) --

Vanda Pharmaceuticals

(VNDA) - Get Report

tells a pretty compelling story about the development of tasimelteon for the treatment of a rare sleep-pattern disorder that affects blind people. Dig into the company's clinical work on tasimelteon, however, and you'll find a disturbingly large number of irregularities and red flags which should ring alarm bells for any investor betting on the drug's approval.

Tasimelteon belongs to the melatonin family of compounds which helps regulate the 24-hour sleep-wake cycle, also known as circadian rhythm. For some people who are totally blind, having no light perception at all knocks the normal, 24-hour circadian rhythm off kilter. Their sleep patterns start to shift -- less sleep at night, more during the day.

The medical term for this condition is "non-24 sleep-wake disorder" (non-24) and there are no drugs approved to treat it. Ordinary, over-the-counter melatonin works for many non-24 sufferers (so does an alarm clock) but FDA classifies melatonin as a nutritional supplement, not a drug.

Based on results from two phase III studies, Vanda believes tasimelteon is the first drug shown to "entrain" non-24 patients back into a normal circadian rhythm. Tasimelteon also improves sleep performance, the company says.

In May, Vanda submitted a U.S. approval application for tasimelteon to treat non-24 disorder. The U.S. Food and Drug Administration has not yet accepted the tasimelteon submission for review or set an official approval decision date.

This is where Vanda's tasimelteon story runs off the rails.

  • The design of Vanda's primary phase III study changed numerous times, including a complete replacement of the primary endpoint just one month before study results were announced.
  • The replacement primary endpoint installed to assess tasimelteon's benefit was created by Vanda and has never been used before in sleep-drug clinical trials, nor was it endorsed by the FDA.
  • Vanda was forced to cut in half the patient enrollment into the tasimelteon clinical trials because totally blind patients with non-24 could not be identified. Even then, Vanda was only able to enroll patients by stretching the clinical definition of non-24.
  • Tasimelteon was only able to demonstrate a benefit for non-24 patients by combining data from two phase III studies. Despite Vanda's claims to the contrary, the phase III studies may have actually failed on their own.

Let's dig into these Vanda red flags in more detail. But first, I'll note Vanda executives chose not to respond to questions about tasimelteon's clinical development. The company arranged an interview with Stephen Lockley, a Harvard professor and sleep expert who oversaw the tasimelteon clinical trials. On Tuesday, hours before the interview with Lockley was to take place, a Vanda spokeswoman emailed to say Lockley was called away by an emergency and would be unavailable. Moreover, Lockley was unable to reschedule because he was traveling through the middle of July.

If you're intending to seek FDA approval for a drug to treat patients who can't sleep when they should (at night) and sleep too much when they shouldn't (during the day), designing a clinical trial that measures improvement in total sleep time at night makes a lot of sense.

Vanda's original study design did just that. Go back to

July 2010 when Vanda first started enrolling patients

in the phase III "SET" study: Patients diagnosed with non-24 were to be randomized to treatment with either tasimelteon or a placebo. The study's primary endpoint was a comparison of average

"nighttime total sleep time"

between the two arms. Secondarily, the study was designed to assess tasimelteon's ability to improve other measures of sleep performance, including reductions in day time sleep and overall patient quality of life.

In 2012,

Vanda's investor presentation

began to include a change in the tasimelteon study. Entrainment -- defined as the "the ability to reset the body clock and maintain a 24-hour clock" -- was listed as a co-primary endpoint of the phase III "SET" study along with nighttime total sleep. The change in the tasimelteon study design to include co-primary endpoints was "under discussion with the US FDA," according to Vanda's investor slides, despite the study being underway for two years.

In June 2012, entrainment was listed as the sole primary endpoint of the tasimelteon "SET" study. Nighttime total sleep was deleted entirely.

In November of the same year, the

tasimelteon study design was changed again

to analyze co-primary endpoints: 1) entrainment; and 2) a measure of response defined as a "significant improvement... in key clinical measures."

This final change was put in place one month before the "SET" study results were analyzed and announced by Vanda in December 2012.

The timing of the changes to the tasimelteon study in the middle of 2012 were pushed through at the same time Vanda already had data in hand from the study suggesting the original primary endpoint -- improvement in night time total sleep -- was likely to fail.

In mid-2012,

Vanda conducted an analysis of 161 blind people with self-reported sleep problems

who participated in screening for the phase III "SET" study. Of these patients interested in enrolling into the tasimelteon study, 107 were found to have non-24 and would therefore be eligible, while 54 had "normal" circadian rhythms and therefore could not participate. Sleep performance data, including night time total sleep, was collected for all the patients.

The goal of this analysis, according to

Vanda's own poster

, was to confirm that night time total sleep -- the endpoint already chosen for the ongoing phase III study -- was an appropriate measure for assessing the clinical benefit of a drug to treat non-24.

And here is how Vanda presented the results of this endpoint analysis:

Nightime total sleep time (nTST) measures the total amount of time an individual is asleep over the entire sleep opportunity. Individuals with N24HSWD non-24 get, on average, 53 minutes less nighttime sleep than the individuals without this disorder (entrained) during the worst quartile of all of their sleep. This difference was statistically significant." Bolded emphasis mine.

Did you catch the bait and switch? Vanda doesn't present data on nighttime total sleep, even though this information was collected for all patients. Instead, Vanda only discloses sleep time for 25 percent of nights when sleep times were lowest (the worst quartile.)

This suggests there was no discernible difference in nighttime total sleep between those blind patients deemed to have non-24 and blind patients with a normal circadian rhythm. If there was a significant difference, why not disclose? Vanda doesn't, even though nighttime total sleep is the primary endpoint of the phase III study.

Vanda presented a poster on this endpoint analysis in June 2012. That same month, Vanda eliminated nighttime total sleep as the primary endpoint of the tasimelteon phase III study. Coincidence? Seems unlikely.

Changing the endpoints of an ongoing clinical trial -- especially without the FDA's blessing -- is a sign of trouble.

Speaking at an investor conference in July 2012, right after the tasimelteon study was changed, Vanda Chief Commercial Officer Robert Repella said, "We do not, at this point, have concurrence with the FDA on the primary and secondary endpoints for the study. We've been in active dialogue with the agency now for some time and we continue to move in the direction of clarification around how the endpoints will be characterized for the filing."

Even worse for Vanda: The "entrainment" co-primary endpoint finally selected for the tasimelteon phase III study was a urine-based laboratory measurement which couldn't be used on its own for an FDA drug approval. The second co-primary endpoint of the tasimelteon study relied on a "non-24 clinical response scale" created by Vanda and never validated or used by any other company or sleep expert conducting clinical trials.

In the phase III study, Vanda counted patients as responders if they passed three of four sleep-performance tests that made up the "non-24 clinical response scale." Why wasn't passing all four tests required for a response? Vanda has never explained.

Everything about the "non-24 clinical response scale" was arbitrary.

As an example, one of the four tests measured patient quality of life or global functioning known as CGI-C. A positive score was defined by Vanda as "less than or equal to 2 from the average of Day 112 and Day 183 compared to baseline."

The significance of measuring quality of life only from days 112 to 183 of the study has not been explained.

Vanda announced "positive" results from the phase III "SET" study of tasimelteon on Dec. 18, 2012. The study was designed originally to enroll 160 totally blind people (self reported, by the way) diagnosed with non-24. In August 2011, enrollment was reduced from 160 to 100 patients. In January 2012, enrollment was cut again from 100 patients to 84 patients.

Of the 84 patients enrolled into the study, 22 patients (26 percent) dropped out.

By Vanda's own admission, a majority of the patients enrolled in the tasimelteon study did not suffer from non-24, according to the "textbook definition" of the disease.

"Less than 5% of individuals matched the traditional textbook definition of N24HSWD with a clearly non-24 hour sleep period," Vanda states in a

scientific poster presented in June 2012

which characterized patients screened for entry into the phase III study.


Vanda poster states


"Only 30% of participants exhibited cyclic sleep-wake patterns despite all participants having confirmed non-24 melatonin rhythm. A minority (less than 10%) of individuals exhibited no evidence of sleep disturbance and the majority exhibited sleep complaints but with no evidence of a cyclic disturbance."

The slide below is Vanda's characterization of the "positive" results from the "SET" study.

Check out the footnote attached to the clinical response co-primary endpoint. It reads, "Entrainment status from the randomized portion of the SET study and/or the screening portion of the RESET study."

In other words, the SET study only achieved one of its two co-primary endpoints by adding in patients who participated in the second phase III trial. Without these extra patients, the SET study likely failed.

The story Vanda tells of its sleep-disorder drug doesn't have a happy ending.

-- Reported by Adam Feuerstein in Boston.

Follow @AdamFeuerstein

Adam Feuerstein writes regularly for TheStreet. In keeping with company editorial policy, he doesn't own or short individual stocks, although he owns stock in TheStreet. He also doesn't invest in hedge funds or other private investment partnerships. Feuerstein appreciates your feedback;

click here

to send him an email.