Updated: 6 days ago
I am a psychiatrist in private practice. A common situation I find myself in is this - a patient (usually, but not always, a young woman) is referred to me by their primary care physician, who has started them on an antidepressant (usually for the first time, and usually an serotonin reuptake inhibitor, SRI). The patient has failed to improve, prompting a psychiatric assessment. The history reveals past episodes of probable hypomania and/or cyclothymic temperament, and of course there is often a family history of bipolar illness or recurrent severe depressive illness.
What is your usual approach to these situations? Generally I will offer patients a choice of lurasidone, lithium, or lamotrigine (although I am aware that the evidence of the acute antidepressant efficacy of lamotrigine is quite limited), with a plan to taper and cease the antidepressant at a later date. I have found that, following a description of the relative risks and benefits of the three medications, people will most often opt for lamotrigine (especially young women). And, in my clinical experience, those in this group who opt for lamotrigine will often end up doing very, very well. Furthermore, many of these people are adamant that the antidepressant is having an acute effect, despite my clear advice that it the evidence for its acute efficacy is very limited.
Should I be advocating for lithium or lurasidone (or something else) more strongly, despite my clinical experience of lamotrigine in this context being highly favourable? And should I be more aggressive in tapering antidepressants early on?
PL: This scenario is increasingly common in the experience of the PL editor also. Once you begin to look for affective temperaments, you begin to see it everywhere. What used to be "MDD" and "GAD" or "ADD" or bipolar spectrum or type II bipolar illness often turns out to be cyclothymia. Mood temperaments explain a lot. If someone has cyclothymia, for instance, she will have mild constant mood swings, not severe enough to diagnose bipolar illness - but it's there. She will be anxious frequently, as mood symptoms go with anxiety. She won't be able to concentrate most of the time. And cyclothymia predisposes to depressive episodes, so she will have depressive episodes now and then. Antidepressants are more likely to cause mania/hypomanic switches in such persons also. And yet the baseline cyclothymic temperament seems to respond well to simple interventions like low-dose lithium or valproate. Full doses often are not needed. In contrast, such patients fail or do not respond fully to multiple trials of full dose SRIs and/or amphetamines. In some cases, such patients also seem to respond to low dose dopamine blockers, like aripiprazole.
In this field, the research exists to tell us what's wrong and how to diagnose it. The research is not there to tell us what to do in any definitive way. There are some observational reports on low-dose lithium and valproate, and that's it. That doesn't mean that low dose dopamine blockers might not work, or that lamotrigine might not work or even carbamazepine. We don't know. When we don't know, the bad news is that you cannot be told what to do. And the good news is that you cannot be told what to do. So the PL view would be to follow your clinical observation here and continue to use lamotrigine if it seems helpful, carefully of course, with low and slow dosing.
Regarding the patient experience that antidepressants are helpful, the placebo-controlled randomized studies in bipolar illness prove that this is natural history, not drug effect, and the patient conclusion is correct clinically (they're better) but false scientifically (they're not better because of serotonin reuptake inhibition). The best approach is not to prescribe the antidepressant at all so as not to have to explain why it should be stopped. Although even here, the research is not clear in mood temperaments. What we know for now is that antidepressants can cause mania; we don't know if they're effective or not. Again, not a reason to prohibit them, but certainly a reason not to assume they should be used.
Follow-up: [From the clinician who initially asked the above question] Interestingly the patient I had in mind when I wrote to you originally has ended up responding very well to low-dose lithium (in combination with lamotrigine). The more I got to know the patient, the more obvious it became that she had fairly clearcut bipolar disorder, and not “just” cyclothymia.
PL: This is an important point too. Sometimes we underestimate the symptoms, and the patient may underreport, and then we are dealing with standard bipolar illness, not a mood temperament. Other times, we overreport the mood or associated anxiety or cognitive symptoms, and usually it is labeled "MDD" or "ADD" or "GAD"; then it is misdiagnosed altogether and the underlying mood temperament is missed.
Anecdote: On the inpatient unit, I educated our team that a recent increase in sertraline (Zoloft) may have worsened the delusions and auditory hallucinations in a patient with bipolar depression.
PL: Sertraline has dopamine reuptake blockade, which is potent, and thus could cause or worsen delusions or hallucinations. Another reason why SRIs are not "selective" (hence not "SSRIs").
Question: How would PL approach this case and what does it think of the plan taken?
A 14 year old female is in the midst of a likely resolving acute depressive episode (onset spring 2019). Around this one-year mark, neurovegetative symptoms including energy are beginning to improve which is consistent with natural history of the disease. She has bipolar genetics, as a maternal uncle has bipolar disorder; brother also attempted suicide. She is at increased risk for such given her young age of presentation and family history. A SRI may be helpful in the short-term for this acute episode, though it also carries risks of activation, and at this time, as the episode seems to be resolving, is not recommended. Time will reveal whether she may need longer term treatment for prevention of episodes with a mood stabilizer. Of note, her symptoms are on the background of years of low mood which emerged following parents' divorce, though no other trauma, for which psychotherapy is indicated.
Plan: For acute depressive episode, consider SRI (fluoxetine) though will likely not use given resolving symptoms.
Continue weekly psychotherapy to process impacts of divorce plus the experience of an acute depressive episode, and to support resolving symptoms.
PL: PL agrees with this overall approach. In the PL view, it’s best not to give SRIs in such cases. You might highlight the suicide risk and change the word “activation” to “mania” or “manic-like worsening”, to be clear on the problem. One could summarize this way:
“This patient with early-onset depression in the setting of bipolar family history has a high likelihood of eventually having a manic episode and thus bipolar illness, which usually begins in late adolescence or by age 20. Also SRIs increase suicide risk in this population, especially with bipolar genetics. For those reasons, antidepressants are best avoided initially….etc”
Follow-up Question: To be clear to generalize the concept, if a clinician ended up treating this same patient (say the episode was not resolving), would he/she be more likely to use an SRI in short-term with careful monitoring, or go ahead w/ dopamine blocker (for short-term) or even mood stabilizer?
PL: This is a tough question. In the US, it has important legal implications. Since clinicians need to be aware of their legal risks, they should realize that they need to justify why they would not do the standard of care, which is what most clinicians do, ie, use SRIs.
So it would depend on the patient’s willingness, but PL would prefer to treat with low dose dopamine blocker, like abilify, or low dose mood stabilizer, like lithium or lamotrigine. But clinicians should make sure they document that they discussed SRIs, and the pros and cons including suicidality, and that the patient agreed with the idea of using something else initially.
Another option would be to use a low dose SRI that would be a near-placebo, like 10 mg of citalopram, and then use aripiprazole or low dose mood stabilizer with it. Then one could taper the SRI if she improves later and leave her off if she remains well. That way clinicians would be more protected legally as well.
A 21 year, female in mid-March 2020, had a severe manic episode. She had been diagnosed two years earlier. She was hospitalized for 3 weeks and stabilized on Latuda (Lurasidone) 80 mg qhs – as she refused any mood stabilizer. She won’t take Lithium as she claims it left her numb and “ a robot.” She won’t take Lamictal for uncertain reasons. I’ve been adding Tegretol (carbamazepine) and it has been very, very useful. Everything has quieted down, pretty nicely. She’s now on 250 mg qam and 500 mg qhs, where we’ll stay until getting a carbamazepine level next week. She continues on Latuda 80 mgs qhs, which is her “go to” drug during most of 2019 (when she was moderately stable) and a drug she has a lot of faith in. Curbside questions:
Do you worry about an interaction between Tegretol and Latuda that is problematic? The combination over these past several weeks seems to be quite good.
Should the Tegretol level be a “trough” level – early in the morning, before her morning dose of Tegretol?
Should I wait until we hit your suggested “target level” of Tegretol, of 8, before reducing the Latuda?
At the “target level” should the Tegretol dose continue to be split, a.m. & p.m., or can it all be moved to the qhs dose?
There will be lower levels of latuda with tegretol, but since you’re dosing latuda higher, that is a way to manage that issue, as long as she's tolerating it fine.
There may be no need to reduce latuda doses for the reasons given above. The target level really is not well correlated with clinical efficacy. Dose to clinical effect, not to the level.
On pharmacokinetic grounds, the dose should be given twice daily since the half life is about 12 hours. However clinicians have given it at night in some patients and have not noticed any difference. If it's more convenient, in the long term, it can be given at night in one dose.
Question: What is your take on the Vasquez et al 2013 study that a bipolar expert likes to quote in support of using antidepressants in bipolar depression?
PL: This is pseudoscience. It reports responder rates (50% improvement), not actual change in depression symptom scores. For instance, it gives a relative risk of improvement with antidepressants in bipolar depression of 1.46, which means a relative 46% improvement. The important word is "relative". This is the same trick that was used to claim that antidepressants are highly effective in a recent widely-cited meta-analysis in the Lancet, previously analyzed in the PL newsletter.
Let's take the bipolar depression review above. A meaningless absolute difference between drug and placebo can be transmuted from dross to gold if we convert a small absolute difference to an apparently larger relative benefit. So if we have 100 patients in this study, and we define treatment response as 50% relative decline in symptoms, we may have some patients on drug who improve 16 or 17 points from a baseline of 30, and are responders; and we may have fewer who improve at that level on placebo, but may only improve 13 or 14 points from a baseline of 30, so they are not responders. Then when we calculate the one group versus the other, we'll have more responders in the antidepressant group (45% in this review) than placebo (34%). Dividing these precentages (45/34) gives you the relative risk of 1.46, which seems so impressive. That 46% relative benefit is actually a 11% absolute benefit, which clearly is much smaller.
The authors try to address this issue statistically with a "number needed to treat" analysis, reporting a benefit of 6.2, which is a good effect size (NNT of 5-10 is considered a moderate benefit). NNT still relies on the definition of treatment responder, which as noted above, could overstate benefit. But for some reason, they assiduously avoid looking at the absolute benefit in a much more simple way, which is the absolute scores on the depression rating scales. Perhaps they avoid this approach because when done, one sees repeatedly that there is very little benefit.
Here's the rub. If your depression is terrible - let's say a score of 30 on the Montgomery Asberg depression scale - and it improves to 15, then that's a 15 point improvement, which is clinically notable. You're still depressed but one-third less so than previously. Your sleep is bad, but not as bad; your appetite is low but not as low; your interests are diminished, but not as diminished. That's with the drug. In these randomized trials, there also is placebo. Let's say placebo leads to improvement from 30 to 16; that's still good benefit; your symptoms are notably better though present. The absolute benefit is 15 points for drug and 14 points for placebo. The absolute difference is only one point, which is clinically meaningless between the two interventions, meaning the drug actually isn't doing anything beneficial that can be attributed to the pharmacological properties of taking the drug.
What was the absolute difference between antidepressant and placebo in these bipolar depression studies? The review does not say, but in our more updated 2016 meta-analysis, we found that it was about one point on standard depression rating scales (as in the example above). For context, the average difference between antidepressants and placebo in "major depressive disorder" studies is about two points. The minimum cut-off that the FDA accepts for marketing a drug is three points. Anything lower would be considered clinically meaningless by the FDA. In other words, antidepressants in bipolar depression (and most also in unipolar depression) have clinically meaningless benefits, if any.
What explains the unwillingness of so many experts to accept the weak effects, if any, of standard "antidepressants"?
Question: I wanted to ask about the first confrontation of the patient with a depressive condition (not related to the subtype of depression), but rather the criteria that you consider fundamental to say that a person is in clinical depression.
PL: PL likes the term "clinical" depression (rather than "major" which was invented by DSM-III to contrast with "minor," later dropping the latter. The term "clinical" implies that the depressive symptoms are severe enough to reach clinical attention, meaning visiting some kind of health professional. This clinical attention could involve intervention with drugs, psychotherapy, or neither (watchful waiting). DSM terminology tends to emphasize the concept of functional impairment, but sometimes people can function normally at work or at home, and still have severe depression, even receiving clinical attention. So for PL, the concept of clinical depression is, in part, a proxy for severity: a depression severe enough to be beyond the bounds of typical sadness in human experience. PL would not emphasize the number of symptoms, nor the 2 week minimum of DSM. Rather, such symptoms do tend to last months when severe, and usually involve more than just one or two symptoms.
Question: Can we hear your a comment on this new Cochrane review of valproate in acute mania?
PL: This meta-analysis concludes that valproate is effective in acute mania, which is obvious based on multiple studies. There are few studies in adolescents, so efficacy there is not proven. It is claimed that valproate is less effective than dopamine blockers, like olanzapine and risperidone, which may be true, but the olanzapine studies are misleading. One was conducted by Eli Lilly, the maker of olanzapine, and it dosed valproate low; it found more efficacy with olanzapine. Another was conducted by Abbott Laboratories, and it dosed valproate high; it found more efficacy with valproate. These important clinical nuances get lost in meta-analyses. The mean level of valproate that was effective in its FDA-labeled trials for acute mania was in the mid 90s. At such high levels it is effective, and it may be similar to neuroleptics. However, some patients may have trouble reaching those doses. Thus, the PL view is that high dose valproate is effective in acute mania by itself, but in most people, dopamine blockers remain easier to dose and are very effective for acute mania. There is an old literature, also, that neuroleptics were more effective than lithium in acute mania. PL does not doubt that dopamine blockers are rapid beneficial treatments for acute mania. But they are not "mood stabilizers," for the reasons provided on the PL March 2016 newsletter, and do not have the longer term benefits seen with lithium and valproate.
Comment: One impression from the local scene is that over-reliance on “atypical” monotherapy for maintenance is increasing. (Antidepressants continue to be mistakenly prescribed, as well.) Prescribers are unaware of the problems with the enriched design. Of even greater concern: they have an irrational fear of lithium and appear to know little about its ability to prevent suicide and relapse into depression.
PL: Thanks for this excellent comment. The link above explains why PL does not think that dopamine blockers are "mood stabilizers" and thus should not be used in long-term treatment of bipolar illness by themselves. The problem of the "enriched" design is key: patients are preselected to be treatment responders to the dopamine blockers before the maintenance study begins. This biases the study from the beginning and produces invalid results, as explained in the PL newsletter here.
On the irrational fear of lithium, what can PL say? It is irrational itself.
Question: A patient of mine who is on oxcarbazine is concerned about the article that just came on in JAMA about anticholinergic medications and dementia. I am not aware that oxcarbazine has prominent anticholinergic properties. Can you comment on this please?
PL: Anticholinergic agents have direct impairment on cognition; whether this leads to dementia long-term is much more questionable, and doubtful in the PL view. Oxcarbazepine has no anticholinergic effects; unfortunately it's also basically ineffective for any psychiatric uses.
Question: You recommend the TEMPS scale to assess mood temperaments. Do you know where the questionnaire can be obtained? I can not find it online.
PL: The TEMPS scale has been published in multiple different papers in different lengths. PL prefers the 50 item scale which is reasonably brief and patients can fill it out in the waiting room. The original 50-item paper can be found here, with the scale in the appendix. However, the paper doesn't provide the specific scoring guidelines. PL has prepared its own version of the TEMPS-50 scale with scoring guidelines included, along with the PL approach to diagnosis based on 75% cut-off on each scale as the strict definition, and 50% or higher cut-off as a broader definition. The scale is copyrighted by the Journal of Affective Disorders in the original publication, so for any commercial or broader use, that journal should be contacted for permission to use it.