(Page 3 of 4)
turned away by SAFER interviews. Placebo response rates in those studies dropped as well, ranging from 13 percent to 27 percent, which is below the typical 30 to 40 percent in antidepressant studies.
It’s not just unconscious bias that warps study recruitment, according to the paper’s authors. Recruiters often face “financial and systemic pressures” to enroll patients as fast as possible, the authors write. “These burdens… may influence the suitability and appropriateness of enrolled study participants.”
What’s more, some participants—“professional patients”—are motivated by cash and other gains. “Therefore, quality assurance measures are considered a necessary element to enhance the precision of patient selection” for depression studies, they write.
In explaining the success of the Sage Therapeutics (NASDAQ: SAGE) treatment brexanolone (Zulresso) for postpartum depression, which won FDA approval earlier this year, chief medical officer Steve Kanes underscores the need to monitor study sites, “train and retrain” the staff, and make sure to enroll “bona fide” patients who meet the trial criteria.
Sage didn’t use a complicated trial design. In fact, it started with open-label studies—no hiding from the patients what they’re receiving, and no placebo—and advanced into placebo-controlled trials only after seeing a strong response. “When you have drugs with large effect sizes early on, it’s relatively straightforward to demonstrate those effects in subsequent trials,” says Kanes.
While that worked in postpartum depression, Zulresso is an outlier that requires a 60-hour infusion under medical supervision. Those are difficult logistics for a very specific patient population. (Zulresso also failed a Phase 3 placebo-controlled trial in a rare, severe form of epilepsy.)
A broader test of Sage’s “strong-response” philosophy will come with its first pill, SAGE-217, in postpartum depression and major depressive disorder, the most common form of depression. Sage could ask the FDA for approval pending the topline results by year’s end from one of the Phase 3 studies.
While every psychiatric drug trial stresses quality control measures, but with the rise in placebo response over the years there’s obviously room for improvement. Could high-tech tools come into play? Eddie Martucci, whose firm Akili Interactive is developing digital (not pharmaceutical) therapies to treat behavioral disorders, says the field needs “functional measurements,” not just questions about how you’re feeling. (Your optometrist doesn’t ask “Are you seeing better?” says Martucci. She gives you eye tests.)
They’re out there under development. The question is whether they can help boost current practices—adding another guardian at the gate, so to speak. Or can they go farther, as some in the field would like, and reshape the way psychiatric drugs are conceived and developed?
Bananas and Clues
In his office, which he shares with other health-related startups in Boston’s busy Back Bay neighborhood, Jim Harper opens an app on his mobile phone and says the word “banana” several times.
Then he taps the app again, pinches his nose to mimic bad congestion, and repeats it again.
Bananas are funny, from slippery peels to knock-knock jokes, but Harper’s amusing repetition has a purpose. The word’s N sound, obviously altered by holding the nose shut, helps demonstrate Sonde’s voice analysis technology. Using artificial intelligence, the firm wants to detect clues from subtle, involuntary vocal changes—not just if you have a cold, but early signs of mental illness like depression that are perhaps undetectable to, well, the naked ear.
Sonde isn’t there yet. It has collected speech samples from 15,000 volunteers. One goal is to have its technology predict who would qualify for a depression diagnosis on a common questionnaire. First it would augment the questionnaire, Harper says. Then it might one day replace it, helping drug researchers recruit those “bona fide” patients and gather more accurate data during trials.
Pharma moves in baby steps, however, which poses what Harper calls “a fundamental mismatch” for an AI-driven program. “To get enough data and create a model, it’s a tall ask with [a study of] 100 people,” says Harper. “Yet that’s the mode we’re in with pharma: Do things small before you move on.”
Another firm, Winterlight Labs in Vancouver, BC, has similar goals with … Next Page »