Replies to post #249915 on Anavex Life Sciences Corp (AVXL)

[georgejjl]

powerwalker, I agree.

The following should lead to very simple analysis using a statistical database.

The advantage of our trial is that there is a – two questions to the – two assessment of the physician – by the physician and directly towards the patient and also questions towards the caregiver and these questions are very the same if you would do it in person or if you would do it over audio/video or WebEx function or over telephone. The other point I like to make that in Parkinson's disease dementia, the assessment is done over an objective computer-based assessment, which is also not dependent on the interaction or the interface of the interaction.

Good luck and GOD bless,

[XenaLives]

Perhaps the delay is for human interpretation and application of the AI results.

[nidan7500]

PW, have had similar thoughts. The VIRTUAL TRIAL of the future … cannot be far off. BTW, if we must use the current FDA 5+year protocols we will all be dead and broke before any effective treatment is developed.


https://www.sciencedirect.com/science/article/pii/S235287371830026X

[Biostockclub]

Hi Power, and others,

The use of AI/machine learning/supercomputing (high performance) is really not necessary to process results of our trial.

AI/etc is a power tool used to detect hidden correlations from unstructured data. In contrast, our results should be (and are) Direct correlations between very structured data.

In very simplified essence, someone is going to take 120-150 participant names, and go across a list of primary endpoints and secondary endpoints. A result (or several data points) will be measured for each participant for each endpoint.
The dosages will be revealed and trends will be apparent via obvious graphs of patients’ endpoints which can show trends in doses over the trial time. The comparison against placebo to determine statistical significance is just a computation - no hidden correlations here. All straightforward.

Here’s an article (first one googled) which adequately explains AI/machine learning use in finding hidden correlations within groups of unstructured data - the real power of the machines (forget Watson!) is in designing(!) trials. Correlations can enable selection of patients or predict which patients should respond. It’s how you get to the significant results.

https://hitconsultant.net/2020/02/28/many-faces-ai-clinical-trials/#.XrbHHuRq2Ec

There are hundreds or more articles discussing ways these machines find amazing hidden correlations. I am working in one of 3 teams currently modeling the pandemic at PSC. Big, unstructured data - trying to gain meaningful/reliable projections of subgroups at risk, locations, behaviors - etc. This is all hands on deck situation.

My only caveat about using AI to further process our data is if we use the results as a basis for further stratification in order to prepare for a ph3 which is even more powered to patients which will succeed. (That may take longer to gain those and to enroll for that trial due to very precise screening)

Other than that, I don’t see much benefit or added “speed” in the process by employing the use of AI.
If High Performance Computing were required, all CRO’s, sponsors, and the FDA would have them. They don’t.


Analogy: AI can help you analyze your golf swing and adjust certain metrics to fine tune your game. It does nothing extra to help keep score better which a scorecard and that tiny pencil can’t do already just fine.

Btw, data mining is not data fishing nor cherry picking. Anyone who thinks these are synonymous doesn’t comprehend the field - but it’s a tough field to fully comprehend, granted.

Bio