There is a strange disconnect between what we psychiatrists do in daily practice and what official antidepressant (AD) treatment guidelines recommend. Treatment guidelines typically say, essentially, that all antidepressants are equal in efficacy, but real psychiatrists have strong personal preferences, based on some combination of the scientific literature, the advice of experts, our clinical experience, and perhaps even the personalities of the last drug reps we saw in the office. In this article we review a range of evidence to come up with suggestions for which antidepressants to start with, as well as adding bonus material on how to start meds that many of us may have little experience with, namely the tricyclics and the MAOIs.

Are certain AD medications more effective than others?

Were there a clear winner here, all psychiatrists would already know of it; indeed, many have sought to tease out even the slightest edge that one agent might have among the many that are available. For a time, venlafaxine was thought to have a significant advantage, but estimates of this advantage have narrowed over time. A recent estimate of the NNT (Number Needed to Treat) to see a benefit with venlafaxine over SSRIs is 17, meaning that you have to treat 17 patients with venlafaxine XR to find one additional patient who would not have responded to an SSRI. Generally, any NNT above 10 is considered to be clinically insignificant. (Nemeroff C, BiolPsychiatry 2008 Feb 15;63(4):424-34. Epub 2007 Sep 24; see also TCPR Jan 2007 for a discussion with Michael Thase on this topic).

The search for an advantage continues, however. The latest paper to challenge the idea that all antidepressants are created equal is the “multiple treatments meta-analysis” by Cipriani et al., which compares the results of 117 placebocontrolled, randomized trials. They concluded that venlafaxine, mirtazapine, sertraline and escitalopram were slightly better than the eight other new-generation ADM examined. Among these escitalopram and sertraline had the best tolerability, while sertraline was the most economical (Cipriani A, Lancet 2009;373: 746-758). However, the methodology of this paper is in dispute and further work will be needed before an unequivocal winner can be declared (see the interview with Erick Turner in this month’s issue).

Which antidepressant should you start with?

The right medication is a matter of judgment, and will vary with the patient. Here are TCPR’s common sense guidelines.

1. For a patient with uncomplicated major depression and no comorbid anxiety disorder, one could argue that generic bupropion SR (Wellbutrin SR) should be considered first. Bupropion remains as effective for depression as SSRIs and it almost never causes the two most common SSRI side effects: sexual dysfunction and fatigue/apathy.

2. Depression with comorbid anxiety disorder. Choose an SSRI over bupropion. Which SSRI? We suggest sertraline, for the following reasons: like paroxetine, it has indications for a range of anxiety disorders, but unlike paroxetine it does not inhibit cytochrome liver enzymes, and it is less likely to cause sedation, weight gain, sexual dysfunction, or discontinuation side effects. In addition, Paxil is the SSRI with the poorest safety data in pregnancy (pregnancy category D).

3. Major depression comorbid with pain. The only antidepressant with a pain syndrome indication is duloxetine (Cymbalta), so many practitioners use this as a first line drug for patients with depression plus either Fibromyalgia or diabetic neuropathic pain. However, don’t get fooled into believing that duloxetine is the go-to antidepressant for all patients with any neuropathic pain. A recent Cochrane review that found tricyclics and venlafaxine (Effexor) were very effective for neuropathic pain of any cause (NNT ≈3), while data for SSRIs was too scant to evaluate rigorously (Saarto T and Wiffen PJ,Cochrane Database Syst Rev 2007;(4): CD005454). Though duloxetine has been marketed heavily for its pain indications a recent post-hoc analysis of three trials of duloxetine for diabetic neuropathic pain showed a NNT of 5.2—a respectable result, but apparently less effective than tricyclics or venlafaxine (Kajdasz DK et al., Clin Ther 2007;29 Suppl:2536-2546).

For depressed patients who suffer with migraine or tension headaches, the first choice is a tricyclic (Koch HJ et al., Drugs 2009;69:1-19). Amitriptyline (AMI) has the longest track record and the best data from well-conducted trials, though it can be very difficult to tolerate in antidepressant doses. Nortriptyline (NT) is better tolerated, although it has not been extensively evaluated for headache treatment. To use nortriptyline, start with 25- 50 mg at bedtime and gradually titrate up to the usual effective dose of 75-150 mg per day. The utility of getting blood levels is debatable, but is reasonable if the patient is taking a medication that interacts with NT, or if the patient has a history of cardiac problems. The usual recommended NT blood level is 50-150 ng/L. For amitriptyline, you can use the same initial dosing as you would with NT (25-50 mg at bedtime), but the usual effective dose is higher, typically in the 150-250 mg per day range. If you like checking serum levels, shoot for a total level of less than 300 ng/L of AMI + NT. Because of the risk of TCAs interfering with cardiac conduction, some authorities recommend checking an EKG before treatment in any patient over 40.

Finally, if you don’t want your patient to have to deal with tricyclic side effects, you can try venlafaxine, which has some positive data both for the treatment of headaches and for the vague somatic pain accompanying depression (Koch HJ et al., Drugs 2009;69:1-19).

4. Depression in an underweight patient with insomnia. Our first choice here is mirtazapine (Remeron), with paroxetine a close second. Mirtazapine has strong antihistamine properties, and therefore causes both sedation and increased appetite at low doses. Start at 7.5-15 mg at bedtime. At higher doses (often needed to fully treat depression) there is often less sedation because norepinephrine uptake inhibition kicks into higher gear. For paroxetine, start at 10- 20 mg per day and titrate gradually upwards if needed.

5. “Atypical” depression symptoms. Although we often think of MAOIs as being particularly effective for atypical depression (depression with increased appetite, increased sleep, leaden paralysis, and rejection sensitivity), a recent meta-analysis (Henkel et al., Psychiatry Res 2006;89-101) found that MAOIs are no more effective than SSRIs for such symptoms (MAOIs are more effective than tricyclics for atypical symptoms, however). If you do choose an MAOI, we prefer tranylcypromine (Parnate) because it tends to cause less weight gain and sedation than the other MAOIs. Start tranylcypromine at 10 mg BID, keeping the doses in the morning and early afternoon to avoid insomnia. Gradually increase up to 30 mg BID as needed. For details on drug interactions and food restrictions, see the November 2006 issue of TCPR.

6. Depression comorbid with substance abuse. Use bupropion if the patient wants to quit smoking. A meta analysis showed that the average oneyear quit rate on bupropion was 20% vs. 10% on placebo (Eisenberg MJ et al., CMAJ 2008;179:135-144). Not dazzling, but we’ll take what we can get. For patients addicted to alcohol or illicit drugs, there is no evidence-based guideline for which AD to choose.

7. Depression and osteoporosis or GI bleeding. Try to avoid SSRIs and SNRIs in patients with these problems, as any medications that block serotonin reuptake can also contribute to osteoporosis and risk of bleeding. Medications such as tricyclics or buproprion are a safer bet (see Haney EM et al., Arch Intern Med 2007;167:1246-51, also see Diem SJ et al., Arch Intern Med 2007;167:1240-5).

8. Should we consider a family history of success with an AD? As a crude “pharmacogenetic test,” many clinicians use a family history of response to a given AD to guide treatment selection. This is not a new idea; retrospective research done in the 1960s and 1970s found that if patients’ first-degree relatives had good results with either an MAOI or tricyclic, the patient was far more likely to respond to that class of medication (Pare CM et al., J Med Genet 1971;8:306-309). Unfortunately, few studies have examined the predictive value of family response for the newer ADs, though one study did find that fluvoxamine response tended to cluster in families at a higher rate than predicted by chance (Franchini L et al., J Psychiatr
Res 1998;32:255-259). Bottom line is that while we have little solid evidence to go on, choosing an AD based on first-degree family history of a response is reasonable.

9. Avoiding drug-drug interactions. The cleanest ADs in terms of drug-drug interactions are (in alphabetical order) citalopram, escitalopram, and sertraline.