Neurocrine Biosciences has released the PhII results for NBI-98854, their VMAT2 inhibitor for Tardive Dyskinesia. The Company reported that a single site (of the eight US sites involved) threw off the overall data analysis because their AIMS testing scores did not line up with the videotaped interview data, and that when the seven other sites were analyzed on their own, the results were strongly supportive of NBI-98854’s efficacy. In this post hoc analysis of the crossover design, patients receiving the higher, 50mg dose showed a mean decrease of 9.2 points from their baseline AIMS scores (mean 14.7), compared with a 4.7 point decrease when on placebo. Clinician and self-rating patterns were similarly indicative of a strong treatment effect.
The initial market reaction to the post hoc data analysis revision was negative, as is not unexpected; ‘ad hoc’ and ‘post hoc’ are generally warning signs in a data report. But in this case, for a POC/HOC study that had very small treatment cohorts, and is being used to confirm a signal of effect, Neurocrine’s revised analysis came across as valid, and their conclusion that the drug does provide a signal of therapeutic effect is reasonable. The fact that it took about six months to fully enroll a total of 37 TD patients in eight sites suggests something else that NeuroPerspective has previously raised: There are not that many TD patients out there. NBI-98854’s future prospects could rest on its potential utility in other relevant disorders, like Huntington’s.
The anomaly that led to the deletion of data from what we will here call the ‘rogue site’ was apparently rather dramatic, with a complete lack of correlation between AIMS scores and the videotaped record of the patient’s actual clinical presentation. Patients at that site sometimes received much better AIMS scores off drug than on drug, something not seen at any other site. This seems very unlikely to be the product of poor training, such randomness either reflects a complete disregard for correct data entry, or a mishandling of data after recording. Neurocrine does not intend to disclose the identity of this site, which is unfortunate, because this lack of quality control could have been fatal to the program, had Neurocrine not maintained this backup system of videotapes. But then again, there will be a way to know: Neurocrine intends to start a Phase IIb this summer, and a quick inspection of sites to be utilized will identify the one previous site that did not make the Phase IIb list. Given what appears to have been gross malpractice on its part, considerable proof-of-remediation will be needed before anyone should consider entrusting them with another clinical trial.