Medication Treatment Challenges in Mood Disorders: A Discussion with Jim Phelps
Reevaluating the emphasis on efficacy vs risks
Jim Phelps, MD, is a psychiatrist who specializes in the treatment of complex mood disorders, with over 20 years of clinical experience across a range of settings. He served on the International Society for Bipolar Disorders’ Diagnosis and Chronotherapy Task Forces. He founded the website PsychEducation.org in 2000 and recently a simpler version aimed at primary care physicians and patients, DepressionEducation.org. He is the author of Why Am I Still Depressed? (2006) for patients and families and A Spectrum Approach to Mood Disorders (2016) for professionals. Currently, Dr. Phelps is completing YouTube/@PsychEducation, presenting the same material in 30 five-minute videos. He is particularly excited to announce a professional documentary on the bipolar spectrum coming in June/July 2025; preview and podcast up now at BrainstormTheFilm.com.
Aftab: Let’s start by talking about issues related to the management of depression and anxiety that arise in the primary care context and general psychiatric practice. I have a couple of complaints from my side that I want to mention. First, “depression” has become a very bloated diagnosis, and clinicians end up diagnosing basically any psychiatric presentation with dysphoria, sadness, low mood, etc., as “major depression,” even though it may be better conceptualized, for instance, as adjustment disorder, generalized anxiety, complicated grief, substance-induced, or something else. A second issue is that patients are being prescribed antidepressants without any transparency regarding the side effects and potential harms due to which medications are considered less preferable for milder and self-limiting conditions. Third, patients are being started on antidepressants with no plan to take them off them, with many patients staying on them by default, and down the road when patients do want to come off the medication, the prescribers lack the competence to safely taper them off the medication. All this in aggregate creates an alarming state of affairs. Curious to hear your perspectives on these and other additional issues.
Phelps: Thanks very much, Awais, for the invitation to discuss these things—with one of the best minds in psychiatry these days.
“Diagnostic bloating”: may I assert that this is largely due to our inability to divide the presentations you mention—dysphoria, sadness, low mood; plus malaise, ennui, existential angst, profound anergy, hyperfocus on negative thoughts—along some kind of “natural fault lines” outside our diagnostic system. These phenomena blend with one another in myriad ways. We have been trying for decades to “carve nature at its joints” with regard to mood problems, without success. So I think it’s time to accept this inability, at least for now, and diagnose accordingly, namely: accepting diagnostic uncertainty as the norm rather than the exception.
Regarding antidepressants in primary care: yes, there’s plenty of room to help there, with patient education and their shared decision-making regarding treatment options for depression. At minimum, we should take every public opportunity to present: a) the efficacy of psychotherapy as an alternative treatment for depression and anxiety, and its increasing availability in the form of digital therapies such as MoodGym and E-Couch; b) the frequency of sexual problems and weight gain with SSRIs; and c) the increasing evidence that some patients have extreme difficulty getting off antidepressants—at the very least, that’s 1 person in 100 (10-20 times higher than an allergic reaction to lamotrigine, which we always clearly describe) and some would say it’s more like 1 person in 50, or even more common, depending on how one defines “extreme.”
Aftab: How does conceptualizing mood disorders as existing on a spectrum of bipolarity influence the way we approach their management?
Phelps: Philosophically, you know that a short answer to this question is … challenging, shall we say? My attempt: with the spectrum approach, the diagnostic question changes from “could this depression be bipolar?” to “how bipolar is it?”. In querying for hypomania/mania, subthreshold answers should not be discounted.
If DSM criteria are met, fine. If not, more data can help: family history, age of onset, course of illness (episodic? postpartum?), and response to treatment (mainly adverse reactions to antidepressants). Then one can say to one’s patient “you don’t have ‘bipolar disorder’. But you don’t have a pure/plain depression either, because you have a, b, c, and d.” Then follows a discussion of the pro’s and con’s of mood stabilizers versus antidepressants, admitting uncertainty as to which route is warranted.
For this mid-spectrum group of patients, antidepressants carry more risk than for those who have no bipolar markers: not just causing mania, which is not very common even in higher-risk patients, but also inducing mixed state symptoms (recently found to most commonly include “the 4 A’s”: anxiety, anger, agitation, and attention problems). Often depression improves but these other symptoms become such a problem that another medication is added to handle them: an antipsychotic, leading to much greater long-term risks and tolerability problems.
Compare lamotrigine: properly managed, it has near-zero long-term risks (the allergy risk goes essentially to zero after titration) and most often zero side effects. So my short answer concludes: a spectrum approach to diagnosis leads to the consideration of lamotrigine as an alternative to antidepressants for patients who prefer a medication to a psychotherapy approach for their depression.
Aftab: But the evidence for using lamotrigine for unipolar-ish depression is fairly limited, right? CANMAT mentions lamotrigine as a third-line adjunctive agent in major depressive disorder, and I’m not sure if we have data for lamotrigine monotherapy. I haven’t used lamotrigine in unipolar depression, but my clinical experience using it in the acute treatment of depressive symptoms on the bipolar spectrum or “unspecified mood disorder” hasn’t left me very impressed. (It does seem more effective for maintenance treatment.)
Phelps: Thank you for that, Awais, as here’s another point of divergence for me and the treatment guidelines. They emphasize efficacy and give relatively little weight to risks and side effects. For example, the recent VA/DoD guidelines put quetiapine first for bipolar depression. Based on years of talking with patients, I think the weighting should be the other way around: risks and side effects should outweigh efficacy.
Phelps: Treatment guidelines emphasize efficacy and give relatively little weight to risks and side effects. Based on years of talking with patients, I think the weighting should be the other way around: risks and side effects should outweigh efficacy.
That’s like swearing in church, isn’t it? Let’s go further. I’d suggest that the need for efficacy evidence is proportional to risks, costs, and side effects. That is, treatments that have very low costs, risks, and side effects need only a minimum of evidence for efficacy, and vice versa.
That’s because placebos have such power. For example, perhaps I overestimate lamotrigine’s efficacy because of confirmation, expectancy, recall, or other biases. But if I am a believer, my placebo response rates will be substantial (supporting my biases). But these are responses, nonetheless, with the lowest risks of any prescription psychotropic.
Of course I believe I have not thus misled myself, and that lamotrigine really is far better than a placebo. In this I am supported by conversations with clinician colleagues at the International Society for Bipolar Disorders meetings, where every one of us seems to have arrived independently at the same conclusion. Not the academics, interestingly; the differences of opinion are really striking. I’m left wondering if one simply has to treat hundreds and hundreds of the patients best suited for lamotrigine (e.g., Bipolar II, with no antidepressant in recent history) to become so enamored of it.
But my larger point: let’s question “the hierarchy of evidence” as the primary determinant of a medication’s position in treatment guidelines. It’s just one part of the story. Listen to patients. They’ll tell you that risks, costs, and side effects are more important when they know that if the treatment works, they’ll be doing it for a long time. Pick a loser the first time, and you can pick another. Pick the med with the greatest efficacy, and it may be more likely to work, but now the patient faces weight gain and tardive dyskinesia and diabetes. I think I’m closer to common patient preferences with my weighting.
Aftab: You’ve previously written about the thorny problem of differential diagnosis of mixed states, that bipolar mixed states are nearly impossible to differentiate from depression that is comorbid with posttraumatic stress disorder, or generalized anxiety disorder, or attention-deficit disorder. I have to say that I encounter this challenge a lot in my own clinical practice as well, and the uncertainty persists despite a rigorous psychiatric evaluation. Primary care physicians and psychiatric nurse practitioners are often far less equipped for this diagnostic challenge, and I shudder at the thought of putting the responsibility of diagnosing mixed features on primary care docs. What are your recommendations for approaching possible mixed states?
Phelps: Humility. We have to be more open and comfortable with uncertainty, with ourselves and with our patients. It’s okay, we can manage, including by:
1. Getting more data. A systematic query of the non-manic markers (family history, age of onset, course of illness, response to antidepressants) should be routine.
2. Enlisting the patient and perhaps family. Explain the mood spectrum and mixed states. The more they understand about what we’re trying to determine (“how bipolar are you, if at all?”) the more they can help.
3. Presenting the pro’s and con’s of several treatment alternatives (e.g., a select antidepressant and lamotrigine or low-dose lithium or lurasidone) and then taking time for shared decision-making.
As for primary care… For years they drew a line: “bipolar” is beyond what should be asked of us, which was appropriate for Bipolar I. But Bipolar II complicated diagnosis and treatment, and now a research-informed definition of mixed states (including “the 4 A’s”) complicates them further. If psychiatry made it easier to refer complex patients, PCPs might be willing to screen for bipolarity more routinely (e.g. every PHQ-9 >10 gets a Rapid Mood Screener and any value >1 gets referred). Collaborative Care is one model for that.
Aftab: I’m curious about your personal views about antidepressant use in bipolar disorder informed by your clinical practice. I am familiar with the research literature: use of antidepressants in bipolar disorder is not supported by clinical trials, with medications failing to separate from placebo and, in some cases, causing mood instability, and guidelines do not recommend using them as first-line. Yet, in practice, antidepressants remain the most prescribed medication class for bipolar disorder in the US. I accept the research evidence. At the same time, in my clinical experience, I frequently encounter situations where first-line medications for bipolar depression do not sufficiently work, but an antidepressant does. Some of it is perhaps related to antidepressant actions on anxiety and neuroticism, and these would be expected to improve even in bipolar depression. I was recently, for example, working with a patient with a clear history of Bipolar I, who had had a good response to fluoxetine in the years prior to his first manic episode. Afterwards, when he had another episode of depression, lurasidone, quetiapine, and lamotrigine didn’t sufficiently address the depression, but adding fluoxetine to lamotrigine led to substantial improvement. I see these kinds of situations even more frequently in bipolar II. I am not saying that clinical experience should override evidence from high-quality RCTs, and I don’t advocate using antidepressants as first-line treatments for bipolar disorder, but at the same time, I find it hard to accept that they lack efficacy as a second or third-line treatment based on what I see in the clinic. Your thoughts on this?
Phelps: Thanks, Awais, that was well put: research data lean in one direction, clinical experience (yes, I’ve seen what you’ve seen) in another. One partial explanation is time frame. The longer I practiced, the more people I saw who did well on an antidepressant for a while, sometimes quite a while, and then started having what I’d now call a mixed depression.
Phelps: The longer I practiced, the more people I saw with bipolarity who did well on an antidepressant for a while, sometimes quite a while, and then started having what I’d now call a mixed depression.
For an extreme example, I published a case report of a patient who did really well on sertraline for 7 years (“I joined the human race,” she said, after she’d not improved on several other antidepressants). She presented with severe agitation, anxiety, and racing thoughts, which began when her dose was increased by her OB-GYN from 100 to 150 mg. Several atypicals were tried, and ultimately 400 mg of quetiapine stopped these extreme symptoms. After a year on quetiapine alone, she restarted sertraline on her own at 25 mg. After two doses, she called me saying, “I’ve joined the human race again!” I said, “Great, come see me tomorrow.” After taking a third dose that morning, she arrived with the same severe symptoms with which she presented initially—at a quarter of her previous effective dose.
Ultimately I ended up sort of specializing in antidepressant discontinuation (what else are you gonna do when patients being referred have “tried everything”?) because it seemed to lead to improvement in many patients. For example, a resident colleague and I published a case series of 12 patients whose extreme anxiety remitted with cessation of their antidepressant (admittedly, we had added lamotrigine and/or low-dose lithium, though none of them met criteria for bipolar disorder).
So I’m nervous about how long any improvement on antidepressants might last, even though it might be seven years. I’m nervous about your guy with Bipolar I on fluoxetine and lamotrigine. Yeah, I’m basically nervous about starting anyone with bipolarity on an antidepressant; heck, I’m nervous about starting anyone on an antidepressant, given what we’re learning about withdrawal.
I spent a lot of time with patients working through alternatives, including psychotherapies, low-dose lithium, and so-called “triple chronotherapy”: light therapy, shorter and earlier sleep using dawn simulators and virtual darkness, and sometimes overnight sleep deprivation. Notice the theme here: these are all treatments that have less evidence for efficacy than quetiapine, for example, but have little or no long-term risk.
Aftab: Do you think the antidepressant withdrawal issues are not more manageable with things like gradual hyperbolic tapering?
Phelps: Oh definitely, yes they are. I add a tweak to hyperbolic tapering, namely to make the first step as small as the patient can easily engineer (most can halve pills, some can quarter them; using a new prescription for the smallest available dose, which they’ll need at the end anyway). Because if the first step goes badly, the patient may think they “need” the medication, or that getting off will be too difficult. Then they’re really stuck on it.
So the graph looks more like an S-curve than exponential decay: slow, fast, slow. Some patients will discover they can go quite fast in the middle; some can even skip the tapering at the end. But a small step at the beginning helps identify those who will struggle with each decrement. They need a slower course and perhaps a liquid formula at the end. (Mark Horowitz, whose work on hyperbolic tapering brought it to wide attention, admitted in personal conversation that he does much the same with his patients.)
Phelps: I add a tweak to hyperbolic tapering, namely to make the first step as small as the patient can easily engineer. So the graph looks more like an S-curve than exponential decay: slow, fast, slow.
But here’s the concern, Awais: there are many patients who, despite such measures, will have extreme symptoms every time they reduce the dose, even by small amounts. I had one PhD psychologist patient who took nearly a year to get from 20 mg of citalopram to 5 mg. At 3 mg using a liquid form, she said, “just put me in the hospital and stop it.” She knew she’d be having suicidal thoughts.
And finally, the data that really drive my thinking are the accounts at SurvivingAntidepressants.org: thousands of what amount to case reports of profound withdrawal symptoms, so many as to overwhelm any skepticism about credibility. For example, some users describe counting the individual beads in venlafaxine capsules so as to reduce their dose by a few beads at a time. Based on such experiences, the site recommends tapering by 10% of one’s current dose each month. Do the math? This takes years.
Of course the question is: how common is such an experience? To which I say: does it matter if it’s 1 in 100 or 1 in 1000? These reactions can be life-limiting for years. Yet look at the Cochrane review of antidepressants for the treatment of generalized anxiety disorder that just came out: “We have high confidence that antidepressants are more effective than placebo at improving treatment response and that antidepressants have similar acceptability to placebo.” (emphasis mine)
If the withdrawal reactions described on SurvivingAntidepressants.org occurred during the randomized trials, rather than years later, I doubt psychiatry would be so sanguine about antidepressant prescription. Isn’t it shocking that we still don’t have a widely accepted estimate of the incidence of severe withdrawal?
Lest I sound like an extremist, I’ll point out that our group published data suggesting that 90% of primary care patients who stopped their antidepressant did so without even receiving the lowest available dose of their medication to facilitate a taper. These results suggest that a substantial majority of patients can discontinue their antidepressant without much difficulty. But it’s the rest of the patients I’m worried about. I’m still nursing the hypothesis that the more vulnerable subset is highly enriched with bipolarity, but that’s for another day.
Aftab: Second-generation antipsychotics in mood disorders… what worries you about their prevalent use? What are we neglecting and overlooking?
Phelps: That’s easy; two things. First, the TRIO study demonstrated an antidepressant effect of metformin. Metformin? Right; patients with treatment-resistant bipolar depression were randomized to receive metformin or placebo. Everyone entering the study had insulin resistance. Those whose resistance reversed had a 16-point drop in their MADRS scores—same as seen in randomized trials of quetiapine for bipolar depression. The effect size of reversing insulin resistance (versus those in which the resistance was not reversed) was a whopping 1.17. Compare the effect size of antidepressants vs placebo: roughly 0.3, in the meta-analysis of meta-analyses by Munkholm et al.
I hope the implication is obvious: insulin resistance is a common consequence of metabolic syndrome. Metabolic syndrome is a common consequence of most second-generation antipsychotics. If this study is replicated, that will strongly suggest that we are creating treatment resistance with our antipsychotics in at least some patients. Less so with lurasidone, perhaps? Yes, but not zero in my experience: rapid weight gain still can occur.
Secondly: what does it feel like to take an antipsychotic? A residency colleague thought we should know. “What a waste of a weekend,” he said after taking 2 mg of risperidone. “I just sat there.” By contrast, many patients express surprise about lamotrigine after being on other medications, “it doesn’t feel like I’m taking anything.” That’s the point: a medication you’re going to take for a long time shouldn’t feel like anything.
Phelps: Many patients express surprise about lamotrigine after being on other medications, “it doesn’t feel like I’m taking anything.” That’s the point: a medication you’re going to take for a long time shouldn’t feel like anything.
Similarly, in my hands, “low-dose lithium” means titrate from zero by 150 mg steps until improvement or something bad happens (then lower the dose and watch for another week or two to be sure nothing has improved). Blood levels are drawn only to insure they’re not over 0.8 mmol/L where sustained close monitoring and attempts to reduce the dose are needed. Using this approach assures that lithium is side-effect free. True, some people can’t take it, usually because of dulling, about 1 person in 10 in my experience. But the rest get benefits with no downside, assuming proper monitoring of creatinine, polyuria, and thyroid function.
Aftab: Is part of the issue with antidepressants and antipsychotics in mood disorders that we are overusing them for maintenance treatment and that they may offer a more favorable risk-benefit ratio if we encourage using them mainly for acute treatment?
Phelps: Well, one could frame it that way, which is probably how inpatient psychiatrists think. But those of us on the outpatient “receiving” end know how hard it is to stop these medications. Go too fast and one can precipitate the very symptoms one is trying to forestall.
Inpatient providers often regard lamotrigine as “too slow,” yet they are not unwilling to start antidepressants, which can take 4-6 weeks to kick in, the same time frame in which we could expect a response to lamotrigine. A trial of low-dose lithium is far faster. So dividing medications into “acute” and “maintenance” options, well, at this point it’s obvious that I don’t think that way. What’s the phrase, “begin with the end in mind”?
Aftab: There are far too many people with depressive and anxiety disorders, and we don’t have enough psychiatrists available to evaluate and treat them. There are also not enough psychotherapists to provide the requisite psychotherapeutic treatments. The healthcare system has relied on primary care to treat depression and anxiety and has expanded the psychiatric workforce by including nurse practitioners, but now we have the resulting challenge that psychotherapies still remain difficult to access and medications are accessible without the careful evaluation and appropriate consideration of iatrogenic concerns that is required. What do you see as the path forward?
Phelps: Two things here as well. First, I discovered long ago that online psychoeducation for primary care providers is far more likely to be viewed by patients and families than physicians. So while I haven’t given up on reaching PCPs directly, I’m working on making the information patients need to make diagnostic and treatment decisions readable (shorter!) and accessible (videos!); thus evolved DepressionEducation.org and YouTube/@PsychEducation.
Secondly: As I’ve probably already horrified some of your readers, I shouldn’t fear horrifying more of them. Very well…
An optimistic physician supply projection estimates that the number of psychiatrists might finally be sufficient to meet demand by 2050. In the interim, the current shortage worsens, peaking around 2030 (retirements exceeding residency graduations).
We need reinforcements. Psychiatric physician’s assistants (PAs) receive no specific post-graduate training in psychopharmacology, nor training in evidence-based psychotherapies beyond perhaps some motivational interviewing. Yet they are now prescribing psychotropics with varying levels of supervision. (For example, an advertised “fully remote” PA position describes “seeing and treating patients independently”.)
By contrast, to be a prescribing psychologist requires 2 years of post-graduate training in psychopharmacology on top of a PhD or PsyD, and usually (in the 7 states now authorizing them) further supervised practice thereafter. Because they are fully trained to offer psychotherapy, prescribing psychologists could fill the gap between primary care and psychiatry, providing skilled differential diagnosis of depression and anxiety plus a full array of treatment options. Yet our APA actively resists this idea. Indeed, in response to intense lobbying by organized psychiatry in Oregon, governors have twice vetoed psychologist prescribing bills that passed both houses of the legislature.
I’m ashamed of this guild protectionism, rationalized as protecting patients while we leave them without services entirely. So I’m beginning discussions with the American Psychological Association’s Division 55 on how I might assist them in other states.
Thank you, Awais, for the invitation to express these views and share them with your readers. I apologize for any flack you may receive for doing so! But I hope to prompt further thought, which I know is always your goal as well.
Aftab: Thank you, Jim!
This post is part of a series featuring interviews and discussions intended to foster a re-examination of philosophical and scientific debates in the psy-sciences. See prior interviews here.
Comments by Michael Ostacher, MD, MPH on X/Twitter [who was involved in the development of the VA/DOD guidelines]: https://x.com/RecoveryDoctor/status/1891176349374820387
"Jim Phelps is a lovely and thoughtful man whose work with patients I deeply respect, but he should read the whole VA/DOD guidelines before suggesting that we didn’t raise concerns about risks over benefits, or take into consideration patient preferences. See pages 54-55 https://www.healthquality.va.gov/guidelines/MH/bd/VA-DoD-CPG-BD-Full-CPGFinal508.pdf
Second, Jim sees people at the margins, so it fits you, but don’t mistake his clinical practice, with his admitted bias (mostly confirmation bias) about the results of his practice. People see him because they haven’t responded to treatment but the are not average patients.
Evidence is by definition limited; only some things are studied, and only in some populations. Guidelines are not a substitute for clinical decision making. Reasonable people can disagree, but our guideline does not put effectiveness over risk. At all.
Our synopsis is here, but it behooves clinicians to read the whole guideline. As an an important caveat: those factors that experts say imply bipolar disorder (family history, poor response to antidepressant, etc.) have very poor predictive value. They cannot be used to make a diagnosis.
https://www.psychiatrist.com/jcp/synopsis-2023-us-department-veterans-affairs-department-of-defense-clinical-practice-guideline-management-bipolar-disorder/"
"We have to be more open and comfortable with uncertainty, with ourselves and with our patients." ~ Very much respect for you, Dr. Phelps, and your approach to the overwhelming challenge of supporting suffering people from within a framework that a) struggles to achieve a level of understanding that can justifiably confer confidence, and b) typically pressures the practitioner to behave misleadingly as though certainty rules. Thank you, on behalf of your patients and all the others you have and will influence, for stepping into that more vulnerable space of not-knowing.