|Home | About | Journals | Submit | Contact Us | Français|
An imperfect measure of a modest predictor of response to antidepressants may not be ready for clinical application
Patients' responses to treatment with antidepressants vary greatly. The largest study on the effectiveness of antidepressants to date suggested that roughly one third of patients will recover fully given a long enough trial, one third will improve substantially, and one third will fail to respond.1 A subset in all three groups will have adverse effects such as sexual dysfunction, insomnia, nausea, and weight gain. Side effects are rarely dangerous, but they cause many patients to discontinue treatment, often after a single prescription.2
Similar concerns exist for many drugs, not just antidepressants. What makes antidepressants especially frustrating for clinicians and patients is the lack of factors to predict how individual patients will respond to a given treatment. We know that one third of patients will have minimal improvement, but we have few data to guide selection of alternative treatments. Indeed, while the psychiatric literature abounds with reports of clinical predictors, the findings are rarely replicated.3 One of the only replicated predictors, that atypical depression responds better to monoamine oxidase inhibitors than to tricyclics, does not seem to apply to newer antidepressants. Worse, many of the widely touted and taught predictors turn out to be invalid on closer inspection. For example, conventional wisdom holds that antidepressants with “activating” properties, such as bupropion, are less likely to be effective in patients with anxiety or insomnia. In a clinical trial, however, this was not the case.4 5
Any predictive test, especially one with the shiny lustre of genetics and the promise of rapid and highly accurate results, therefore causes excitement among psychiatric clinicians. The marketing of a test that can define a person's genetic profile in terms of two important P450 enzymes has generated just this sort of excitement. One journal devoted its cover to the headline, “New tool: genotyping makes prescribing safer, more effective!” However, a recent report by the US Agency for Healthcare Research and Quality6 found little evidence to support a role for cytochrome P450 genotyping when prescribing antidepressants.
The argument for P450 genotyping is straightforward. Selective serotonin reuptake inhibitors and other newer antidepressants are metabolised by enzymes in the cytochrome P450 system, so variation in the encoding genes would be expected to influence concentrations of these inhibitors in the blood. In theory, people who metabolise these inhibitors poorly might develop supratherapeutic concentrations and be more likely to have adverse effects; conversely, those who metabolise them rapidly might develop subtherapeutic concentrations and be less likely to respond well to treatment.
As the agency report highlights, both aspects of this argument are suspect. The 11 studies that examined the relation between P450 genotype and antidepressant concentrations found only the suggestion of an association. A key point here is that P450 genotype is just one of many factors that influence drug concentrations. Other factors include variables that can change over time, such as diet, smoking, and cotreatment with other P450 substrates or inhibitors.7 Furthermore, newer generations of antidepressants do not exhibit clear dose-response relations, and concentrations of antidepressants in the blood are rarely informative, except at the extremes. Measuring drug concentrations may be more useful in populations at higher risk; one study of depressed elderly patients did suggest that therapeutic drug monitoring led to changes in treatment about half the time.8
So P450 genotype is not an accurate predictor of drug concentration in the blood, which in turn is not a strong predictor of outcome. Still, even modestly accurate predictors might have benefit. Unfortunately, no adequately powered studies have investigated this question directly. The suggestion that people who metabolise antidepressants rapidly may have a worse response and that those who metabolise them slowly may have more side effects is encouraging but not convincing. Notably, the largest study to look at this question failed to find an effect on adverse effects.9
Therefore, before P450 genotyping can be recommended when prescribing antidepressants, we need to establish that this test can help improve outcomes, either in terms of tolerability or effectiveness. Studies that examine the effects of P450 in populations where the consequences of adverse effects might be greater (such as elderly patients) or suspicion of metabolic differences is high, or in patients who fail to respond to multiple antidepressant trials, would be particularly valuable. In the meantime, what should individual clinicians do? Where concern for drug interactions or toxicity is high, the simplest approach is to begin with antidepressants minimally metabolised by P450 isoforms CYP2D6, CYP3A, and CYP2C19. If this is not practical, the most direct and informative approach is to check concentrations of antidepressants in the blood.
How quickly we forget. In the mid-1980s, the dexamethasone suppression test was widely touted as a sensitive test for identifying major depressive disorder. Only later did it become clear that it is neither sensitive nor specific.10 In a specialty where biomarkers of disease are rare, any biomarker is welcome—provided it is truly clinically useful. Unfortunately, at least as far as antidepressant prescribing is concerned, insufficient evidence is available to support the routine use of P450 genotyping. Ultimately, as with any other diagnostic tool, the value of pharmacogenetic tests needs to be determined by well designed and adequately powered trials before they are used in practice.
Competing interests: RHP has received consulting fees or honorariums from AstraZeneca, Bristol-Myers Squibb, Eli Lilly and Co, GlaxoSmithKline, Pfizer, and Shire Pharmaceuticals.
Provenance and peer review: Commissioned; not externally peer reviewed.