Which antidepressants should be used for the initial treatment of depression?

Andrea: Family physicians regularly encounter patients with depression. With the ever-increasing number of new-generation antidepressants available and their competing claims of tolerability and effectiveness, it's hard to determine which antidepressant to use for initial treatment of major depression. The authors of this multiple-treatments meta-analysis used direct and indirect comparisons of 12 antidepressants to help determine which were effective and tolerable as first-line agents for the treatment of depression in adults.

What does this article say?

Andrea: The authors of this study reviewed 117 randomized controlled trials comparing one antidepressant with another; placebo-controlled trials were excluded. The outcomes analyzed were effectiveness (whether depression scores were reduced by at least 50 percent, or a clinical global impression of “much” or “very much improved” was achieved) and tolerability (defined by all-cause withdrawals from the trial at eight weeks).

Multiple-treatments meta-analyses are not often seen in the literature. This type of analysis allows comparison of treatments directly (e.g., head-to-head trials) or indirectly (e.g., comparing the medications against a first-line treatment; in this case, fluoxetine [Prozac]), thus allowing physicians to determine if one treatment offers benefit over another. Standard meta-analyses only perform pairwise comparisons of treatment A versus treatment B. Often, large numbers of similar treatments are available for a given disease; therefore, conclusions have to be extrapolated from several different meta-analyses to make decisions about effectiveness.¹

I got a little cross-eyed interpreting these data tables. Simply stated, for effectiveness (or response rate), an odds ratio (OR) greater than 1 favored fluoxetine, and for tolerability (or drop-out rate), an OR less than 1 favored fluoxetine. Based on their comparisons, the authors conclude that sertraline (Zoloft) and escitalopram (Lexapro) are more effective (OR = 0.80; 95% confidence interval [CI], 0.69 to 0.93; P < .05 and OR = 0.76; 95% CI, 0.65 to 0.89; P < .05, respectively) and better tolerated (OR = 1.14; 95% CI, 0.96 to 1.36 and OR = 1.19; 95% CI, 0.99 to 1.44, respectively) than fluoxetine. Mirtazapine (Remeron) and venlafaxine (Effexor) perform well in effectiveness (OR = 0.73; 95% CI, 0.60 to 0.88; P < .05 and OR = 0.78; 95% CI, 0.68 to 0.90; P < .05, respectively), but are not as well tolerated as sertraline or escitalopram. Paroxetine (Paxil) and bupropion (Wellbutrin) are similar to fluoxetine in effectiveness and tolerability. Fluvoxamine and duloxetine (Cymbalta) are less effective than fluoxetine. See Table 1 for a comparison of antidepressants available in the United States.²

Should we believe this study?

Andrea: The authors used two outcomes that we care about: effectiveness (how well the medication works) and tolerability (whether the patients actually took the medication). Adverse effects and cost of the medication can affect drop-out rates. The reality is that it doesn't matter why people stop taking their medications—we know that the medications won't work if they aren't ingested! It's notable that 18 to 28 percent of patients stopped these medications within eight weeks of prescription. The drop-out rate is likely higher in our community practices.

Bob: The study does have limitations that are common to nearly all meta-analyses. Pharmaceutical companies sponsored most of the studies, and many of the studies were of small size. Unfortunately, it is unlikely that we will have a similar head-to-head, non-industry–sponsored study of substantial size that will allow us to do these comparisons directly.

Mark: A meta-analysis can help us overcome small size by increasing the number of patients receiving any given treatment; thus, improving the strength of the conclusion. Nearly 25,000 patients participated in these trials. The mean duration of the studies was only six weeks, but that is within the window that we typically expect patients to see some effect (positive and negative) from these medications.

Andrea: Meta-analyses have an insidious downside, however. They are only as good as the studies included (the “garbage in, garbage out” rule). So you shouldn't always trust the results of a meta-analysis. The validity of the results is highly influenced by the quality of the studies included, as well as by how carefully a search was done for studies to include in the meta-analysis. These authors collected studies from all available sources, including unpublished studies from pharmaceutical manufacturers, and all studies were conducted in a similar manner. Those that did not meet quality criteria were excluded from the final meta-analysis. It is reasonable to believe that the data represent the best available for prescribing decisions in the treatment of depression.

What should the family physician do?

Bob: This study gives me a reasonable starting point for choosing antidepressants for my patients. I would likely start with sertraline over escitalopram, because it is one fourth the cost.³ Fluvoxamine and duloxetine will remain low on my list of first-line antidepressants.

Andrea: I'm more likely to use sertraline as a first-line agent based on this information. I think it's a toss-up between citalopram (Celexa) and escitalopram based on the available data, so the cheapest medication will win in that competition. I'll certainly consider mirtazapine for my patients with depression who are also having difficulty sleeping.

Mark: Based on this study and the 2006 STAR⋆D trial, it appears that bupropion may be better as an augmenting agent than as a primary treatment for depression. In the STAR⋆D trial, some patients who did not respond to initial treatment with citalopram showed improvement in their depression scores with the addition of bupropion to their regimen.⁴ Interestingly, citalopram and bupropion had similar effectiveness in this multiple-treatments meta-analysis; citalopram was better tolerated than bupropion.

Andrea: I'd like to see more multiple-treatments meta-analyses for other medications that we use often, including statins, beta blockers, diuretics, and benign prostatic hyperplasia medications. Physicians and their patients are often tasked with choosing among multiple treatment options for any given condition. A multiple-treatments meta-analysis will allow us to be better informed about the options that are available and will allow development of treatment guidelines that reflect the best available evidence.