Moreover, they were significantly less likely to achieve clinical responses or remissions compared with the 22.2% who did meet such criteria, even after adjusting for baseline differences in such factors as duration of illness, presence of anxiety symptoms, history of suicide attempts, and demographic and socioeconomic variables.
"Phase III trials do not recruit representative treatment-seeking depressed patients," the researchers concluded.
They recommended that pivotal drug trials use broader inclusion criteria, which would increase their generalizability to real-world practice.
The strategy would reduce the number of failed trials, they argued, by cutting down responses and remissions with placebo.
But the researchers also acknowledged that such a strategy would probably increase rates of adverse events, insofar as safety profiles of investigational drugs are inherently not fully known.
The National Institute of Mental Health funded STAR*D -- Sequenced Treatment Alternatives to Relieve Depression -- precisely to examine the efficacy of various antidepressants in patients most likely to receive them in clinical practice.
Its inclusion criteria were relatively loose: a DSM-IV diagnosis of single or recurrent nonpsychotic major depressive disorder and a minimum depressive symptom score for moderate severity (>14) on the 17-item Hamilton Rating Scale for Depression.
In contrast, said Dr. Wisniewski and colleagues, most company-sponsored phase III trials typically bar patients with Hamilton scores lower than 19, more than one concurrent general medical condition or axis I psychiatric disorder in addition to depression, or current episodes lasting more than two years. Those with obsessive-compulsive disorder are also usually excluded.
The researchers stratified the 2,855 STAR*D patients according to these more restrictive criteria. The 635 who met the typical criteria for entry into phase III clinical trials were called the "efficacy sample," the remaining 2,220 were the "nonefficacy sample."
Dr. Wisniewski and colleagues found important differences in responses to initial treatment with citalopram (Celexa) in the trial:
- Response rates (defined as at least 50% reduction in Hamilton score): 51.6% in the efficacy sample, 39.1% in the nonefficacy sample (P<0.0001)
- Remission rate (Hamilton score of 5 or less): 34.4% in the efficacy sample, 24.7% in the nonefficacy sample (P<0.0001)
After adjusting for baseline factors, the differences were changed only slightly.
The adjusted odds ratio for clinical responses in the efficacy versus nonefficacy samples was 1.37 (95% CI 1.12 to 1.68).
For remission, it was 1.33 (95% CI 1.07 to 1.65).
Psychiatric hospitalization was required in 2.5% of the nonefficacy group versus 0.3% of the efficacy sample (P<0.001).
Severely or intolerably intense side effects were also significantly less common in the efficacy group, despite similar levels of drug exposure.
"Given these between-group differences, the smaller efficacy sample is clearly not representative of the more inclusive, treatment-seeking population," the researchers said.
Dr. Wisniewski and colleagues noted that previous studies had also found that a majority of depressed patients would be excluded from registration trials of antidepressants.
However, the researchers said, those trials examined only baseline characteristics. They said theirs was "the first to examine the differences in treatment outcome."
Their findings that outcomes are generally better with tight inclusion criteria mean that registration trials likely paint a rosier picture than should be expected in ordinary practice.
"The duration of adequate treatment suggested by data from efficacy trials may be too short," they added.
Dr. Wisniewski and colleagues acknowledged several limitations to their analysis: STAR*D relied on participants' self-reports of comorbid somatic and psychiatric conditions. Also, the current analysis used STAR*D data from a single drug that was given open-label.
Moreover, they said, although the investigators have previously helped design antidepressant registration trials, the inclusion-exclusion criteria they identified as typical for such trials may differ from those of actual individual studies.
No comments:
Post a Comment