Wang Yaning: Hope to eliminate the misunderstanding of aducanumab approval IBIWS

✔ According to Dr.
Yaning Wang, Bojian’s decision to stop research is wrong
.

The reason is that the two assumptions on which the decision is based are not valid
.

✔ Combining all the data, even if it only depends on the clinical endpoint, there is enough evidence to prove the effectiveness of the drug
.

✔ The Quantitative Pharmacology Review Office of the FDA Clinical Pharmacology Department and the clinical review department have collected data on other monoclonal antibody drugs that also target β amyloid for analysis, clearly revealing the previous failure of β amyloid drugs Reason: The reduction of β amyloid is not enough
.

"I hope that through a more objective evaluation of the data, some misunderstandings about aducanumab approval can be eliminated.
" "In my opinion, they (the experts of the FDA aducanumab review team) are a group of heroes, and they have withstood all kinds of pressures.
, To bring an effective drug to the market in time and benefit millions of patients
.

" At the "Yangtze River Delta Summit: Innovation Strategy for Biomedicine R&D", one of the many forums of Shanghai International Biomedical Industry Week, Wuhan Langlai Dr.
Yaning Wang, CEO of the technology company, made his point at the beginning of the speech
.

Dr.
Yaning Wang resigned from the position of Director of the Quantitative Pharmacology Evaluation Office of the Clinical Pharmacology Evaluation Department of the US FDA Center for Drug Evaluation and Research (CDER) more than a month ago
.

He is very proud of being able to participate in the review of aducanumab, "It is no exaggeration to say that this case is the most influential in my (FDA) 18-year career, and it makes me most proud and proud.
One of the cases
.

"Almost rarely the approval of a drug will cause such controversy
.

In June of this year, after experts from the external advisory committee (AdCom) voted overwhelmingly against the approval, the FDA still decided to grant aducanumab conditional approval, making it a since 2003, the first to be approved for a new type of therapy for Alzheimer's disease
.

since then, the controversy aducanumab approval did not stop
.

shortly after drug approval, and around three FDA Advisory Committee of experts on the central nervous system They expressed their dissatisfaction with the approval decision by announcing their withdrawal from the advisory committee
.

Public Citizen, a non-profit organization, subsequently wrote to the Secretary of HHS in the United States, saying that the review process of the drug "undermined the FDA's standards for reviewing new drugs" and requested the FDA Senior officials resigned immediately
.

On June 22, the FDA, which was in the center of the whirlpool of public opinion, published an 83-page memo detailing the evidence supporting aducanumab approval and the approval process
.
A
week later, it issued a further release including clinical pharmacology ( 147 pages), medicine (392 pages) and statistics (113 pages), a total of more than 650 pages of review documents
.

However, after the documents were made public, the situation has not changed
.

In early July, the drug developer Bojian announced that it would reduce the use of aducanumab , Limited to patients with mild symptoms, which is also the research population of drug clinical trials
.

However, due to concerns about the efficacy of the drug, many well-known large medical centers in the United States chose not to use the drug
.

Under the controversy, the sales of the drug Also very depressed
.

After being approved in early June to the end of the month, aducanumab's sales were only US$1.
6 million, far lower than the previously forecasted US$3.
2 million
.

Why does the approval of a drug cause such controversy? Has the FDA really lowered the usual new drug review standards during the review? Dr.
Yaning Wang gave a detailed explanation on the public's doubts about the review by analyzing the data of the FDA review report
.

Why is the review based on data that has been discontinued? After completing the Phase II study, on the basis of communicating with the FDA, Bojian initiated two large Phase III studies 301 and 302 almost simultaneously in the second half of 2015
.

In March 2019, the results of the interim invalidity analysis showed that the study is unlikely to reach the predetermined study endpoint, so Bojian chose to stop the study
.

But then, in further analysis of the data, Bojian found that the data obtained had a strong signal that the drug might be effective
.

Therefore, in June of that year, the company held a communication meeting with the FDA, and the FDA recommended that Bojian conduct a further detailed analysis of the data
.

Four months later, after seeing the preliminary results, the FDA stated that Biologics should submit all the materials to initiate a formal new biological drug marketing application (Biologics License Application, BLA)
.

According to Dr.
Wang Yaning, in July 2020, Bojian completed the submission of all review materials
.

However, the accelerated review, which was originally scheduled to be completed within eight months, was postponed due to the negative voting results of the expert committee in November 2020 and more information submitted by Bojian after the expert committee
.

During this period, the FDA has done a lot of additional data collection and analysis after fully listening to the feedback from the expert committee
.

After many internal meetings, the FDA did not make a conditional approval decision until June 2021.
Bojian still needs to carry out confirmatory clinical studies after the drug is marketed
.

According to Dr.
Wang Yaning, Bojian’s mid-term invalidity analysis is based on two assumptions: one is that the results of the two studies 301 and 302 are similar, and the other is that the test results of the interim analysis are similar to future results
.

The results found that, with the clinical dementia rating scale (CDR-SB) as the endpoint, with the exception of the high-dose group of the 301 study, the other three groups showed clinical improvement (changes were negative)
.

The "accident" of the data in the 301 study high-dose group directly led to the very low probability of success calculated based on the hypothesis, and Bojian therefore made the decision to stop research and development
.

In Dr.
Wang Yaning's view, Bojian's decision to stop research was wrong
.

The reason is that the two assumptions on which the decision is based are not valid
.

Not only are the results of the two studies 301 and 302 not similar, but as the study continues, the results of the high-dose group of the follow-up study 301 are getting closer to those of other groups, showing a trend of improvement
.

Two contradictory findings? Another key point where the approval of aducanumab has been questioned is that Bojian’s subsequent reanalysis of a larger data set showed that although the 302 study showed an improvement in the condition based on the scale, the results of the 301 study showed that it was comparable to placebo.
Compared with patients receiving high-dose aducanumab treatment, the average condition of the disease did not improve but slightly worsened
.

Bojian’s explanation for this is that the difference in the results of the study is mainly due to the ApoE ε4 positive patients in the high-dose group of the 301 study that only used 6mg/kg instead of 10mg/kg and there were more in the high-dose group.
The "rapidly progressing population"
.

But such an explanation obviously cannot convince the public
.

Dr.
Yaning Wang gave more evidence in his speech
.

First of all, the analysis of the complete data before the suspension of the study in March 2019 shows that although the high-dose group of the 301 study failed to show the disease on the two evaluation indicators of CDR-SB and the Mini Mental State Examination Scale (MMSE) Improved, but it is shown on the cognitive scale (ADAS-Cog13) and the activity of daily living scale (ADCS-ADL-MCI), which have been used in the past as the primary clinical endpoint to approve other Alzheimer’s disease drugs Make positive improvements
.

Moreover, even the poor CDR-SB score, compared with the previous interim analysis results, the complete data set also showed a trend of getting better
.

At the same time, the low-dose group showed a high degree of consistency in the data of more than 301 and 302 clinical endpoints
.

Although it did not reach statistical significance, the test results generally support that the drug slowed the progression of the disease
.

So the results of 301 and 302 are only partially inconsistent
.

On the other hand, Dr.
Wang Yaning introduced that when Bojian designed the 301 and 302 studies (the two studies have the same design), patients with Alzheimer's disease were given different doses based on the mutation of the apolipoprotein E (ApoE) ε4 gene
.

In the high-dose group, the initial dose of ApoE ε4 negative patients can reach 10 mg/kg, and the dose of ApoE ε4 positive patients is 6 mg/kg
.

However, in subsequent studies, Bojian found that patients with ApoE ε4 positive can control adverse events (AEs) while administering higher doses as long as the dose climbing speed is controlled
.

Therefore, about halfway through the study, Bojian carried out an important revision of the research protocol, adjusting the dose of ApoE ε4 positive patients in the high-dose group to a higher level (10mg/kg), which is the fourth version of the protocol.
(PV4)
.

Based on this, the entire study is divided into two important time periods before (Pre-PV4) and after (Post-PV4) revision of the fourth edition of the program
.

The FDA Clinical Pharmacology Review Department analyzed the correlation between the blood drug concentration and clinical efficacy of the patients collected by Bojian in the study.
Both the 301 and 302 studies showed a significant dose-effect relationship.
The medication group was compared with the placebo.
The group had significant disease improvement, and the improvement of the high-dose group was better than that of the low-dose group
.

However, in terms of the degree of improvement, the results of the 301 study are obviously inferior to the 302 study
.

After further analysis according to the two time segments of Pre-PV4 and Post-PV4, it was found that taking the CDR-SB endpoint as an example, the high-dose group in the 301 study did not achieve disease improvement at Pre-PV4, but at Post-PV4 The result is similar to the 302 study
.

Dr.
Yaning Wang also showed the results of quantitative pharmacological analysis on other disease endpoint indicators, and they all showed the same trend
.

In addition to analytical methods based on quantitative pharmacology, Dr.
Yaning Wang also made the same argument from the perspective of statistical analysis methods
.

The failure of the high-dose group in the 301 study was completely driven by the Pre-PV4 data, while the post-PV4 data in the high-dose group in the 301 study were completely consistent with the 302 study
.

If the data of all doses and all time points in all studies are put together for a comprehensive analysis, for the CDR-SB endpoint, the only data in the 13 groups that does not support the efficacy of the drug is the Pre-PV4 data of the high-dose group in the 301 study (red in the figure below) Box), while all other data supports the effectiveness of the drug
.

Dr.
Yaning Wang also showed the analysis results on other disease endpoint indicators, and they all showed the same pattern
.

Why is there such a big difference in the data of Pre-PV4 in the high-dose group of the 301 study? Dr.
Yaning Wang said that, on the one hand, it may be because the first half of the study was too low due to the poor results, and the randomization happened to have the disease progressed quickly, and the more serious patients were assigned to the high-dose group of the 301 study
.

On the other hand, from the perspective of statistical second-class errors (that is, multiple verifications of an effective drug, such as multiple doses, multiple trials, multiple populations, and multiple time points, allow a certain percentage of failures to occur), Dr.
Yaning Wang It is believed that the occurrence of this deviation value can be explained as a typical second-class error
.

Therefore, Dr.
Wang Yaning stated that the different results of the 301 and 302 studies can be explained no matter from a quantitative pharmacology or a statistical point of view
.

In fact, if all the evidence (four clinical endpoints, two doses, and three time points) is considered comprehensively, the consistency of the results of the 301 and 302 studies is far greater than the inconsistency, and the efficacy of the drug can be well demonstrated based on the clinical endpoints
.

Can beta amyloid be used as a surrogate endpoint? According to the FDA, the approval of aducanumab is based on the surrogate end point of reducing beta-amyloid plaques in the brain, and stated that there is "reason to believe that the drug will bring clinical benefits
.
"
The FDA also pointed out that Bojian’s research proved for the first time that reducing these plaques will slow down clinical symptoms
.

However, many previous drugs aimed at lowering β-amyloid protein have all been compromised in clinical development
.

Moreover, although some drugs did show a statistically significant decrease in beta amyloid, they did not achieve improvement in clinical symptoms
.

Therefore, it is widely challenged whether to approve based on the surrogate end point of amyloid beta.

.

At the expert review meeting, Bojian showed the results on the right side of the figure below
.

When looking at the results (changes relative to the placebo group) of all the dosing groups of several studies (301, 302, and Phase II 103 studies), it can be seen that, on the whole, the symptoms of aducanumab improved ( (Assessed by CDR-SB) and SUVr (the ratio of the standard uptake value of β-amyloid protein) show a linear relationship
.

Dr.
Yaning Wang further explained that because not all subjects have imaging data (SUVr), there are only more than 100 people in each subgroup that can be used for analysis (studies 301 and 302), and about 80% of them are patients with Pre-PV4
.

This also explains that in this analysis, the high-dose group of the 301 study has particularly poor data, even inferior to placebo
.

Therefore, Dr.
Wang Yaning said that the above picture clearly reflects that the high-dose group of the 301 study is a set of abnormal results that deviate from normal
.

So why did the previous drugs that also target beta amyloid fail? According to Dr.
Yaning Wang, the Quantitative Pharmacology Review Office of the FDA's Clinical Pharmacology Department cooperated with the clinical review department to collect data on other monoclonal antibody drugs that also target β amyloid for analysis
.

These drugs under research or discontinued must have been carried out in a large randomized double-blind study, lasting more than one year, and have both clinical symptoms and imaging data of amyloid beta
.

Among them, three drugs (see the green box in the figure above) have relatively large improvements based on clinical endpoints
.

In addition to the approved aducanumab, BAN2401 is also a drug jointly developed by Bojian and Eisai.
Another Eli Lilly's donanemab also intends to submit a marketing application to the FDA based on beta amyloid as an alternative endpoint
.

In addition, Roche's gantenerumab (see blue box above) previously announced that two large phase III studies had failed
.

However, the development of this drug has not stopped
.

According to Dr.
Yaning Wang, the company’s further analysis showed that the previous poor clinical results were because the dose of the drug was too low, so that the effect on beta amyloid was not enough.
So next, Roche increased the dose of the drug by 10 times to continue the study
.

The remaining three drugs (see the red box in the figure above) completely terminated the research and development after the phase III study failed
.

Integrating the clinical (CDR-SB) and β-amyloid protein (SUVr) data of these drugs can be seen.
Although these compounds are completely different, the relationship between the β-amyloid protein and the improvement of clinical symptoms is related to these compounds.
Aducanumab is very similar, which is a broad-spectrum quantitative relationship between biomarkers and clinical endpoints that does not depend on specific drugs in the field of quantitative pharmacology
.

Moreover, the three drugs that were discontinued due to poor clinical efficacy and Roche's gantenerumab were significantly inferior to aducanumab and BAN2401 in reducing β amyloid.
Compared with placebo, the reduction in SUVr value of the drug group was less than 0.
1 unit
.

Dr.
Yaning Wang believes that this analysis clearly reveals the reason for the failure of these drugs: the degree of reduction of beta amyloid is not enough
.

302 What is the probability of a false positive in the study? According to Dr.
Yaning Wang, the expert committee in November 2020 was changed to an online meeting due to the new crown epidemic.
The pre-recorded report of the FDA clinical and statistical department was uploaded to the FDA official website for experts and the public to watch and study before the meeting.
The day of the meeting Due to time constraints, the two reports were not scheduled to be broadcast, and the oral report of Billy Dunn, the director of clinical medicine on behalf of the FDA, was interpreted by the outside world as only emphasizing information that is beneficial to the drug
.

As we all know, the recorded FDA clinical department report (supporting approval) and the statistical department report (not supporting approval) are completely opposite opinions
.

Although all the background information published by the expert committee and the recorded reports include the different opinions of the FDA statistics department, the participating experts still believe that the FDA only presents information that is beneficial to the drug, or believes that the results of the 302 study represent the truth (the drug is effective).
Blindly explain why 301 should not have a negative result
.

Therefore, an expert asked the FDA at the meeting, if it is assumed that the 301 study result (the drug is invalid) represents the truth, what is the probability of a 302 positive result
?
To answer this question, the Quantitative Pharmacology Review Office of the FDA's Department of Clinical Pharmacology designed a virtual clinical trial
.

The virtual study extracts data from the placebo group of the aducanumab-related study, and randomly assigns it to the placebo, low-dose and high-dose groups, and observes the probability of a positive result in the 302 study
.

In order to maintain the correlation between the data, the data includes data for all clinical endpoints when there are four follow-up points (0, 26, 50, and 78 weeks)
.

Dr.
Yaning Wang told everyone that combining all the data and considering four clinical endpoints, namely CDR-SB, MMSE, ADAS-Cog13 and ADCS-ADL-MCI, it was finally found that the probability of false positive results in Study 302 was less than 10 million points.
One
.

Even in the 301 study with poor results due to various reasons, the false positive probability is only one in ten thousand, because it still includes about 75% of the positive data
.

Even the 103 study with a small sample size has a false positive probability of only one in ten thousand
.

If the truth is that the drug is ineffective, the probability that the positive data observed in the three clinical trials can be obtained at the same time due to random reasons is almost zero, which further indirectly proves the effectiveness of the drug
.

Does the reduction of the evaluation scale have clinical significance? Some experts believe that the range of CDR-SB is between 0 and 18, so it needs to be improved by at least 1 to 2 units relative to placebo to prove that the drug does cause a clinically significant improvement
.

In response to this statement, Dr.
Yaning Wang believes that these experts did not understand the design of studies 301 and 302 and the conditions of patients with mild symptoms of Alzheimer's disease
.

From the research data, the mild patients in the placebo group, after 78 weeks, the average CDR-SB worsened by only 1.
5 to 1.
7 units
.

Using this as a baseline value, the value of the 302 study high-dose group compared with the placebo group to reduce the deterioration of CDR-SB is 0.
39, which is a 25% improvement relative to the placebo (calculated at 1.
6), which is in line with the improvement planned in the trial.
The proportions are exactly the same
.

The 25% improvement is the consensus reached by the FDA and Bojian in the phase III trial design stage, and it has passed the FDA's highest-level Special Protocol Assessment (SPA)
.

Objections from the FDA Statistics Department Although the FDA’s clinical and clinical pharmacology departments support the approval of aducanumab, the statistics department has always opposed it.
This view, especially the data and reports presented by the FDA statistics team in the external advisory committee’s materials Objectively affect the final decision of external experts
.

So, what are the reasons for the opposition of the statistics department? Dr.
Yaning Wang told everyone that there are three main reasons for the statistical team’s objection: First, the 302 study was a false positive result, which was caused by the deterioration of the placebo group; second, there was no relationship between SUVr and CDR-SB; third, the primary endpoint The improvement of CDR-SB did not reach the statistical significance threshold of p-value<0.
05 in the low-dose group, so the statistically significant results of the other three secondary endpoints at the high-dose should not be considered meaningful
.

"A lot of mistakes have been made here
.

" Dr.
Wang Yaning said, and clarified item by item
.

First of all, the statistical team showed wrong chart data to the external expert team in order to prove that the 302 study was a false positive result
.

The above figure is the result of ApoE ε4 positive patients, and the number marked "overall" at the bottom is the data of all patients with ApoE ε4 positive and negative, and it does not correspond, and the data placement is wrong
.

In the above figure, in the 302 study, comparing the subgroups of the Pre-PV4 and Post-PV4 periods, the CDR-SB value of the placebo group changed from 1.
51 to 1.
75, an increase of 0.
24; while the high-dose group changed from 1.
17 to 1.
25, only With an increase of 0.
08, the gap between the high-dose group and the placebo group becomes larger
.

Therefore, the statistical team believes that the deterioration of the placebo group is the cause of the positive result, so the 302 study is a false positive result
.

But in fact, Dr.
Yaning Wang explained that the statistical team misplaced the data, and the Post-PV4 data (1.
37 and 1.
25) of the low-dose and high-dose groups of the 302 study were inverted
.

The correct data should be 1.
25 for the low-dose group and 1.
37 for the high-dose group
.

After correcting the data, it can be found that the change in the CDR-SB value of the high-dose group and the placebo group from Pre-PV4 to Post-PV4 is in line with randomization, so whether it is Pre-PV4 (1.
17-1.
51=-0.
34) or Post-PV4 In the PV4 stage (1.
37-1.
75=-0.
38), the difference between the high-dose group and the placebo group is the same
.

In order to amplify the signal of "deterioration in the placebo group", the statistical team only selected the results of the subgroup of ApoE ε4 positive patients for mapping
.

In the above figure, from Pre-PV4 to Post-PV4 in Study 302, the CDR-SB value of the placebo group increased by 0.
6 at 78 weeks (pointed by the black arrow in the figure), while the high-dose group did not change at all , Than the “overall” result provides stronger evidence to support the conclusion that “the worsening of the placebo group caused a positive result”
.

However, the quantitative pharmacology team found that the above picture could not be reproduced at all during the data analysis.
The correct picture (the small picture in the lower right corner of the picture below) showed that the CDR-SB value of the high-dose group from Pre-PV4 to Post-PV4 also increased by a considerable amount.
Therefore, the difference between the high-dose group and the placebo group is consistent regardless of whether it is in the Pre-PV4 or Post-PV4 stage, which is in line with the expectations of randomization
.

In the process of communicating the results with the statistical team, I discovered that the above figure is based on data that has not been fully verified in the early stage, rather than the final reliable data that should be used for review
.

The "overall" result (although the Post-PV4 data of the high-dose group and the low-dose group of Study 302 are inverted) is based on the correct data set
.

In other words, the data set used by the graphical result in the above figure and the "overall" result are not consistent
.

Note: APOE=N represents ApoE ε4 negative patients, APOE=Y represents ApoE ε4 positive patients, dose=0 represents the placebo group, dose=1 represents the low-dose group, and dose=2 represents the high-dose group
.

Second, when proving that there is no relationship between the reduction of SUVr and CDR-SB, the statistical team did not choose to analyze the group average based on random grouping, but only used the 301 and 302 studies.
The individual data of the high-dose group was subjected to univariate regression analysis
.

Due to the different conditions of each individual patient, there are multiple confounding factors.
According to Dr.
Yaning Wang, this wrong analysis method leads to the complete disappearance of the linear relationship of the obvious positive correlation or even the negative correlation.

.

Such mistakes are common mistakes that are highly vigilant in the field of quantitative pharmacology to avoid
.

Regarding the last reason put forward by the statistical team, Dr.
Yaning Wang believes that this is a misunderstanding of the FDA's statistical team on Bojian's statistical analysis methods
.

After the expert committee, Bojian submitted additional materials for the misunderstanding of statistical analysis methods by the FDA statistical team, and provided detailed explanations and illustrations on the correct understanding of statistical analysis methods
.

Moreover, the same statistical analysis method has been accepted and used by the FDA statistical team in the past new drug approval process, and there is no misunderstanding
.

For detailed information about the above questioning of the methodology of the FDA statistics team, please refer to the FDA Clinical Pharmacology Review Report (pages 134-143): https://—this article was written by Yaning Wang Thank you for proofreading and reviewing by the doctor himself! —Edit Layout | Mr.
Yao Jia No.
1430 Visit the developer website to browse more articles