here are 2 pieces by j. birchenough on HGSI: the first after yesterday's results; the second before yesterday's results,explaining why the trial will likely fail. the second is interesting since it discusses the differences between the mechanism of action in rituxan and benlysta.
any thoughts on the price of HGSI if bliss-76 does not achieve statistical significance in november?
thanks.
sales
Equity Research
July 20, 2009
Human Genome Sciences (HGSI - US$ 3.32) 2-Equal Weight
Change of Price Target
BLISS52 Positive - BLISS76 Risk Remains
Investment Conclusion
We are increasing our price target on HGSI to $7
but maintaining our 2-Equal Weight rating
following release of phase III data for Benlysta in
Lupus (SLE). While results from the first BLISS-52
study are positive and HGSI has done a good job
to optimize study design, we see little margin for
error in BLISS-76 and would advise caution given
potential patient heterogeneity.
Summary
?? HGSI reported positive top-line phase III results
for BLyS antibody Benlysta in patients with SLE.
Results from the 865 pt BLISS-52 study suggest a
statistically significant improvement in the primary
endpoint of composite SLE response index (SRI)
at 57.6% for Benlysta 10mg/kg vs 43.6% for
placebo. Improvement in SRI components were
also observed with SELENA-SLEDAI 4 point
reduction in 58.3% vs 46% for placebo.
?? With BLA filing now contingent on confirmatory
data from BLISS-76 in November, we believe that
a placebo response approaching 50%, differences
in study geography and SLE patient heterogeneity
make it difficult to extrapolate results across
studies. With 300,000 Lupus patients in the US we
estimate 5% with poor disease control and peak
global sales of $600M and incremental NPV of $6.
Adjustment to Valuation Methodology
In arriving at our new $7 price target we have employed an NPV methodology as opposed to our prior 2.5x 2010 revenue multiple approach.
With longer-term cash flows now possible with Benlysta, we believe that an NPV methodology is more appropriate. Our NPV estimate of $7
assumes $1/share in cash and $6/share for Benlysta based on peak global sales of $600M, 50:50 profit split and 70% operating margins.
July 06, 2009
Human Genome Sciences (HGSI - US$ 2.92) 2-Equal Weight
Company Update
Benlysta Likely To Fail BLISS Studies
Investment Conclusion
We are reiterating our 2-Equal Weight rating on
shares of HGSI ahead BLISS-52 data for BlyS
antibody Benlysta in patients with SLE. With prior
phase II failure, questionable retrospective subset
analysis and with modest biologic effects we
believe likelihood of success is low. Upside
potential to $7 exceeds downside risk to $1 but
with 85% likelihood of failure we see fair value at
$2 and would advise caution at current levels.
Summary
?? HGSI has scheduled a call on Monday, July 20 to
release phase III BLISS-52 results for Benlysta in
patients with systemic lupus erythrematosis (SLE).
Prior phase II data failed to meet a 1ary endpoint
of SELENA-SLEDAI improvement but HGSI has
suggested a benefit in ANA sero-positive patients
with a 50% improvement in a retrospectively
defined composite endpoint of SELENA-SLEDAI,
BILAG and PGA improvements.
?? While HGSI has sought to optimize BLISS trial
design we believe phase II data suggests a weak
biologic effect, incomplete dose range,
inconsistent effects on key B-cell populations
delayed effects on autoantibodies beyond 52 wks.
Expert feedback suggests a $600M opportunity
globally with NPV estimate of $825M or $6/share.
United States of America
Healthcare
Biotechnology
Reuters HGSI
Bloomberg HGSI
ADR
EPS (US$) (FY Dec)
2008 2009 2010 % Change
Actual Old New St. Est. Old New St. Est. 2009 2010
1Q -0.35A 0.85A 0.85A 0.85A N/A N/A -0.44E 343% N/A
2Q -0.55A -0.20E -0.20E -0.33E N/A N/A -0.39E 64% N/A
3Q -0.50A -0.14E -0.14E -0.39E N/A N/A -0.22E 72% N/A
4Q -0.41A -0.15E -0.15E -0.31E N/A N/A 0.07E 63% N/A
Year -1.81A 0.36E 0.36E -0.09E N/A N/A -1.16E 120% N/A
P/E 8.1 N/A
Market Data
Market Cap (Mil.) 397
Dividend Yield 0.00
52 Week Range 8.00 - 0.45
Financial Summary
Revenue TTM (Mil.) 213.4
Stock Overview
HUMAN GENOME SCIENCES - 7/ 2/ 2009
Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun
Source: LehmanLive
0.75
2.25
3.75
5.25
6.75
8.25
Volume
0
20M
Stock Rating Target Price
New: 2-Equal Weight New: US$ 2.00
Old: 2-Equal Weight Old: US$ 2.00
Sector View: 1-Positive
??
Equity Research
2
Investment Thesis
We believe that the BLISS-52 and -76 trials are likely to fail and with questionable prospects for other programs like Albuferon in hepatitis C
we see downside risk to $1/share. In the unlikely event of Benlysta success we estimate peak sales of $600M globally, an incremental NPV
of $825M or $6/share and upside potential to $7/share. While the absolute magnitude of upside potential outweighs the magnitude of
downside risk we estimate a probability of success of only 15% and believe that current valuation slightly over-estimates the risk-reward
proposition heading into BLISS-52 data on July 20.
Overall we believe that failure of Benlysta to meet the SELENA-SLEDAI primary endpoint in phase II, lack of statistical significance in
responder analysis overall, retrospective nature of subset analysis in sero-positive patients, uncertain biologic implications of BlyS inhibition
and recent failures of more potent anti-B cell therapies support a more skeptical view towards likelihood of phase III success in BLISS-52
and -76 studies. HGSI has been diligent in pursuing its phase II hypothesis and in attempting to optimize a phase III patient population most
apt to respond to Benlysta. We believe, however, that the hypothesis is flawed, that patient heterogeneity will confound efforts to optimize
for Benlysta responders and that the hurdle in phase III will prove too onerous.
Overview of Systemic Lupus Erythematosis (SLE)
Systemic Lupus Erythematosis (SLE) is a systemic autoimmune disease that causes significant morbidity across many organ systems. The
absence of a single target organ and the presence of numerous types of anti-nuclear and double stranded antibodies differentiate lupus
from other autoimmune diseases. The ability of these auto-antibodies to initiate organ damage is thought to be central to the morbidity
associated with this disease. Lupus disease flares are thought to be associated with high circulating levels of pathogenic auto-antibodies
while disease remission is associated with their reduction. These auto-antibodies target self antigens, in the case of SLE specifically, nucleic
acids, and are produced by antibody secreting plasma cells that have escaped the normal deletion process responsible for removing such
cells in healthy individuals. Given that nucleic acid is ubiquitous and present in every cell, auto-antibodies targeting nucleic acids have the
potential to target any organ. The formation of immune complexes between these auto-antibodies and other immune system components
such as complement presents a focus for immune-mediated organ damage. Immune complexes consume complement and thus low levels
of complement are also proposed to be a feature of active SLE.
All antibodies are secreted by plasma cells which represent the terminally differentiated state of the B cell lineage. B cell ontogeny is very
complex and involves numerous steps as cells differentiate from stem cells in the bone marrow and undergo differentiation and education in
peripheral lymphoid organs. During this process the presence or absence of various cell surface markers, the so called cluster of
differentiation (CD) antigens, can be used to categorize the numerous stages of B cell development and differentiation. The most widely
known B cell marker is CD20 as this is the target of the antibody Rituxan. As depicted below, CD20 appears on the cell surface as B cell
progenitors exit the pro-B cell stage to become pre-B cells and remains on the cell surface as cells become either memory B cells or
plasmacytoid B cells.
Illustration of B-Cell Differentiation and B-Cell Inhibition
CD20 +ve CD20 -ve CD38 BRIGHT
Stem
Cell
Pro/Pre
B cell
Immature/
Mature B cell Memory
B cell
Plasmacytoid
B cell
Plasma
Cell
SLE plasma
Cell
Benlysta
Activity
Rituxan
Activity
Source PubMed, Company Documents and Barclays Capital Research
Equity Research
3
As depicted above, Benlysta is designed to target a different range of junctions along the B cell development pathway. Specifically, Benlysta
is designed to prevent BlyS from binding to the B cell receptor which is present from the immature B cell to certain plasma cell. Unlike CD20
which has no known function, BlyS is one of the survival factors for B cells and the absence of BlyS is thought to lead to a reduction in B cell
activity over time and ultimately in the absence of salvage pathways, B cell apoptosis. While CD20 has no known function, binding Rituxan
to CD20 engages immunologic machinery that targets the B cell for destruction. Thus one observes that Rituxan rapidly reduces the number
of circulating CD20+ve B cells, but has no immediate affect on plasma cell numbers and thus autoantibody production. Longer term, Rituxan
strangles the production of new plasma cells, but given the longevity of plasma cell survival, autoantibody levels may remain elevated for
many months to years. In contrast, Benlysta is designed to reduce B cell and plasma cell activation states on a more gradual manner
leading to a reduction in certain plasma cells and various autoantibody levels over time depending on the level of addiction to BlyS.
Monitoring lupus disease activity has been a challenge for investigators and clinicians alike and unlike many other autoimmune diseases
there is not one or even two standard evaluations of disease activity. Lupus disease activity indices (DAIs) that have been used in the clinic
include BILAG, SELENA-SLEDAI, SLAM and SLICC, however the first two indices have become the most commonly used tools for recent
clinical trials. HGSI previously elected to use the SELENA-SLEDAI (SS) disease activity indexin phase II and has maintained the measure
as a component of its composite SLE response index (SRI) in phase III. SELENA-SLEDAI (SS) is a cumulative numerical system, assigning
points to certain physical and serological disease attributes over a range of 0-105 points, with higher scores representing more active
disease. The key goal for any lupus therapy is to reduce the incidence of flares which can lead to permanent end-organ damage. Petri has
suggested that an increase in the SS of 3 points reflects a disease flare, while Fortin has suggested that flare be declared if the intensity of
lupus therapy is increased, which using SS, translates to a 6-point change. Flare is also a binary outcome and investigators have sought to
provide additional detail to characterize lupus disease activity. Relevant to the BLISS trials is the work of Gladman and in particular a paper
published in the Journal of Rheumatology, 27: 377-379, 2000 correlating clinically defined changes in disease activity with SS scores in
lupus. A retrospective analysis of charts was undertaken in this study from 230 lupus patients assigning a score for the level of disease
activity for over 1500 visits correlating this to a computer generated SS score (see graph below).
Cumulative clinical activity and median SS score of clinic visits.
694
323
144
220
162
2
4
6
8 8
0
100
200
300
400
500
600
700
800
0- No activity 1- Mild activity
with no Rx
2- Active dx, but
improvement
3- Refractory dx 4- Flare
0
1
2
3
4
5
6
7
8
9
Number SSScore
Source. Gladman et al and Barclays Capital Research
Across 1500 visits in the Gladman study, 45% of patients were categorized as having no clinical disease activity and this correlated with a
median SS score of 2. Mild disease not requiring therapeutic intervention was observed in 20% of patient visits with a corresponding
median SS score of 4. Therapeutic intervention was required for 35% of visits, and in 10% of the time this reflected an increase in therapy
from the prior period due to a flare and a further 10% of the time an improvement in disease activity was noted while the patient was
receiving therapy. In 14% of visits no change in disease activity was noted despite therapeutic intervention. The median SS scores for
refractory patients and those with flares was 8, while an improvement in disease activity reduced SS score to 6. Gladman also evaluated
which organs systems contributed to the varying levels of disease activity. The graph below shows the distribution of affected organs and
serological abnormalities according to SS score. Feedback from our expert consultants suggests a lower 5% annual rate of disease flare
requiring more potent disease modifying drugs, and we would suggest that data from expert centers may actually be biased towards more
severely affected patients to begin with, perhaps over-stating the true percent of the total SLE population in need of newer therapies.
Equity Research
4
Cumulative percentage of organ specific symptoms and serological abnormalities according to SS score
0
50
100
150
200
250
SS2 SS4 SS 6 SS8 SS 10
DNA auto abs Complement Arthritis Alopecia Rash Renal CNS Vaculitis
Source. Gladman et al and Barclays Capital Research
Based on these data auto-antibody and complement abnormalities appear consistent and unrelated to disease activity. According to the
Gladman data, these serological abnormalities contribute: 60-80% of the total SS score in the none-to-mild disease activity level, 60% of the
SS score in patients with a flare and 50% of the SS score in patients that require therapy but did not suffer a flare. Alopecia and rash appear
to be limited to patients with low levels of disease activity. Conversely, CNS and vasculitis are restricted to patients with more severe
disease. Arthritis and renal disease are persistent across all levels of disease activity but appear to be more common with more serious
disease. Based on this analysis, Gladman determined that a flare was associated with a SS score increase of >3 points and that significant
improvement in disease activity is associated with a decrease of >3 points.
Systemic lupus erythematosus (SLE) is a chronic waxing and waning autoimmune disease which predominantly afflicts younger women
(late teens to early 40s) and those of African ancestry. Although it is commonly recognized by skin abnormalities (e.g., malar butterfly rash
on the face), lupus is a multi-organ disease. Patients can develop manifestations across multiple organ systems, including arthritis, skin
rashes, decreased renal function, mild mental dysfunction, seizures, blood clotting disorders, and cardiopulmonary effusions. Different
patients may have dominant pathology in different organs (e.g., lupus nephritis is a dominant form in which patients have substantial kidney
disease). A notable feature of lupus is alternating periods of quiescence and symptom flaring. Durable remissions, whether occurring
naturally or because of therapy, are rare, and many patients suffer lifelong morbidity. Lupus is rarely fatal as the 5-year survival is greater
than 90% in patients with SLE and 80% for patients with lupus nephritis (LN). While multi-organ system disease is common place we would
note that over half of deaths are due to cardiovascular and cerebrovascular disease.
As suggested previously Lupus is characterized by multiple immunologic abnormalities, including autoantibodies, hypergammaglobulinemia,
decreased serum complement, and increased levels of immune complexes in the circulation. Production of pathogenic antibodies and
deposition of immune complexes in tissues (e.g., renal glomeruli, blood vessel walls) are key features of lupus pathology. Renal biopsy
specimens have shown that glomerular deposits consist of immunoglobulin, complement, and DNA. The precise etiology of lupus is
unclear, however, autoantibodies are present on average 2.7 years before onset of disease symptoms. A hallmark feature of lupus is the
presence of antibodies to double-stranded DNA and there are several other auto-antibodies that have been identified, including antinuclear
(95% incidence), anti-DNA (70%), antihistone (70%), antiRNP (40%), antiRo (30%), antiLa (10%), antiSm (30%), and antiphospholipid
antibodies. Examination of the U.S. Department of Defense serum repository reveals that antiSM and antiRNP antibodies appear shortly
before diagnosis, suggesting increases of autoimmune activity to a level sufficient to cause clinical signs and symptoms.
Overview of Humanized Anti-BlyS Antibody Benlysta
Benlysta (belimumab) is a human monoclonal antibody that specifically recognizes and inhibits the biological activity of B-lymphocyte
stimulator, or BlyS. Elevated levels of BlyS were discovered by HGSI in patients with lupus, rheumatoid arthritis (RA) and certain other
autoimmune diseases. Biologically, elevated levels of BLyS are believed to contribute to the over production of auto-antibodies by B cells
that lead to an attack and subsequent damage to the body’s own healthy tissues. B cell targeted therapies have been an area of intense
research focus, and the subject of numerous positive publications albeit from non-randomized case series. Data from randomized studies
have not supported these open label observations however: Roche/Genentech’s B cell clearing antibody, Rituxan, failed in both lupus and
lupus nephritis phase III studies; La Jolla Pharmaceuticals’ Riquent targeted a specific subset of B-cells for depletion in lupus nephritis but
the phase III ASPEN trial was terminated for futility. Safety has also been a problem as a phase II/III study of ZGEN/Merck-Serono’s
BlyS/APRIL neutralizing antibody, atacicept, was halted due to a high rate of infections when combined with Cellcept in lupus nephritis.
Equity Research
5
HGSI and partner GSK are completing two phase III studies of Benlysta in patients with systemic lupus erythematosus (SLE). Each study
enrolled >800 patients and evaluates two dose levels of Benlysta, 1mg/kg and 10 mg/kg, compared to placebo. Benlysta is dosed
intravenously (IV) on days 0, 14, 28 and then every 28 days for the balance of each study. The studies, designated BLISS-52 and BLISS-
76, will follow patients in a randomized fashion for 52 and 76 weeks respectively, but in each case the primary endpoint is response rate at
week 52 compared to day 0. Based on prior phase II results, enrolment in BLISS-52 and BLISS-76 was restricted to sero-positive patients
with active disease. The BLISS-76 study recruited patients in North America, Western and Central Europe, while BLISS-52 recruited
patients in Western and Central Europe, Latin America and Asia Pacific.
The primary endpoint of each BLISS phase III study is a novel composite response, the SLE Responder Index (SRI) defined by
achievement at week 52 compared to day 0 of:
1) >4 point improvement in SELENA/ SLEDAI disease activity index;
2) absence of BILAG organ domain score worsening categorized as a new BILAG A and no more than 2 new BILAG B and
3) no worsening of physician’s global assessment (PGA) by a score of 0.3 or more.
To qualify as a responder in BLISS-52 and BLISS-76 all three criteria must be met. Secondary endpoints in the trials include percent of
patients with prednisone dose reduction between weeks 40-52, measures of fatigue and Health Survey Component Summary Score as well
as response rate at 76 weeks in BLISS-76. Finally, the studies have been designed with 90% power to detect a 40% relative improvement
in response rate at 52 weeks, which is in line with post hoc analysis of earlier phase II results.
Previously reported phase II data for Benlysta, demonstrated modest efficacy trends overall with more significant results in the subset of
patients defined as sero-positive at baseline. The phase II study itself was a 449 patient randomized, placebo controlled study, designed to
evaluate multiple Benlysta doses over 24-52 weeks of follow-up with co-primary endpoints of 1) improvement in SELENA SLEDAI (SS)
score at week 24 and 2) increase in time to renal flare over 52 weeks. Patients enrolled in the phase II study were required to have a
baseline disease activity score >4 and indeed averaged an SS score of 10 at baseline. While neither of the co-primary endpoints in the
phase II study was met with statistical significance, evaluation of 72% of patients with ANA sero-positivity did reveal potential benefits in this
subset of patients with more active disease. In particular, results at 52 weeks in sero-positive patients demonstrated 1) 49% response rate
in SELENA SLEDAI vs. 39% with placebo (NS), 2) 90% of patients with no worsening of PGA vs. 77% with placebo and 3) BILAG, no new
A or 2B were observed in 81% and 91% of placebo and Benlysta treated patients respectively.
Phase II Benlysta study Demographics: Updates from EULAR 2009
Baseline Demographics
n 449
Female Sex 93% Selena-Sledai Score, mean 9.6
Age, mean 42 PGA mean 1.5
Race BILAG >/=A or 2B 68%
White 70% Positive autoantibodies
Black 24% ANA >1:80 72%
Hispanic/Latino 19% Anti-dsDNA >30 IU/mL 50%
Dx duration, mean 8.8yr Anti-RNP 47%
Prednisone Anti-Ro/SSA 39%
all 69% Anti-cardiolipin 28%
>7.5mg 35% Anti-Sm 19%
Immunosuppresant use 50% Complement C4<LLN (16mg/dL) 41%
Source: EULAR and Barclays Capital Research
As noted previously, 72% of patients in phase 2 were defined as being seropositive at baseline. An exploratory subset analysis showed that
clinical response as defined by the SRI was more pronounced in the Benlysta treated cohort than the placebo-treated cohort. A not
unreasonable assumption is that the seropositive subset has more active disease. In evaluating this assumption, HGSI noted that more
seropositive patients were taking steroids at baseline ~ 70% vs ~60% for seronegative patients, further, a more significant difference was
observed with respect to patients taking a high dose of steroids, defined as >7.5mg prednisone/day, ~40% vs ~20%. Low complement C4 is
also associated with active disease and this held for the seropositive vs negative subset analysis, 16.9mg/dL vs 24.8mg/dL. In terms of
affected organ systems, there was significantly more renal involvement, 34% vs 19%, hematologic involvement 59% vs 33% and
immunologic involvement, 84% vs 43%, observed in the seropositive vs seronegative subsets. No differences were observed between the
two cohorts for the incidence of malar rash, discoid rash, photosensitivity, oral ulcers, arthritis, serositis and neurological disorders. Finally
using a cut-off of 9% above the limit of detection, 51% of seropositive patients met the criterion for excess BlyS compared to 24% of the
seronegative group.
During the first year of dosing in this phase 2 study, patients were randomized 1:1:1:1 to one of three doses of Benlysta (1mg/kg, 4mg/kg or
10mg/kg) or a matching placebo. Between weeks 52 and 208, patients on the placebo arm switched to 10mg/kg Benlysta while patients
originally randomized to Benlysta continued on their assigned dose for a further 24 weeks before switching to the 10mg/kg dose at week 76.
Equity Research
6
Of the 449 patients originally enrolled into the study, 345 (77%) entered the 24 week follow-up with 296 (66%) entering the long term followup.
At EULAR 09, HGSI reported that 214 patients (72%) who entered the long-term dosing study remained in the study. Overall 47% of
patients originally enrolled in the randomized portion of the study remained on the study through 4 years introducing a potential selection
bias to results beyond 52 weeks with patients electing to complete the study representing those most likely to respond.
Efficacy analyses have largely been restricted to the 72% subset of ANA positive patients with a focus on the SRI and its 3 components. At
week 52, the absolute difference in SRI response rate between the Benlysta arm and placebo was 15%, 46% vs 31% p=0.0056. As has
been previously reported, SRI response rates for patients in the placebo arm tracked those in the Benlysta arm for the first few months of
the study before plateauing at about 30%. In contrast, improvement in response rates continue to accrue to the Benlysta arm through 52
weeks by which time the response rate is 46%. The Benlysta response rate continues to increase during the 24 week follow-up period
(weeks 52-76) where patients continue at their randomized Benlysta dose before reaching a plateau of ~55%, which is subsequently
maintained through year 4. We would note however that only 77% of patients (345 of 449) elected to continue in the 24 week follow-up and
only 48% (214 of 449) of patients remained on study through 4 years, introducing the strong possibility of selection bias.
After week 52, placebo patients switched to Benlysta 10mg/kg and an inflection in response rate was noted. The slope of response rate
improvement in the placebo switch group between weeks 52-76 is very similar to that observed in the early period of the Benlysta group.
This robust increase in response rate allows the placebo-switch group to intersect with the Benlysta group at week 148. Between weeks 148
and 208, the response rate in the placebo-switch group decreased notably, in contrast to the Benlysta group. This decrease in SRI is
probably due to small patient numbers since the placebo-switch group represented 25% of the original enrollees. While the cross-over effect
between week 52 to week 76 would appear to confirm Benlysta activity we would note the same potential for selection bias beyond 76
weeks of dosing.
Analyzing the Benlysta response to individual components of the SRI shows that the weakest component is clearly the SELENA-SLEDAI.
• A SELENA-SLEDAI decrease of 4 or more points was observed in 49% of Benlysta treated patients and 39% in placebo
(p=0.1169);
• BILAG, no new A or 2B flares were observed in 81% and 91% of placebo and Benlysta treated patients respectively (p=0.0152);
• PGA, Over 90% of patients treated with Lymphostat had less than a 0.3 point decrease in PGA compared to 77% for placebo
(p=0.0027).
The most sensitive SRI component is Physician Global Assessment (PGA) where the magnitude of difference between placebo and
Benlysta is the greatest in absolute terms and where the difference becomes statistically significant by week 4. Over the first 52 weeks,
PGA improved on average by ~15% in the placebo group reaching a plateau after week 16. In contrast, the Benlysta group benefits from a
continuous linear change in PGA reaching an average 35% improvement by week 52 and continuing to a 45% improvement by week 76. As
with SELENA-SLEDAI, switching from placebo to Benlysta resulted in an improvement in PGA, but unlike the SELENA-SLEDAI, the
placebo-switch group did not achieve the same magnitude of benefit as seen with Benlysta patients during the randomized portion of the
study.
Flares are graphed below and include all flares and severe flares using the SELENA-SLEDAI flare index as well as the SRI-relevant new
BILAG A or 2B flares, for Benlysta-treated patients. During the first 12 months of the study, these flare rates can be compared to those
observed on the placebo arm, data not shown below, and suggest a lower rate of flares in favor of Benlysta. Over months 0-6 and 6-12,
new BILAG flares were noted in 33% and 30% and 25% and 23% of placebo and Benlysta treated patients respectively.
Distribution of SELENA-SLEDAI and BILAG Flares in Phase II
0
10
20
30
40
50
60
70
80
Selena-Sledai Flares Severe Flares New BILAG A/ 2B
0-6mo 6-12mo 12-18mo 18-24mo 24-30mo 30-36mo 36-42mo 42-48mo
Source: EULAR 2009 and Barclays Capital Research
Equity Research
7
Validity of ANA+ Sero-Positive Subset Analysis?
In identifying a sero-positive subset of patients as most likely to respond to Benlysta HGSI is seeking to optimize for a group of patients with
elevated BlyS, heightened B-cell activity, heightened auto-antibodies like anti-dsDNA and with active disease amenable to BlyS inhibition
and less likely to have spurious placebo responses. While detailed phase II data suggests some correlation between sero-positive status
and certain markers of SLE we believe that correlations are weak and that clinical relevance of markers are not completely understood. In
particular differences in complement C4 levels of 16.9mg/dl and 24.8mg/dl are of questionable significance given that both levels are
actually elevated above the lower limit of normal at 16mg/dl. While differences were seen in terms of extent of renal and hematologic
involvement in sero-positive patients we would note that no differences were seen in the incidence of malar rash, discoid rash,
photosensitivity, oral ulcers, arthritis, serositis and neurologic involvement. Characterization of the sero-positive subgroup as having more
active disease than sero-negative patients may be fair but with only 19% of placebo patients flaring within this subgroup over 52 weeks we
would question how active a subgroup this really is.
Ultimately there may be a subgroup of SLE patients with highly active disease, low placebo responses and optimal response to BlyS
inhibition, however given the significant heterogeneity of the disease we believe that a 72% subset is too broad to isolate a homogenous
subgroup with reproducible characteristics in phase III. We would highlight the Anti-dsDNA subgroup at 50% of patients in phase II, the low
complement C4 subgroup at 41% of the patients in phase II, the anti-RNP subgroup at 47% of patients in phase II, the Anti-Ro/SSA
subgroup at 39%, the anti-cardiolipin subgroup at 28%, and the Anti-SM subgroup at 19% as all equally plausible auto-antibody subset
populations amenable to BlyS inhibition. Within the sero-positive subset in phase II that there may be other subsets of patients that are
actually driving the true benefit of Benlysta and that remain unidentified and unlikely to be reproduced in phase III.
Ability of SRI Composite to capture Benlysta clinical activity?
Failure of Lupus studies has often been attributed to trial design, choice of endpoints and the difficulty in establishing benefit in a
heterogeneous population with waxing and waning disease. In the phase II experience where Benlysta failed to meet the primary endpoint
of SELENA/SLEDAI improvement in disease activity HGSI has suggested that a more comprehensive composite responder index is better
able to discern a drug benefit over placebo. Beyond the fact that this SLE Responder Index (SRI) did not establish a benefit in the
prospectively defined overall unselected population and beyond the fact that the SLE responder index (SRI) was not identified prospectively
with specific alpha allocation we believe that examination of the individual components of the SRI suggest modest differences for Benlysta
over placebo and limited cushion for more robust evaluation in phase III.
Component Responses in Phase II SRI Composite Index
Placebo Benlysta Diff. vs
plc
Relative
increase
n 86 235
SELENA-SLEDAI 39.5 49.4 9.9 25%
BILAG 81.4 91.5 10.1 12%
PGA 76.7 90.2 13.5 18%
SRI Response 29.1 46 16.9 58%
Week 52
SRI Component
Source ACR2007 and Barclays Capital Research
In assessing the internal consistency of the composite SLE responder index (SRI) we would note that absolute difference between Benlysta
and placebo for each of the components is relatively small with 9.9% absolute difference in SELENA-SLEDAI, 10.1% absolute difference in
BILAG A and B and 13.5% absolute difference in physician global assessment (PGA). Interestingly the most statistically robust component
of the SLE responder index (SLE) and the biggest driver of Benlysta composite results is perhaps the most subjective measure of disease
activity in the physician global assessment (PGA). We would note as well a high placebo response in both BILAG and PGA at 81.4% and
76.7% respectively that provides limited margin for improvement with Benlysta in patients with active disease. The SELENA-SLEDAI
remains the domain with greatest opportunity for benefit from Benlysta but unfortunately is also the domain where no benefit is seen with
the drug even in the more select sero-positive population in phase II.
Implications of Lack of Dose Response
The absence of a dose response across a 10-fold dose range in a phase II study is another problematic aspect of the Benlysta proof-ofconcept
data that should be considered. While it is possible that there is a threshold effect in terms of BlyS inhibition, B-cell suppression,
and dsDNA reduction above the 1mg/kg dose it is also equally plausible that the 1mg/kg to 10mg/kg dose range is so far below a
therapeutically effective range as to be indistinguishable across doses and from placebo. Indeed phase I data published in Arthritis
Research and Therapy by Furie et al. (10:R109, 2008) suggests that reductions in dsDNA were more significant for a higher 20mg/kg dose
as compared to lower doses in patients with dsDNA levels >10 IU/ml. Additional data from phase II does suggest a dose response in terms
of steroid reduction where steroid dose reductions <7.5mg/day were seen with increasing frequency between the 1mg/kg and 10mg/kg dose
and where requirement for steroid dose increase >7.5mg was seen with decreasing frequency between the 1mg/kg and 10mg/kg dose. Of
Equity Research
8
particular note, only the 10mg/kg dose of Benlysta was seen to result in reduction of steroid requirements <7.5mg/day suggesting that this
dose may be at the beginning of the effective dose range and in terms of requirements for steroid dose increase >7.5mg/day the range of
3.5-11% was not much different than the 17% with placebo.
While reductions in dsDNA and steroid requirements would suggest the beginning of a dose response between 10mg/kg and 20mg/kg,
other data suggest limited dose response across other parameters including reduction in CD20+ve B-cells. While reductions in CD20+ve Bcells
were observed across Benlysta dose ranges of 1mg/kg to 10mg/kg in published phase I data we would note that CD20+ve B-cells do
not produce auto-antibodies like dsDNA and are involved in pathogenesis of SLE as are CD20-ve plasma cells. Indeed, we believe that part
of the problem in identifying an effective dose range in earlier phase I/II studies is a lack of clear understanding in terms of key parameters
of disease. Recent discussions with HGSI have highlighted the role of CD20-ve B-cells in auto-antibody production, however phase II
experience suggests that CD20-ve B-cell counts actually go up in response to Benlysta treatment. More recently HGSI has highlighted the
CD38 (bright) cells as the key target of interest in SLE, however this revelation appears to come somewhat late to inform an appropriate
dose selection for phase III.
Ultimately we believe that the lack of Benlysta dose response between 1 mg/kg and 10mg/kg and the emergence of modest dose response
between 10mg/kg and 20mg/kg suggests that the dose range selected for phase III is at the low end of an effective dose range and is likely
responsible for the lack of benefit of Benlysta on SELENA/SLEDAI, modest effect on BILAG and relatively undifferentiated effect on PGA.
Similarly we believe that the relatively benign safety and tolerability profile for a potent immunomodulatory drug may also be best explained
on the basis of sub-therapeutic dosing. In particular when one considers the severe adverse events with related BlyS inhibitor Atacicept and
excess infections that resulted in discontinuation of that program we believe once again that Benlysta is operating at the very low end of the
effective dose range. HGSI has acknowledged that Benlysta benefit may increase up to doses of 50mg/kg however the protein
requirements would be prohibitive, perhaps constraining fuller review of dose ranging.
Implications of Variable B-Cell Effects With Benlysta?
As noted previously, CD20+ve B cell levels are reduced by 50% over a 52 week Benlysta dosing period. While CD20+ve B cells can be
involved in antigen presentation and perpetuation of an inappropriate immune response, they are not the immunologic trigger for the SLE.
Instead, the pathogenic role of ds-DNA and ANAs implicates the antibody secreting CD20-ve plasma cell as the key effector cell of disease.
In considering the failure of anti-CD20+ antibody Rituxan in the EXPLORER SLE trial we would note that Rituxan has no impact on CD20-
ve plasma cells.
Interestingly, plasma cells actually increased by 67% in the Benlysta arms compared to a 30% increase in the control arm over 52 weeks in
phase II. Given the central role of BlyS in B cell and plasma cell survival this observation is perhaps counterintuitive. HGSI has suggested
that patients with SLE appear to have a unique population of plasma cells, the so-called CD19dim CD38bright plasma cells and this sub
population decreased by 44% compared to 30% in the Benlysta and control arms respectively.
In trying to understand apparent discordances between an increase in CD20-ve plasma cells and decrease in CD19 and CD38
subpopulations we surmise that APRIL signaling through BCMA may be a more important survival signal for plasma cells than BlyS, but
that BlyS signaling is more important in the CD38 subset. Supportive of this notion are changes in IgG levels where treatment with Benlysta
led to a modest 10% reduction in IgG levels, however, a 30% decrease in ds-DNA IgG was observed suggesting a preferential Benlysta
effect on the ds-DNA secreting CD38bright plasma cells.
While HGSI has presented an analysis of SRI responder correlation with reductions in CD20+ve B-cells the company has not presented an
analysis of SRI response by changes in either plasma cell or the SLE-plasma cell numbers which especially in the case of the latter may be
expected to correlate to efficacy. The absence of plasma cell correlation analyses would seem to be puzzling given their pivotal role in the
disease. However, we understand that at the time of the phase II study, the technology to accurately elucidate changes in numbers of these
relatively rare cells was not well developed and too much noise existed in the data. For the BLISS studies, B cell populations will be
monitored with 6-color FACS instead of the 4-color analysis used in phase 2 and this may provide additional detail regarding B cell subsets.
Overall, in reviewing the data for Benlysta effects on B-cell reductions we are not convinced that correlations between anti-CD20+ B-cell
reductions and SRI responder rates are instructive given the lack of primary role of anti-CD20+ve B-cells and the abject failure of potent
anti-CD20+ve B-cell antibody Rituxan. With actual increases in antibody producing CD20-ve B-cells we believe that there is theoretical
concern that Benlysta could actually worsen disease in some patients, however HGSI's argument regarding the central role of CD38 (bright)
B-cells in dsDNA production seems compelling. The difficulty we have with the development of an Anti-CD38 bright B-cell thesis so late in
development is that a) there is no correlation data between CD38 bright B-cell reduction and drug activity b) there has been no ability to
optimize dose against CD38 bright B-cell response and c) the reductions in CD38 bright B-cells seem modest with 44% reduction with
Benlysta and 30% reduction with placebo.
Equity Research
9
Correlations Between Sero-Conversion and Clinical Response
In attempting to better understand the impact of Benlysta on SLE pathogenesis HGSI investigators have also attempted to correlate
composite SLE Responder Index (SRI) responses with reductions in dsDNA, various dsDNA subsets and increases in complement C4. At a
high level HGSI has suggested that roughly 50% of sero-positive patients with ANA positivity, dsDNA, anti-Sm, anti-Ro, anti-RNP and anticardiolipin
auto-antibides at baseline have an SRI response to Benlysta, however we would note that by the same token 50% of these
patients do not have an SRI response to the drug.
Sero-Conversions Rates – Benlysta vs Placebo
Placebo Benlysta
Response at week 24 4/50 (8%) 35/151 (23%)
Response at week 52 7/45 (16%) 41/138 (30%)
PLc to 10mg/kg
Response at week 76 5/40 (13%) 41/122 (34%)
Yes No
dsDNA response 53% 38%
SRI Response
Source: EULAR 2007 and Barclays Capital Research
Given the proposed pathogenic role of dsDNA auto-antibodies Benlysta investigators have provided further detail on correlations between
dsDNA reduction of >50% or sero-conversion to sero-negative status with SRI response. In the previously discussed phase II study
investigators have noted that within the 72% of patients who were sero-positive that 30% had reductions in dsDNA of >50% or became
sero-negative as compared to 16% of placebo patients at 52 weeks and that overall in patients that sero-converted 53% had an SRI
response while in those that didn't sero-convert 38% had an SRI response. Overall we believe that data suggest a weak association
between SRI response and sero-conversion of patients with elevated dsDNA auto-antibody levels and with no separate analysis of SRI
responder rates in Benlysta and placebo sero-converters it is difficult to make any conclusions from this data. Despite a low sero-conversion
rate of 30% with Benlysta, clearly twice as many patients in phase II sero-converted with Benlysta as compared to placebo, however we do
not know whether that correlated with SRI responses within each treatment group. In addition we would note that for SELENA-SLEDAI the
apparent benefit of "sero-converting" is an average 2 point decrease which is half the required 4 point decrease necessary to be defined as
a responder.
Correlations Between dsDNA Changes and SRI Response
SRI Responder SRI Non-responder Relative Increase
Week 52 35% 19% 84%
Week 76 35% 20% 75%
Week 96 47% 32% 47%
Week 128 57% 29% 97%
Week 160 52% 45% 16%
Source EULAR 2008 and Barclays Capital Research
Longer term follow-up of dsDNA auto-antibody levels suggested continued decreases although with different kinetics in different subtypes of
dsDNA auto-antibodies. In particular anti-Sm and anti-RNP levels show no change at 52 weeks in either Benlysta or placebo groups
although by week 76 we see a 20% reduction in anti-Sm with no change still in anti-RNP until week 112 where a 20% reduction is also
observed. While investigators have noted that maximal reductions in dsDNA auto-antibodies occur after 2 years and that this correlates with
optimal SRI response we would note that a) correlations between SRI response and anti-dsDNA are weak b) that SRI responses at 2 years
are plagued by selection bias and that c) the primary endpoint in phase III is a 52 week endpoint.
Correlations Between Complement C4 Changes and SRI Response
SRI Responder SRI Non-responder Relative Increase
Week 52 45% 30% 50%
Week 76 54% 31% 74%
Week 96 64% 44% 45%
Week 128 70% 46% 52%
Week 160 75% 42% 79%
Source EULAR 2008 and Barclays Capital Research
Similarly assessments of complement C4 elevations over the longer term experience with Benlysta are difficult to interpret because of
selection bias although once again there seems to be some correlation suggested between SRI responses over the longer term and optimal
increases in complement C4 levels. With 41% of patients enrolled in the phase II described as having abnormally low levels of complement
pathway protein C4 at baseline investigators have noted that between 52 weeks and 160 weeks SRI responder rates increased from 45% to
75% suggesting a higher correlation than with dsDNA. Beyond the aforementioned biases we would note that plasma cells responsible for
auto-antibody production have long half-lives and it may be that effects on BlyS may not appear until well beyond 52 weeks.
Equity Research
10
Optimization of Phase III Clinical Trial Design – Is It Enough?
In the event that retrospective subset analysis from phase II is valid and that failure of prospectively defined primary endpoints was in fact
driven by sub-optimal trial design we believe that efforts to replicate phase II results will in large part rely on optimization of phase III design.
• Baseline steroid requirements: In the phase II study, 39% of patients enrolled were taking at least 7.5mg/d of steroids. In the phase
III study, the lower limit for enrollment has been raised from 5mg/d to 7.5mg/d with the upper limit being maintained at 40mg/d. The
intention for this change is to enroll patients that have a greater requirement for steroids indicative of more active disease. While it
is unlikely that the response rate to Benlysta increases as a result of this change, HGSI hopes that the differential between
Benlysta and placebo increases. Allied to this, HGSI has built in restrictions to therapy changes between months 10 and 12. In the
phase 2 study changes in background therapy were permitted throughout the study and HGSI felt that this permitted bias in the
study which favored the placebo arm. In phase III HGSI has added strict restrictions to define what changes in therapy are
permitted between months 10 and 12 with the goal of eliminating any bias in favor of placebo.
• Baseline SELENA-SLEDAI: In phase III baseline SS score requirement has been increased to 6. While the average SS at baseline
in the phase II study was >9, patients were eligible if they had an SS score of 4. However, during the run-in period changes in
serum parameters could occur which remained cryptic to the treating physician at the time of first dosing. When the day 0 serum
analyses had been completed a patients’ baseline SS score could have declined by 1 or 2 points. If their baseline score was in fact
3 at the time of dosing instead of 4, the patient could not achieve the necessary 4 point reduction in SS score required to achieve
responder status even if the disease emission occurred. By increasing the baseline SS to 6, HGSI has built into phase III the ability
to absorb a 2 point decrease in SS that is expected to occur in a few percent of patients enrolled during the 35 day run in period
after randomization but prior to dosing.
• Out of HGSI’s control was the need to take the BLISS studies to a global stage rather than restrict enrollment to North America.
Epidemiological studies for lupus are lacking, however patients of African Americana and Indigenous American origin may have
more aggressive disease. Given this, HGSI decided to segregate each of these high risk groups into separate studies, African
Americans into BLISS-76 and Indigenous Americans into BLISS-52.
While efforts by HGSI to optimize Benlysta benefit and reduce spurious placebo responses in phase III are commendable we believe that
the heterogeneity of SLE precludes optimization with crude measures like ANA and dsDNA sero-positivity as well as baseline steroid use.
There is no evidence from phase II that Benlysta benefit was confounded by imbalances in high dose steroid use and overall steroid doses
used in phase II appear relatively low compared to enrollment criteria in phase III.
Conclusion
Ultimately we believe that the likelihood of Benlysta success in BLISS-52 and BLISS-76 studies is exceedingly small. With uncertainty on
markers of optimal efficacy it appears likely that Benlysta dose has not been fully optimized with hints of dose response appearing above
the dose range in phase III and with no dose response within the phase III dose range. Biologically the effect of Benlysta appears modest
on mediators of disease with uncertain and variable contribution to a heterogeneous patient population. Reductions in CD20+ve B-cell are
predictable based on the role of BlyS in B-cell maturation but may be of limited significance given the abject failure of potent anti-CD20+ve
antibody Rituxan in EXPLORER and LUNAR studies. With CD20-ve plasma cells as more important elaborators of SLE autoantibodies we
believe that an actual neutral affect with Rituxan and increase with Benlysta is more consistent with EXPLORER and LUNAR failures as
well as absent affect of Benlysta on SELENA-SLEDAI measures in phase II. While CD38 bright cell thesis is intriguing we would note that
with only 5% of B-cell population accounted for by CD38 bright cells and with no correlation data between CD38 bright cell reduction and
SRI improvement with Benlysta we remain skeptical that Benlysta has been developed optimally as a CD38 bright cell positive inhibitor.
With most compelling biologic data surrounding partial sero-conversion of dsDNA autoantibody levels we believe that an overall 30% rate
and 14% difference from placebo is insufficient to support a significant clinical benefit from placebo. In particular with effects on anti-Ro,
anti-RNP, anti-cardiolipin, and anti-Sm not appearing until after 52 months we see little reason for any therapeutic benefit with Benlysta.
Overall we believe that retrospective subset data from the Benlysta phase II study has significant limitations and that with failure of the
SELENA-SLEDAI primary endpoint clinical benefit of Benlysta is questionable. Component data from the SLE responder index (SRI)
suggests no impact on SELENA-SLEDAI and modest absolute differences in BILAG A/B and PGA domains of 10% and 13% respectively.
In a disease as heterogeneous as SLE we believe that selection of a 72% ANA/dsDNA sero-positive subset is both arbitrary and too broad
with unidentified subcomponents that are likely more important determinants of drug response. We have seen few, if any, retrospective
subset analyses of failed phase II trials produce subsequent success in prospectively controlled phase III trials and believe that the
combination of a seropositive subset with a composite metric analysis is more fraught with error than most retrospective analyses that we
are confronted with. In addition, with phase II data typically over-estimating drug benefit we believe that there is little cushion going into
phase III with a 50% benefit claimed in phase II and a 40% benefit sought in BLISS-52.
Equity Research
11
Considering Risk-Reward Proposition for BLISS-52
In the event that Benlysta were to succeed where Rituxan, Riquent and multiple other therapies have failed we believe that the available
opportunity in SLE has been over-estimated. Recent publication suggests a market in the US of roughly 300,000 sufferers with expert
feedback suggesting roughly 5% of patients each year with worsening disease requiring rescue with immunosuppressives or newer
biologics like Benlysta. Assuming pricing of $20,000 per year we estimate a peak opportunity in the US of $300M with comparable sales of
$300M ex-US. With HGSI sharing economics 50:50 with GSK we estimate $300M of global value accruing to HGSI. We estimate the
incremental value of positive BLISS-52 data at an NPV of $825M or $6/share with upside potential to $7/share. In the more likely event (in
our view) of BLISS-52 failure we would expect HGSI to test prior lows of $1/share. While absolute magnitude of upside potential far exceeds
downside risk in absolute terms we believe that with a 85% likelihood of failure and 15% likelihood of success that fair value is closer to
$2/share. |
