After a psychiatric drug is approved by the Food and Drug Administration (FDA), a marketing juggernaut often follows, trying to convince us that the newly approved drug offers substantial benefits for treating a mental disorder.
But how does the FDA determine whether to approve a drug? What follows is a breakdown of the process.
Phases of Clinical Trials
When determining whether to approve a psychiatric drug, the FDA evaluates evidence from a series of trials, conducted in “phases.” In a phase I trial, a drug is tested in normal volunteers and/or people with a relevant psychiatric diagnosis to examine potential adverse events and how the drug is absorbed, distributed, excreted, and metabolized.
Phase II studies look for initial evidence of efficacy. These are typically small and are frequently single-blind (ie, the researchers know what’s being given but the subjects do not) or open-label (ie, everyone knows what’s being given and the goals of the study). These studies should also provide an idea of the appropriate dose range.
If a drug progresses to phase III, it typically undergoes a series of doubleblind, placebo-controlled trials. These may also involve an active treatment comparison, such as an already approved medication for treating the condition, although most psychiatric trials still compare the new drug to placebo.
Upon completion of phase III trials—or sometimes as the data are being gathered—the FDA may convene an “advisory panel” of external experts who interpret and provide advice upon on the relevant evidence regarding a drug’s efficacy and safety. Typically, the FDA heeds the advice of such panels.
Phase III trials are key to determining FDA approval. Two positive studies that are “adequate and well-controlled,” as defined by the FDA, are needed to demonstrate acceptable efficacy for approval. A positive study shows a statistically significant advantage for the new drug over placebo on the primary outcome
measure, whereas a negative study finds no such benefit on the primary outcome.
This comes with a catch. Because only two positive studies are required—seemingly irrespective of the number of negative studies—pharmaceutical companies often conduct a slew of studies to secure regulatory approval. In fact, for several antidepressants, including citalopram (Celexa), sertraline (Zoloft), and bupropion SR (Wellbutrin), the number of negative trials actually outnumbered the number of positive trials. This has led one FDA historian to note that the FDA efficacy requirements can be placed anywhere between a “scintilla and a preponderance.”
Also, while a statistically significant benefit in multiple studies is generally required, keep in mind that a statistically significant effect (the drug’s effect is greater than zero) is not necessarily the same as a clinically significant effect (drug’s effect is substantial). (See accompanying articles in this issue for more on these distinctions.)
In psychiatric drug trials, the primary outcome is nearly always a clinician-rated measure of symptom severity. Some of the most common include the Hamilton Rating Scale for Depression (HAM-D) and the Positive and Negative Syndrome Scale (PANSS) for schizophrenia. Patient-rated scales, if they’re even included in the trial at all, are generally not considered important in the approval process.
Broader constructs like the patient’s everyday functioning or quality of life are rarely considered as critical outcomes. Consider the three FDA-approved atypical antipsychotics for treatment-resistant
depression: aripiprazole (Abilify), olanzapine/fluoxetine (Symbyax), and quetiapine (Seroquel XR). On a quality-of-life measure in placebo-controlled trials, aripiprazole yielded only a very small benefit, which was not statistically significant when patients who violated study protocol were removed from the final analysis. Further, there was only a marginal benefit when patients rated their own subjective response to the drug.
No self-report depression scales were administered in trials of quetiapine or olanzapine/fluoxetine, and subjects showed no overall benefit on quality-of-life measures with these drugs. Thus, even though three drugs gave rise to statistically significant (though modest) benefit on clinician-rated measures, these benefits did not carry over to important outcomes relevant to the patient (Spielmans GI et al, PLoS Med 2013;10:e1001403).
Assessing the safety of new medications is another challenge altogether. Some adverse events or side effects, such as weight gain, can be observed easily and reliably, but many adverse events are more difficult to measure. In most clinical trials, participants answer nonspecific open-ended questions regarding the experience of side effects, rather than completing a structured checklist of particular adverse events. This might underestimate
Consider the case of SSRI-related sexual dysfunction. The controlled trials upon which the FDA approved these medications found low rates of sexual side effects: only 5% to 10% (or fewer) of all subjects.
Subsequent trials, which systematically assessed the potential onset of sexual adverse events and which included only patients without prior sexual dysfunction, found that over half of participants taking SSRIs reported at least one such event (Serretti A et al, J Clin Psychopharmacol 2009;29:259–266).
This problem is now acknowledged on the drug label for all SSRIs, which says that “…estimates of the incidence of untoward sexual experience and performance cited in product labeling are likely to underestimate their actual incidence.”
Other adverse events may also be similarly underreported. One study found much higher rates of adverse events via a structured checklist completed by patients, as opposed to the treating psychiatrist’s observations (Zimmerman M et al, J Clin Psychiatry 2010;71(4):484–490). The psychiatrists in this particular study used open-ended questioning to assess for potential adverse events—quite similar to the methods used in psychopharmacology clinical trials. Even though serious adverse events are probably reported a bit more accurately in trials, as participants are likely more motivated to report them, the rates of adverse events on drug labels should be considered as low estimates.
Recommendations for Change
To ensure that the FDA approves only those drugs with a high likelihood of benefit and low potential for harm, the following changes may be beneficial:
- Negative trials should be more strongly considered when the FDA assesses drug efficacy. The current system requires two positive studies from a pool of several other trials that may be negative.
- Measures of quality of life or daily functioning should be taken more seriously in the drug review process. If a drug does not improve one’s overall quality of life, then the drug may not be worthy of approval.
- Self-reports of patient symptoms should be considered. Though clinician-rated measures are useful, the information provided directly by subjects in a clinical trial may be more relevant to a patient’s experience with a drug.
- The clinical significance of drug benefits should be considered. A drug that offers a statistically significant, but very small, benefit may not be worthy of approval, particularly if its adverse event profile is problematic.
- Adverse events should be assessed systematically using structured checklists. This would help to provide a more accurate picture of drug side effects.
For more on the FDA’s approval process for psychiatric drugs, see Spielmans GI and Kirsch I, Annu Rev Clin Psychol 2014;10:741–766.
TCPR’s Verdict: Federal law as enforced by the FDA does not set a particularly high bar for psychiatric medicines to win approval. At least a minimal degree of efficacy compared to placebo and evidence of safety are required. Improvements to the process may result in better and safer medications but fewer new drugs may meet more stringent standards.