stordeur, Junhua Zhang, Richard Delorme, Samuele Cortese (2019). https://doi.org/10.1016/j.neubiorev.2019.05.022

This was recorded on 4 April 2019 at the ACAMH Southern Branch Research Day.

ACAMH members can now receive a CPD certificate for watching this recorded lecture. Simply email membership@acamh.org with the day and time you watch it, so we can check the analytics, and we’ll email you your certificate.

Professor Samuele Cortese
Professor Samuele Cortese

Prof Samuele Cortese, Child and Adolescent Psychiatrist, is Professor of Child and Adolescent Psychiatry at the University of Southampton and Honorary Consultant Child Psychiatrist at the Solent NHS Trust. He is also Adjunct Associate Professor at the University of New York, USA. His main clinical and research interests focus on the epidemiology, neurobiology and treatment of neurodevelopmental disorders (in particular on ADHD) and on evidence-based practice in child psychiatry. Dr. Cortese is author/co-author of more than 200 papers in international peer-reviewed journals. He is on the editorial board of the Journal of the American Academy of Child and Adolescent Psychiatry, (JAACAP), Child and Adolescent Mental Health (CAMH), Evidence Based Mental Health (EBMH), CNS Drugs and on the advisory board of the Journal of Child Psychology and Psychiatry. Dr. Cortese is a member of the European Network for Hyperkinetic Disorder (EUNETHYDIS), the European ADHD Guidelines Group (EAGG), and the Child & Adolescent Neuropsychopharmacology Network of the European College of Neuropsychopharmacology (ECNP).


Hi, my name is Samuele Cortese. I’m a child and adolescent psychiatrist, currently Associate Professor at the University of Southampton and an Honorary Consultant, child and adolescent psychiatrist with Solent NHS Trust.

Today, I will present a study on the association between suicidal spectrum behaviours and ADHD, and this is a systematic review and a meta-analysis which is based on collaboration with colleagues from Paris, in particular Dr Methilde Septier and Coline Stordeur who really did the bulk of the work, and my colleague, Professor Richard Delorme, alongside with another colleague from China. I would like to acknowledge Junhua Zhang, from Yangcheng University, China. Before starting my presentation these are my disclosures of potential conflict of interest.

I have a number of collaborations in organisations such as ACAMH, so I received an honoraria for talks mainly on ADHD and I do not have any other disclosures to declare today. Right. So as a general introduction to this topic I think it’s important to highlight how there is, of course, a large body of evidence on the association between ADHD and other mental health conditions. There is also an increasing body of evidence on the relationship between ADHD and somatic conditions. However, we lack a systematic evidence on the association between ADHD and other relevant outcomes and one of these is suicide and suicidal spectrum behaviours.

So this work really aimed to fill this gap in the literature and to provide a systematic overview and meta-analysis to provide a quantitative data on the relationship between ADHD and suicidal spectrum behaviours, which is, of course highly relevant in the clinical practice. So before moving to our findings, I would like to provide a general overview on suicidal spectrum behaviours. So as the name suggests this is really a continuum and includes suicidal ideations as well as suicidal upturns and suicidal plans and finally completed the suicide.

And there is a link amongst these behaviours in adolescents, up to one third of those with suicidal ideation have been found to go on to develop suicidal plans, and 33% of these make a suicidal attempt, and eventually it has been reported in the literature, according to this recent study, by [inaudible 00:03:29] colleagues, that 60% of those with a plan versus 20% of those without a plan make a suicidal attempt, and 60% of first attempts have been found to be planned.

So in terms of prevalence, clearly these figures highlight the clinical and the public health relevance of this topic. The cross nation lifetime prevalence for suicidal ideation has been found to be around 9%, suicidal plans 9.2 and suicidal attempt 2.7. In terms of risk factors, of course there are a number of risk factors for suicidal behaviours which are related to several mental health conditions and other situations. However, if we look at the strongest evidence it turns out that mood disorder is really the main important risk factors, at least in developed countries. While interestingly in developing countries the risk factor with the highest, the strongest link with these behaviours has been found to be related to the presence of impulse and control disorders.

So this brings us to, of course, the link between ADHD and suicidal spectrum behaviours because of the impulsive component in particular.

However, when we look at the latest findings across studies are quite mixed in terms of a significant association between ADHD and suicidal behaviours, and also what is not clear is to which extent comorbid disorders, such as mood disorder or substance use disorder and other important psychosocial variables do account for the link between ADHD and suicidal behaviours, or to which extent they act as a mediator or moderator in this way.

So we thought that it was important to provide a comprehensive overview of the literature and to use meta-analytic methods to address these questions, to provide the field with a clear and solid answer to this matter. So we conducted a systematic review and a meta-analysis according to the state of our recommendations in the field, we followed the standard guidelines, which are also endorsed by Cochrane and other groups.

So the PRISMA Guidelines, and importantly, as it is a good practice in this field of research, we pre-registered the protocol in advance so that the outcomes and the aims and the methods were clear from the beginning to avoid bias in reporting results later on.

So in terms of such a strategy to retrieve the evidence, to include the studies, we look at a comprehensive set of electronic databases, including the most common ones, which are PubMed, a set of all the databases, including [inaudible 00:07:14], Nightline and BASE and BASE classic, and also all the set of databases of knowledge. So we think that this strategy was quite comprehensive. However, as it is common practice, we also checked and hence searched all the references or relevant reviews, previous reviews in the field.

And importantly, and this was quite a time consuming process, but I think something that added value to and rigour to our study was that we contacted the corresponding authors of these studies to retrieve additional information and to clarify aspects that were not clear from the published paper.

Right. So this is just an example of a search terms and search syntax that we use for PubMed. As you can see, a broad set of terms to include hopefully all and to retrieve all available studies, and the search was adapted for our databases.

So in terms of the type of studies that we included in our systematic review, we basically looked at all observational and empirical studies that reported and yes, data on the association between ADHD and suicidal spectrum behaviours or also those study that without directly reporting this data, provided data that could be useful for us in terms of calculating the association.

We did not include, of course, case reports and also we excluded, of course, the randomised control trials on ADHD because usually suicide, fortunately, is quite a rare event in trials. So it will not have been possible to capture the prevalence from these type of studies, and we did include the longitudinal studies, but of course for these ones we just extracted data at the baseline.

Right. So in terms of the types of participants we included studies on children, adolescents or adults with ADHD defined according to standardised criteria as per the DSM or ICD. Also studies where ADHD was defined as the presence of a number of symptoms and symptom severity above a cut-off. Of course this is a less rigorous definition of ADHD, but it is something that is quite often implemented in epidemiological studies. So in order to address the possible bias of including these studies we also plan a sensitivity analysis where we removed studies where ADHD was not defined according to standardised criteria.

And in terms of the outcomes we were interested in, of course, the primary outcome was a number of statistical measure expressing the association between ADHD and suicidal behaviours, and this is usually the other ratio which is used commonly in studies addressing association between variables. It’s not clinically straightforward, but is the most commonly reported measure.

A secondary outcome which we think actually is very important was the odd ratio from each study adjusted for a number of confounding factors. So when you run a study on the association between two variables, you can just assess the accrued association, but you can also adjust statistically to assess the impact of possible confounding factors.

So the aim of the secondary analysis on the secondary outcome was really to check to which extent the results from the primary analysis were solid after controlling for possible confounding factors. Needless to say, each study and control for different confounding factors. So we don’t have a common set of confounding factors that have been adjusted for consistently across the studies. So this is the plan of the statistical analysis that we use basically in quite standard methods in the field, a random effect model.

We also did a subgroup analysis and meta-regression to assess the impact of important variables, and I will highlight these during the presentation of the results. We assess importantly something that should be done every time you run a meta-analysis. So the heterogeneity. So to which extent of the studies that we pooled where homogeneous or not. This is very important because if there is evidence that we pull studies that are impacted by possible publication bias, the conclusions should be taken very cautious. So comprehensive meta-analysis was the software that we use to run the analysis. And finally, this is now really part of the statistical analysis, per say, but it is an important component of a systematic review and meta-analysis.

We assess the quality of the studies. This is important because it is really fundamental to appreciate to which extent the studies that we included were of good quality or not, and we use a tool which is commonly used for epidemiological studies, which is called Newcastle-Ottawa Scale, which addresses a number of aspects related to the quality of the study, mainly in terms of the comparison between individuals with that particular disorder and individuals without the disorder.

Right. So this is what is called usually the PRISMA flow-chart, and this is a graphic representation to show really the selection process in terms of screening of studies. So after we considered and we screen across all the databases, we had almost 3,000 references to screen and then eventually, after discarding the non-relevant references and studies we ended up with 57 original studies that were put in our meta-analysis, and this gave us a total of 90,000 participants with ADHD across all these studies and more than 200,000 individuals without ADHD. So, of course, this gave us a relevant statistical power to address the questions that we wanted to answer.

Right. So I will present now in a series of slides the results of the main analysis. So for those who are not familiar with this way of representing the results, this is what is called a forest plot. In this case this is the analysis on the unadjusted occlude or the ratio for the outcomes suicidal at times, and for each analysis we actually address each individual suicidal behaviour. So this graph relates to the relationship between ADHD and suicidal attempts.

Each study is represented here and each study has a particular odd ratio which is represented on the right side as an average effect with its confidence interval, and when the confidence interval crosses the line one, in this case, it means that that particular study suggests that there is not a significant association between ADHD and suicidal behaviours. When all the confidence interval is higher than one such as in this case, for instance, well, this means that this study suggests that definitely there is an increased risk of suicidal behaviour in individuals with ADHD, and when it is on this side, it means that actually there is a decrease of risk of suicide.

So you can see just from examining each confidence interval, from each study, that there is a lot of variability in terms of results across studies. So this was really the purpose of the meta-analysis to combine them in order to have higher statistical power and to see if there is a direction in terms of a possible effect, and actually what we found is that, yes, there is a significant association, which means that there is an increased risk of suicidal at times in this case in ADHD, because the final effect which is derived from the meta-analysis is represented at the bottom here in red, and as you can see the confidence interval is above one, which clearly states that there is a statistical significant association between increased risk of ADHD and suicidal at times.

This was the case also for when we considered the odd ratio from each study unadjusted to ratio for suicidal ideation and also with a larger effect for complete suicide.

Now, these are important values, but they need to be taken into the context of additional data. So first of all this is another table which summarises all the results.

So first of all we need to consider i-squared, which as I mentioned, is an index of heterogeneity across studies and in this table is represented here. So as you can see in all the analysis we ran, the i-square was higher than 50%. This means that we cannot be completely confident in these results because the studies were quite heterogeneous. So when you pull, of course, different studies you need to pay attention in interpreting the results. This does not mean that our results are not valid, but it simply means that the final estimate that we derived cannot fully represent all the heterogeneity across studies.

Then what I presented previously, as you may remember, are the results for the unadjusted odds ratio. For each analysis, as you may remember, we also calculated as a secondary outcome the pool adjusted of the ratio. Adjusted for a number of confounders, and we think that this is more in a way is a more reliable measure of the association because it takes into account possible confounding factors. So one may think, well, there is an association between ADHD and suicidal behaviours, which is significant but it may be accounted for by another factor which is related to both to ADHD and suicide. For example, low socio-economic status. So after taking into account socio-economic status, is this association still valid?

So this is why we also pulled the adjusted odd ratio from each study, and you can see these on the left side in green. So for each outcome, suicidal attempts, suicidal ideation, suicide and specified behaviours and suicidal plans, we calculated where available also the pool adjusted ratio. So actually, this was available just for suicidal at times and suicidal ideation.

Unfortunately for other, these other outcomes here adjusted odd ratios were not available, and what we can see here is that even when we consider adjusted odd ratio, the association is still significant. A quick way to look at this is to look at the P values here and when they are lower than 0.5, this means that there is a significant association.

So to summarise, when we consider both an adjusted and adjusted odd ratio when we pull them from each study there is a significant increase risk of suicidal attempts and suicidal ideation and suicide and suicidal plans in individuals with ADHD. The only outcome for which we did not find this significant association and I have highlighted this right here, is what this study has coded as unspecified suicidal behaviour. However, this is quite a vague term. So we should consider this with caution.

Right. So this table refers to all the studies pulled together, but we were also interested in performing sub-group analysis, just retaining some type of studies to see if the association was still significant. So the risk will still increase, and for instance, we wanted to know is the risk increase in a population base or epidemiological studies as well as in clinical studies?

Actually, what we found was a quite interesting pattern. So when it comes to an adjusted odd ratio, both for clinical and epidemiological studies split. We did find this significant association. So once again, the way to know if the association was significant is to look at the p value here. However, when we consider the adjusted odd ratio both for clinical and epidemiological studies, the association was significant just for the epidemiological study, not for the clinical one, and likewise, as you can see here, when we consider the unspecified suicidal behaviours the association was not significant for clinical, but it was for epidemiological.

So it looks like the setting of this study is a factor that may influence the association. We do not know really the exact reason for that. We suspect is related to the severity of the clinical situation which, of course, is more evident in a clinical setting.

Right. So in another subgroup analysis we also wanted to look at the association in terms of current versus lifetime prevalence of suicidal behaviour. So, of course, when we assess the variables in a study we can look at the prevalence of these variable in a specific time point or across life, basically. So current versus lifetime and basically this subgroup analysis told us that basically results were quite solid and didn’t change according to the temporal type of the variable.

Then, of course, we also wanted to look if other characteristics of the study design had an impact. So we are considering this subgroup analysis. We split the cross-sectional versus retrospective or prospective study, and also because this is relevant in terms of the reporting of the outcome in a retrospective study this outcome is a report retrospectively. So we may wonder to which extent it is accurate.

However, it looks like this variable did not have a significant effect on the results. So the results stayed. We’re still there when we considered the type of design.

Finally, as I mentioned, we conducted a specific analysis which is called meta-regression, and this is meant to assess the impact of specific variables on the results. So we may wonder, is this association the same regardless of the age of the patient? Is it the same regardless of the year the study was published or is the same across continents, or is it impacted by the quality of this study?

Well, actually, we did find that none of these variables had a significant impact on the association, which means that the association between ADHD and suicidal behaviours most of them was still there, regardless of, you know, the age of the gender, the year of study publication, the study continent and the study quality, and in terms actually of the study quality, as I mentioned previously, we did use a rigorous method to assess these and overall it turned out that the quality of the study was in general, quite good in most of the parameters that we considered. So to rate the quality of the study, we can do this on this scale from zero to four and in many parameters, 2.5 or three or higher than this value.

So in terms of the discussion of these findings, I think that overall we believe that our study is relevant because it stands in the way the conclusion of previous studies were only non-systematic reviews or individual studies and they provide us with numbers, with figures that we can use in our clinical practice when we discuss, of course, with patients. Also previous studies did not provide us with a clear answer on which variables could influence the relationship between ADHD and suicidal behaviours.

And I think, I hope I made it clear that we did perform an extensive set of analysis to assess the impact of these variables, and also we need to point out, of course, that we cannot be fully confident in our results because, as I mentioned, each study adjusted for different variables.

So ideally, we would like to see all studies controlling for the same confounding factors. This will provide us with more solid evidence, but as it always happens, a meta-analysis is never the final answer to a question. Often times, it points us to the right direction for future studies. So we think that our meta-analysis should also inform future research in the field.

In terms of the limitation, of course, of our study I like to point out that, of course, this is a meta-analysis assessing the so-called cross-sectional association. So it does not tell us anything in terms of the cause effect relationship and the temporality. So we just established that if you look at individuals with ADHD and if you compare them to individuals without ADHD those with ADHD have an increased risk of suicide, but we don’t know what are the causes for that and we don’t know the temporal link.

It is possible, there are possibly a number of possible explanations. I will highlight here in the interest of time, just two of these. The first is that ADHD per say contributes. So causal is a risk factor is a causal factor to suicidal behaviours via the impulsive components and more generally via the deficits in all of the executive functions. However, it is also possible that both ADHD and suicidal spectrum behaviours are underpinned by common neurobiological and psychopathological mechanisms. For instance, there is evidence of a shared genetic vulnerability between suicidal spectrum behaviours and impulsive behaviour, and you can see here some of the genes that have been implicated.

Of course, an important factor that we should consider when we try to explain the association between ADHD and suicidal behaviours is the possible role of the pharmacological treatment, the medication for ADHD and this is clearly important, especially for clinicians, for prescribers is an extremely important question to address. Now, the problem is that in general, this is not easy to understand. If you consider, for example, if you compare your epidemiological study population based study and you compare the rates of suicidal behaviours in a group treated with ADHD medication and in another group of individuals not treated with ADHD medication, and you find that the risk of suicidal behaviour is higher in the group treated with the medication.

You cannot conclude that this is due to the medication because the two groups may be different. We may think, for example, that they treat the group is more severe. So per say they may present with more psychopathology and more risk. So there is, however, clever type of design study design, which may be informative in this regard, which is called the self-control, or we think CASA design whereby basically each individual is considered when they are on and off medication.

And you can compare the prevalence of suicidal behaviours when they are on medication and off medication. So you control for a series, you adjust naturally for a series of confounding factors. The only aspect that this design cannot control is related to those factors that changes over time, but there are some also statistical approach to partially control for this. So we found just two studies that use these self-controlled methodology, and the first one is this study that was published by colleagues from Sweden in the BMJ and actually they found that there is no significant association between the use of psycho stimulants for ADHD and increase the risk of suicide. If anything, as suggested by the other ratio, which was below one actually, if anything, stimulants seem to have rather a protective factor. While intense on non-stimulants there was not really association at all because the confidence interval of the other risk was crossing one.

So basically it means that there is no protective and there is no increased risk of suicidal behaviour with non-stimulants. The second study provided a slightly different results, a very interesting one. So this was a study from Hong Kong published in JAMA Psychiatry and what they found was that when they look at the risk of suicidal behaviours in those treated with methylphenidate, actually in the period 90 days before starting the treatment, the risk was higher than one clearly. So it was significant and it stayed higher than one.

So it remained significant even during the first 90 days following the initial prescription of the medication. However, then the risk became not significant later on, as you can see here, the confidence interval crosses the line on one.

So this may suggest that, of course, there’s many explanations for this, but what the authors seems to suggest is that the period before and immediately during the initial treatment is a very critical period and is when these patients are in distress and they go and see their psychiatrist or prescriber to seek a prescription. However, once the treatment starts being effective the distress and the risk may decrease.

Of course, we need additional evidence, but so far it will be quite tricky to argue based on the available evidence that medications for ADHD per say have a causal role in increasing the risk of suicidal behaviours. Right, other limitations that we should point out in our studies, as I mentioned, i-square, which is a measure of heterogeneity, was quite high, which means that our estimate, our figures do not capture the variability across studies.

And also I mentioned previously that in general the study quality was quite good. However, there were some items of this scale that we use to assess the study quality which indicated some issues in terms of possible bias in the methodology of the studies.

So moving to the possible clinical implications of this study. Well, I do think that it is important because it suggests that we should systematically screen for suicidal spectrum behaviours in patients with ADHD. Of course, at the first assessment but also the follow up and this I think is particularly relevant because when we look at available rating scales and questionnaires on ADHD, they do not in general include the specific items on this. So this may be missed if we perform it quite quick and standard assessment of ADHD. While of course, given the importance of these screening for these aspects seems extremely important.

I think that was all in terms of my presentation. So I thank you for your attention.

Add a comment

Your email address will not be published. Required fields are marked *