TheStreet) -- Did you catch the juicy nugget of statistical jujitsu that
(AZN - Get Report) and
(NKTR - Get Report) buried in Monday's press release announcing "positive" results from the two pivotal trials of its opioid constipation drug naloxegol?
One of those naloxegol studies -- dubbed KODIAC-05 -- might have failed.
Here's the key statement about the conduct of the KODIAC-05 study that AstraZeneca and Nektar sunk deep in
After initial locking of the database for KODIAC-05, data associated with one patient that was previously assessed as non-retrievable was found to be retrievable. These data were added to the database and the database was again locked and underwent a final analysis.
Once a clinical trial database is locked, it needs to stay locked. Locking and unlocking a database raises the risk that data can be improperly altered or manipulated. AstraZeneca insists proper procedures were followed and the naloxegol analysis presented Monday is accurate. But it's also plausible to believe AstraZeneca found no statistically significant difference in response rates between naloxegol at either the 12.5 mg or 25 mg dose compared to placebo.
AstraZeneca admits the 12.5 mg dose of naloxegol showed no benefit.
Faced with a failed phase III trial of its critically important opioid constipation drug, AstraZeneca may have embarked on a hunt to retrieve additional data from the "non-retrievable" patient. "Retrieved" data found. Database unlocked. "Retrieved" data added. Database locked a second time. Trial analyzed (again.) Winner!
Before you ask, yes, data from a single patient can make a difference in the outcome of KODIAC-05 trial. Look at the results announced Monday: The 12.5 mg dose of naloxegol failed badly with a p value of 0.202 while the 25 mg dose barely squeaked into victory lane with a p value of 0.021.
The trial required a p value less than 0.025 to reach statistical significance.
Think about it another way: If the retrieved data from the single patient had no material impact on the outcome of the KODIAC-05 study, then AstraZeneca would have had no reason to disclose anything about its data-retrieval hunting expedition.
Yet, the company did.
Again, AstraZeneca and Nektar claim both naloxegol studies -- KODIAC-04 and KODIAC-05 -- achieved their primary endpoints of increasing bowel movements compared to placebo in patients taking opioids for non-cancer pain. But very little actual data from the studies was disclosed Monday, making it impossible to verify the company's claims independently. We may not find out the truth about the studies' real outcomes until naloxegol works its way through the FDA review process.
"The topline results announced today reflect an analysis of a dataset that includes all the accessible patient records," said AstraZeneca spokesman Esra Erkal-Paler. "Ultimately, that's the only analysis that matters and that is why we only report the result based on the most accurate and complete dataset. Any earlier analyses would have been based on an incomplete dataset and we therefore would not consider them meaningful."
Referring to the post-database lockup addition of the single-patient data, Erkal-Paler added, "Although unusual, this is not unprecedented and entirely appropriate under the circumstances to ensure data from all patients who took part in the trial are included where possible."