DESCRIPTION: Do antidepressants increase the risk of suicide? What the FDA approval data on SSRIs in adolescents and young adults with depression reveal, and why observational studies are of little use here.

Antidepressants and suicide: What the figures show and what ‘confounding by indication’ means – a non-statistic

Do antidepressants increase the risk of suicide? This question has been asked for four decades. In September 2026, Peter Gøtzsche published an article on Mad in America which described antidepressants as a major driver of suicides. The claim predates the text itself. With each new round of discussion, it becomes more extreme.

The available research allows neither an ‘all-clear’ nor the assertion that antidepressants are generally a major cause of suicide. The most robust part of the data concerns young patients: in randomised trials, the risk of suicidal thoughts and suicidal behaviour increases in those under 25 during the early stages of treatment. The number of incidents in clinical trials is insufficient to conclude completed suicides. Observational studies provide particularly unreliable answers in this regard, because people are prescribed antidepressants precisely because they are depressed and often at greater risk.

The figure: Hospitals make people ill

People in hospital are more likely to die than those at home. This is a valid observation. It is not an argument against hospitals. People go to hospital because they are ill; the illness explains the difference.

The same pattern is evident in many figures relating to antidepressants and suicide. People are prescribed antidepressants because they are depressed, have anxiety disorders, suffer from chronic pain or are in some other distressing situation. Depression is one of the strongest risk factors for suicide. Anyone comparing patients on medication with those not on medication is therefore, first and foremost, comparing groups with different baseline levels of risk.

In epidemiology, this is known as ‘confounding by indication’. The indication for a medication is itself linked to the outcome being studied. In the case of antidepressants, the chain is as follows: more severe depression or an acute worsening of symptoms is more likely to lead to a prescription; at the same time, this same severity increases the likelihood of suicidal thoughts, suicide attempts and suicide.

Observational data can mitigate this problem. They can compare those prescribed the drug with those who were not, use propensity scores, perform time-dependent analyses, or compare participants with themselves at different points in time. However, routine data often record baseline severity only incompletely. People rarely start treatment on a random day.

Antidepressants and suicidality in regulatory data

What the study in question actually compares

The study compares two groups that were as similar as possible at the start. One group received an antidepressant, the other a placebo – that is, a tablet that looked identical but contained no active ingredient. Neither the participants nor the treating doctors knew during the study who was receiving what. Allocation was randomised.

This is important because people do not receive antidepressants at random in everyday life. Those who are particularly severely depressed, acutely distressed, suffering from insomnia, restless or at risk of suicide are more likely to be prescribed medication. When comparing those who were subsequently prescribed medication with those who were not, the first step is therefore to assess differences in condition severity. In a randomised study, this severity is intended to be distributed equally across both groups through random allocation. Any subsequent difference can then be more readily attributed to the medication.

Marc Stone and colleagues analysed 372 such studies for the FDA. A total of 99,231 adults took part. These were not exclusively patients with severe depression. Just under 46 per cent were being treated for major depression. The remainder were taking antidepressants for other mental health conditions or for physical symptoms, such as chronic pain. The average age was 43.

How to interpret the figures

The study does not primarily investigate whether someone commits suicide as a result of taking an antidepressant. Completed suicides are very rare in clinical trials. The researchers therefore grouped several events: suicidal thoughts, preparatory actions and suicide attempts. In the study, this combined measure is referred to as ‘suicidal thoughts or more serious events’.

An odds ratio of 1 means that the event occurred with the same frequency in the treatment group as in the placebo group.

- A value above 1 means: the event occurred more frequently whilst taking the medication.

- A value below 1 means: the event occurred less frequently whilst taking the medication.

- A value of 2 does not mean that half of the participants were affected. It means that the event occurred approximately twice as often in the active treatment group as in the placebo group.

The confidence interval describes the statistical uncertainty. For a value of 1.62 with an interval ranging from 0.97 to 2.71, the best estimate is 1.62. However, the data are also consistent with a very small difference or with a risk more than twice as high. Because the range includes 1, this result is not considered unequivocally confirmed according to standard statistical rules. For a value of 2.30 with an interval ranging from 1.04 to 5.09, the entire range lies above 1. This finding is therefore considered statistically significant. Nevertheless, it remains imprecise due to the wide range: the risk could be slightly elevated or around five times as high.

What happened in young patients

Among participants under the age of 25, suicidal thoughts or more serious events occurred more frequently whilst taking antidepressants than whilst taking a placebo. The odds ratio was 1.62. For suicidal behaviour alone, it was 2.30.

The relative figures sound high. That is why we also need the absolute figures. On average, for every 1,000 young participants taking antidepressants, there were:

- around five additional cases of suicidal thoughts or more serious incidents;

- around four additional cases of suicidal behaviour.

This does not mean that five or four per cent of those treated were affected. It means that, for every 1,000 comparable young people given an antidepressant or a placebo, there were approximately five additional events of the first type and just under four additional events of the second type in the group taking the medication during the study period.

These figures justify the warning for young patients. They do not mean that antidepressants cause suicidal thoughts in every young person, nor that starting treatment is dangerous for everyone. They mean the first few weeks and any dosage changes must be monitored more closely.

Why age is a decisive factor

The risk was not the same across all age groups.

The figures are risk ratios. The comparison indicates a trend, not a personal prediction. A 24-year-old man with mild depression and good social support has a different baseline risk than a 24-year-old woman with severe depression, insomnia, alcohol problems and a history of suicide attempts. The study describes an average across many people and many medicines.

The statement that the odds ratio decreases by 2.6 per cent or 4.6 per cent per year of age means that the difference between the medication and the placebo gradually decreased with increasing age and eventually reversed. A medication that was associated with more suicidal events in younger participants was associated with fewer such events in older participants. In this analysis, age was the strongest factor determining the direction of the difference. [pmc.ncbi.nlm.nih](https://pmc.ncbi.nlm.nih.gov/articles/PMC2725270/)

The important caveat

The study highlights a risk of suicidal thoughts and behaviour in certain age groups. It does not prove how many people actually die by suicide as a result of taking antidepressants.

Across all 372 studies, there were eight completed suicides. Eight incidents amongst nearly 100,000 participants are not sufficient to draw a reliable conclusion about completed suicides. The estimate was too uncertain. Therefore, the study cannot support either the claim that antidepressants definitely increase the number of completed suicides in adults, nor the opposite claim that they definitely reduce it.

What the figures mean for adolescents and young adults

The regulatory data support a clear conclusion: in young patients, the risk of suicidal thoughts and suicidal behaviour increases measurably at the start of antidepressant treatment. Anyone who disputes this is disputing the data on which the regulatory warnings are based.

Following paediatric analyses, the FDA issued a warning for children and adolescents; it later extended the warning to young adults up to age 24. There were no completed suicides in the paediatric studies. The strength of the signal related to thoughts and behavioural events, not evidence of completed suicides.

Two caveats are essential.

Firstly: clinical trials almost always measure suicidal thoughts, suicide attempts and preparatory actions. Completed suicides are so rare that even large study cohorts scarcely allow for reliable conclusions. In the Stone analysis, there were eight completed suicides amongst 99,231 participants. The estimated odds ratio was 2.13, but the confidence interval ranged from 0.41 to 10.99.

Secondly: The FDA data do not relate exclusively to severe depression. More than half of the participants were prescribed antidepressants for other depressive, psychiatric, behavioural or non-psychiatric conditions. The findings therefore represent an analysis of antidepressant drug trials in adults, not a straightforward answer to the question of what typically happens to an individual with severe depression in standard clinical practice.

The 2019 re-analysis of the FDA data

In 2019, Michael Hengartner and Martin Plöderl re-analysed the same FDA database. They corrected two suicides that had been allocated to the placebo group rather than the active-drug group in the paroxetine data. They also included events from follow-up periods after the end of randomised treatment.

They reported more than twice as many suicides among adults in the active-drug group as in the placebo group. Their calculated odds ratio was 2.48. This figure became central to the critical literature on antidepressants and suicide.

The re-analysis makes a strong case. Incorrect classification of suicides is no trivial matter. With rare events, every single case has a significant impact on the result. The question of whether events were recorded fully and correctly therefore remains valid.

However, their conclusion remains methodologically open to criticism. The follow-up period was not randomised. Those who withdraw from the randomised study phase prematurely or switch to open-label follow-up treatment often differ from the participants who remain in their original allocation. Reasons for withdrawal, switching and follow-up may be related to side effects, disease severity, treatment expectations and suicide risk. During this period, the very problem that randomised trials are designed to avoid resurfaces.

The reanalysis is also statistically controversial. A published response ran several meta-analysis methods for rare events and found estimates significantly lower than the odds ratio reported by Hengartner and Plöderl. None of these methods definitively ruled out clinically significant harm; the confidence intervals were wide. The rebuttal therefore establishes one thing above all: the data are too imprecise to support a far-reaching claim regarding completed suicides.

An odds ratio of 2.48 sounds large. However, for a very rare event, the absolute difference may still be small. This is precisely why just a few dozen events can visibly shift the results of individual classification or modelling decisions.

The key evidence against his own thesis

Gøtzsche cites a Danish register-based study by Carsten Hjorthøj and colleagues from 2014. It reports a sharply increased risk of suicide following psychiatric treatment and a higher risk associated with more intensive psychiatric contact. Gøtzsche cites a 44-fold increased risk associated with inpatient psychiatric treatment, whilst at the same time acknowledging that patients admitted as inpatients are more seriously ill and would therefore be expected to be at higher risk.

This is precisely what is known as ‘confounding by indication’. People who are acutely suicidal, psychotic, severely depressed or can no longer be safely cared for are more likely to be admitted to hospital. The fact that hospitalised patients are more likely to die by suicide later on than those who were not admitted does not prove that the admission caused the suicide. It primarily reflects the crisis that led to the admission.

This example does not refute every criticism of psychiatric care. It shows how strictly the same methodological rule must be applied to all observational data. The rules must not change simply because a result points in the desired direction.

What observational studies contribute to our understanding of SSRIs

Observational studies are not worthless. They capture groups that rarely feature in regulatory trials: older patients, people with multiple conditions, long-term treatments, repeated changes of medication and more severe everyday crises. They can indicate rare risks and raise questions that randomised trials must later examine.

However, when it comes to antidepressants and suicide, their informative value remains particularly limited. The severity of depression, the current course of the crisis, previous suicide attempts, acute relationship conflicts, alcohol and substance use, access to social support, and the reason for a prescription are often not adequately reflected in routine data. These same factors influence prescribing practices and suicide risk.

Hengartner and colleagues' systematic review of observational studies highlights this limitation. The authors themselves treat ‘confounding by indication’ as a central problem. An earlier review in the Canadian Medical Association Journal also concluded that bias due to disease severity cannot be reliably removed from such data.

The age gradient observed in randomised trials carries particular weight here. For confounding alone to explain the difference between younger and older adults, the bias would have to increase or decrease systematically with age in the same direction. Whilst this is conceivable, it is not an obvious explanation for the pattern observed in the FDA data.

Publication and interests

Martin Plöderl and colleagues reported in 2023 that observational studies showing an increased risk of suicide whilst on antidepressants are published less frequently in psychiatric journals. This highlights a serious problem. Selective publication can skew the literature in favour of a reassuring message.

This observation does not exempt the critical side from the same standards. The opposing camp, too, has journals, blogs, reach, readers and institutional incentives. An author who has been publishing as an opponent of antidepressant treatment for years holds a recognisable position in the field. This is not an argument against his figures. It is a reason to read them with the same care as manufacturer studies, guidelines and regulatory statements.

Conflicts of interest are no substitute for methodological scrutiny. They explain why methodological scrutiny remains particularly necessary for certain claims.

What is established, what remains unclear

The age difference is evident in the randomised regulatory data. People under the age of 25 require particular attention during the first weeks of treatment for increases in restlessness, hyperactivity, agitation, akathisia, sleep disturbance, despair, suicidal thoughts and impulsivity. Providing information and maintaining close contact during this phase is not an overcautious response. It follows from the data.

The limitations of the data on completed suicides are also established. Randomised trials contain too few such events. They can neither reliably prove that antidepressants increase the rate of completed suicides in adults, nor that they definitively reduce this risk.

The magnitude of any potential risk of completed suicides in adults remains unclear. The 2019 re-analysis suggests an increased risk; however, its inclusion of non-randomised follow-up periods and the small number of events limit the conclusion. Other analytical methods find no statistically significant increase, but do not rule out clinically significant harm either.

It also remains unclear to what extent withdrawal phenomena play a role. Withdrawal symptoms can be significant. Akathisia, insomnia, anxiety symptoms and a rapid deterioration following dose changes warrant particular attention. This clinically important issue has not been adequately investigated.

What this post does not say

This post does not contain any recommendation to stop taking antidepressants on one’s own initiative or abruptly. Abrupt discontinuation can trigger withdrawal symptoms; at the same time, the condition being treated may return or worsen. Discuss any changes to your current medication with your treating doctor.

Nor does it constitute an ‘all-clear’ signal. Warnings issued by medicines regulatory authorities regarding children, adolescents, and young adults are based on a genuine signal. They are not outdated simply because one side of the debate considers them to be inconvenient regulation.

In the event of acute suicidal thoughts, immediate danger or the feeling that you are not safe on your own, rapid help is needed: in Germany, call 112, the medical out-of-hours service on 116117, a psychiatric A&E department or the Telephone Counselling Service on 0800 1110111, 0800 1110222 or 116123. Further information can be found in the article on suicide prevention.

Key points in brief

•            People are often prescribed antidepressants for a condition that itself increases the risk of suicide. Comparisons between those who are prescribed antidepressants and those who are not therefore first take into account the indication and the severity of the condition.

•            This bias is known as ‘confounding by indication’.

•            Stone and colleagues analysed 372 randomised, placebo-controlled trials involving 99,231 adults. The medication's effect varied significantly by age.

•            Among those under 25, the odds ratio was 1.62 for suicidal thoughts or more serious events and 2.30 for suicidal behaviour. The absolute risk difference was 5.34 and 3.64 events per 1,000 participants, respectively.

•            Among those aged 25 to 64, there was no statistically significant difference in suicidal behaviour between the active drug and the placebo; among those aged 65 and over, the rates were lower in the active drug group in this analysis.

•            Completed suicides were too rare in randomised trials to allow for a reliable conclusion. Stone and colleagues identified eight such events.

•            The re-analysis by Hengartner and Plöderl from 2019 reported an odds ratio of 2.48 for completed suicides in adults. The inclusion of non-randomised follow-up periods and the small number of events limit the validity of this finding.

•            Further statistical analyses found no confirmed increase but also could not rule out clinically significant harm.

•            In young patients, counselling and close monitoring are particularly indicated at the start of treatment and when doses are adjusted.

•            Antidepressants currently being taken should not be discontinued abruptly or without medical supervision.

Sources

·        Gøtzsche, P. C. (2026, 12 September). Antidepressants are major drivers of suicides. Mad in America. https://www.madinamerica.com/2026/09/antidepressants-are-major-drivers-of-suicides/

·        Barbui, C., Esposito, E., & Cipriani, A. (2009). Selective serotonin reuptake inhibitors and risk of suicide: A systematic review of observational studies. Canadian Medical Association Journal, 180, 291–297.

·        Hengartner, M. P., & Plöderl, M. (2019). Newer-generation antidepressants and suicide risk in randomised controlled trials: A re-analysis of the FDA database. Psychotherapy and Psychosomatics, 88, 247–248.

·        Hengartner, M. P., Amendola, S., Kaminski, J. A., Kindler, S., Bschor, T., & Plöderl, M. (2021). Suicide risk with selective serotonin reuptake inhibitors and other new-generation antidepressants in adults: A systematic review and meta-analysis of observational studies. Journal of Epidemiology and Community Health, 75, 523–530.

·        Hjorthøj, C. R., Madsen, T., Agerbo, E., & Nordentoft, M. (2014). Risk of suicide according to level of psychiatric treatment: A nationwide nested case-control study. Social Psychiatry and Psychiatric Epidemiology, 49, 1357–1365.

·        Plöderl, M., Amendola, S., & Hengartner, M. P. (2023). Observational studies of antidepressant use and suicide risk are selectively published in psychiatric journals. Journal of Clinical Epidemiology, 162, 10–18.

·        Stone, M., Laughren, T., Jones, M. L., Levenson, M., Holland, P. C., Hughes, A., et al. (2009). Risk of suicidality in clinical trials of antidepressants in adults: Analysis of proprietary data submitted to the US Food and Drug Administration. BMJ, 339, b2880.

Related

• Stopping antidepressants: What the data really says about withdrawal symptoms

• Depression: causes, symptoms and treatment options

• Highly functional depression in men: symptoms, causes and solutions

• Childhood Trauma: Live – overcoming suicidal thoughts in cPTSD.

• Suicide prevention: help in life crises for young people and adults