009, 117, 171 & 175 Trial Results

[irashorr]

My thanks too for the analysis.
I'd be interested in your take on Red Acres recent point about the results:

"Whether or not MNKD can land a partner who will make the commercial investment necessary to cause a paradigm shift in diabetes treatment remains to be seen. If the Type 2 HbA1c advantage over placebo was much greater (in the 1% to 1.5% range) we would be much more bullish on MNKD's prospects. The data from these trials just do not seem, in our view, compelling enough to attract a partnership on great terms."

[karma1]

OPC,

Well done on the synopsis! Thank You!

[johnnicholas14]

I have some concerns about your interpretation of the data.


"8. As for the primary endpoint, Afrezza came in at .21, which is within the .02-.36 range and therefore met the non-inferior threshold."


HbA1c reduction difference between the means: 0.40-0.21=0.19%

I’m not sure that you understand the results: 0.02-0.36 is the 95% CL of the difference between the groups, i.e. 0.19+/- 0.17%; the lower bound, 0.02, would mean they were nearly identical; the upper bound, 0.36, meant that the control group was better at affecting HbA1c reduction. Since 0.36 was less than the margin, 0.40, Dreamboat-Afrezza met the NI end-point.

That 0.21 is between 0.02-0.36 is irrelevant. Had the Dreamboat-Afrezza arm lowered HbA1c by an average of 0.36, would you expect 0.36 to lie in the 95% CI of the difference between the groups?



"2. a1c results: -.21% vs -.49 (Afrezza vs Comparator). The confidence interval (CI) was -.09% to -.40%. The difference, -.28, fell within the CI thereby establishing non-inferiority. The final mean a1c's were 8.20% vs 7.99% (Afrezza vs Comparator) or a difference of .21."

I have no idea where you got your numbers from, but looking at the EASD poster of the TI-009 data, you made some transcription errors.

HbA1c Reduction: -0.13 (Afrezza) v -0.37 (Comparator); the 95% CI for the mean difference between the groups was 0.08-0.40; the margin for NI was 0.40. The upper limit equals the margin. Therefore, the non-inferior end-point was not met. If you notice in the conclusions section of the EASD poster, you’ll never see the words “non-inferiority, etc.”
If you read the transcripts where MNKD mgmt discuss the 009 data, you’ll not see them describe the data as “non-inferior, etc.” Other descriptors, e.g. ‘comparable,’ are used, but are meaningless in this context.

HTH

jn

[qqueue83]

TI-009 did meet its non-inferior endpoint.

"Study 009 meets its primary endpoint of non-inferiority to a rapid-acting insulin analog"

www.news.mannkindcorp.com/phoenix.zhtml?c=147953&p=irol-newsArticle&ID=1197424&highlight

[StevieRay]

"Over the 52-week period of this study, A1C levels decreased comparably in the two treatment groups, with a between-group difference of -0.24%. The 95% confidence interval (-0.09% to -0.40%) did not exceed the predetermined threshold of 0.40%, thereby establishing non-inferiority between Technosphere Insulin and insulin aspart."

[opc]

Hi John,

Glad to have your assistance. I will be the first to admit this is new to me. It's one of the reasons I welcome corrections.

Allow me to deal w/ your second point first. We need to be dealing with the same data points. If you look in the "Results" section of the poster, the 2nd & 3rd bullet points contain the data I referenced. Note final a1c's were 8.20 vs 7.99. Those same bullet points state baselines were 8.41 vs 8.48. Hence, -.21 vs -.49. The 95% CI number I used was from the company's PR (-.09 to -.40). (i read these before looking for additional info). No issue w/ someone using the posters CI (-.11 to -.38) as it doesn't change the picture.

However, I disagree that the non-inferiority endpoint was not met. -.40 is the upper threshold. -.09 is the lower threshold. Data (differences) falling within this range would be considered non-inferior. Our data fell within this range. Their failure to state "non-inferiority" in the transcripts doesn't automatically mean it wasn't. (It doesn't say the powder was white. Does that mean it was black??)

In fact, I would direct you to the company's PR on the data where they specifically note non-inferiority: "A1C levels decreased comparably in the two treatment groups, with a between-group difference of -0.24%. The 95% confidence interval (-0.09% to -0.40%) did not exceed the predetermined threshold of 0.40%, thereby establishing non-inferiority between Technosphere Insulin and insulin aspart."

It would appear this exercise has uncovered the fact that results get fine tuned as time progresses IMO. Something we probably shouldn't forget.

As to your first point. I think your quote suggests I have a decent grasp on CI's: "-0.13 (Afrezza) v -0.37 (Comparator); the 95% CI for the mean DIFFERENCE between the groups was 0.08-0.40;" I did make a mistake and appreciate you catching it. It should have said the difference, -.19 (.40 - .21), was within the CI. I used the result, not the difference in results.


Regards,

[johnnicholas14]

First of all Peter Richardson no longer is employed by MNKD.

Second:

www.fda.gov/downloads/Drugs/.../Guidances/UCM202140.pdf

From pg 10/66

"As described above, the NI study seeks to show that the difference in response between the
active control (C) and the test drug (T), (C-T), the amount by which the control is superior to
test drug, is less than some pre-specified non-inferiority margin (M)."

Note that it says "less than," not less than or equal to. If Study 009 met its end-point, Afrezza would have had no clinical utility issues, and would have been approved.

HTH

jn

[opc]

not sure what Richardson's tenure has to do with this discussion.

I don't feel the need to continue this discussion on CI's. If CI didn't need a range, why quote it as such?? Sorry. You're right that .40 is the max. However, you're wrong in assuming the lower threshold is of no importance. I suggest you google "confidence intervals".

I'll also assume since you did not respond to any of my other retorts that you agree.

[StevieRay]

From pg 11/66

"Because the consequence of choosing a margin greater than the actual treatment effect of the
active control in the study is the false conclusion that a new drug is effective (a very bad public health outcome), there is a powerful tendency to be conservative in the choice of margin and in the statistical analysis that seeks to rule out a degree of inferiority of the test drug to the active control of more than that margin. This is generally done by ensuring that the upper bound of the 95% two-sided confidence interval for C-T is smaller than M1. The upper bound of the confidence interval for C-T is not, however, the only measurement of interest, just as the lower bound of a 95% confidence interval for effect size of drug versus placebo is not the only value of relevance in a placebo-controlled trial. The point estimate of the treatment effect and the distribution of estimates of C-T smaller than the 95% upper bound are also relevant. Nonetheless, the upper bound of the 95% CI is typically used to judge the effectiveness of the test drug in the NI study, just as a two-sided p-value of 0.05 or less is traditionally the standard used for defining success in a superiority trial. The 95% CI upper bound for C-T is used to provide a reasonably high level of assurance that the test drug does, in fact, have an effect greater than zero (i.e., that it has not lost all of the effect of the active control)."

[opc]

Stevie, thanks for doing the legwork. I had no intentions of doing it. But thanks for removing all doubt from the discussion.

[johnnicholas14]

“…Nonetheless, the upper bound of the 95% CI [0.40 for Study 009] is typically used to judge the effectiveness of the test drug in the NI study, just as a two-sided p-value of 0.05 or less is traditionally the standard used for defining success in a superiority trial. The 95% CI upper bound for C-T is used to provide a reasonably high level of assurance that the test drug does, in fact, have an effect greater than zero (i.e., that it has not lost all of the effect of the active control)."

If you go to pg 12/66 of the guidance, you'll note Study 009 most closely resembles scenario 4, where the upper CI crosses M1. In Study 009, the upper bound would touch M1, where it needs to be ‘less than’ M1 to claim non-inferiority.

HTH

jn

[StevieRay]

I see John's point but I'm not sure I would agree. The graph and the text clearly show it must exceed the line to be rejected. And that fact that it is on the line (.40) and not over puts it clearly in the acceptable range.

[opc]

I don't see John's point.

The difference was -.28. The CI was -.11 to -.38 (if you reference the poster) or -.09 to -.40 if you reference mnkd's initial PR. But the results fall between both.

Why is it being suggested it was on the line?

[johnnicholas14]

“I see John's point but I'm not sure I would agree. The graph and the text clearly show it must exceed the line to be rejected. And that fact that it is on the line (.40) and not over puts it clearly in the acceptable range.”

Read the guidance further; I believe you need to be ‘less than.’ Also, if the comparator lowered A1c by less than the margin (0.37<0.40) , then this statement may apply:

“Because the consequence of choosing a margin greater than the actual treatment effect of the active control in the study is the false conclusion that a new drug is effective (a very bad public health outcome).”

HTH

jn

[johnnicholas14]

Opc

The difference between the groups (C-T) was 0.37-0.13=0.24.

The CI from the poster was 0.08 to 0.40, which 0.24+/-0.16. The margin was 0.40. The upper CI bound = 0.40. Hence, 0.40=0.40. I believe it needed to be less than the margin, not less than or equal to the margin. Read the guidance.

The fact that 0.24 falls between 0.08 to 0.40 is irrelevant. The mean always falls between its CI bounds J.
HTH

jn