Thinking Through Antidepressants and Suicidality
And the ubiquity of risk in medicine
While most depressed and suicidal people who start an antidepressant experience a reduction in suicidality over the course of treatment as the depression improves, in some cases antidepressants such as SSRIs can lead to suicidal thoughts, suicide attempts, or suicide death. The signal for suicidal ideation is clearer for children and adolescents in clinical trials (hence the FDA black box warning), but the concern is not absent for adults.
SSRI-related suicide deaths in adults at the population level seem to be uncommon enough that observational studies of adults do not show a conclusive link. In RCTs and observational studies, to oversimplify a rather complicated and mixed body of literature, SSRIs seem to neither increase nor decrease suicide rates in adults overall—which could either mean that there is no substantial effect either way or that the two effects are roughly equal in magnitude and cancel out.
In my clinical work, I frequently encounter individuals who tell me that they have experienced suicidal thoughts on antidepressants. This experience is often restricted to a particular antidepressant, and a different one is tolerated well, but not always. In a small number of cases (usually with significant baseline mood dysregulation), all monoaminergic antidepressant medications seem to provoke suicidality.
The scenarios of SSRI-related suicidality I have observed and heard about from patients tend to fall into the following:
Initiation of antidepressant leads to intense anxiety or dysphoria that produces (or exacerbates existing) suicidality
Treatment with antidepressant leads to manic or mixed symptoms, which lead to suicidality
Antidepressants make the person feel emotionally indifferent, and in the context of existing suicidal thoughts, it leads them to act on the suicidal thoughts because the usual emotional aversion to suicidal thoughts gets replaced by, “Eh, so what if I die?”
Stopping an antidepressant abruptly (in someone who has been on them for a prolonged period) leads to a sudden exacerbation of mood and suicidality (withdrawal-related suicidality)
Such things make physicians uncomfortable, but I’ll restate it clearly: on starting an antidepressant (as well as going off them abruptly), some patients can experience worsening anxiety, agitation, restlessness, irritability, mood lability, or akathisia. These can subsequently induce or exacerbate suicidal thoughts, which in some cases can lead to suicide attempts and, rarely, suicide deaths. In my clinical experience of working with people with complicated mood and anxiety disorders, these issues with antidepressants are common enough that I try to see patients soon after starting an SSRI so that I can address early issues with tolerability. The idea of just starting an antidepressant and seeing them a month (or months) later worries me. As a psychiatrist, I get to see more complicated cases, and my experience doesn’t generalize to primary care, but I do think that even in primary care, it should be anticipated that a substantial number will struggle with SSRI-initiation and that a close follow-up after starting an SSRI is a useful general strategy.
SSRIs perform best, IMO, when the problem is chronic—generalized anxiety, dysthymia, adjustment disorder with high neuroticism—or sufficiently severe, e.g., major depression. Stress-related anxiety or depression that is mild and anticipated to be self-limiting is best addressed via supportive interventions or short-term use of quick-acting medications, such as sedatives. SSRIs can still be reasonable options in such situations, but the risk-benefit considerations are not quite straightforward.
We accept all kinds of risks in our lives. We drive cars on roads, we fly on planes, we drink alcohol socially, we use tobacco and cannabis, we invest our lifesavings in stocks, we undergo general anesthesia for surgical procedures, etc., etc. When my son was an infant, he loved being in a particular bouncer. We used it all the time. This is despite the fact that there was a warning prominently displayed on it. “Fall Hazard: Babies have suffered skull fractures falling while in and from bouncers.” Essentially, we accepted a very small risk of a catastrophic event because of the tremendous everyday convenience the bouncer offered us. Such considerations are ubiquitous in healthcare.
Every time a person uses an antidepressant (or any medication, for that matter), there is a small possibility that their experience may go very badly, sometimes in life-altering ways. Risk is ubiquitous in medicine and in life; the possibility of iatrogenic harm is unavoidable. This will not stop people from using antidepressants and other psychiatric medications. I have worked with many patients who experienced serious adverse effects with an antidepressant in the past but were still interested in trying others because they were desperate for relief.
But being transparent with the public about these risks accomplishes two things.
First, if patients and clinicians are aware of the possibility of harm, they can recognize the problem early and stop the medication before it progresses.
Second, it serves as a reminder that the condition being treated should be distressing enough or severe enough for the person or the benefit of the medication advantageous enough that they are willing to accept the possibility of uncommon but serious adverse events. The threshold for what counts as distressing enough or severe enough or advantageous enough will be different for different people, but what that threshold is in any particular situation for any particular person is worth giving some thought to.
This discussion was originally published as part of a post (Dec 2024) about Thomas Kingston’s death and has been edited to function as a standalone post.
See also:
Maybe this new research finding points towards a likely scenario: https://www.nature.com/articles/s41380-026-03644-x
They observe two different neuron populations that react in opposing directions and different timescales to SSRIs. One increases short-term stress, and with that potentially suicidality, the other provides the long-term antidepressive effect.
Still, this is a mouse study with only fluoxetine tested. So, it is unclear how this would impact different patients and age groups differently.
This is sad topic for me because it highlights several problems I see in modern psychiatry.
The issue is whether particular medications can “cause” a patient’s suicide, and therefore does the prescribing physician have some kind of liability for a patient’s death.
The reason I find this topic sad is that the “original sin” is for psychiatry to imagine that it is going to deliver care without consequences. Suicide is a rare event relative to the total number of patient encounters psychiatrists engage in. Not all suicides are preceded by contact with a psychiatrist who has personal responsibility for a patient’s treatment, for example suicides maye not interact with a professional , they may interact transiently with a professional for example a PCP who refers him to a psychiatrist, or he may go to an urgent care or ER and receive care that is understood to end at the patient’s departure from the care location.
In any event rarely, a psyuchiatrist will have personal responsibility for the patient’s ongoing care and the patient may commit suicide. Most of these patients had psychiatric complaints preceding their suicide. The psychiatrist makes recommendations the patient adheres to or doesn’t, and even more rarely, the psychiarist will make a “wrong” decision that will be followed by a suicide. By “wrong”, I mean that he psychiatrist opted for a treatment that did not make the patient better fast enough to avert his taking his life.
The subject at hand of course isn’t just that a medication didn’t help, but that it made a patient worse than he was in the first place. That is the implication of the discussion anyway. And Aftab and others want to chip away at the concept that a psychiatric medication could “make” someone commit suicide.
What I would like to point out is that this is a distinction without a difference, and rather than rooted in actual concern for the patient’s welfare, it is rooted in concern for the psychiatrist’s liability. And that makes me sad.
Whether the medication “didn’t help”, “made the patient worse”, or “didn’t help enough”, makes no difference to the patient, he just wants to feel better and his loved ones want him to remain alive and intact. Only psychiatrists engage in such a discussion, I don’t think any other field attempts to parce out these distinctions to avoid liability, and given the already extremely small liability psychiatrists face compared to other specialties, the argument seems petty and misdirected.
I see no reason why a particular medication could not “trigger” someone’s suicide, just like a traffic ticket, an alchoholic binge, an argument, or a financial notice, might, amongst others. To suggest that this could never be possible seems ridiculous. I know that the NAS has attempted to stratify the “statistical signal” of suicidality by age to suggest that certain age groups might be more vulnerable to this effect and that is fine but doesn’t help us clinically because a busy clinician will still see patients from each of the age groups with depression and potential suicidality, as I do on a near daily basis.
The real question everyone should be asking is, is the strategy of prescribing antidepressants to depressed or anxious people, protective against suicide generally: that is the only patient-centered question. And of course, absolutely no one has investigated this in a prospective way, in fact I have not even seen a chart review meta-analysis (the “garbage” of statistical analysis) that answered this specific question. I would argue this is because psychiatrists are more concerned with liability in this instance than the welfare of their patients (collectively, i don’t mean to criticize Aftab or any other individual).
I PERSONALLY (meaning this approach works for me but may not work for other people) don’t see the “big deal”. Lots of my patients complain of suicidality or emotional problems distressing enough one could imagine suicide to result. I make serious clinical decisions all day every day. I am not going to get it right every time. My solution is to know my medications extremely well and to recommend follow-up appropriate to the selected treatments.
I have never, thankfully, had a patient commit suicide while I was responsible for his treatment. This could happen at any time of course, I am not claiming special powers. But if I have decided to practice psychiatry, already one of the slower-paced, less risky specialties, I think to imagine I could never make a patient worse with a treatment choice strikes me as cowardly and self-serving. Of course I make “bad decisions” every day, hopefully the good exceeds the bad. Also, because I have ongoing responsibility, I am very focused on appropriate follow-up, I believe this remains patient-centered while protecting me from liability issues. But I won’t lie. I could get things wrong at any time. I am not different from any other doctor practicing clinical medicine. If psychiatrists have no liability for their decisions, I am not sure that they are making decisions that have any value.