Sunday, November 22, 2015

NEJM Review of Generalized Anxiety Disorder

There was a review of Generalized Anxiety Disorder (GAD) in this week's New England Journal of Medicine by Stein and Sareen (1).  I just did a bit of a critical review of the concept here and thought I would look at what these authors had to say.  

They start the review with a clinical vignette of a 46 year old married woman with insomnia, headaches, back pain, and excessive worry about a number of daily stressors.  She is also drinking alcohol on a daily basis to "self-medicate".  She is described as a person who comes in frequently for appointments.  After reviewing the phenomenology,  comorbidity, and differential diagnosis - the authors come back to this case and apply what is in the review.

Their review of the diagnosis does highlight a few things that are problematic about the diagnosis.  The key diagnostic feature is chronic excessive worry.  The worry has to be there for at least 6 months.  In their review of other psychiatric causes of anxiety they omit diagnoses that can cause short term worry or anxiety - the adjustment disorders.  They point out that GAD is more common in primary care clinics where it usually presents with a chief compliant of somatic problems rather than excessive worry.  They discuss major depression as a common co-occuring condition and suggest that anhedonia may be a distinguishing symptom for depression.  They also describe anxious depression as episodic depression superimposed on chronic anxiety.  There is no mention of the low diagnostic reliability of the disorder and why that might occur.  I think that any psychiatrist who sees GAD over time experiences the same problem that occurred in the DSM-5 field trials, the diagnosis can seem to change between visits from GAD to major depression, even in the absence of any new stressful life events.  Critics of psychiatry frequently cite this as a problem with DSM-5.  I think that DSM-5 does a good job with the symptom descriptors, but we don't know why this change occurs and I have not heard anyone talk about it like it is a real phenomenon.

Alcohol use is described as a common co-morbidity with 35% of people with GAD "self-medicating."  I put that term in quotes because it suggests that alcohol can actually be used for the purpose of medication.  What really occurs is that over time the person becomes more anxious and sleep deprived because of the negative effects of alcohol on sleep, baseline anxiety, and baseline mood.  Practically everyone I talk with who has an alcohol use disorder can recognize this pattern and modify any remarks about self-medication to "feel better for a few hours" or "knock myself out and forget about my problems".  There is also the issue of alcohol use being the cause of an anxiety disorder rather than temporary relief.  While I am on the topic of substance use and GAD, at one point the authors make the statement: "Data are also lacking on the use, usefulness, and safety of medicinal marijuana for generalized anxiety disorder" (p. 2066).  Many if not most anxious people are averse to the use of marijuana for anxiety.  Initial use of marijuana typically causes a drop in blood pressure with a compensatory tachycardia.  Tachycardia especially if there is a noticeable accentuation of heart beats is not tolerated well by patients with anxiety.  Many have had panic attacks.  Others have cardiac awareness and are sensitive to any changes in heart rate or intensity.  Many people tell me they thought that marijuana was effective for anxiety, but over time it seemed to make them more and more anxious, they developed panic attacks, and they had to stop using it.  These features combined with a tendency of patients to stop talking to their primary care physicians about substance use are good reasons to heavily educate them about these problems at the earliest possible time.

The authors take a risk factor analysis approach to looking at historical features that can also be associated with the diagnosis.  They point out that they are nonspecific and amy be associated with other psychiatric diagnoses.  I would encourage a more developmental approach, looking back at the first recollection of anxiety - usually at some point in childhood and how that developed in the childhood environment.  It is fairly common for the patient to describe one or both parents being anxious and how that was transmitted to them  eg. ) "I started to worry about the same things my  mother worried about" or "I started to worry about my mother because she was worried all of the time - I worried that something was going to happen to her."  Those learning patterns associated with adult anxiety are fairly common and may explain the low heritability (15-20%) of the disorder.  The authors do discuss one feature that is important in this context and that is intolerance of uncertainty.  Clinically that translates to excessive and at times catastrophic worry about uncertain situations.  They are unsure about the biological or experiential origins of the symptom.  I think the important part is that with a careful enough history and sometimes collateral information the learning aspects of this bias can be examined and it can be unlearned in therapy.

The authors advocate for a stepped approach to treatment and I certainly agree.  This approach would include an initial medical assessment to look for common medical conditions that can cause anxiety followed by education about anxiety and lifestyle changes to address sleep, exercise, caffeine intake and alcohol use with monitoring response to those interventions.  Those first two phases could be accomplished at the initial visit.  If those initial interventions don't help moving on to "low intensity psychological interventions" like self-help books, computer-assisted psychotherapy, and support groups.  The next step up is more intensive psychological interventions like individualized cognitive behavioral therapy (CBT) or pharmacological management based on the patient's preference.  The highest level of care would include pharmacotherapy and more intensive CBT alone or in combination with other therapies (psychodynamic or acceptance and commitment therapy (ACT)).  The practical issue with this 4 step algorithmic approach to care is that it is generally not available in primary care settings.  In many of those settings, the patient is screened with the Generalized Anxiety Disorder 7-item questionnaire (GAD-7) and the patient is treated with a medication.  This is viewed as "cost-effective" care by managed care systems because an inexpensive prescription and a 20 minute appointment with a physician is apparently much more "cost effective" to the organization than maintaining computerized psychotherapy or educational and monitoring systems.  There is also the largely undetermined effect of the patient taking a completely passive role in their care.  There is a significant difference between a patient who is actively engaged in lifestyle changes and self education and one who expects a complete cure from a pill.  The actively participating patient has better outcomes.   

The authors include a table of 16 medications used to treat GAD.  They point out that the effects of medication are modest at best and no single medication has better efficacy.  They discuss vilazodone as a promising medication in clinical trials and do not include it in the list.  My current prescribing information says that it is FDA approved only for major depression, but only 4 of the 16 drugs on the list are approved for GAD: paroxetine, venlafaxine XR, duloxetine, and buspirone.   The authors comment on the practice of using hydroxyzine for GAD and suggest not to use it.  I am in complete agreement with that recommendation and think that any anti-anxiety effect comes from the non-specific sedating effect of antihistamines.  The side effect profile is also not very favorable.  They point out the benzodiazepine paradox with GAD - they are recommended for short term (3-6 month) use but the condition is chronic.  There is even more subtlety there.  Some early studies of GAD treated with antidepressants suggests that patients needed to take the medication only 30% of the time over ten years of treatment.  I don't think you will see a similar study with benzodiazepines and I think it has to do with the behavioral pharmacology of the drug.  The single-most important issue when it comes to benzodiazepines is the informed consent and letting the patient know that they are taking a potentially addictive drug.  

The  authors are silent about the fact that GAD may be the most heterogenous of all of the DSM-5 categories.  In October and November of this year, I went to three excellent conferences.  One of the central themes was phenotypic diversity in DSM-5 categories and what it implies for biology and genetics.  GAD seems to offer some of the best clinical features for distinguishing intermediate phenotypes and I outline a few in my previous post.  There are problems with a diagnostic category that says "excessive worry" is a discriminating feature and ignores real physiological markers like persistent tachycardia, hypertension, body mass index, and hyperarousal at the time of sleep.   This also points out how basic science can drive clinical diagnoses in psychiatry and hopefully at some point in the near future we will see this kind of research.
I think that we have gotten as much as we can out of the GAD diagnosis at this point and it is time to break it down into what can be more reliably observed. 

George Dawson, MD, DFAPA


1: Stein MB, Sareen J. Generalized Anxiety Disorder. N Engl J Med. 2015 Nov 19;373(21):2059-68. doi: 10.1056/NEJMcp1502514. PubMed PMID: 26580998.


  1. Excellent post and astute observation about the familial/development history often seen in this illness.

  2. Hello,

    Thanks for the informative review. I suffer with GAD NOS and refractory MDD and am interested in a point you mentioned in this commentary.

    "They point out the benzodiazepine paradox with GAD - they are recommended for short term (3-6 month) use but the condition is chronic. There is even more subtlety there. Some early studies of GAD treated with antidepressants suggests that patients needed to take the medication only 30% of the time over ten years of treatment. I don't think you will see a similar study with benzodiazepines and I think it has to do with the behavioral pharmacology of the drug."

    As part of my medical care for the aforementioned Dx I have been treated for years with low dose Klonopin PRN and recognize it's highly addictive potential and hence use it very sparingly under my doctor supervision. I do not think any of my medications are terribly effective for all the money I spend on them however the cheap benzos are a fail safe anytime I feel that I am in a crisis and they are most effective only when used occasionally for short duration, exactly in the fashion I am doing.

    What are you referring to in the behavioral pharmacology of the benzos that do you not foresee long term intermittent studies similar to those done with SSRI's?


    1. Thanks for your reply. I was referring to the addictive potential of the drug. Any drug that reinforces its own use and in the process creates a state where the person needs to take it in order to not get rebound anxiety or insomnia is more problematic to study on a long term basis. We do not know the true portion of the population at risk, but I think most experts would consider it to be substantial. We are just now getting to the point where epidemiological data is being collected that might give us a better handle on the scope of the problem.