PsychEducation.org (home)

Antidepressants in Bipolar Disorder: The Controversies
(update controversy 3, 9/2009)

This page has been maintained for nearly 5 years. Unfortunately, most of the controversies below remain controversial. I even had to add one, controversy "zero", when new data arrived suggesting that antidepressants may not really be effective at all in bipolar depression.

The bottom line overall here: antidepressants may carry much more risk for people with bipolar disorder than is generally recognized. That antidepressants can cause "switching", bringing on a manic or hypomanic phase, is generally accepted, although how often this occurs is still hotly debated (somewhere between 4% and 40% of the time?).

However, antidepressants may pose bigger risks in the long term. Substantial evidence suggests that antidepressants can induce "rapid cycling". Indeed, it is a standard recommendation for the treatment of rapid cycling to gradually withdraw any antidepressant. In addition, more subtle "destabilizing" effects are possible. Antidepressants may make it more difficult to get a good outcome from an otherwise effective mood stabilizer treatment. There is even a concern that antidepressants may permanently alter the course of a person's bipolar illness, through a phenomenon called "kindling".

Therefore, considerable caution should be used before starting an antidepressant in a patient with bipolar disorder. Because data suggest that antidepressants may not be effective at all in this condition, one can wonder why they should be used at all, given the accepted risks (switching, rapid cycling), let alone the more difficult-to-prove concerns about kindling.

Finally, some patients clearly do better if they stay on an antidepressant. However, new evidence raises further questions about just how many patients meet this description. In my view, nearly every patient with bipolar disorder deserves at least a significant trial, if not several, of a treatment approach without an antidepressant to see if this might work as well as when the antidepressant is included -- or, quite possibly, better.

Page outline

Controversy "zero": Do antidepressants even work in bipolar depression?

Controversy 1: Antidepressants can cause "switching" from depression to hypomania or mania, but how common is this? Can we identify in advance likely "switchers"? Do mood stabilizers protect against it?

Controversy 2: "Mood destabilization": A longer-term risk that goes beyond the short-term "switch" risk?

Controversy 2a: Do antidepressants cause rapid cycling? Or rather, is there anyone who thinks they don't?

Controversy 2b: "Kindling": Could antidepressants cause a permanent worsening in the course of a person's bipolar disorder? (no one -- or perhaps at most a few other extremists -- appears to worry about this but me, fortunately).

An opposing view, well expressed, with links to yet more such thought

Controversy 3: For patients already on an antidepressant, and who are doing well -- should they taper off or stay on?

Controversy "zero" : Do antidepressants even work in bipolar depression?
(Added May 2007)

If antidepressants are not clearly effective in bipolar disorder, then how much risk they carry is moot, right? In other words, if antidepressants don't work, why would one want to use them? If they are not effective, it doesn't really matter how much risk they pose: even if that risk was very low, there would still be no point in using them.

(Sorry, I know I just said the same thing three times in a row, but the question is so fundamental, isn't it? Obviously it is the first question we should ask, and yet it hardly ever seems to be asked, except by those who question the use of antidepressants. Even I did not begin with this question when I wrote this page; thus the backtracking to "controversy zero")

However, at the moment there is no clear yes-or-no answer regarding whether antidepressants actually work in bipolar depression. The assumption for years has been that they do indeed work, just as they often do in unipolar depression. However, the evidence for their effectiveness in bipolar depression is actually surprisingly limited. On the other hand, most clinicians are quite convinced that antidepressants can treat bipolar depression. They have often seen rapid and obvious response to these medications in patients whom they believe to have bipolar disorder. So how could it be that the research evidence does not support this? In the face of their clinical experience, they usually assume there must be something wrong with the research studies.

I myself have seen quite convincing improvements. Could they really have been placebo responses? Theoretically, we clinicians should not be convinced by what we have seen with our own eyes, but rather by the randomized trial data. Yet there is an old joke about "evidence-based medicine": when the study agrees with your point of view, you cite it. When it disagrees, you criticize the study design! The following study is a good example, trumpeted by those who agree with its conclusion, examined for flaws by many others.

In April 2007 The New England Journal reported a study by Sachs and colleagues which strongly suggests that antidepressants do not have sustained the effectiveness in patients with bipolar depression who are already on an adequate level of a mood stabilizer. to understand this study, it is important to understand how they measured response to the medication. Instead of the usual approach, which typically looks for a 50% reduction in depression scale scores as a criterion for effectiveness, in this study they used "durable response": patients had to have at least an 8-week period during the six-month study in which they had neither depression nor significant manic symptoms. Using this measure, here was their result:

Of course this is only one study. However, it is consistent with many previous studies showing the same result. In general, of the studies looking at long-term response to antidepressants, in patients with bipolar disorder, there are very few that suggest long-term benefit. These studies were summarized in a review by Ghaemi, available online. Here's the summary table:

However, one research group has produced several studies showing sustained benefit from antidepressants in patients with bipolar II, even without a mood stabilizer on board. Their 2008 study is probably the most dramatic yet.Amsterdam In patients with Bipolar II, they compared venlafaxine (Effexor) versus Lithium as the sole treatment. Patients on venlafaxine clearly improved more than those on lithium. Note that these results directly contradict those from the Sachs/New England Journal study above. One of the reasons for this might be the "open label" design of the Amsterdam study, in which both patients and investigators knew what medications people were receiving. Since this research group clearly has an enthusiasm for antidepressants in the treatment of bipolar disorder, that enthusiasm could have been conveyed to the patients. By comparison, the Sachs study was "double blinded": neither the patients nor the researchers knew what any given patient was receiving, until the code was broken after the study.

In theory, a double blinded study is a more powerful research tool than an "open label trial"; however, for now I think we should conclude that even the most basic question about antidepressants for bipolar disorder -- do they work? -- remains unclear. They probably work in the short run for some people with Bipolar II. In my view, the benefits from antidepressants are probably less than they seem, but there are some patients with bipolar disorder out there who do well on them -- even without a mood stabilizer. For complex patients, I think the downside of antidepressants, presented below, generally outweighs the potential for benefit. But this must be individualized.

Controversy 1 : Switching into Mania or Hypomania
(revised 7/2008)

As of late 2007, three recent studies have addressed the question as to whether antidepressants can induce switches into mania or hypomania. The bottom line: switching happens, but now often seems to vary widely, depending on the patient, and to some extent on the antidepressant chosen.

1. Leverich et al, Am J Psych 2006

How common is the development of hypomania or mania in a depressed patient with bipolar disorder, who is given an antidepressant (a phenomenon generally referred to as "switching" from depressed into hypo/mania)? Until recently this was quite debated. In a 2007 study designed specifically to examine switching, the rates match older, higher estimates: 20% in Bipolar II, 30% in Bipolar I.Leverich (A more conservative estimate from the same data, using different criteria for mania, yielded rates of 4-15%,Post ) All of the patients in this study were taking a mood stabilizer before the antidepressant was added. Unfortunately, there was no control group receiving a placebo so we don't know what the spontaneous switch rate might have been.

2. Sachs et al, New England Journal of Medicine 2007

In the Sachs New England Journal of Medicine study described above, adding an antidepressant or a placebo to a mood stabilizer for patients with bipolar depression, switching was also deliberately measured. In that study, the switch rate was no greater on antidepressants than on placebo. This is completely at odds with the results from the prospective study by Leverich and colleagues above. Some have speculated that the Sachs study did not find switching because clinicians in that study did not refer patients to participate who are at high risk of switching -- such as those who had a history of switching in the past. Note then a separate analysis of the same population from which this study was drawn, the STEP-BD, did find evidence of antidepressant-induced switching.Truman

3. Truman et al, J Clin Psych 11/2007

The authors of this study point out that for this particular question about switch risk, a randomized trial may not be the best source of information. Instead, asking hundreds of individuals to report on their experience with antidepressants, in a standardized fashion, avoids the potential for selecting patients who are not representative of bipolar disorder in general -- as may have been the case in the Sachs study above. They selected the entire set of patients entering a large, multi-site study of bipolar disorder (STEP-BD) and looked at patients experience on antidepressant medications.

Excluding only patients whose data was not complete, they examined a set of 338 participants. 150 of them (44%) reported a history of at least one manic switch on a standardized intake form for this major study. Two other important findings: patients who had taken multiple antidepressants were 75% more likely to have a history of switching. Secondly, patients who had switched once war much more likely to switch again when given another antidepressant. Interestingly, in this study, no single antidepressant or class of antidepressants was found to produce a more or less switching than any other, except for MAOIs, which were associated with much less switching (which runs counter to a previous study indicating more switching with MAOIs. Boerlin

(An older oft-cited study by Gjisman is discussed below.)

Conclusion: doubt remains about whether the risk of antidepressants is justified in patients with bipolar disorder.

Before we move on, let me show you one of the reasons why I worry about all this. A research group that has published several papers endorsing antidepressants for the treatment of bipolar depression has repeatedly found very little evidence of "switching". In their 2008 study, venlafaxine (Effexor) seemed clearly better than lithium in terms of antidepressant effect. I mentioned this study in the Controversy Zero section above.Amsterdam In that paper, they devote two paragraphs to explaining why the findings below are not significant .



Shown here are individual patient results on the Young Mania Rating Scale (YMRS). The YMRS is a standard measure of manic symptoms. Most research studies regard scores over 12 as evidence of continued manic symptoms warranting further treatment. As you can see, one patient in the lithium group on the right had a brief increase in manic symptoms, in a single week out of the 12. On the left, you can clearly recognize another single patient who in the second week of the study became significantly manic, and discontinued the research study at about week nine, because of this shift.

My concern is this: a research team which regards the curve on the left as insignificant, and indeed devotes two paragraphs to explaining why statistically it is not significant, is trying very hard to ignore the possibility that their antidepressant might have induced this mood shift. Granted, bipolar disorder by its nature includes episodes of mood shift. One cannot know whether the shift shown above was really "caused" by the antidepressant. On the other hand, having scores take off in the second week of treatment -- isn't that precisely what we would be looking for if we were worried about a "switch"?

Further Conclusion: you should look hard for bias in any account of "switching". Psychiatrists seem to be finding and reporting evidence which matches what they believed before the research was done -- including me, as I hope you have noticed. You should be very skeptical about the account you just read here. Obviously I am one of those people who is very concerned about antidepressant-induced switching (and other negative effects, about which you will read in the next section). For me, I think the weight of evidence clearly demonstrates a reason for that concern. We would need numerous studies showing that there is no reason to worry before we let go of this worry. As shown in the graph above, even when a researcher says that his data do not support a concern about switching, with close examination one can find evidence to the contrary -- at least if you are wearing my glasses!



The next section presents my analysis prior to 2007, including a number of relevant literature citations. Most readers should skip ahead to controversy 2 unless highly interested in this issue about switching.

In Sept 2004, Gijsman and colleagues reviewed randomized trials in which patients with bipolar disorder were given antidepressants or placebo. They found an overall "switch rate" of 3%, very low compared to another recent analysis by Drs. Goldberg and Truman, detailed below, which found switch rates varying from 20 to 40%. What's going on here?

First, the Gijsman report was a "meta-analysis". Usually this is a much stronger design than a retrospective literature review, such as that by Goldberg and Truman. But in this case, although the abstract of the Gijsman report says 12 studies were evaluated, on the issue of "switching" only three were analyzable. Suprisingly, the abstract reports the very low switch rate but leaves the impression that this comes from the 12 studies, not just 3. A meta-analysis of only three studies is not very powerful. By comparison, Goldberg and Truman looked at an extensive body of literature. Statistical sophisticates will want to note a recent commentary (full text link) which holds that observational data, such as another recent retrospective review which showed a switch rate of 49%Ghaemi, are more accurate for assessing the frequency of such effects than randomized trials.

For more on the unusual findings and style of reporting in the Gijsman report, here are more details.

The dilemma as to whether to give an antidepressant to someone who might have an undetected, underlying bipolar disorder, is even worse in children, of course. Recently an entire issue of the main Child/Adolescent psych' journal was devoted to this switching issue, all of which is available in full text. See the masterful summary editorial by Gabrielle Carlson, entitled "The Bottom Line". Then see a review of how complicated it can be sorting out rapid cycling from continuous symptoms, with or without antidepressants, by Dimitri Papalos. At the bottom is research led by Melissa DelBello which suggests that at least in the first 1-2 years after exposure to antidepressants, in depressed kids at high risk for bipolar disorder, there is no increase in "switching" into mania -- perhaps even a protective effect!

An emerging view is that the risk of antidepressant-induced switching into hypo/mania varies depending on the patient. Risk factors include hyperthymic temperament Henry; as well as multiple previous antidepressants, female gender, and thyroid abnormalities.Pies For more details on these factors, including conflicting data about the "female gender" factor, see the previously cited review by Drs. Goldberg and Truman. They note that patients in their Cornell Bipolar Cohort who "switched" while taking an antidepressant had had more antidepressant trials per year than those who did not switch. This could be interpreted as one more tiny bit of evidence supporting the idea that antidepressant exposure, in the long-term, might carry some risk in bipolar disorder. For example, it might make a person more "switch prone". That matches my experience with patients. People who have had a lot of antidepressants seem "touchier" in terms of what it takes to destabilize them. Of course, they could have had a "touchier" illness in the first place, and that was why they got all those antidepressants! We'll return to this theme in Controversy 2b below.

In summary on this first controversy: we have to recognize that it will be a long time before we know exactly how much risk antidepressants pose, in the short term, and to which people. However, there is no longer any doubt that there is some risk. Even the FDA now requires doctors to tell patients about the risk of causing agitation and other bipolar-like symptoms (using the word "should"; here's a summary on that) . Now let's look at the possible longer-term risk of antidepressants.

Controversy 2: Mood Destabilization

Quoting from an editorial in the American Journal of Psychiatry, March 2008, by Nassir Ghaemi:

Mood destabilization with antidepressants should be distinguished from an acute manic "switch." Antidepressant-induced mania, or switch, is a short-term phenomenon; one might define it as happening within 2 months of the beginning of antidepressant treatment. Mood destabilization is a long-term phenomenon, reflecting more mood episodes over time than would have occurred by natural history. Antidepressants may cause long-term mood destabilization without a short-term manic switch, and vice versa. Although some agents may have low rates of acute manic switch, especially when used with mood stabilizers, the data from STEP-BD [the world's largest research study on bipolar disorder] suggest that even the new generation of antidepressants can produce long-term mood destabilization.

In that editorial, Dr. Ghaemi also emphasizes an approach I've been espousing for years: if a mood stabilizer is tried with an antidepressant also in use at the same time, and the mood stabilizer "doesn't work", that was an unfair trial of the mood stabilizer. It will need to be tried again later with no antidepressant in the picture at the time.

Want to see more data on this? Dr. Ghaemi refers to two randomized trials, presented below. One is small, and the other one is convincing only for rapid cycling bipolar disorder, but that's what we have to go on.

Study One
From back in 1988Wehr:

Each color represents a different patient's cycle lengths (note that's in days -- we're watching multiple cycles over a period of many, many months here) These folks are not your average person with bipolar disorder. They were at the NIMH for research and were "treatment-resistant" to be sure, highly selected, and further selected for antidepressant-induced cycling by the design of this study. But the important result here is to recognize a clear reduction in cycle length when the patients went onto an antidepressant (a TCA) in addition to their ongoing lithium. Think about it: a reduction in cycle length. We're not talking about switching here, we're talking about changing the course of bipolar disorder, at least while they're on an antidepressant. Not just one switch, but more switches than they would have had were they not on the antidepressant.

Study Two

This one is less obvious. But because antidepressants are so widely used in bipolar disorder, the idea that they can actually make people worse, or at least keep them from getting better, has tremendous implications. In that spirit then, I would like to show you these new data (presented but not yet published; used by permission from Dr. Nassir Ghaemi).

Seventy patients with bipolar disorder who had been on an antidepressant for at least two months were randomly assigned to either continue their antidepressant, or stop it. Here's what happened in the following year (solid line, antidepressants stopped; dotted line, antidepressants continued. :



This graph shows the patient's scores on their "Clinical Monitoring Form" (CMF). This form is used by the doctor to rate the patient's symptoms, either manic or depressed. A zero would mean complete mood stability. Higher scores, like the 3-6 range you see here, mean the patients are having symptoms. You cannot tell whether those symptoms are mania or depression from this graph; it contains both such mood changes. What you can see is that patients in both groups wobble up and down, but their average scores over the year are in the same range -- except for the big initial spike in the patients taken off antidepressants. You might guess that this was an increase in depression scores, and you would be right, as you can see in the next graph.

(I am not certain how quickly these antidepressants were tapered, but I believe one of the authors told me it was over one month or less. A much slower taper rate, closer to four months, might have eliminated that spike entirely. I have used that rate for years now after hearing it recommended by Harvard's Dr. Sachs, and I think it has allowed me to taper patients off antidepressants without seeing an increase in depression very often). UPDATE 2008: I finally found a reference supporting this practice, sort of anyway. If you want it, read this page on Antidepressant Withdrawal.

In the following graph, you see the CMF scores for depression only. Now that early black spike is even more obvious. However, note that both groups show roughly the same results thereafter, over the year. If antidepressants really kept patients from being depressed, you'd expect that the group which stayed on them would look better in this graph: the dotted line would be significantly lower than the solid line, right?



As you can see, the two lines are roughly similar. This is the first of two very important results from this study. Here we see data suggesting that continuing an antidepressant does not make patients experience less depression. In other words, if you pull an antidepressant out fairly quickly, many patients may have an initial increase in their depression scores, but this worsening does not persist. Over the course of a year, that group will be indistinguishable from a group of patients who continued their antidepressant. At minimum, this suggests that long-term antidepressants provide no sustained benefit. But might they present sustained risk?

In this same study, we see results that suggest patients actually do worse if they stay on their antidepressant. I will show you those in a moment. However, caution is needed here: these are preliminary results from a rather small study. Before we draw any firm conclusions, it would be great to have a larger study showing the same results. Unfortunately, we are likely to be waiting a long time. Replicating this study is likely to be very difficult, because it took a huge sample of research patients to set the stage for this small group to be studied like this.

That's a problem, because even though this is the best study we have, it has some serious drawbacks. For example: patients were selected to enter this study knowing that they might be taken off their antidepressant. Those who were completely convinced the antidepressant was helping them were probably reluctant to participate. Their physicians were probably reluctant to refer them to this study. Thus we have a selection bias: patients who entered this study may have been much more likely to be able to do well without an antidepressant than the general group of patients with bipolar disorder. This is a serious problem with this study.

Worse yet, this was not a "blinded" study. Both the investigators and the patients knew who was in which group. They may have had expectations of what would happen. Those expectations could easily have influenced the numbers which were recorded on the Clinical Monitoring Form (CMF).

Finally, although the results you're about to see look impressive, suggesting that antidepressants make outcomes worse, they are not statistically significant. They illustrate a trend. The trend is in exactly the direction one would predict if antidepressants really do cause mood instability. With a larger sample size, if the same results were obtained, they would be statistically significant. Thus the small size of this study is very limiting.

Therefore, the following graph must be interpreted with great caution. I present it because it supports my point of view! To my knowledge there are no data to the contrary from any study of similar design. Again, this is the best we have right now, and as such it is relevant for patients who need a decision right now. Okay, enough warnings. Here are the results, this time looking just at CMF manic symptoms (solid line, antidepressants stopped; dotted line, antidepressants continued):

Although blurry (I scanned this from a poster handout the authors presented recently), you can see now that the dotted line is mapping out a different course than the solid line. (This is not as big a difference as it looks, because the scale has changed: the difference between these two groups is really quite tiny, only one unit on the CMF; in the previous graph, the scale was about three times as big). These data suggest that patients on antidepressants seemed to have more manic symptoms in the following year than those whose antidepressant was discontinued. But this is a weak result, not to be relied upon as the basis for an opinion -- by itself, anyway. Remember, the earlier finding: patients who were maintained on antidepressants didn't do better than those whose antidepressant was discontinued. Actually, even that is overstated: in a separate analysis, the authors found a slight advantage for those patients who stayed on their antidepressants. But the difference was very small, and if the patient had rapid cycling, they were clearly better off to have stopped their antidepressant. A more skeptical reader might find that the two results presented here cancel one another: some very slight evidence of increased risk of mania amongst patients who stayed on their antidepressant, versus a benefit of slightly less depression. Clearly what is needed is another study, a bigger one. But my patients need to know what to do now.

Below you will find other thoughts on this issue, from my analysis prior to 2007. Nothing more definitive, though. Skip ahead to controversy 3 unless you want to see a whole lot of gritty detail on this issue.

Phelps' Assessment Before 2007

Antidepressants may cause an increase in cycling rate and a need for more medication to prevent cycling. This has been referred to as mood destabilization. Concern about causing a "switch" often gets much more emphasis than this longer-term issue, which if true actually represents a greater risk. In my experience, physicians are not asking this question -- sometimes at all. Yet there are studies out there which suggest this is indeed a risk.

In fact, one of the world's experts on this subject believes this very strongly. Dr. Nassir Ghaemi, a Harvard assistant professor and head of the Bipolar Disorder Research Program at Cambridge Hospital, recently guest-edited a review of the risks of antidepressants in bipolar disorder, to which an entire issue of the journal Bipolar Disorder was devoted (Dec. 2003). His co-authors include some of the best-known bipolar experts. He clearly has a "big picture" view spanning years of mood disorder research that can be used to examine this issue. In other words, this guy is not an extremist. He represents a keep-the-risks-down, thoroughly data-based viewpoint.

His review, with Drs. Hsu, Soldani, and F. Goodwin, entitled "Antidepressants in Bipolar Disorder: the case for caution"2003 , is worth reading for anyone who thinks we've all gone a little too anti-antidepressant. Their article was structured as a response to a nearly opposite point of view, from an earlier European essay.Moller A more recent view similar to the Europeans' was presented by Dr. Altshuler in a Point/Counterpoint pair of essays where again Dr. Ghaemi was selected to present a short version of his 2003 cautionary review. Below you'll find a summary table of these contrary views. I hope the summary makes you just itch to see the data they each cite as the basis for the assumptions and conclusions, which obviously differ considerably. If you're able to do so, you will find that the Ghaemi group citations span a broader range of the available literature, both in time and scope.
Moller and Grunze 2000; Altshuler 2004 Ghaemi et al, 2003

The risk of antidepressant-induced cycling is not high.
Antidepressants reduce the risk of suicidality.
Antidepressants are effective in treating bipolar depression.
Mood stabilizers have not been shown to be effective in bipolar depression.



The risk of antidepressant induced cycling is high.
Antidepressants have not been shown to definitively prevent completed suicides and reduce mortality, whereas lithium has.
Antidepressants have not been shown to be more effective than mood stabilizers in acute bipolar depression and have been shown to be less effective than mood stabilizers in preventing depressive relapse in bipolar disorder.
Mood stabilizers, especially lithium and lamotrigine, have been shown to be effective in acute and prophylactic treatment of bipolar depressive episodes.

Conclusions and recommendations: Conclusions and recommendations:

Antidepressant-associated risks are exaggerated.
Antidepressants should be used frequently along with mood stabilizers.
Antidepressant treatment should be continued long-term (ideally 12 months or more) to avoid depressive relapse.




There are significant risks of mania and long-term worsening of illness with antidepressants.
Antidepressants should generally be reserved for severe cases of acute bipolar depression and not routinely used in mild to moderate cases.
Antidepressants should be discontinued after recovery from the depressive episode and maintained only in those who repeatedly relapse soon after antidepressant discontinuation.

In my travels it appears that the Altshuler point of view is more common amongst community psychiatrists -- which matches the data above on the frequency of antidepressant use in bipolar disorder.

This leads to the other major point stressed by Drs. Ghaemi and colleagues: there is no good (randomized, blinded trial) evidence for sustained mood benefit from going on an antidepressant, versus management with mood stabilizers [Update 2007: now we have stronger evidence supporting Ghaemi's stance -- see "Controversy Zero" above]. And remember, the latter strategy will not cause short or long-term destabilization. Here is a slightly paraphrased summary from Ghaemi et al's review of the two existing studies of this issue:

In a large study, the addition of an antidepressant (paroxetine or imipramine) to lithium was not more effective than lithium-plus-placebo -- at least in patients with therapeutic lithium levels (> 0.8 ng/dL).Nemeroff In another smaller (but randomized and blinded) study, the addition of paroxetine to a mood stabilizer (lithium or valproate) was not more effective than the continuation of two mood stabilizers. Although the two-mood-stabilizer group experienced more side effects, because full doses of each stabilizer were used, adding paroxetine instead of an additional mood stabilizer did not produce a better outcome in terms of mood.Young (It should be acknowledged, however, that in the latter study, a true difference could have been missed because of the small sample size).

There is no doubt that antidepressants can treat depressive symptoms in bipolar disorder. The issue is whether they do it better than mood stabilizers alone. And there should be some good evidence to that effect, because we know that they carry more risk of worsening the condition.

In conclusion on this second controversy, in my opinion (matching the recommendations from Ghaemi et al above), it is best to restrict the use of antidepressants as follows:

The patient is already on a mood stabilizer with antidepressant effects (lithium, lamotrigine, perhaps even fish oil; or perhaps Zyprexa, Seroquel, or risperidone -- which may not be "mood stabilizers" in the purest long-term sense, we don't know yet, but all of which have data supporting their antidepressant potential); and
for some reason, adding another medication from the above list is not a good idea; and
she/he is still very depressed, with no evidence of slow improvement; and
her/his depression has no clear features of hypo/mania or instability. In other words, the patient:

is sleeping too much, or at least without repeated interruptions or too little total sleep; and
shows very low energy, not too much, such as agitation or anxiety; and
is "stably depressed", with no evidence for rapid cycling; and
has no clear history of antidepressant-worsening of cycling. And,
there is no mood lability, a strange reaction where tears or anger or laughter are extremely easily provoked, such as by a TV commercial or poor driver or bad joke.

Controversy 2a: Do antidepressants cause rapid cycling? Or rather, is there anyone who thinks they don't?

Rapid cycling is a particular form of "Mood Destabilization". What data we have on this are presented in this section just above, including the new data from Dr. Ghaemi and colleagues showing that patients with rapid cycling did better when their antidepressant was discontinued. On this particular topic, you should be familiar with his study. Go back and make sure you understand the following graph. Here is a link that will take you back to the introduction to that study.

Since we have no data to the contrary, and since this is not regarded as a controversial topic, I think that just about ends this discussion. The good news is that there continued symptoms can be treated not by adding another medication, but by removing one! Of course this will not work for everyone, but it is an obvious strategy to be considered.

Nevertheless, to my surprise, there is a contrary point of view. I have only heard it arise from one source, but in the name of maintaining fair balance, I have kept the links below, from back in the days when this was potentially an issue open for discussion. Skip to Controversy 2b unless you want the gory details.

Recently a "pro/con" presentation on antidepressant controversies included an essay by Dr. Coryell of Iowa. This is perhaps a relevant detail in that Dr. Coryell worked for years with in Iowa with Dr. Winokur, whose had very strong views on antidepressant risk in bipolar disorder. Coryell's position in this debate is that antidepressants do not cause rapid cycling (J Bipolar Disorders Sept 2004, not online).

This view came as a surprise to me, as for years I had thought this particular point was conceded by all mood experts -- not necessarily from a wealth of data, which Dr. Coryell's essay makes clear do not exist, but instead from clinical experience: i.e. if one uses antidepressants, one will likely have repeated experiences of seeing them induce rapid cycling; and clear experiences of seeing rapid cycling resolve when they were tapered off -- and thus there would be less reason to be looking at the literature to see if this had been demonstrated to be "causal".

Granted that this clinical experience is subject to both observer and sample bias, i.e. one can see what one expects to see, and some patient populations may be more likely to have rapid cycling than others. That is why randomized trials are so valuable! But as Dr. Coryell, and his counterpart in this pro/con format, Dr. Ghaemi, both point out: the whole problem on this topic is the lack of randomized trial data, and likewise how hard it would be to actually get such data now for ethical reasons. The two experts end up focusing on the antidepressant-induced shortening of cycle length data presented above, as that is one of the few controlled (on-off comparisons) studies we have to go on at this point.

If you are concerned about the data supporting this widespread clinical impression or wondering just how widespread it is, then the accompanying commentary on this pair of essays by Dr. Akiskal will be of interest as well. His broad perspective and his years of experience watching such debates are evident. Have your librarian hunt up this issue: The Journal of Bipolar Disorders: Reviews and Commentaries; Volume III, Number 2.

"Bottom line": there are some psychiatrists out there who do not believe antidepressants induce rapid cycling, even though the recommendation on treating rapid cycling -- "taper the antidepressant" -- is clearly widespread in the literature. The problem is not so much disbelief in this principle as lack of randomized trial data to support it. Again, Dr. Ghaemi's reviews and citations seem to be much broader than those pitted to dispute him, so for now I find his point of view holding the high ground in these debates.

Controversy 2b: Kindling and Long-term Worsening

Could antidepressants cause "kindling"? The phrase "kindling" is borrowed from neurology, where it has been used to describe forms of epilepsy, which appear to worsen with time. In this model, it is as though each episode of illness makes later episodes both more likely and more severe. It is clear that some patients' bipolar disorder worsens as they get older, with more frequent and more severe episodes. Could this kind of pattern be triggered by antidepressants, at least in some susceptible patients?

Here is good visual example of the phenomenon we're talking about here. The graph shows the mood episodes of a man whose bipolar disorder seemed to clearly worsen with time (his age is shown at the bottom of the timeline; red means hospitalized, up is manic and down is depressed, of course):

Note the pattern: after each episode, the next episodes tend to come sooner and become more severe. This is the "kindling" pattern, though this man's experience alone of course does not prove that the illness itself can do this. There could have been some other factors, such as alcohol or other drugs, etc.

However, suppose some forms of bipolar disorder really do "kindle" themselves. If that is so, then any worsening has the potential to be a "permanent" worsening. What if the patient above had been treated with psychotherapy at age 18, during that first depression? Compare what might have happened if he had been given an antidepressant, triggering a manic episode: might his graph have changed from the course we saw just above (a real patient's experience):

to this (a hypothetical example):

The difference, as you can see, is that this hypothetical patient lost 5 years of symptom free life. And he arrives at a nearly continuous course of illness by age 35, instead of age 40.

This "kindling" concern is very rarely raised in the bipolar literature, at least as regards the risks of antidepressants. One of its early proponents, Dr. Robert Post of the NIMH, is still describing this phenomenon in terms akin to epilepsy;Post but he is not as concerned about antidepressants as some.Post(b) However, the logic still seems compelling to me, as a possible, plausible risk. Here is an example of at least one such case (direct quote from an email, used by permission):

Before my first use of an antidepressant, I had never suffered mania. I had been diagnosed with depression and anxiety, but not bipolar disorder. I was prescribed Lexapro for anxiety (I had never used psychiatric medication before) and used it for five or six days, taking a small dose (half tablet each day). It induced mania so I was hospitalized for a week or so.

Since then, I have steadily had irrational grandiose thoughts. In hindsight, I can see that I had some irrational grandiose thoughts before my Lexapro use, but since my Lexapro use they are far stronger. As far as permanence goes, so far I have not noticed any improvement at all coming simply from time passing (although therapy and other active approaches have been helpful). I had a second manic episode less than a year later (I was not on any medication at the time).

So there's one case. There is also one published example of a patient given steroids for colitis, with a similar course.Pies Here's another widely regarded expert expressing the same concern -- in a different contex, but same resulting worry: quoting a 9/2008 news article...

Kiki Chang, director of the pediatric bipolar-disorders program at Stanford, has embraced the kindling theory. “We are interested in looking at medication not just to treat and prevent future episodes, but also to get in early and — this is the controversial part — to prevent the manic episode,” he told me. “Once you’ve had a manic episode, you’ve already crossed the threshold, you’ve jumped off the bridge: it’s done. The chances that you’re going to have another episode are extremely high.”

Continuing, looking at evidence supporting the kindling concern:

Dr. Rif El-Mallakh, head of the Mood Disorders Clinic at the University of Louisville, Kentucky, published concern about "loss of response" (sometimes called "Prozac poop-out", or more Greek: antidepressant tachyphylaxis) to antidepressants several years ago: "Can long-term antidepressant use be depressogenic?"El-Mallakh Another of his papers on antidepressant-associated worseningEl-Mallakh,b includes an excellent literature review on this entire subject. However, even Dr. El-Mallakh seems to stop short of saying that antidepressants might permanently worsen a person's mood disorder. One of the only published descriptions of this concern I've found, though I've heard it voiced many times (generally by psychiatrists with a lot of clinical experience and willingness to speak out against prevailing beliefs), states directly my concern: namely that antidepressants could make people "treatment-resistant", even when they started out with a good response, and even when they started out looking "unipolar", not bipolar.Sharma 2006

I think I've seen at least one such case in my practice, which was so striking I published the woman's experience as a case report.Phelps This 60-year old woman had done very well on sertaline (Zoloft) for 7 years, for a recurrence of depression she'd had numerous times before. She then developed severe agitation and insomnia in the absence of any factor we could find to account for this change, and these symptoms only stopped after the sertraline was tapered off. But a year later her old depression was back again, so on her own she tried a quarter-dose of the sertraline, and improved dramatically that same day; so she repeated that dose the next day and called me to say how well she was doing. I said that this was nice to hear but would she please come see me tomorrow, by which time, after one more 25 mg dose, she had the same severe agitation for which she had been referred a year earlier, including suicidal thinking. Of course we stopped the sertaline, but it took a month for the agitation to again taper off. We added a very low dose of lithium to the Seroquel she had improved on earlier, and her depression lifted. She ultimately added tap dancing to the activities her excellent therapist had encouraged her to pursue. She remains well, nearly symptom-free, 3 years later (5/2006).

My opinion? Obviously I'm expressing it by including this concern on this page and providing these references, as scant as they are. I fear there may be some people who are indeed "kindled" by antidepressants, so that they experience a form of bipolar disorder they might not have experienced until later in their lives, or perhaps might never have experienced at all. To my mind, this is just a matter of adding two plus two: antidepressants can clearly induce hypomania and/or mania; and manic episodes, at least -- and perhaps hypomania as well -- may in some patients cause later episodes to be more severe and more likely to recur (the "kindling" concept). Therefore shouldn't we worry that antidepressant-induced mania or hypomania could "bump" some people forward along the curve of their illness?

I'm not alone in this concern. Dr. Fred Goodwin, one of the most respected bipolar experts in the world, echoed it in a recent interview (Primary Psychiatry, July 2005):

[Primary care doctors] tend to think of antidepressants as light, easy, uncomplicated drugs, and mood stabilizers as heavy drugs that should be reserved for a last resort. But in fact, recent data suggests that we may have to reverse that order of preference, or at least put them on an equal plane. [emphasis mine]

He then concludes:

When it comes to selecting treatment, the "first do no harm" principle should govern all of medicine... If you create [by giving an antidepressant] a bipolar patient, that patient is likely to have recurrences of bipolar illness even if the offending antidepressant is discontinued.

An Opposing View

A colleague reading this webpage suggested that another view be presented, to give a broader look at these controversies. With appreciation for his passion as well as his intellect, here is Dr. Simon Sobo:

"Although my experience has been that antidepressants can cause rapid cycling, others have had a very different experience and argue, not without merit, that since depression in mania can be very incapacitating, and since, with the exception of lamotrigine and to some extent lithium, the other mood stabilizers don't seem to do much for depression, antidepressants are indicated. Their effectiveness is also controversial, but there is no doubt that some bipolar patients with depression do respond well to them. Moreover, some patients are depressed far more often and longer and more intensely than they are manic. Some bipolar II patients have hypomanic episodes that are not that severe and it never goes beyond that, but their depressions are a big problem. If some patients labeled bipolar II don't really have the disorder, avoiding antidepressants for them may be avoiding a potentially helpful treatment for a serious depression. Even when the diagnosis is clear, if mania has been very infrequent the argument for using antidepressants becomes stronger. (Of course if suddenly mania is becoming frequent, so much for the antidepressants, unless everything else has failed.)"

Dr. Sobo has written extensively on problems of bipolar diagnosis and treatment. For more of his strong skepticism, try his essay on Mood Stabilizers and Mood Swings: In Search of a Definition (see his website, scroll to the title) and the update on that essay.



Controversy 3: Your patient is better (and on a mood stabilizer). Now, should you continue or taper off the antidepressant?
(revised 9/2009)

Three studies address this issue directly -- and they have different conclusions! Make sure you know about the second one, the results of which are a more reliable guidepost by standard criteria for judgement (randomized trials trump naturalistic studies).
•1. Altshuler et al ,Am J Psych 2003
– naturalistic
•2. Ghaemi STEP-BD; 2007 ISBD presentation, not yet published to my knowledge
– randomized

3. Altshuler et al ,J Clin Psych 2009
randomized? no, even though it looks like it

.

1. The "Altshuler" study (lead author, among many contributors, Lori Altshuler)Altshuler

Among psychiatrists who know something about the research on this question, this is the study they have probably heard of. Yet despite how frequently it is cited, the Altshuler report was not a randomized trial; i.e. not a fair comparison between patients who were kept on their antidepressant versus those who were deliberately taken off. (It may appear so from the graph unless one looks closely). Many patients were screened out before arriving at the final results, far more than were followed through the reporting period.

The bottom line: in this study, patients to stay on antidepressants did substantially better than patients who -- for whatever reason -- were taken off their antidepressants. Yet because the patients were not randomly assigned to one or the other approach, we don't know if taking the antidepressant out is what caused them to do less well. They might have done less well because something else was going wrong, and that was the basis for coming off the antidepressant. What we really need is a true randomized trial; and that is what we have, albeit quite small, in the second study below.

If you are really interested in the details, you might wish to see a breakdown of the structure and results of the Altshuler study. It is amazing how often this study is cited, and how firmly, given the way it is structured (a naturalistic follow-up of a select group of patients).

2. The Ghaemi study (lead author for a team of investigators was Nassir Ghaemi)

Data from this study have already been described in detail under question 2a above. The conclusion: for patients with a history of rapid cycling, coming off the antidepressant leads to a better outcome. For those who did not have rapid cycling, the data are equivocal; staying on might be just slightly better, but coming off does not leave people clearly worse. If you have already worked your way through the Ghaemi paper description above, you will recognize the graph below, which supports these conclusions. If you haven't seen the Ghaemi paper, or if the graph doesn't readily make sense as speaking to this third controversy, here's a link back to the introduction to the Ghaemi study.

My opinion, as regards this third controversy, ? Here's what I think: just about every patient with bipolar disorder needs a trial off of antidepressants (with the antidepressant very slowly tapered, to get there). A few patients clearly do better when they are maintained on an antidepressant, almost always in conjunction with a mood stabilizer medication. These patients are a distinct minority, perhaps 10% at best in my practice (about 20% in some bipolar specialty clinics Ghaemi).

3. The "new" Altshuler studyAltshuler

This team of investigators includes some of the most widely respected authors in psychiatry. I don't mean to dismiss their work. At the same time, there is still complete agreement in psychiatry that randomized trials are a better guide than naturalistic studies (the former both an experimental group and a placebo or control group; the latter are based on observations of a group as a whole). So it's important to look carefully at this study because although it is billed as a randomized trial, it is effectively closer to a naturalistic design.

This team of authors put together the comparison trial in which the switch-potential of antidepressants, for patients on mood stabilizers, was compared. That studyLeverich has been described briefly in Controversy 1. After that trial was completed, participants were given the opportunity to continue the treatment to which they had been randomly assigned. That's why the authors call this new report a "randomized trial", but the randomization was not to continuing or stopping antidepressants. It was simply a continuation study, and that is basically very like the first "Altshuler" study described above in this section.

The results: of those who did very well when they first got on an antidepressant in addition to their mood stabilizer ("acute positive response"), 70% continued to do well up to 1 year. The others: 13% relapsed into depression, and 17% had a manic episode (not thought to be related to the antidepressant; the authors note that this "switch" rate is similar to that for patients on mood stabilizers alone. Recall that this trial did not have a placebo group for a direct switch rate comparison).

By comparison, of those who had only a partial improvement when they first got on the antidepressant in addition to their mood stabilizer ("acute partial response"), 6 out of 22 (27%) ended up with a more positive response later.

Bottom line here: that's useful and does help reassure that staying on one's antidepressant, if it works well, is probably okay for up to a year. Beyond that we don't know. The trial does not comment on whether it's better overall to stay on an antidepressant long term or not. The Ghaemi trial (#2 just above in this section) is a direct examination of that question.



IN SUM regarding all three controversies (my opinions):

1. Do not use antidepressants if rapid cycling or severe insomnia/agitation (suggesting bipolar mixed state) is already present.

2. If a patient becomes hypomanic when given an antidepressant, use a mood stabilizer (see guidelines on their use in primary care) as opposed to another antidepressant trial.

3. If in doubt, refer for a diagnostic consultation (if no psychiatrist is available, use a therapist whose diagnostic skills you trust) before proceeding. At minimum, use a screening tool to "rule out bipolar", and/or have the patient read about bipolar II to enlist her/his efforts re: accurate diagnosis and understanding of risk.

4. Almost every patient with bipolar disorder who is taking an antidepressant deserves a trial off of that antidepressant to see if things are more stable (or at least, no worse). When trying this, taper off the antidepressant very slowly: four months, 25% per month, is a good rate.



Here is Dr. Ghaemi's summary, from the introduction to a recent analysis (I haven't linked all the references; see the original article, online).
Recent randomized controlled trials (RCTs) indicate that lamotrigine can reduce long-term risk of depressive recurrences in BPD (12, 13), olanzapine combined with fluoxetine (14), as well as quetiapine (15, 16) are effective short-term in acute BP depression and quetiapine appears to be helpful long-term (16). In addition, lithium has long-term beneficial effects in all phases of BPD (1, 2, 17) and appears to have unique and substantial anti-suicide effects (18, 19). Remarkably, however, antidepressants (ADs) lack specific regulatory approval for use in BP depression, and their clinical value and safety in BPD, particularly with modern agents and type II BPD, remain uncertain and strikingly little studied (20). Several trials of ADs in acute depressive phases of BPD indicate some short-term benefit vs. placebo (21), although two comparisons with mood stabilizers (MSs)-alone did not find superior short-term efficacy (22, 23). Benefits of AD treatment in BPD also may be inferior and less sustained than in non-bipolar major depression (24). Compared with these limited short-term studies, there are even fewer long-term studies of ADs in BPD (25). Nevertheless, in the US currently, they are the most frequently prescribed, and among the longest sustained, of all psychotropic agents for BPD patients of all types (26–28).



Below are some more links on these controversies if you wish. I will try to keep this up to date, adding articles as they appear, on both sides of the discussion.

Important Article Links

For psychiatrists, a technical review of this subject by Dr. Joseph Goldberg, whose research in this area is well-known, can be found in his essay "Antidepressant-Induced Switching to Hypomania, Mixed Mania, or Rapid Cycling Mania", Fall 2006. You'll find that he cites many of the same studies I've noted above, and arrives at nearly the same conclusions.

Another technical review just came out in 2007: Dr. El-Mallakh, whose work is cited above, authored a comprehensive examination of antidepressants in bipolar disorder. see his chapter in Bipolar Depression: a Comprehensive Guide.

Articles/Experts Suggesting Use Caution (updated 3/2008)

In the world's largest series of bipolar patients followed over time, antidepressants were associated with rapid cycling. The authors conclude: "cycling is on a continuum and that prevention of recurrences may require early intervention and restricted use of antidepressants." (Schneck et al, Am J Psych 2008). See the accompanying editorial by Nassir Ghaemi: "Stopping antidepressants thus is the sine qua non of treating rapid-cycling bipolar disorder."

Looking back at records for bipolar patients placed on antidepressants while on mood stabilizers, in a recent review of 20 studies of rapid cyclingKupka , 46% of the patients seemed to have their rapid cycling begin after an antidepressant, though the authors emphasize that a "causal" effect was not established.

Ghaemi N:
See a series of essays by Dr. Ghaemi on this subject, all of which echo the main themes above:
Is Bipolar Disorder Misdiagnosed?
Antidepressants and Bipolar Depression: First, Do No Harm
Antidepressants in Bipolar Disorder: Are They Safe? Are They Effective?
The Bipolar Spectrum: A Valid Forme Fruste?
Mixed states: Broad or Narrow?

Ghaemi SN, Lenox MS, Baldessarini RJ. Effectiveness and safety of long-term antidepressant treatment in bipolar disorder. J Clin Psychiatry 2001 Jul;62(7):565-569 (their review of 7 double-blind studies: conclusion -- increased caution warranted but too few data overall).

Antidepressant-associated mania or psychosis: 10% of all psychiatric admissions in one study

Dr. Gary Sachs at Harvard: a case illustration

Dr. Ron Pies' experience and advice (a regular columnist, now editor, for Psychiatric Times); October 2003 alsoPies

Antidepressant "poop-out" associated with severe, treatment resistant illness:
"METHOD: I describe 15 cases who had a loss of response to repeated trials of antidepressants before developing a chronic and severe, refractory depression. RESULTS: ...Following discontinuation of antidepressants and treatment with mood stabilizers, there was a sustained improvement." J Affect Disord 2001 Apr;64(1):99-106.


Less caution necessary

Randomized comparison of an antidepressant, citalopram, versus lamotrigine in patients with bipolar II: equal improvement in both groups, one patient in 10 in each group had an episode of hypomania.Schaffer et al, 2006

Study suggests no increase in mania for those on antidepressants (insurance data-base mining): Fu et al. (posted 11/2007)

No increase in manic episodes in a large, definitive study of antidepressants versus placebo (Sachs et al, NEJM) -- although a subsequent second look at that same dataset did find a moderate increase in manic symptoms (Goldberg et al, Am J Psych 2007)

Recent Bipolar Network News summary (lead article)
See their analysis of why we see differing observations of antidepressant risk

Young et al: Adding a antidepressant to a mood stabilizer (versus adding another mood stabilizer)

Amsterdam et al have published at least three studies showing patients with bipolar disorder doing well on antidepressants and low rates of worsening (and a nice review of their results and perspective)

Fluoxetine 2004
venlafaxine (Effexor)
Fluoxetine 1998

Update 2006: Now comes another from this group, similar conclusion:
Fluoxetine 2005

In one study (Fluoxetine 2004) the assessment of worsening was the Young Mania Rating Scale (YMRS). If it did not increase more than 7 points, this was scored as "no switch" into hypo/mania. If you look closely at this scale, you'll note that you could get no sleep at all (4 points) and feel extremely irritable (as long as you don't act it out with the rating person; thus 2 points), for a total of 6 points, and this would not classify as hypomania on the YMRS. The test clearly looks more for bipolar I features (e.g. delusions, lack of insight, disorganized thought) than bipolar II features.

In the 2005 study, again we see the YMRS as perhaps the key in understanding the perpective of Amsterdam et al, who state "The mean increase in YMR score in study phase II was slightly higher in the fluoxetine-treated patients ... (P=0.01). However, this difference was not clinically meaningful." In other words, even though they saw a three-point increase in the YMRS in the patients on the antidepressant, with no such increase in the placebo group, they don't think that three points on the YMRS is significant. Okay. It certainly does not declare "manic", that's true. But is it meaningful? I think that depends on how much one fears making someone worse by inducing manic symptoms. I fear this a lot, so 3 points sounds meaningful to me, when the placebo group shows no such shift. I think that Dr. Amsterdam probably just worries less than I do about such shifts, so if the patient is diagnosably more manic than before, he is not inclined to find this concerning.

Moller et al: The European view is also less cautious.

Coryell et al openly question the idea that antidepressants cause rapid cycling. (However, looking closely at their results (namely, Table 2), they actually do find a very small association between rapid cycling and antidepressant use ("The proportion of weeks with SSRI treatment was low in both groups but significantly higher for those with rapid cycling").

2006: Gordon Parker and colleagues in Australia presented 10 cases of patients with Bipolar II treated with an antidepressant alone, and in at least 4, there appears to be a rather striking mood stabilizer effect. That's a paradox we cannot overlook.

For another summary of recent presentations and thinking -- though not a whole lot more data to go on -- see an excellent review on this issue by John McManamy on his well-organized and frequently updated site Depression and Bipolar Web.