Frequency Therapeutics — Hearing Loss Regeneration

@Mr Mister: both the FX-322 and the placebo group had the same test and both of the groups may have had learning curve effects. But at the end the FX-322 group scored significantly better than the placebo group.

 

It wasn't just words, it was words in noise. Higher frequencies are needed to distinguish the noise from some of the 'ssssss' and 't' and 'p' sounds that our mouths make.

https://www.ecophon.com/en/knowledg...coustics/generating-and-understanding-speech/

 

I can somewhat hear up to 1 kHz and nothing past. I can hear human voice, but I can only understand about 20 to 30 percent of what is being said. The 1 kHz to 16 kHz range is the difference between distinguishing words that sound alike such as like and bike and hike and Mike, etc etc. For every word I hear I can guess 10 words but only one is right. The higher frequencies help to better hear and understand the specific letters being pronounced.

It could also be that an audiogram is not the end all of human hearing and understanding. There could be other factors that we aren't aware of. Maybe FX-322 addressed those issues somehow. It did something to help with the better words scores, but what? We can guess that hearing got better above 8 kHz, but I"m not sure that is enough to get the data that they observed.

 

I thought the test result was significant? Which is very much different from significantly better. It’s very possible to get an inflated P-value depending on the mode that is being used. In all reality, the result could be entirely meaningless. We won’t know anything unless the model used to measure the p-value is released.

 

I was showing ways how word scores may be unreliable especially when you have low sample size. Some patients may move faster on the learning curve than others or maybe there is something else going on. There must be a reason why word scores improved but the reason can’t be improved hearing as audiograms were practically unchanged. They clearly state that they assessed audiometric changes from 250 Hz to 8000 Hz and there was no statistical difference between the treatment groups.

 

100% agree and I invested in biotech on a small scale/what I could afford for 10 years. They all read like that.

 

I think what you are missing here though is the statistical difference between control (placebo) and the drug was at a 0.01 p value. We can assume both groups got the same test so any subjective variable would affect both groups equally. The 0.01 p value says it didn't.

 

This is an incorrect interpretation (in a practical real world sense) of how speech recognition works.

Google the "speech banana" and you can see the frequencies that audiologists use to evaluate speech comprehension and then make adjustments in hearing aids (tl;dr: the frequencies are way higher than the wiki article). The reason for that is that speech (and all sound) follows a sine wave pattern where the resonances (i forgot the exact term audiologists use for thus but i could probably find out if you are interested) on the higher frequencies make it clearer. What this means is you can technically *hear* speech at the lower frequencies but the clarity comes from hearing the resonance at the higher frequency. So speech is heard but it sounds jumbled and unclear. Even localizing sound requires higher frequencies.

https://www.ncbi.nlm.nih.gov/pubmed/16119356/

 

Didn't they say previously the full study would be released sometime in 3rd quarter 2019?

 

Hi bro, could you elaborate on the IPO... ouch, that went over my head?

I was looking at my audiogram last night and got really bluesy.
70 dB loss at 8000 Hz, 60 dB at 6000 Hz, 45 dB at 4000 Hz: 30 dB at 3000 Hz, 25 dB at 2000 Hz, and then up to reasonable patterns. I know your hearing is bad... I ask you, anybody here, for advice. Should I get hearing aids?

I just lost my closest friend and my anxieties are raw and stumbled across my audiogram. I had been doing really well dealing with the 24/7 eeeeeeeee. I’ve just been knocked off my stride and looking for some advice... anybody?

 

Man, I am really sorry to hear that you lost your closest friend. God loves you and hang in there.

I don't have my audiogram handy and will check it later, but I'm 99% certain my hearing loss in one ear is less bad than that, at least in some frequencies, and my ENT and audiologist recommended a hearing aid right away. Which I'm trialing now.

I haven't read the IPO disclosures so I can't judge them specifically, but every IPO must disclose factors that threaten the investment and they throw every single thing in there they can think of.

While I understand people being disappointed about the safety trial showing much more modest hearing gains than all of us wanted, I think the reaction has been too negative at this stage. For one thing, the safety trial could have shown it's not safe! There's some good commentary above about reasons for hope, and I think "cautious optimism" is completely justified.

 

I think you should try them.

 

I am not missing anything. You did not not read what I just wrote. I only explained that there may be chance involved in word scores.

You seem to have a deep belief in p-values. Can you explain what p-value means? What is your take on the sample size?

 

Can you guess why Frequency Therapeutics completed the audiogram tests in the clinical study up to 8000 Hz and not up to 16000 Hz?

 

You are suggesting that audiogram no longer is the standard method for measuring hearing loss. Extraordinary claims demand extraordinary proof. Where is the extraordinary proof?

 

To anyone knowledgeable with investing, when do you predict we will able to buy stock of Frequency Therapeutics?

 

Sure, happy to. The p value is the chance that the results were due to a chance occurance rather than an effect in the hypothesis. So the smaller the p value, the less likely observations were due to chance.

P value takes into account sample size, so the smaller the sample size, the greater the effect has to be to get a small p value.

What this means is, though it's not impossible for the differences between placebo and the drug groups to be due to chance, it's unlikely because of the small p value. So if we are going by odds, it's much more statistically likely there were significant meaningful differences in word scores before and after the drug vs differences in placebo.

 

I believe it's because they didn't have a baseline (before injection) of the very high frequemcy measurements because they were just testing safety.

A safety study just requires your hearing not to get worse on a standard audiogram to not report an adverse event. They happened to get an improvement in word score on the safety study, which is exciting so they published. They probably also hoped it would diffuse further but the full dose has potential to diffuse a lot further.

Phase 2 is where efficacy is tested and up to 16000 Hz is included in Frequency's exploratory arm of phase 2. That's not a coincidence. I suspect they believe improvements will be seen there based on the lack of "standard" audiogram changes yet having the increased word score and speech in noise.

Interestingly they are also including THI as well as up to 16000 Hz audiogram. This will be a full efficacy study.

 

Feel free to message me if you need someone to commiserate with. Without hyperbole, I lost everything to this.

 

That information does not appear to be public yet.

 

Yes, you should get hearing aids. At those levels they would help with understanding people and with tinnitus. Individual results vary and they help some people while others not. When I was at those levels they did help a lot though.

The IPO had a lot of risks listed, but I guess that is pretty standard from what everyone said.

 

I'm just suggesting that if word scores can be improved with no improvement in the audiogram maybe there is more to it than we know. I don't know either though. It could be that their hearing improved beyond 8 kHz. It could be that the sample size was too small. It could be the people learned the words and they got better. We are all just speculating.

Some people hear better in noisy environments than others. An audiogram won't tell you that.

 

Not to be pedantic, but isn’t the p value measuring the odds that the model is significant, not the actual result? So you can say that with the given model, the result mean X. The real world application is really pretty disconnected from the model. So says the bayesian to the frequentist.

 

In the sense that real world odds are essentially either it is correlated (to x degree) or it's not, sure. Something doesn't have a 99% chance of being relevant: either it is or it's not a factor.

The way i was taught is that it gives you the "chance of chance" on those results and the interpretation ("since it doesn't seem like chance, here is a possibility of why...") is not straight from a p value.

 

Run that by me again?

 

Yes, you probably should. Best case scenario is that by the time the treatment comes, you'll be ready for your second set of aids. That's a few years of compromised hearing if you do nothing. Be aware that the longer your hearing loss goes unaided, the more your brain forgets how to interpret speech.

If money is an issue and you live near a Costco, go there first.

 

If you had bought it you would have lost a lot of money. Be realistic. It doesn’t work.

 

Just curious, but in your opinion why would Astellas partner with them? They gave them a lot of money up front (not for stock in return but for a return on profits after licensing) after reading the same results we did. They are an established pharma company and this isn't their first rodeo.

 

Sure. I always taught that a p-value only measures the validity of an outcome against a statistical model. Not against the real world. This is one of the primary gripes of Bayesian statistics against frequentist. See this comic for an explanation:

https://xkcd.com/1132/

Hope that helps!

 

I’ve thought a lot about Astella after reading the IPO document. Here are some of my thoughts:

1) I call bullshit on Frequency Therapeutics not testing hearing out to 20 kHz. Langer, Karp, McLean, et al, they all know an extraordinary amount about hearing biology, enough so that they were able to potentially create a drug to combat this ailment. I know the IPO paper stated they only tested out to 8 kHz... but come on. There’s no way they sat back and let that happen. They have this data. I also guarantee Astella has this info too. Why else would Astella pony up another $540mil upon successful endpoints? They would have to have serious reason to write up a contract for an additional $540mil. Astella has more information than we do and they put their money where their mouth is based on that information.

2) Liabilities and lawsuits are the norm in this day and age. In a document like the IPO we’ve all been dwelling on, any smart company isn’t going to say more than they have to in writing. You don’t want to provide too much up front because people can turn around and sue for false claims if things go awry, therefore, you utilize jargon like “has the potential” and you leave certain information out to give yourself some wiggle room going forward.

3) The PCA platform is a new technology being pushed through an antiquated review process (FDA). I guarantee whatever dose Frequency Therapeutics deemed to be therapeutically significant the FDA forced them to cut it in half to test it in a safety trial. There’s no previous example of this technology being tested in the auditory system, therefore, there’s no baseline to set guidelines to. This forces the FDA to be more cautious than necessary. This gives me hope that the dose in the High Dose group from 1B isn’t the actual high dose that can be safely used. All that being said, we still achieved a 10 dB increase at 8 kHz and WR scores were nearly doubled for the 4 individuals who benefited from this experiment.

4) The sample size is way too small and the variance is way too great for the 15 people who actually received the drug in the 23 person trial. We know nothing about these 15 people and a reliable regression model can’t be built on a sample population of 15. 4 had 10 dB improvement, but how many had 5 dB improvement at these safety doses? Yes 5 dB isn't clinically significant, but how many on average throughout the 90 days of testing scored 5 dB higher multiple times than their baseline? Of the few who didn’t show any improvement how old are they? What medications do they take on a daily basis? How loud is their work environment? How healthy are they? What are their lifestyle choices? Do they have a family history of hearing loss? Are they diabetic? Etcetera, etcetera, etcetera... all this gets ironed out when your sample populations are in the thousands... not 23.