Frequency Therapeutics — Hearing Loss Regeneration

[Guywithapug]

It's 3/22 today Frequency Therapeutics. No better day than today. So come on and throw us a bone... You know you want to.

 

[Zugzug]

I've been reading a few papers on how statistical and clinical significance of TFI is measured. Since tinnitus is only a secondary end point, I have a feeling the analysis may be simple.

I have seen repeated measure ANOVA tests done (more rigorous) for a controlled study, looking at the impact of hearing aids with (treated group) and without (placebo group) "noise features" to treat tinnitus for people with hearing loss.

This study proposal uses an early-start, delayed-start scheme to look for the impact of conservative TMJ treatments on TMD induced tinnitus. They also use repeated measure ANOVA.

However, in Otonomy's OTO-313 top-line results seemed to break down the key results with p-values into categorical variables. In other words, people were either responders or not based on various clinical guidelines. In all likelihood, they performed a Fisher Exact Test (same test Frequency Therapeutics used for "responders"), which is essentially the same thing as a Chi-square test.

As a secondary endpoint, and the sheer amount of important data for the primary end points, I suspect their (at least initial) top-line report will be categorical. There are 4 total cohorts, which generates 6 comparisons. That's a lot to analyze with regards to the primary end points (WR, WIN, standard PTA).

They will probably do this a few times, even with EHF (responder or not) as well.

EDIT: On page 188 of the patent, they talk about Phase 2a. Tinnitus (TFI) will be measured on days 0, 60, 120, and 210. Hence, each person will only have two data points: base line and day 60.

 

[Diesel]

If I count all 6, you get 82 words improved across 6 people, which amounts to 27% absolute improvement.

If I count all 5, you get 78 words improved across 5 people, which amounts to 30% absolute improvement.

Maybe I'm bad at math... but looking at the Phase 1/2:

The baseline for the group of 6 responders was: 104/300
At 90-days, the 6 collectively improved to: 186/300
This is an improvement of about 78% from group baseline.

Am I missing something?

 

[twa]

Very good point. If I understand the idea of what you're saying, it's that if I take a drug, though the half-life elimination rate is roughly fixed, the efficacy per amount improves with time. This is definitely true and if I understand correctly, a lot of how drugs work. For example, if I take an Ibuprofen, I am "overdosing" because a bunch will be metabolized, but what remains after a while is more effective. Is that the idea?

Completely agree. Actually, I nearly had a part III exploring this very thing. I didn't know a great way of studying it with such small sample sizes.

The message is clear though. Responders respond and it's not a mistake.

Thank you for the bottom line. :=)

Maybe I'm bad at math... but looking at the Phase 1/2:

The baseline for the group of 6 responders was: 104/300
At 90-days, the 6 collectively improved to: 186/300
This is an improvement of about 78% from group baseline.

Am I missing something?

Have the career mathematician analyze the numbers.

 

[Zugzug]

Maybe I'm bad at math... but looking at the Phase 1/2:

The baseline for the group of 6 responders was: 104/300
At 90-days, the 6 collectively improved to: 186/300
This is an improvement of about 78% from group baseline.

Am I missing something?

The confusion (this has confused me too at times) is that the test is given as percentage score, but "absolute percentage improvements" literally means the difference, as opposed to the ratio. For example, if I score 25/50 (50%) at baseline and 30/50 (60%) at day 90, my absolute percentage improvement is 60-50=10%.

On the other hand, my percentage improvement (classical meaning) would be 100*(30/25-1)=20%. I guess this is not different from person A scoring 95% on a test and saying they beat their friend (who scored 85%) by 10%. It's really an abuse of the word percentage, which is annoying and confusing.

For the Phase 1 analysis at the individual level (95% CI per Thornton and Raffin), the confidence intervals for the day 90 score are not symmetric about the baseline score (i.e. baseline score is not in the middle). See page 515 of the paper. The first column provides the baseline score and the second column provides the lower and upper bounds of the 95% confidence interval.

To put a technical idea in simple terms, this is because the angular transform function is arcsin and square root functions, which are highly nonlinear.

Now, at the group level, first of all, they aren't dropping the people who scored >90%. Because of this, all of the analysis is done based on comparing average percentage improvements (actual meaning, ratios). However, the "average percentage improvement" in a group is not as simple as the fraction with numerator as sum of scores and denominator being 50*(number of people).

We actually have to calculate the percentage (ratio) improvement for each person. Then average them and compare groups across all time periods.

 

[dd314]

I've been reading a few papers on how statistical and clinical significance of TFI is measured. Since tinnitus is only a secondary end point, I have a feeling the analysis may be simple.

I have seen repeated measure ANOVA tests done (more rigorous) for a controlled study, looking at the impact of hearing aids with (treated group) and without (placebo group) "noise features" to treat tinnitus for people with hearing loss.

This study proposal uses an early-start, delayed-start scheme to look for the impact of conservative TMJ treatments on TMD induced tinnitus. They also use repeated measure ANOVA.

However, in Otonomy's OTO-313 top-line results seemed to break down the key results with p-values into categorical variables. In other words, people were either responders or not based on various clinical guidelines. In all likelihood, they performed a Fisher Exact Test (same test Frequency Therapeutics used for "responders"), which is essentially the same thing as a Chi-square test.

As a secondary endpoint, and the sheer amount of important data for the primary end points, I suspect their (at least initial) top-line report will be categorical. There are 4 total cohorts, which generates 6 comparisons. That's a lot to analyze with regards to the primary end points (WR, WIN, standard PTA).

They will probably do this a few times, even with EHF (responder or not) as well.

EDIT: On page 188 of the patent, they talk about Phase 2a. Tinnitus (TFI) will be measured on days 0, 60, 120, and 210. Hence, each person will only have two data points: base line and day 60.

IIRC, the ANOVA test challenges the null hypothesis that all groups have the same average, and it alone doesn't tell you which subgroups differ from which. In the same way, a chi2 test of independence just tells you whether categorical variables are independent of one another. All are based on some fundamentally arbitrary cutoff along the probably density function.

 

[Chetan]

Maybe I'm bad at math... but looking at the Phase 1/2:

The baseline for the group of 6 responders was: 104/300
At 90-days, the 6 collectively improved to: 186/300
This is an improvement of about 78% from group baseline.

Am I missing something?

82 words improved collectively across 6 people.

Average improvement is 82/6.

Multiple by 2 to get a score improvement (since 50 word test).

That gets you 27%.

Remove the 4 word improvement 6th patient.

Becomes 78 improvement, divide by 5 patients for average word improvement, multiple by 2 to get score improvement.

It's 30%ish.

@Zugzug, worth considering that for the group that is four doses, day 60 is really equivalent to day 90 given 4 weeks of dosing with first dose starting day -28 I think, then it starts day 0 and then day 60, hence it's really day 88.

For others, it's variant of day 74 or day 67.

I agree they've got so much to analyze and do a really thorough job to ensure people don't think they are a scam anymore, especially since their Phase 1 readout was very generic and little detail disclosed. They NEED to share the data transparently and with full diligence to persuade patients, insurance, regulators, and industry that this is real (especially given the inherent skepticism of value of EHF by industry).

 

[Diesel]

The confusion (this has confused me too at times) is that the test is given as percentage score, but "absolute percentage improvements" literally means the difference, as opposed to the ratio. For example, if I score 25/50 (50%) at baseline and 30/50 (60%) at day 90, my absolute percentage improvement is 60-50=10%.

On the other hand, my percentage improvement (classical meaning) would be 100*(30/25-1)=20%. I guess this is not different from person A scoring 95% on a test and saying they beat their friend (who scored 85%) by 10%. It's really an abuse of the word percentage, which is annoying and confusing.

For the Phase 1 analysis at the individual level (95% CI per Thornton and Raffin), the confidence intervals for the day 90 score are not symmetric about the baseline score (i.e. baseline score is not in the middle). See page 515 of the paper. The first column provides the baseline score and the second column provides the lower and upper bounds of the 95% confidence interval.

To put a technical idea in simple terms, this is because the angular transform function is arcsin and square root functions, which are highly nonlinear.

Now, at the group level, first of all, they aren't dropping the people who scored >90%. Because of this, all of the analysis is done based on comparing average percentage improvements (actual meaning, ratios). However, the "average percentage improvement" in a group is not as simple as the fraction with numerator as sum of scores and denominator being 50*(number of people).

We actually have to calculate the percentage (ratio) improvement for each person. Then average them and compare groups across all time periods.

Ahhh... ok... this is helpful. Thanks.

 

[Diesel]

@Zugzug, worth considering that for the group that is four doses, day 60 is really equivalent to day 90 given 4 weeks of dosing with first dose starting day -28 I think, then it starts day 0 and then day 60, hence it's really day 88.

For others, it's variant of day 74 or day 67.

Yeah, it seems as though "Day 90" in the Phase 2A is essentially 120 days from the 1st FX-322 dose, and 90 days from the 4th dose. Should help further weed out any placebo effect.

 

[Zugzug]

IIRC, the ANOVA test challenges the null hypothesis that all groups have the same average, and it alone doesn't tell you which subgroups differ from which. In the same way, a chi2 test of independence just tells you whether categorical variables are independent of one another. All are based on some fundamentally arbitrary cutoff along the probably density function.

Great point. I'm not even sure if they will use a repeated measures ANOVA. They say they use a MMRM (Mixed Model for Repeated Measures), but I wish they disclosed more details.

Fortunately, the SIN test provides simple clinical guidelines, based on Wilson and McArdle, 2007. All they do is apply the Spearman-Karber formula to each patients SIN results to estimate the SNR level that leads to 50% correctness.
For comparing within groups (i.e looking for individual clinical improvements), they simply perform a matched t-test with 1 data point being baseline score and the second being the day 90 score. Then to compare groups, they just do an unmatched t-test for the differences. Pretty straightforward and Wilson and McArdle even verified that the formula preserves normality of data so a t-test is reasonable.

You make a great point about the multiple comparisons though. I really don't know what they will do. I have a suspicion they may use the Bonferroni correction for the multiple comparisons because it's a nice, easy, conservative test that reduces the likelihood of at least 1 type I error in all of the comparisons.

I've also wondered if they will, at times, group treaters vs placebo. In other words, drive home the point by comparing the 24 placebos vs 72 in treatment group.

My prediction is that there will be stratification in all groups, but only the following comparisons will be statistically significant:

• Placebo vs 1x, 2x, or 4x
• 1x vs 4x

The only test where I think there could be statistical significance between 1x vs 2x and 2x vs 4x is in the 8 kHz PTA because it's right at the border of action.

 

[GlennS]

They will release the results tomorrow.

Here we go again

 

[Toby1972]

I just had an inspiration that the tide was turning for the better. I can verify it in maybe 1 hour.

 

[Keith Handy]

Here we go again

"Does anybody remember laughter?"

 

[Pero1234]

All hail the Bobb!

 

[Zugzug]

I just had an inspiration that the tide was turning for the better. I can verify it in maybe 1 hour.

There are two low tides and two high tides in every day for every coastline. Maybe your inspiration was this.

 

[Toby1972]

It takes a little bit more time with the inspiration.

I would like to have hope, but you just have to look at the chart.

The chart shows the comparison of the highly speculative Otonomy share BEFORE the bad news was released (in mid-February, the market value halved in one day) and Frequency Therapeutics' just before the news now.

Well who notices something?

Are these completely different courses, quite similar or even almost identical?
You tell me, otherwise I look like a know-it-all.

 

[dd314]

It takes a little bit more time with the inspiration.

I would like to have hope, but you just have to look at the chart.

View attachment 44172

The chart shows the comparison of the highly speculative Otonomy share BEFORE the bad news was released (in mid-February, the market value halved in one day) and Frequency Therapeutics' just before the news now.

Well who notices something?

Are these completely different courses, quite similar or even almost identical?
You tell me, otherwise I look like a know-it-all.

It's basically following the biotech index. Look at $LABU and you'll see it basically mirrors FREQ.
edit- LABU is the bull index but they are basically identical

Edit2-FREQ was just upgraded to "buy" at the upswing, a simple enough explanation

 

[Christine2222]

What's more likely, an aftermarket or premarket news release? Anyone know?

 

[Aaron91]

It takes a little bit more time with the inspiration.

I would like to have hope, but you just have to look at the chart.

View attachment 44172

The chart shows the comparison of the highly speculative Otonomy share BEFORE the bad news was released (in mid-February, the market value halved in one day) and Frequency Therapeutics' just before the news now.

Well who notices something?

Are these completely different courses, quite similar or even almost identical?
You tell me, otherwise I look like a know-it-all.

I'm cringing. Please stop.

 

[Zugzug]

I'm a stock market noob. Does anyone understand this picture? It's 1 month of data, mapping percentage changes between the NBI (dark blue) and FREQ (light blue) lines. In particular, it seems like there's a drop around March 3 to March 8, then the curves are approximately parallel.

 

[Diesel]

Again, I have learned nothing.

 

[Tezcatlipoca]

It takes a little bit more time with the inspiration.

I would like to have hope, but you just have to look at the chart.

View attachment 44172

The chart shows the comparison of the highly speculative Otonomy share BEFORE the bad news was released (in mid-February, the market value halved in one day) and Frequency Therapeutics' just before the news now.

Well who notices something?

Are these completely different courses, quite similar or even almost identical?
You tell me, otherwise I look like a know-it-all.

Don't worry Toby, the last thing we think of you is a know-it-all.

 

[Gb3]

Just an update.

I called regarding the severe hearing loss trial that is ongoing.

I spoke to the clinical researcher.

Based on my conversation with them it was my impression that a Phase 3 trial is almost certain.

Just saying.

 

[Aaron91]

What's more likely, an aftermarket or premarket news release? Anyone know?

Most likely: pre-market if good news (early in week), after-market if bad news (late in week).

 

[Chetan]

Folks concerned about recent weakness.

It's all part of biotech ETF rebalancing.

I confirmed my statement Friday when I made this claim today, XBI added 1.5%, over 520k shares to their existing 345k position to get to 872k shares.

I'm sure other institutes were front running and making money on that dip to $32.

The bounce you see today is people letting it go now that artificially suppressing to load the shares is done.

In fact, today's option activity is so incredibly bullish I can't even believe the dichotomy between biotwitter's bearishness and what is clearly smart institutional money placing hundreds of thousands on each trade, with millions of dollars cumulative today on bullish option positions (meaning this stock is expected to go up, at least to the $60-$80 range minimum).

I am shocked I've rarely seen this, but I will always bet on Langer and Perceptive Advisors over people who have done very little diligence and have nothing at risk on this trial going wrong.

I believe the trial will be positive. And apparently some people think it'll be unambiguously good enough that the stock goes at least to the $80-90 range by April end.

 

[Lucifer]

Based on my conversation with them it was my impression that a Phase 3 trial is almost certain.

What did they say that made you think a Phase 3 trial was almost certain?

 

[FGG]

It takes a little bit more time with the inspiration.

This is the language psychics use. Do you feel like you are "psychic"?

 

[RichieTheKid]

Based on my conversation with them it was my impression that a Phase 3 trial is almost certain.

When is this trial starting? I need to get in.

 

[Diesel]

I'm not worried.

 

[HootOwl]

Don't worry Toby, the last thing we think of you is a know-it-all.