No Doubt that

Genfit's Drug cures Nash
On the basis of the analysis of data from clinical trials, Genfit has consistently argued that GFT505,
now named Elafibranor, shows good efficacy to cure Nash. However, not all investors or analysts
take the time to research the documents to build up a complete picture of the elements of doubt and
conviction that help to make the case for a balanced opinion.
So let's look at the some of these elements that contribute to doubt or certitude, not only on the basis
of statistical analysis but also on the basis of human conception of doubt trust, confidence and
conviction.
Reliable and trustworthy data
In this attempt to define areas of doubt and conviction, I refer to a poster published at the AASLD
conference in November 2015, signed by 17 authors from 14 different institutions and presented for
discussion by the leading experts in liver diseases. This considerably increases the trust we can have
in the data. Some of this data is also given in Genfit's presentation for Investors and is certainly
easier to read ( http://www.genfit.com/wp-content/uploads/2015/11/20151116AASLD_Investor_Event.pdf )
Histological scores versus Biochemical tests
In the upcoming Phase III trial for Elafibranor, the primary outcome criterion is the resolution of
NASH without worsening of fibrosis, based solely on histological measurements of liver biopsies in
Elafibranor-treated patients versus placebo. The new consensual definition of the reversion of
NASH is a score of 0 for ballooning (on a scale of 0-2) and 0-1 for inflammation (on a scale of 0-3).
Steatosis is no longer used to define the reversion of NASH, but remains a parameter in selecting
patients with a score of at least 1 in steatosis (on a scale of 0-3). The previous phase IIB trial data
have been analysed on the basis of these new criteria and compared to extensive biochemical tests
from the same patients.
The aim of this article is to define if and where these sources of data converge or diverge and hence
establish where lies the conviction or doubt. I will demonstrate that the biochemical data is more
relevant, more consistent, more precise and more reliable than the histological data in measuring the
beneficial effects of Elafibranor in the treatment of NASH which lead to its reversion.
Relevance
Liver biopsies are images of a stained slice of liver tissue at a given time. Since they are expensive
and carry a health risk, they are taken at the beginning and end of a large time interval in order to
identify changes in the physical state of the liver. They provide no information on the functional
state of the liver. Since the liver is an organ whose job is to clean up toxins and store, convert and
transfer sugars and lipids, its physical appearance is less relevant to the health of the patient than its
capacity to function. Furthermore its physical state develops as a result of its incapacity to manage
an overload of sugars and lipids over a very long period of time.
As with many chronic diseases, changes in physical appearance lag behind changes in functional
capacity. NASH typically takes 15-20 years to develop during which time the functional capacity of
the liver slowly declines. The reversion of NASH can only begin after the functional state starts to
improve. The biopsy is lagging indicator of the progression of NASH whatever the direction of
progression. Furthermore, the choice of “Reversion”, which is considered to be equivalent to “cure”
is an odd choice for a chronic disease. For patients who suffer from a chronic disease, what matters
is whether or not they are “getting better”. A patient whose treatment improves his health is usually
quite satisfied with his progress. No one expects chronically ill patients to be cured by taking a pill.

So is it relevant to use a cure test to see if a drug works on NASH ? What the patient wants to know
is if the treatment is making him better.
Biochemical tests give information on the functional state of organs and on the overall health of the
patient. NASH patients don't only have NASH. Many are also dyslipidemic and insulino-resistant.
They are at high risk of developing diabetes and cardiovascular diseases. Indeed, NASH patients are
at higher risk of dying from cardiovascular events than from liver disease. For the patient's overall
health, treating NASH also means reducing the risk of cardiovascular disease and diabetes. The
Chairman of Genfit repeated that in an interview this morning (4 December 2015)
http://ow.ly/Vt5AV . Liver biopsies give no information on these which is also why Genfit has
developed its bio-marker programme. The biochemical test reviewed in this data give information
on the real-time progression of liver enzymes, diabetes, and cardiovascular risk. This is relevant to
the NASH patient. These are the tests that practitioners will apply to NASH patients to check
whether they are getting better or worse.
Precision
The measurement scale of biopsies for the phase III trials to check the progress of NASH is crude
and misleading. Requiring a score of 0 for ballooning is an on/off result that takes no account of the
starting point. Requiring a score of 0 or 1 for inflammation is a cut-off that ignores any partial
progression.
The problem with biopsy data is that it is very susceptible to sampling variation. A typical biopsy
represents 1/50 000 of the volume of the liver and NASH varies depending on the sampling site. A
study designed to estimate sampling variation by taking two biopsies from the same patient, carried
out by V Ratziu, one of the leading experts on NASH, entitled « Sampling variability of liver
biopsy in nonalcoholic fatty liver disease » drew the following conclusions :
- Histological lesions of NASH are unevenly distributed throughout the liver parenchyma;
therefore, sampling error of liver biopsy can result in substantial misdiagnosis and staging
inaccuracies.
- Overall, the discordance rate for the presence of hepatocyte ballooning was 18%, and
ballooning would have been missed in 24% of patients had only 1 biopsy been performed.
- The negative predictive value of a single biopsy for the diagnosis of NASH was at best 0.74.
- Discordance of 1 stage or more was 41%.
- Six of 17 patients with bridging fibrosis (35%) on 1 sample had only mild or no fibrosis on
the other and therefore could have been under staged with only 1 biopsy.
Since 41% of biopsies could be in error by at least one NASH stage due to sampling variation, it is
not surprising that many low NASH patients treated with placebo were recorded as cured when
second biopsies were taken at the end of the Genfit trial. These patients were in no way cured.
Indeed all the plasma tests revealed that their condition was unchanged. These results were
misleading since these patients were still just as much at risk from diabetes and CVD events.
Let's look at the statistical issues that this raises. If we take biopsies from a large group of patients
with an average score of NASH of 4,5, you will get a spread of results from say 2 to 6, each patient
with his proper score. If you repeat this with a new biopsy on the same group of patients with a
double blind analysis you will come close to the same average score of 4,5 but because of the errors
of sampling, some patients will have a different score. Patient A could have changed from 3 to 2,

patient B could have changed from 2 to 3. Patient C could have changed from 4 to 2. Does this
mean that Patient B has got worse and Patient C has been cured ? Of course not, this is just a result
of the sampling error of the technique.
To get a clear picture of whether a drug is effective in treating NASH you have to look at the whole
spread of the scores before and after treatment irrespective of the response of individual patients. An
effective drug will cause the whole envelope of results to move to lower levels. The use of a cut-off
point which designates individual patients as being cured as in the original phase IIB analysis (NAS
>3) is bound to lead to distortion of the results. It does not take into account that the results of other
individual patients will have recorded higher levels. This was very clearly demonstrated on the
results for patients on placebo (see below).
Biochemical blood tests are far more precise than biopsies. They are measured on precise extended
scales and the patient variations and measurement errors are known to the analysts. There is no need
to make a choice between two integer values. Interpretation can be made in terms of a trend towards
higher or lower levels and correlated with other biochemical tests known to be related. Furthermore
such tests can be repeated at short intervals to confirm any doubtful results. One may argue that
they are only indirectly related to the state of NASH, but they are directly related to the patient's
health and the risk of complication. Small changes are often meaningful and send a clear message to
the practitioner.
Convergence and divergence
The document below “THE HEPATIC AND EXTRA-HEPATIC PROFILE OF RESOLUTION OF
STEATOHEPATITIS INDUCED BY GFT-505 (ELAFIBRANOR) » was presented at the AASLD
conference in November 2015. It is signed by 17 authors from 14 different institutions and
presented for discussion by the leading experts in liver diseases. It is also published on the Genfit
web site on the Elafibranor page.
This poster shows results comparing
histological data and biochemical data
taken during the Genfit phase IIB trial.
Columns 2 and 3 present these results
for two groups of patients, so-called
“Responders” that passed the new
histological cut-off for reversion of
NASH and Non-Responders that failed
to meet the cut-off histological criteria.
For the Responder group, we have
excellent convergence between
histological results and all the
biochemical data for liver enzymes,
plasma lipids, glucose homeostasis and markers of inflammation (16 biological measurements). The
biochemical markers clearly show that the patients' liver function has very significantly improved in
parallel with the improvement in the physical appearance of the liver biopsies. This is accompanied
by improvements in plasma lipids and insulin resistance. The histological data and the biochemical
data are consistent and convergent for this group of patients and clearly indicate that Elafibranor has
efficacy in the treatment of NASH.
For the Non-responder patients, the data shows low convergence between histology and

biochemical tests. There is however high consistency between all four groups of biological tests:
liver enzymes, plasma lipids, glucose homeostasis and markers of inflammation, most of which
record between 40 and 70% improvement compared to the responder group, whereas the
histological test show no significant improvement. This gives credence to significant improvement
in the overall health of the patients in this group and a consequent reduction of risk from Nash
related diseases, diabetes and cardiovascular disease.
What can we deduce from this low convergence between histology and biochemical tests ?
For these patients, the low relevance of the histological tests in terms of the cut-off level after 52
weeks of treatment could be the cause. Whereas the biological tests are real-time, precise
measurements of relevance to the patients' health, the histological measurements are time-lagged
and subject to a cut-off that may not be sensitive the true development of the disease.
Now let's look at column 4 which looks at the convergence/divergence between Responders treated
with Elafibranor and Responders on placebo.
Once again, the 16 biological measurements are compared in the two groups. In the placebo group,
13% passed the criteria for reversion of NASH according to their histological readouts compared to
22% on Elafibranor. On the basis of histological data, both of these groups are considered to be
cured, either by simply following the restrictions imposed by the clinical trial (diet and exercise) or
as a result of the treatment with Elafibranor.
The biochemical data fails to support this conclusion. There is in fact total divergence between
Placebo responders and Elafibranor responders. Placebo responders show no significant change in
the four conditions measured by the biochemical tests compared to the beginning of the trial. These
patients are just as much at risk of progression of their NASH and remain at high risk for diabetes
and cardiovascular disease. The positive placebo responders are in fact false positives.
This observation concerning the placebo responder group is extremely important. It confirms that
the histological method used to measure the efficacy of Elafibranor carries a high degree of doubt
and the cut-off criteria to be used in the phase III trial may seriously bias the results. These criteria
create a systematic placebo trap that produces false positive results.
No Doubt that Elafibranor cures NASH, but there's real doubt about the histological method.
These results show that there is no doubt that patients treated by Elafibranor whose histological
tests met the trial criteria were effectively cured of NASH over 52 weeks. All the tests converge to
this conclusion.
For hose patients on Elafibranor who did not meet the trial criteria, there low convergence between
the histological data and the biochemical data. This raises doubts about the relevance of the
histological method and cut-off criteria used in the trial.
.
The false positives of the placebo group clearly discredit the histological method and criteria used.
If these same criteria are to be used for the Phase III trials, then Genfit must take into account the
weakness of the histological methods used when selecting patients so as to avoid the systematic
placebo trap. A review of these criteria to include biochemical tests as part of the primary outcome
should be considered.
Albert Wright PhD,
Revised 15 December 2015