Executive Summary – Pamrevlumab • Management appears to share our doubts about pamrevlumab in IPF: FibroGen is currently only running a single Phase 3 trial now even though two Phase 3 trials would be required for approval and each trial is expected to take four years • FibroGen withheld efficacy results of a placebo-controlled study designed to test whether pamrevlumab added any benefit when stacked with current IPF standard- of-care (Esbriet and Ofev); this indicates that pamrevlumab failed to show any benefit • That failure likely informed the decision to structure the Phase 3 trial as placebo- controlled with background Esbriet/Ofev prohibited. This would make pamrevlumab unsellable even with Phase 3 success: • Pamrevlumab would be a branded IV formulation with no evidence of additive benefit, superiority, or even non-inferiority vs. oral Esbriet/Ofev, at least one of which will likely soon go generic • Possibly competing against new drugs for IPF that show additive benefit with Esbriet/Ofev PLAINVIEW Executive Summary 10
FibroGen Overview PLAINVIEW
FibroGen Overview • $3.7bn market capitalization • Founded in 1993 • FibroGen’s flagship drug is roxadustat, an HIF-PH inhibitor designed to treat anemia. We estimate roxadustat makes up 85- 90% of FibroGen’s current enterprise value. FibroGen is scheduled to present its key major adverse cardiovascular event (MACE) data at ASN 2019 (2:00 PM Friday, November 8): a make-or-break moment for the drug and the company • FibroGen is also developing pamrevlumab, an anti-CTGF antibody, for three indications: idiopathic pulmonary fibrosis (IPF), pancreatic cancer, and Duchenne’s Muscular Dystrophy (DMD). We estimate pamrevlumab makes up the remainder of FibroGen’s current enterprise value. Pamrevlumab’s Phase 3 trials are expected to read out in 2022 (pancreatic cancer) and 2023 (IPF) PLAINVIEW FibroGen Overview 12
Roxadustat Overview • Roxadustat is a hypoxia- inducible factor prolyl hydroxylase (“HIF - PH”) inhibitor • HIF is a transcription factor that can stimulate erythropoiesis, while HIF-PH is an enzyme that binds to and triggers the degradation of HIF through hydroxylation and ubiquitination • In normal circumstances, HIF is inhibited by HIF-PH. However, HIF-PH requires oxygen to function, hence HIF activity is “induced” by hypoxia • Roxadustat is designed to inhibit HIF-PH, mimicking the effects of hypoxia and leading to increased erythropoiesis to treat anemia • FibroGen has completed pivotal trials for roxadustat in the treatment of anemia in patients with chronic kidney disease and has guided to submitting an NDA for dialysis-dependent and non-dialysis-dependent chronic kidney disease (DD-CKD & NDD-CKD) in October 2019 and to submitting an MAA in the EU by the end of Q1 2020 PLAINVIEW FibroGen Overview 13
Pamrevlumab Overview • Pamrevlumab (FG-3019) is an antibody designed to inhibit connective tissue growth factor (“CTGF”) • FibroGen presents CTGF as a principal cause of fibrosis in organs as well as a key target in pancreatic cancer • FibroGen has been conducting clinical trials with pamrevlumab since 2003 • Past indications FibroGen has posted pamrevlumab trials for and subsequently scrapped include liver fibrosis, diabetic nephropathy, and focal segmental glomerulosclerosis (FSGS) • Idiopathic pulmonary fibrosis (“IPF”) and locally advanced pancreatic cancer (“LAPC”) are the current lead indications for pamrevlumab • LAPC Phase 3 results expected in 2022 • IPF Phase 3 results expected in 2023 PLAINVIEW FibroGen Overview 14
Roxadustat PLAINVIEW
Roxadustat Table of Contents ❑ MACE/mortality imbalances appear to be a class effect • There is a large MACE/death imbalance favoring control group across roxadustat trials as well as trials for competing HIF- PH inhibitors vadadustat and daprodustat, indicating a class effect ❑ Roxadustat’s MACE HR is inferior to EPO • FibroGen’s claim of numerically lower events in DD -CKD is a product of the much-higher dropout rates seen in roxadustat groups across trials. Based on published clinical data, we believe roxadustat’s MACE risk is significantly higher than EPO, which is why FibroGen has withheld it and provided bizarre excuses for the lack of disclosure ❑ Roxadustat’s MACE HR is inferior to placebo • FibroGen has withheld the NDD-CKD MACE HR and instead told investors that the post-hoc ITT MACE HR was non- inferior to placebo. ITT analysis is inappropriate & misleading because it dilutes the safety signal with dropouts. As with DD-CKD, we believe roxadustat failed to achieve MACE non-inferiority in the NDD-CKD setting ❑ MACE+ inflated by prohibiting IV iron & using placebo comparator • IV iron is a critical part of EPO treatment but is not necessary for HIF-PH inhibitors to achieve efficacy. FibroGen manipulated MACE+ by barring IV iron in its trials and using a placebo comparator in NDD-CKD, driving up anemia-related “+” events in the control groups ❑ HIF-PH inhibitor mechanism of action is flawed • HIF-PH inhibitors (including roxadustat) act upstream of EPO and are less specific to erythropoiesis. HIFs & HIF-PH inhibitors have been implicated in hypertension, fibrosis, immune suppression, and other adverse effects that could increase risk of MACE ❑ Roxadustat does not work in EPO-hyporesponsive patients • FibroGen’s claim that roxadustat can treat EPO -hyporesponsive patients is based on a misleading study where EPO hyporesponders as well as roxadustat non-responders were removed from the analysis PLAINVIEW 16
Point #1: MACE/Mortality Imbalances Appear to be Class Effect • Published data shows large MACE & mortality imbalances across FibroGen’s trials • 11/983 roxadustat patients died compared to 0/419 control patients • 23/983 roxadustat patients experienced MACE incidents compared to 1/419 control patients • These imbalances were also seen in competitor GSK and Akebia’s HIF -PH inhibitor trials • 9/677 daprodustat patients died compared to 1/192 control patients • 14/677 daprodustat patients experienced MACE incidents compared to 2/192 control patients • 5/522 vadadustat patients died compared to 1/243 control patients • Akebia withheld nearly all adverse event data outside of deaths but reported 7/522 MACE incidents in vadadustat patients compared to 1/243 in control patients PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 17
MACE/Mortality Rate Consistently Higher in Roxadustat Group • We compiled all FibroGen’s published clinical data in the table below. Roxadustat consistently yielded worse MACE rates compared to EPO/placebo despite higher dropout rates. This included 11/983 deaths on roxadustat compared to 0/419 deaths on EPO/placebo and 23/983 total MACE incidents compared to 1/419 for EPO/placebo 1 In the poster discussing the Japan Phase 3 DD-CKD trial results, the authors did not provide a complete SAE breakdown, and it is possible that additional MACE incidents occurred that were not disclosed PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 18
Roxadustat Trial References • Sources for the numbers in the prior slide are provided below. The relevant excerpts from each publication (including the Akizawa poster) are provided in the Appendix at the end of this presentation along with brief discussion Trial Source Where to Find MACE in Each Publication China Phase 3 – DD-CKD Chen et al 2019 SAEs on p. 13 of Supp. Appendix Japan Phase 3 – DD-CKD ASN 2018 Akizawa Poster (not available online) Table 5 of poster China Phase 2 – DD-CKD Chen et al 2017 No reported events US Phase 2 – DD-CKD Provenzano et al 2016 SAEs listed in Table 5 of paper US Phase 2b – NDD-CKD Provenzano et al 2016 SAEs listed in Supplemental Table 5 US Phase 2 – ID-CKD Besarab et al 2016 SAEs listed on pp. 15-16 of Appendix US Phase 2a – NDD-CKD Besarab et al 2015 No reported events China Phase 3 – NDD-CKD Chen et al 2019 No reported events China Phase 2 – NDD-CKD Chen et al 2017 No reported events PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 19
Roxadustat MACE Events by Category 1 In the poster discussing the Japan Phase 3 DD-CKD trial results, the authors did not provide a complete SAE breakdown, and it is possible that additional MACE incidents occurred that were not disclosed PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 20
MACE Imbalance Even Worse After Adjusting For Treatment Duration • One concern with dividing total events by total patients is that the varying trial durations and enrollment sizes lead to different treatment durations for the patient groups which could make absolute event rates misleading • However, event count understates how poorly roxadustat faired: • The average treatment duration based on trial design was longer for control group patients than roxadustat patients (see table) • The dropout rate was 40% higher in the roxadustat group compared to the control (see slide 18) – meaning that actual average treatment duration is even lower in roxadustat Note: patients relative to control 1 US Phase 2 DD-CKD and US Phase 2b NDD-CKD trials each included two different trial durations. These are combined in the prior two slides PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 21
Class Effect: Higher MACE/Mortality Rates Also True for Daprodustat • Daprodustat also saw much higher rates of discontinuation, MACE, and death compared to control. See Appendix at the end of this presentation for sources, relevant excerpts & brief discussion PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 22
Class Effect: Higher MACE/Mortality Rates Also True for Vadadustat • Akebia has withheld much of its safety data, but based on what is publicly available, vadadustat presents MACE/death imbalances similar to roxadustat and daprodustat. See Appendix at the end of this presentation for sources, relevant excerpts & brief discussion PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 23
Death Imbalance in Additional Daprodustat Trial • GSK also conducted a trial to reduce ischemic events in patients undergoing aortic aneurysm repair where patients were either dosed with daprodustat or placebo. Daprodustat appeared to have a detrimental effect: 22% (6/27) of daprodustat patients died compared to 7.1% (2/28) of placebo patients • Caveats: • The death imbalance was not statistically significant due to low N • The doses of daprodustat used in this trial (300 mg loading & 100 mg QD) were higher than the doses used for treating anemia, so it is not an apples-to-apples comparison – but it is still certainly troubling PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 24
MACE Inferiority Even with Questionable In-House Adjudications • In FibroGen’s publication Besarab et al 2015, the authors claimed that there were no cardiac SAEs in the roxadustat group. Curiously, one of the non- cardiac SAEs in the roxadustat group was deemed to be “dyspnea secondary to worsening congestive heart failure” while another SAE was ruled “noncardiac chest pain” with no further commentary • Phase 3 safety results are generally worse for drugs with hazardous mechanisms of action because: • Enrollment is expanded to meet powering requirements, frequently leading to sicker patients who are more susceptible to adverse events • Adverse event adjudication switches from local principal investigator/sponsor to a blinded third party that is impartial and not invested in the success of a drug PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 25
Greatest Phase 2 to Phase 3 Turnaround of All Time? • Given the extraordinarily poor Phase 2 safety results reported despite in- house adjudications, roxadustat achieving MACE noninferiority in Phase 3 would be the greatest Phase 2 to Phase 3 turnaround of all time • We find it highly unlikely that roxadustat inexplicably achieved the safety improvements necessary to reach non-inferiority in Phase 3 after reporting much worse safety in its prior trials • Management’s blatantly misleading statements and refusal to disclose the actual MACE hazard ratios greatly reinforce our view. First rule of biotechnology investing: if Management withholds clinical data, that data is always bad PLAINVIEW MACE/Mortality Imbalances Appear to be Class Effect 26
Point #2: Roxadustat’s MACE HR is Inferior to EPO • During May 2019 call to discuss pooled safety results for roxadustat, analysts repeatedly asked about the MACE hazard ratios in DD-CKD. FibroGen Management refused to provide the hazard ratios, but instead said that the total number of MACE incidents was lower for roxadustat in DD-CKD, leading many to erroneously conclude that the hazard ratios were <1 • This conclusion is wrong. The event counts are distorted by the fact that the dropout rate for the roxadustat group was much higher than the dropout rate for the control group. Dropouts become immune to MACE and reduce total possible events but are censored in calculating hazard ratios. Roxadustat dropout rates have been far higher than EPO across FibroGen’s trials: in the China Phase 3 DD-CKD trial, the roxadustat discontinuation rate was 3.4x the EPO discontinuation rate. The gap in dropouts appears to increase as trial duration increases • We believe the hazard ratio was not disclosed because roxadustat is inferior to EPO – not due to Management’s bizarre claims of being restrained by AstraZeneca and needing to discuss statistical analysis with the FDA well after the trial concluded • FibroGen has confirmed that the roxadustat NDA & MAA are for the DD-CKD and NDD-CKD indications; the ID-CKD safety data is too underpowered to show MACE noninferiority and is irrelevant to drug approval PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 27
“My Dog Ate My Pivotal Trial Results” • The initial FibroGen press release revealed that roxadustat achieved non-inferiority for MACE+ and that MACE showed “no clinically meaningful difference” in DD -CKD and NDD-CKD, but did not state that MACE achieved non-inferiority • On the Q1 2019 call discussing the pooled Phase 3 results, FibroGen first claimed that “ we have to yet agree with our regulator on specific [MACE] analyses to be done.” • Minutes later, FibroGen Management implied that AstraZeneca was preventing FibroGen from calculating/sharing the MACE hazard ratios (but mysteriously not the MACE+ hazard ratios or any of the other statistics they provided): “We don't expect that MACE will be particularly different than this [MACE+]. It's just that with U.S., we have an agreement with our partner, AstraZeneca, to evaluate under the totality of evidence basis.” • These excuses are absurd. MACE is a subgroup of MACE+, so if FibroGen had calculated the MACE+ hazard ratios, then FibroGen must have also calculated the MACE hazard ratios. Even suspending reality and assuming FibroGen could not produce a hazard ratio, they could have easily provided a Kaplan-Meier plot or basic statistical measures such as means/medians to show the data – no complex analysis required PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 28
Event Count Misdirection • When pressed on hazard ratios, FibroGen Management switched the discussion from hazard ratios to the misleading number of events: “So with the MACE+ data, I believe we have numeric advantage in these categories. So there’s 5 categories…Every one of them, we have numeric advantage over ESA. Is that clear enough?” • This was repeated multiple times, and seemed to placate some investors and analysts, with the underlying assumption being that if the total number of events was lower, then the hazard ratio must be <1 • This assumption is wrong • The primary safety endpoint for DD-CKD and NDD-CKD is MACE hazard ratio, not total number of MACE events per group PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 29
Much Higher Dropout + Lower Number of Events = HR Inferiority • The number of events can be lower in the roxadustat group despite HR inferiority because roxadustat has consistently seen far more patient dropouts compared to EPO • Looking at FibroGen’s public EPO -controlled DD-CKD trial data, the dropout rate in the roxadustat group is up to 3.4x as high as the dropout rate in the EPO group, with the difference increasing with duration: • Looking at the 19-26 week trials, we see that the median discontinuation rate for roxadustat patients is 21.2% compared to 8.7% for placebo, or 2.43x as high. The dropout discrepancy appears to increase with increased duration and the median duration was significantly higher in the DD-CKD Phase 3 trials: FibroGen reported a mean duration of treatment of 1.9 years in the SIERRAS trial and 1.8 years in HIMALAYAS • The consistently higher dropout rates should also be alarming to investors given that safety is the key endpoint PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 30
Hazard Ratio Refresher • Hazard ratios (HRs) measure the likelihood of a specific event occurring in one population compared to the likelihood of the same event occurring in another population over a given time horizon • HRs are not the total number of events in one group divided by the other group. “Total events” is not a useful measure (particularly in the case of roxadustat) because dropouts become immune to survival endpoints due to censoring and would inappropriately benefit their group by reducing total possible events • Dropouts are addressed by using Kaplan-Meier estimate censoring: once a patient drops out, he/she is censored from future event probability calculation. The operating assumption is that dropouts would have had the same probability of experiencing an event in the future as the rest of the population PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 31
Hazard Ratio Refresher • HRs can be thought of as the tangent slope of the Kaplan-Meier (K-M) event curve for one group divided by the tangent slope of the K-M curve for the other group. The relative slope represents the relative probability of an event occurring: if Survival Curve 1 (SC1) has a slope of -5% (meaning that, on average, 5% of the population experiences the event per time period) and Survival Curve 2 (SC2) has a slope of -10% (meaning that, on average, 10% of the population experiences the event per time period), the event is twice as likely to occur during a given time period in the Survival Curve 2 group. The HR for SC1 would be 0.5 and the HR for SC2 would be 2.0 • The tangent slope is rarely (if ever) constant, and so each K-M curve generates a range of slopes (and thus hazard ratios) which can be used to calculate confidence intervals PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 32
Hazard Ratio Refresher • In a simple example, we look at two populations where one population has an event rate of 5%/year and a dropout rate of 5%/year while the other population has an event rate of 7.5%/year and a dropout rate of 15%/year. Despite having a lower total event count, the hazard ratio for the second population is much higher and the K-M curve plainly illustrates worse survival: Higher Event Rate + Higher Dropout Rate = Lower Event Count & HR>1 PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 33
High Dropout Rate Suppresses MACE Event Count • Another HIF- PH inhibitor, Bayer’s molidustat, reported comparable MACE event rates between the trial group and control in Macdougall et al 2019. However, we think this is attributable to the astronomically higher dropout rate in the molidustat group (38% vs. 11%), not to any difference between molidustat and the other HIF-PH inhibitors. The control completion rate was nearly 50% higher than molidustat (89% control vs. 62% molidustat) PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 34
ID-CKD: Powering Insufficient to Draw Conclusions • FibroGen claims that roxadustat achieved “directionally lower MACE risk than EPO” in the ID -CKD subpopulation • FibroGen’s decision to frame the ID - CKD hazard ratio as showing “directionally lower” MACE risk without any mention of non -inferiority indicates that the confidence interval did not achieve non-inferiority • We can roughly estimate the confidence interval for the ID-CKD data based on prior confidence interval of the CHOIR trial, which enrolled a similar size population (CHOIR n = 1,432, FibroGen ID-CKD n = 1,526) • The hazard ratio in the CHOIR trial was 1.28 with a 95% confidence interval of (1.03, 1.75). With the same confidence interval range of (-0.25,+0.47), roxadustat would have needed to achieve an HR of 0.83 just to prove non-inferiority • The confidence interval for ID-CKD is likely even wider than the CHOIR trial due to 1) ID-CKD patients experiencing more frequent events compared to the stable dialysis patients enrolled in CHOIR and 2) roxadustat’s high dropout rate PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 35
ID-CKD MACE: Not a Primary Safety Endpoint • We find it highly unlikely that HIF-PH inhibitors can achieve superiority or even non-inferiority vs. EPO in any setting given the preponderance of clinical evidence from FibroGen, Akebia, and GSK that HIF-PH inhibitors increase MACE risk relative to EPO • We note that in FibroGen’s prior uncontrolled ID -CKD roxadustat trial, 3% (2/60) of roxadustat patients experienced MACE events (both deaths) within 12 weeks • It is far more likely that the “directionally lower MACE risk” is the product of the higher variance in the ID-CKD pool fueled by more frequent events and a smaller N • Regardless, the ID-CKD results are insufficient to support an NDA in terms of powering. As FibroGen has stated multiple times (including on the most recent August 8, 2019 Q2 2019 conference call), the NDA is to be submitted for the DD- CKD and NDD-CKD indications – not for ID-CKD PLAINVIEW Roxadustat’s MACE HR is Inferior to EPO 36
Point #3: Roxadustat’s MACE HR is Inferior to Placebo • FibroGen Management refused to disclose the NDD-CKD MACE hazard ratio, but instead said that under a “conservative” post -hoc intention-to-treat (ITT) analysis, roxadustat achieved non-inferiority • The difference between standard Kaplan-Meier analysis and ITT analysis is that Kaplan-Meier censors patients after drug discontinuation while ITT does not • Evaluating safety using an ITT analysis is the opposite of conservative: ITT analyses include dropouts who are no longer taking study medication, which dilutes any drug-driven safety signal • If Management calculated the ITT MACE HR (HR including dropouts), then Management must have also calculated the primary safety endpoint: the MACE hazard ratio corrected for study dropouts via censoring. We believe FibroGen withheld the primary safety endpoint because roxadustat failed to achieve non- inferiority to placebo PLAINVIEW Roxadustat’s MACE HR is Inferior to Placebo 37
NDD-CKD MACE HR Failed to Achieve Non-Inferiority • Like DD-CKD, FibroGen declined to provide the MACE HR for NDD-CKD. We chalk this up to roxadustat failing to achieve MACE non-inferiority compared to control • This is not surprising: if roxadustat could not show MACE non-inferiority to EPO and EPO is known to increase MACE risk, one would expect that roxadustat would also be inferior to placebo in terms of MACE risk • FibroGen did not tout a lower event count for roxadustat in NDD-CKD – we believe this is at least in part because the NDD-CKD placebo dropout rates were closer to roxadustat than the DD-CKD EPO dropout rate (see slide 18), and thus the event count for NDD-CKD was not as skewed by dropouts PLAINVIEW Roxadustat’s MACE HR is Inferior to Placebo 38
Intention-To-Treat Vs. Kaplan-Meier • Instead of providing the actual NDD-CKD MACE HR, FibroGen told investors that the NDD-CKD ITT MACE HR was non-inferior to placebo • Intention-to- treat (“ITT”) is typically used for evaluating drug efficacy (particularly in cancer) in order to avoid survivorship bias in determining how effective a drug is: patients who respond to a drug are more likely to continue taking it than patients who do not respond • ITT is completely inappropriate for evaluating safety because it dilutes the safety data by including safety data of patients who are not taking the drug • The standard method for evaluating safety measures such as MACE is a Kaplan-Meier (K-M) analysis. K-M analyses accounts for dropouts by censoring them. Patients are included in the safety population until they stop taking the drug after which they are excluded from analysis – because they are no longer taking the drug PLAINVIEW Roxadustat’s MACE HR is Inferior to Placebo 39
Roxadustat NDD-CKD NDA Based On K-M HR, Not ITT HR • On the Q2 2019 call, Management discussed the results of its pre-NDA FDA meeting (emphasis added): “ And in terms of the way that the time to MACE primary endpoint is being analyzed in nondialysis, this will account for differential drop out between our drug and placebo , whereas you know that because placebo doesn't work in treating anemia, placebo patients had a tendency to drop out earlier. And we have reached agreement on statistical method that accounts for that .” • Management did not mention “ITT” at any point during the Q2 2019 call • Differential dropout rates are accounted for via Kaplan-Meier censoring; ITT is diluted by dropouts. Barring some other nonconventional way of accounting for dropouts, this confirms that the NDA will use the conventional Kaplan-Meier analysis MACE HR rather than the ITT MACE HR PLAINVIEW Roxadustat’s MACE HR is Inferior to Placebo 40
Roxadustat Clinical Data Summary • All evidence points to roxadustat producing significantly higher MACE HR compared to EPO/placebo: Dropout rate is much higher in roxadustat group vs. EPO group across trials for all HIF-PH inhibitors, pointing to serious safety issues and allowing for lower total event rates and statistically significant inferiority at the same time Reported MACE rate of 23/983 in roxadustat vs. 1/419 in control Reported mortality rate of 11/983 in roxadustat vs. 0/419 in Similar MACE/mortality imbalances in competing HIF-PH inhibitors, indicating class effect FibroGen Management made misleading statements & refused to disclose actual MACE HRs PLAINVIEW Roxadustat’s MACE HR is Inferior to Placebo 41
Point #4: MACE+ Inflated by Prohibiting IV Iron & Using Placebo Comparator • The principal safety concern for ESAs/HIF-PH inhibitors is increased risk of thrombosis which manifests as MACE: stroke, myocardial infarction, and death • The “+” in MACE+ is hospitalizations for congestive heart failure/unstable angina – which are frequently driven by anemia rather than thrombosis. Looking at MACE+ instead of MACE masks the real risk (potentially fatal blood clots) with temporary symptoms driven by an unrelated cause • FibroGen prohibited IV iron, a key part of EPO therapy that is not necessary for HIF-PH inhibitors. Doing so increased anemia in control groups and thus artificially inflated the control group MACE+ rates in the DD-CKD trials • In the NDD-CKD trials, a placebo comparator combined with the IV iron ban inflated control group anemia and MACE+ rates • MACE+ noninferiority will not make regulators overlook clear MACE inferiority MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 42 Comparator
EPO Requires IV Iron & Roxadustat Does Not • Iron is a key component of hemoglobin • Using exogenous EPO in patients skips steps of the HIF pathway, including those that enhance iron absorption and transport in the body • Because iron absorption/transport is not enhanced in EPO patients, IV iron supplementation is frequently essential in EPO patients to ensure sufficient iron is available for erythropoiesis • HIF-PH inhibitors (including roxadustat) drastically reduce the need for IV iron by improving iron absorption/transport MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 43 Comparator
Prohibiting IV Iron Inflated Anemia Rate in Control Group • FibroGen noted in its December 2018 topline press release that “ …roxadustat -treated patients had a 33% reduction in the risk of blood transfusion compared to epoetin alfa (HR=0.67) in the time to first blood transfusion during treatment, p=0.0337. ” Note that a 33% reduction in risk in the roxadustat population equates to a ~50% increase in risk in the EPO population (1/.67 = 1.49) • FibroGen attributes the decreased risk of rescue therapy to roxadustat’s ability to treat EPO-resistant patients. This claim is wrong – as we show in Point #6 of this presentation, HIF-PH inhibitors do not work in true EPO hyporesponders • The difference is attributable to the fact that IV iron is a key part of EPO therapy and it was prohibited (except as rescue therapy) in FibroGen’s trials MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 44 Comparator
Prohibiting IV Iron Increases Anemic Events By Reducing EPO Efficacy • In the roxadustat Phase 2 trial (Provenzano et al 2016), the investigators prohibited IV iron use in both groups • The EPO Hb response rate was only 33% compared to 79% Hb response rate in the high-dose roxadustat group.* However, this trial screened out all EPO hyporesponders prior to randomization – the fact that only 33% of EPO patients in the trial achieved Hb responses is reflective of the impact of prohibiting IV iron, not the efficacy of EPO treatment, and demonstrates that restricting IV iron has a far greater impact on EPO patients than on roxadustat patients • This demonstrates the enormous impact that prohibiting IV iron has on successful treatment with EPO: Kharagjitsingh et al 2005 conducted a retrospective study and concluded that the true EPO hypo-response rate is only 3.3% (57/1,677 patients) with an even lower incidence rate of 16.7 per 1,000 patient-years *Roxadustat was dosed at 1.5-2.0 mg/kg in the high- dose group, comparable to roxadustat’s Phase 3 dosing of 100 mg starting dose in patients <60 kg and 120 mg in patients >60 kg) MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 45 Comparator
Anemia Drives CHF/Unstable Angina • In Bhavanadhar et al 2016, the authors found that 58.2% of patients hospitalized for unstable angina were experiencing anemia. The authors noted that anemia drives angina and CHF, citing the textbook Clinical Hematology (Wintrobe 2008): “It is well established in patients with known CVD that anaemia has adverse effects on myocardial oxygenation resulting in provocation or acceleration of angina, and may worsen Congestive Heart Failure (CHF).” • Anemia, unstable angina, and CHF symptoms are all caused by reduced oxygenated blood flow and present with similar symptoms. They also frequently occur concurrently, making it difficult if not impossible for adjudicators to distinguish between hospitalization for anemia and hospitalization for angina/CHF unrelated to anemia MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 46 Comparator
Anemia Vs. Unstable Angina/CHF • Symptoms of anemia, unstable angina, and CHF are all very similar and frequently occur concurrently, raising issues with adjudicating “+” events for MACE+: Anemia Unstable CHF Angina Dyspnea Fatigue Weakness Edema Chest Pain Dizziness MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 47 Comparator
Prohibiting IV Iron Contradicts Real-World Practices • Macdougall 2017 provides a concise summary of prior trials comparing IV iron with oral/no iron in patients on EPO and shows that IV iron enhanced Hb responses with lower EPO requirements compared to oral or no iron • Dosing anemic hemodialysis patients with EPO + IV iron is standard practice: the USRDS 2018 data report found that in May 2017, 78% of dialysis patients were on EPO and 62% of dialysis patients were on IV iron • Iron is required to complete erythropoiesis, and prohibiting parenteral iron paints an unrealistically favorable picture of roxadustat’s relative efficacy MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 48 Comparator
Prohibiting IV Iron Inflates EPO Group “+” in MACE+ • GSK has published results for two 24-week trials of its HIF-PH inhibitor daprodustat vs. EPO: one where IV iron was not restricted (Meadowcroft et al 2018), and one where IV iron was restricted (Holdstock et al 2018) • MACE rates were higher for daprodustat in both trials, but MACE+ was lower for daprodustat in the iron-restricted trial. In the iron-restricted trial, 2/170 daprodustat patients reported “+” events vs. 4/80 EPO patients • In the trial where iron was not restricted, the “+” event rates were similar. 3/177 daprodustat patients reported “+” events vs. 1/39 EPO patients MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 49 Comparator
Prohibiting IV Iron Also Forces Increased EPO Dosing • In its May press release, FibroGen EPO dosing is stable @ 52 weeks in CHOIR trial while prohibiting IV iron touted that in one Phase 3 DD-CKD trial, forced a 57% increase in EPO dosing in EPO dosing increased by 57% over 52 FibroGen’s SIERRAS trial weeks while roxadustat dosing stayed stable: “In SIERRAS, roxadustat dose requirements remained stable over time, while epoetin alfa dose requirements increased by 57% over 52 weeks in the epoetin arm.” • This is evidence of the adverse effect of restricting IV iron on EPO patients; other EPO trials show stable EPO dosing Source: CHOIR Trial (Singh et al 2006) over 52 weeks (and beyond). See right for the EPO dosing from the CHOIR trial: MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 50 Comparator
Increased EPO Dosing Worsens Event Profile • In the CHOIR trial, investigators found that the high Hb group experienced composite events significantly more frequently compared to the low Hb group, with a hazard ratio of 1.34 (1.03 to 1.75) • Patients in the high Hb group were, on average, receiving roughly double the EPO dose of the low Hb Source: CHOIR Trial (Singh et al 2006) group (see prior slide) MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 51 Comparator
MACE is Principal Measure of Safety • FibroGen’s MACE+ results are a red herring driven by FibroGen’s decisions to 1) prohibit patients from using parenteral iron and 2) use placebo comparators in the NDD-CKD trials, which artificially inflated the anemia rates in the control groups and thus control group MACE+ rates • The primary concern with EPO and HIF-PH inhibitors is that they can cause thrombosis leading to MACE; the “+” in MACE+ is particularly irrelevant to this given FibroGen’s trial protocol • We note that the primary safety endpoint for the FDA NDA is MACE and on p. 47 of its Q2 2019 10-Q, FibroGen (for the first time) disclosed that the EMA will be reviewing MACE as well • The entire point of a primary safety endpoint is to weed out unsafe drugs – MACE inferiority would show that roxadustat is not safe regardless of MACE+ MACE+ Inflated by Prohibiting IV Iron & Using Placebo PLAINVIEW 52 Comparator
Point #5: HIF-PH Inhibitor Mechanism of Action is Flawed • EPO is part of the HIF pathway and lies downstream from HIF. HIFs have been shown to induce hypertension, fibrosis, immune suppression, and other adverse effects through EPO-independent mechanisms • HIF-PH inhibitors are flawed compared to EPO because EPO is highly specific to erythropoiesis and HIFs are not • The pro-HIF-PH inhibitor theory is that administering HIF-PH inhibitors lowers total EPO exposure, which improves safety. The anti-HIF-PH inhibitor thesis is that the pleiotropic effects of HIFs are more dangerous than higher EPO exposure • The large MACE & mortality imbalances and higher dropout rates in published HIF- PH inhibitor trial data show that HIFs’ problematic effects vastly outweigh lower systemic exposure to EPO PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 53
Review: The HIF-EPO Pathway • Under normal conditions (normoxia), HIF- α (HIF - 1α and HIF - 2α) is eliminated via ubiquitination by the Von Hippel Lindau E3 ligase (VHL) after being tagged by prolyl hydroxylase domain-containing proteins (PHD) • Under hypoxic conditions, HIF- PHDs cease to function, which allows for HIF- α to enter the nucleus and activate gene transcription: mimicking this by blocking PHDs is the intended mechanism of action of roxadustat Source: Chen & Lou 2017 and other HIF-PH inhibitors PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 54
Review: The HIF-EPO Pathway • With the inhibition of HIF- PHs, HIF stimulates transcription of target genes • In addition to the ESA (EPO) and iron homeostasis genes in the diagram, HIF- α (HIF - 1α & HIF- 2α) also activates transcription of other genes that serve no purpose in erythropoiesis • These include genes linked to hypertension, fibrosis, and other adverse effects Source: FibroGen presentation PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 55
Hypoxia Induces Hypertension • Patients with sleep apnea (a disorder where breathing repeatedly stops and starts while the patient is sleeping) are exposed to repeated hypoxia and offer insight into the effects of chronic activation of the HIF pathway • Studies identify sleep apnea as a major contributor to increased risk of hypertension and MACE: • Peppard et al 2000 found a dose-response association between hypoxic events and hypertension in sleep apnea patients • Lin et al 2017 compared outcomes for 2,699 patients with obstructive sleep apnea (OSA) and 13,490 non-OSA patients, and found that the MACE incidence rate was 1.95-fold in OSA patients compared to non-OSA • Yacoub et al 2018 estimated the prevalence of sleep apnea in drug-resistant hypertension at 80% PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 56
HIF- 1α is Implicated in Hypertension • Yamashita et al 2001: HIF- 1α binds to an ET -1 promoter, which activates ET-1 expression in endothelial cells. ET-1 acts as a potent vasoconstrictor, and is linked to hypertension & heart failure progression • Diebold et al 2010: HIF- 1α activates NOX4 gene, which leads to increased reactive oxygen species and pulmonary arterial smooth muscle cells which have been implicated in pulmonary hypertension • Eckle et al 2014: HIF- 1α is a transcriptional regulator of the A2B adenosine receptor (ADORA2B) which modulates development of pulmonary hypertension by mediating the release of hyaluronan, an extracellular matrix component implicated in development of pulmonary hypertension • Nanduri et al 2015: HIF- 1α upregulation leads to a transcriptional imbalance favoring pro-oxidant genes, and those genes drive increased generation of reactive oxygen species which cause hypertension PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 57
HIF- 2α is Implicated in Pulmonary Hypertension • Cowburn et al 2016: The rise in right ventricular systolic pressure (RVSP) normally observed following chronic hypoxic exposure was absent in mice with pulmonary endothelial HIF- 2α deletion without affecting hypoxic Hb response, implicating HIF- 2α • Kapitsinou et al 2016: Inactivation of PHD2 (a target of roxadustat and other HIF- PH inhibitors) led to a 118% increase in right ventricular systolic pressure, and the development of pulmonary hypertension was dependent on HIF- 2α • Bryant et al 2016: Deletion of HIF in mice protected them from development of pulmonary hypertension following induction of lung fibrosis via bleomycin. HIF deletion also protected mice from pulmonary hypertension following chronic hypoxia • Dai et al 2016: Removing PHD2 (HIF-PH inhibitor target) from mice led to severe HIF- 2α -mediated pulmonary arterial hypertension, with 12/15 mice dying within 5 months vs. 0/15 control mice PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 58
HIFs & HIF-PHis are Implicated in Fibrosis • Higgins et al 2004: HIF-1 α induces CTGF transcription, which may mediate pro-fibrotic effects of TGF- β 1 • Higgins et al 2007: HIF-1 α leads to fibrosis by stimulating epithelial-to- mesenchymal transition by upregulating lysyl oxidase genes and increasing expression of extracellular matrix modifying factors • Ueno et al 2011: TGF- β 1 upregulates 58 genes through an HIF-1 α -dependent mechanism, including pro-fibrotic genes such as PAI-1 • Baumann et al 2016: HIF-1 α stimulates COL1A2 transcription which is implicated in renal fibrosis. HIF-1 α knockout conferred renal protection to mice • Li et al 2019: HIF-PH inhibition led to tubulointerstitial fibrosis via HIF-1 α - induced transcription of KLF5 which promotes fibrosis via TGF- β 1 PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 59
HIF-1 α Implicated in Atherosclerosis through Multiple Pathways • Gao et al 2012 postulated that HIF- 1 α is involved in the development of atherosclerosis via multiple pathways (see diagram) Source: Gao et al 2012 PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 60
HIF-PH Inhibition Has Many Off-Target Effects • Prolyl-hydroxylase domains are not exclusive to regulating HIFs: Strowitzki et al 2019 summarizes other prolyl hydroxylase targets that would be impacted by HIF-PHis, including NF- κ B (via IKKb), p53, FOXO3a, MAPK6, Cep192, and ZHX2. The authors note that “ hundreds of genes have been reported Source: Strowitzki et al 2019 to be sensitive to hypoxia. ” PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 61
…And Roxadustat/HIF -PHis Aren’t Even Specific to HIF -PHs • HIF-PH inhibitors (roxadustat in particular) are not specific to the HIF-PHs: Kiriakidis et al 2017 found that dosing mice with 10 mg/kg roxadustat led to a 50% decline in C1q secretion, negatively impacting immune responses. • The authors concluded that this was likely due to roxadustat also inhibiting the collagen prolyl-4-hydroxylases, which are necessary for formation of C1q PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 62
HIF-PH Inhibition Suppresses Innate Immune System • Kiriakidis et al 2017 also noted that the study Schindler et al 2016 found that HIF- PH inhibition in septic mice led to significantly higher mortality rates. This is likely a product of the HIF-PHi-induced immune suppression via C1q suppression (see prior slide): 3,4-DHB is an HIF-PH inhibitor. Administration of 3,4-DHB in septic (CLP or PCI) mice led to significantly worse survival outcomes compared with control PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 63
Other Adverse Effects of HIFs • Cheng et al 2014 concluded that HIF- 1α collaborates with Notch -1 to induce neuronal death in ischemic strokes, and that inhibiting HIF- 1α decreased neuronal death • While companies developing HIF-PH inhibitors have not admitted to any pro- oncogenic safety signals, HIF is also strongly implicated in cancer • It has also been shown that HIF- 1α and HIF - 2α expression in tumors is associated with worse progression-free survival and overall survival • Peloton Therapeutics’ HIF - 2α inhibitor PT2385 produced a 22% monotherapy ORR in 2L+ ccRCC patients while its HIF- 2α inhibitor PT2977 produced a 24% ORR in 2L+ mRCC patients PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 64
HIF Exposure Greatly Outweighs Reduced EPO Exposure • There are likely multiple factors driving the imbalance in deaths, MACE and dropout rates between HIF-PH inhibitors and EPO • The pro-roxadustat/HIF-PH inhibitor thesis is that EPO exposure is the principal culprit in deaths/MACE in both EPO and HIF-PH inhibitor patients and HIF-PH inhibitors can achieve equivalent increases in Hb with lower EPO exposure; therefore, HIF-PH inhibitors should reduce MACE risk • The dramatic MACE/death imbalance seen in clinical data showed that this theory was wrong . The consistently higher dropout rate in the HIF-PH inhibitor groups across trials also pose additional safety concerns, and the real- world safety difference is even greater than FibroGen’s trials suggest due to FibroGen’s disingenuous decision to prohibit IV iron, which inflated EPO dosing in the control group far beyond where it would have been under real-world conditions PLAINVIEW HIF-PH Inhibitor Mechanism of Action is Flawed 65
Point #6: Roxadustat Does Not Work in EPO Hyporesponders • FibroGen references two pieces of evidence to show that roxadustat can adequately treat EPO hyporesponders. Both are misleading and wrong: • Provenzano et al 2016 showed that roxadustat seemed to have similar impact on patients regardless of CRP levels, with the implicit assumption that CRP is the key determinant of EPO hyporesponsiveness. The problem is that Provenzano 1) screened out EPO hyporesponders during enrollment and 2) did not include the roxadustat non-responders (16%, 9/58) in the analysis. The only real takeaway is that roxadustat works in 100% of the subgroup that roxadustat works in • In multiple studies, roxadustat patients achieved higher Hb levels compared to EPO patients. This was a result of FibroGen prohibiting use of IV iron, a critical part of EPO therapy – not of superior efficacy. Competitor HIF-PH inhibitor trials permitting IV iron showed no difference in Hb • GlaxoSmithKline conducted a study of its HIF-PH inhibitor daprodustat among EPO-hyporesponsive patients, and found that daprodustat produced no Hb response in those patients. This is likely true for the entire HIF-PH inhibitor class PLAINVIEW Roxadustat Does Not Work in EPO Hyporesponders 66
FibroGen’s EPO Hyporesponse Theory • Inflammation drives an increase in hepcidin, which binds to ferroportin and thus suppresses absorption and extracellular release of iron • HIFs are known to decrease hepcidin levels; this improves iron transport for patients • The theory presented by FibroGen is that the principal cause of EPO hyporesponsiveness is inflammation limiting iron absorption and transportation, and using HIF-PH inhibitors helps combat this by decreasing hepcidin levels and increasing iron absorption/transport • FibroGen’s principal support for this theory comes from Provenzano et al 2016, where the authors measured inflammation based on C-reactive protein (CRP) levels and used CRP as a proxy for EPO hyporesponsiveness. The theory is that inflammation (high CRP) causes EPO hyporesponse, so therefore if roxadustat works in patients with high CRP levels, then this means it will work in EPO hyporesponders. As we show, the study was extremely misleading and wrong PLAINVIEW Roxadustat Does Not Work in EPO Hyporesponders 67
Problem #1: Provenzano Excluded Roxadustat Non-Responders • There are two problems with the Provenzano et al 2016 analysis: first, the authors excluded all the patients who discontinued roxadustat due to lack of efficacy: Excluding patients for lack of efficacy makes the subsequent declaration that CRP levels do not affect roxadustat efficacy completely meaningless – of course roxadustat works for all patients after excluding the 16% of patients it did not work for! PLAINVIEW Roxadustat Does Not Work in EPO Hyporesponders 68
Problem #2: Provenzano Excluded EPO Hyporesponders from Enrolling • The second problem is that EPO hyporesponders were explicitly excluded from the trial! Hard to conclude that a drug works • From the S1 Supplement to Provenzano 2016: in EPO hyporesponders when EPO hyporesponders are excluded from enrolling in the trial PLAINVIEW Roxadustat Does Not Work in EPO Hyporesponders 69
HIF-PH Inhibitors Do Not Work in EPO-Hyporesponsive Patients • GlaxoSmithKline conducted an intellectually honest study of its HIF-PH inhibitor daprodustat among erythropoietin- hyporesponsive patients, and saw no change in Hb after transitioning EPO-hyporesponsive patients to daprodustat Virtually no change in Hb at all time points Source: Cizman et al 2018 PLAINVIEW Roxadustat Does Not Work in EPO Hyporesponders 70
Iron- Mediated EPO Hyporesponses Already Solved…With IV Iron • The need for HIF-PH inhibitors to modulate iron transport is an invented problem: the iron transport problem has already been solved with IV iron, which leaves no benefit to using HIF- PH inhibitors except in FibroGen’s unrealistic trials where patients are denied IV iron except as emergency rescue therapy • IV iron avoids the problems with iron transport/absorption by putting the iron directly into the bloodstream. Ueda & Takasawa 2018 conducted a meta-analysis of nine studies comparing use of IV iron and oral iron in anemic CKD patients, and concluded that “ One of the reasons accounting for superiority of IIT [IV iron therapy] over OIT [oral iron therapy] to maintain tHb [target Hemoglobin] in CKD patients may be the difference in the efficacy of IIT and OIT in CKD patients in the presence of concomitant inflammation that increases ferritin and hepcidin. ” PLAINVIEW Roxadustat Does Not Work in EPO Hyporesponders 71
Higher Hb Levels in Roxadustat Group Due to Study Design, Not Efficacy • Another part of FibroGen’s roxadustat pitch is that roxadustat is capable of achieving higher Hb levels than EPO • In reality, this appears to be driven by the previously-discussed decision to prohibit IV iron rather than efficacy • For instance, in Provenzano 2016, the authors concluded that the Hb responder rate for the EPO group was 33% compared to 79% for the pooled high- dose roxadustat group. This doesn’t make sense at all considering that the supplement to Provenzano 2016 (relevant graphic on slide 69) showed that EPO hyporesponders were already screened out of the trial – unless the EPO hyporesponses were driven by prohibition of IV iron PLAINVIEW Roxadustat Does Not Work in EPO Hyporesponders 72
Competitor Trials Show HIF-PHi & EPO Hb Responses Are Similar • Examining competitors’ long -term DD-CKD trials, we see no meaningful difference between Hb levels achieved in drug vs. control groups (see below table). Roxadustat utilizes the same mechanism of action and we believe the claimed benefit is primarily a function of FibroGen suppressing the results of the EPO group by withholding IV iron PLAINVIEW Roxadustat Does Not Work in EPO Hyporesponders 73
Pamrevlumab PLAINVIEW
Pamrevlumab Table of Contents ❑ Pamrevlumab Background: Lots of Trials, Few Results ❑ CTGF is a Biomarker, Not a Principal Mediator ❑ No Value in IPF Even If Phase 3 Trial Succeeds • FibroGen withheld results from a trial comparing pamrevlumab + SoC vs. placebo + SoC, indicating that the trial likely failed. This means that pamrevlumab offers no additive benefit to SoC which is likely going generic before pamrevlumab could possibly make it to market • We have concerns about pamrevlumab’s efficacy and Management appears to share our concerns: FibroGen only initiated a single Phase 3 despite long trial duration (4+ years expected) and a precedent that clearly shows two Phase 3 trials are required for approval. At best, pamrevlumab will get approval for IPF in 9-10 years ❑ No Evidence Pamrevlumab Works in DMD • Provided virtually no data during recent presentation given 3.5 years after initiating trial ❑ Pamrevlumab Doesn’t Work in Pancreatic Cancer • FibroGen’s pancreatic cancer trial results showed no difference in ORR between pamrevlumab and control. PFS and OS were withheld, and the claimed resection benefit is wildly misleading • Targeting the stroma has already been tried via hedgehog inhibitors and has failed PLAINVIEW 75
Pamrevlumab Background • Previously known as FG-3019 • Currently in development for Duchenne muscular dystrophy (DMD), locally advanced pancreatic cancer (LAPC), and idiopathic pulmonary fibrosis (IPF) • Was previously in development for diabetic nephropathy, focal segmental glomerulosclerosis (FSGS), and Hepatitis B • Paucity of published clinical data despite 16 years in the clinic • Currently in Phase 3 trials for IPF and LAPC • LAPC Phase 3 trial expected to reach primary completion in September 2022 • IPF Phase 3 trial expected to reach primary completion in March 2023 PLAINVIEW Pamrevlumab Background 76
Lots of Trials, Few Results Over 16 Years • December 2003: Phase 1 in IPF posted • January 2005: Phase 1 in diabetic nephropathy initiated • March 2008: Phase 1 for pamrevlumab in diabetes mellitus & diabetic nephropathy initiated • April 2008: Phase 1 trial in FSGS initiated • December 2008: Phase 1 trial in pancreatic cancer initiated • February 2009: Phase 2 RCT in CKD initiated (terminated) • August 2010: Phase 2 RCT in Hepatitis B initiated (terminated) • March 2011: Phase 2 for IPF initiated • July 2014: Phase 1/2 controlled trial for pancreatic cancer initiated • November 2015: Phase 2 trial in DMD initiated • May 2019: Phase 3 trial in locally advanced pancreatic cancer initiated • June 2019: Phase 3 trial in IPF initiated PLAINVIEW Pamrevlumab Background 77
Management’s Sleaziness Permeates Into Pamrevlumab • In a February 2016 investor call, Fibrogen Management told investors that they were “encouraged” by Phase 2 results for pamrevlumab for liver fibrosis resulting from Hepatitis B and implied that pamrevlumab was reversing fibrosis: “ The result showed a trend for dose response improvement in liver fibrosis at doses of FG- 3019 similar to those now being tested in other diseases. We are encouraged by these results, and we continue to evaluate options for trial of FG-3019 in subjects with advanced liver fibrosis due to NASH. In addition, we are evaluating the potential for trial of FG-3019 in subjects with advanced liver fibrosis due to hepatitis C, where it appears that the fibrosis may not resolve after cure of the hepatitis with current potent anti-viral therapy. ” • In reality, Management had already prematurely terminated the HBV trial in June 2015 for futility (note that the primary completion date is June 2015)-- eight months before claiming to be “encouraged” by the results. The promised additional trials in HBV/HCV/NASH never came to fruition • Management also ran a six-month Phase 2 trial in diabetic nephropathy and terminated the trial 18 months later (January 2011) – after the scheduled primary completion date. No results or explanation for termination were ever given outside of the termination being a “strategic decision”. Diabetic nephropathy is an enormous indication, so this strategic decision was likely driven by lack of efficacy PLAINVIEW Pamrevlumab Background 78
Lots of Trials, Few Results • Pancreatic Cancer: FibroGen issued a press release touting that 33.3% of pamrevlumab patients achieved clinical resection compared to 7.7% of control, a meaningless product of data-mining. The PR left out the fact that ORR was 23% in the control group compared to 21% in the pamrevlumab group which was published in an ASCO 2018 poster. PFS and OS data were never published • DMD: In a bizarre presentation at PPMD 2019, FibroGen avoided disclosing efficacy endpoints in the uncontrolled trial – instead only stating that certain endpoints performed better than cherry-picked publications • IPF: As part of the PRAISE trial, FibroGen conducted an RCT comparing pamrevlumab + standard-of-care vs. placebo + standard-of-care. No efficacy results were ever announced or published. This trial is key because it informs whether pamrevlumab adds any benefit to standard-of-care, which is expected to go generic before pamrevlumab could possibly make it to market PLAINVIEW Pamrevlumab Background 79
Point #1: CTGF is a Biomarker, Not a Principal Mediator • FibroGen paints connective tissue growth factor (CTGF) as the principal cause of fibrosis; most other research implicates TGF- β 1 as the master regulator of fibrosis and research papers and primers on the subject pay little attention to CTGF • In drug development, validated targets are quickly copied; often before they are even proven in humans. Pamrevlumab has been in the clinic for 16 years with no imitators. Many drug developers are targeting fibrotic diseases but outside of FibroGen, none are targeting CTGF • While we don’t dispute that CTGF mediates some effects, we believe its impact is not clinically significant and that CTGF is more of a biomarker of fibrosis than a key driver of fibrosis PLAINVIEW CTGF is a Biomarker, Not a Principal Mediator 80
What is CTGF? • CTGF is an extracellular matrix protein (“ECM”) and “connective tissue growth factor” is a misnomer: “ the name CTGF (connective tissue growth factor)…implies activities and a mechanism of action akin to those of classical growth factors, a notion that has not been supported by experimental evidence to date. ” (emphasis added, Chen & Lau 2009) • Fibrosis involves the growth and proliferation of many ECM proteins and CTGF is only one of them • Pamrevlumab’s lack of monetizable or even decisive clinical results after 16 years of trials suggests that the impact of targeting CTGF is simply not meaningful for patients suffering from fibrosis • Based on published research, we believe that CTGF is more of a biomarker of fibrosis and pancreatic cancer than a valid target with clinical value PLAINVIEW CTGF is a Biomarker, Not a Principal Mediator 81
How FibroGen Describes CTGF (Central Mediator of Fibrosis) Source: FibroGen 2010 Publication PLAINVIEW CTGF is a Biomarker, Not a Principal Mediator 82
How Fibrosis Researchers Describe CTGF (They Don’t) • CTGF is not thought to be a principal cause of fibrosis and is rarely discussed meaningfully in publications not sponsored by FibroGen. To the right is an illustration of the fibrotic pathway taken from Meng et al 2016; CTGF is nowhere to be found. PLAINVIEW CTGF is a Biomarker, Not a Principal Mediator 83
Point #2: No Value in IPF Even If Phase 3 Trial Succeeds • We question whether pamrevlumab will succeed in its IPF Phase 3 trial given that: • Pamrevlumab showed no dose response in IPF • It is not clear why pamrevlumab would work in IPF after failing in other fibrotic indications • As part of the PRAISE trial, FibroGen conducted an RCT comparing pamrevlumab + SoC vs. placebo + SoC in IPF which concluded in 2017 – and never announced or published the efficacy results, a near-certain sign that pamrevlumab added no benefit to SoC. This likely informed the decision to prohibit use of SoC as background therapy in the Phase 3 trial • The lack of additive benefit to SoC makes pamrevlumab an impossible sell even assuming Phase 3 success: • Pamrevlumab offers no value. At least one of the current SoCs is widely expected to go generic before pamrevlumab makes it to market. With no head-to-head or combination therapy data, there is no reason for payors to cover pamrevlumab as an alternative or for anyone to prescribe it • Pamrevlumab is less convenient. Pamrevlumab requires Q3W IV administration compared to QD oral dosing. Patients are more likely to prefer daily oral dosing • Management’s decision to only run a single four -year Phase 3 trial now when two Phase 3s are required for approval indicates that Management shares our skepticism about IPF PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 84
No Pamrevlumab Dose Response in IPF • The lines, p-value, and blue shading FibroGen drew into the graph are lipstick on a pig: anyone can look at this graph and plainly tell that there is virtually no relationship between change in Forced Vital Capacity (FVC; measure of lung volume) and drug concentration. Source: FibroGen September 13, 2017 8-K PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 85
Missing Trial Results Likely Due to Missing Efficacy • FibroGen conducted an RCT comparing pamrevlumab + pirfenidone/nintedanib compared to placebo + pirfenidone/nintedanib. It was a 24 week study with an additional 4-10 weeks of screening/follow-up. Enrollment completed on December 16, 2016, and the trial should have concluded in mid-2017 • The efficacy endpoint results were never disclosed, indicating that pamrevlumab did not offer any benefit to patients over placebo • This likely informed the Phase 3 trial design in which FibroGen decided to prohibit use of pirfenidone/nintedanib as background therapy PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 86
Prohibiting Standard-of-Care is Blatantly Unethical • There is strong evidence that pirfenidone (and to a lesser extent, nintedanib) significantly increases survival for IPF patients. It is unethical to enroll a placebo group where patients are denied access to these drugs. This will make it very challenging to fully enroll the trial and we think it is very likely that FibroGen’s first Phase 3 trial will take significantly longer than the estimated four years to reach the primary endpoint • Nathan et al 2017 conducted a pooled analysis of pirfenidone Phase 3 trials and found that the all-cause mortality hazard ratio for pirfenidone was 0.52 (0.31, 0.87) • Lancaster et al 2019 conducted a pooled analysis of nintedanib Phase 2/3 trials and found that the all-cause mortality hazard ratio for nintedanib was 0.70 (0.46, 1.08) PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 87
Generic IPF Competition • It is very likely that pirfenidone, nintedanib, or both will be generic by the time the first pamrevlumab Phase 3 trial concludes • Paragraph IV filings have been submitted for both pirfenidone and nintedanib • At least one company has submitted a Paragraph IV filing for nintedanib • Pirfenidone is particularly vulnerable, as the composition of matter patent covering the drug has long been expired. At least 18 generic manufacturers have filed Paragraph IVs for pirfenidone, including: • • • • Laurus Labs Granules Pharmaceuticals Teva Shilpa Medicare • • • • Aurobindo Alembic Pharmaceuticals Accord Healthcare Scigen Pharms • • • • Lupin Hetero Labs Macleods Pharmaceuticals Sandoz • • • Micro Labs Amneal Pharmaceuticals Cipla • • • Apotex Torrent Pharmaceuticals Aizant Drug Research Solutions PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 88
Phase 3 Trial Design is Clinically Irrelevant • While using a placebo control group makes it much easier for pamrevlumab to show benefit, it also makes the results clinically irrelevant • Competing IPF drugs in Phase 3 (PRM-151 and GLPG1690) are testing for efficacy in combination with pirfenidone/nintedanib: without proving superiority or additive benefit to those drugs, there is no selling point PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 89
GLPG1690 Effective in Small IPF Study • GLPG1690 is an oral autotaxin inhibitor. Galapagos theorizes that autotaxin production of lysophosphatidic acid has a pathogenic role in IPF, and that inhibiting autotaxin could slow IPF progression • The FLORA trial was a Phase 2a study in which Galapagos enrolled 23 patients (17 in GLPG1690 group, six in control group) • GLPG1690 achieved a mean increase in FVC of 25 mL compared to a -70 mL decline in FVC in the control group (Maher et al 2018, ppFVC not provided) • While the FLORA trial excluded patients on pirfenidone/nintedanib, the two Phase 3 trials (ISABELA1 and ISABELA2) are enrolling patients on background pirfenidone/nintedanib therapy PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 90
PRM-151 Succeeded in IPF With Background Therapy • PRM-151 is recombinant human serum amyloid P, and the theory behind its use in fibrosis is that PRM-151 can prevent differentiation of cells into M2- macrophages and fibrocytes, reducing fibrosis in patients • The Phase 2 results for PRM-151 (Raghu et al 2018) show that it produced a statistically significant benefit in ppFVC in patients allowed background pirfenidone/nintedanib, with 48% lower decline in ppFVC over 28 weeks (- 2.5% vs -4.8%, p = 0.001) • 78% of enrolled patients were receiving concurrent pirfenidone or nintedanib • 95% (74/78) of PRM-151 patients completed treatment compared to 95% (37/39) of control patients and SAEs occurred in 8% of PRM-151 patients and 10% of control patients, indicating that the drug was well-tolerated PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 91
Needs Two Phase 3 Trials For Approval…Only Running One • As is clear from precedents pirfenidone/nintedanib and GLPG1690, two pivotal trials are required for IPF approval. FibroGen is only running one: the ZEPHYRUS trial • Galapagos is running two concurrent Phase 3 trials (ISABELA1 and ISABELA2) for GLPG1690 • InterMune conducted three Phase 3 trials for pirfenidone in IPF (one of which failed to achieve primary efficacy endpoint) • Boehringer Ingelheim conducted two Phase 3 trials for nintedanib in IPF • Running a single trial now implies that Management will wait for that trial to read out before deciding on a second Phase 3 trial. A projected 2023 initial Phase 3 readout and a 4-year estimated trial timeline puts earliest potential FDA approval for pamrevlumab in IPF at nearly a decade from today • The decision to only run one trial in the face of multiple competitors speaks to a major lack of confidence in pamrevlumab’s IPF prospects PLAINVIEW No Value in IPF Even if Phase 3 Trial Succeeds 92
Point #3: No Evidence Pamrevlumab Works in DMD • FibroGen presented “preliminary” results of its uncontrolled Phase 2 Duchenne’s Muscular Dystrophy (DMD) study at PPMD on June 28, 2019 • The trial began recruiting in November 2015 and enrollment was completed in April 2018. All patients had been enrolled >1 year by June 2019 • Virtually no actual data was shared in the presentation. FibroGen instead shared abstract comparisons with historical precedents without revealing actual trial datapoints from the pamrevlumab DMD trial in an apparent attempt to imply efficacy without proving efficacy PLAINVIEW No Evidence Pamrevlumab Works in DMD 93
Primary Endpoint & Most Secondary Endpoints Not Disclosed • The trial description for pamrevlumab in DMD lists a primary endpoint (annual change in FVC) and 12 secondary endpoints • FibroGen did not disclose the primary endpoint change – other than to say that change in ppFVC was higher than the change in ppFVC seen in a selected prior DMD publication • Only two of the 12 secondary endpoints were disclosed (LVEF and change in performance of upper limbs), though other metrics were said to be higher or lower than arbitrary historical precedents • If the data were favorable, FibroGen would have shared it instead of making abstract comparisons with favorable historical precedents and withholding numerical data. All patients had been enrolled in the trial >1 year as of April 2019; there should have been ample data to share in June 2019 PLAINVIEW No Evidence Pamrevlumab Works in DMD 94
Cherry-Picking Example • It’s been known since 1989 (see Mendell et al 1989) that glucocorticoid steroid treatment delays muscle loss and the onset of respiratory decline in DMD patients • In FibroGen’s trial, 100% of the patients were given steroids (see Slide 7 of the PPMD presentation) • FibroGen compared its ppFEV1 endpoint with the DELOS trial publication (Meier et al 2017), a trial in which patients were prohibited from taking glucocorticoids • Cross-trial differences such as baseline characteristics and background medications make the comparisons FibroGen provided completely meaningless PLAINVIEW No Evidence Pamrevlumab Works in DMD 95
Point #4: Pamrevlumab Doesn’t Work in Pancreatic Cancer • FibroGen’s ASCO 2018 poster showed that patients in the pamrevlumab arm achieved a 20.8% (5/24) ORR vs. 23.1% (3/13) ORR in the control arm • “Eligibility for surgical resection” is not a valid primary endpoint and FibroGen’s focus on the metric is classic data-mining misdirection in a small-N trial. Legitimate endpoints are ORR, PFS, and OS. PFS and OS were listed among the trial efficacy endpoints, but were conspicuously absent from all disclosures indicating that, like ORR, pamrevlumab showed no benefit in PFS and OS • A large meta-analysis (Gillen et al 2010) showed a near-identical rate of surgical resection following chemotherapy as pamrevlumab + chemotherapy achieved in FibroGen’s trial (33.2% vs 33.3%) – cementing that it is highly unlikely pamrevlumab offers any benefit • The scientific thesis behind using pamrevlumab in pancreatic cancer (targeting the stroma) has been tried multiple times in pancreatic cancer and never succeeded PLAINVIEW Pamrevlumab Doesn’t Work in Pancreatic Cancer 96
Clinical Data Shows No ORR Benefit in Pancreatic Cancer • The RECIST ORR was 20.8% (5/24) in the pamrevlumab arm and 23.1% (3/13) in the control arm, as shown in FibroGen’s ASCO 2018 poster • ORR is especially significant considering that the trial enrolled patients with localized pancreatic cancer – if a drug doesn’t appear to have an additive impact on the target tumor in a localized cancer then it is almost certainly accomplishing nothing • PFS and OS were listed among the efficacy endpoints but were never disclosed, strongly indicating that, like ORR, pamrevlumab showed no benefit in PFS or OS • Instead of disclosing ORR/PFS/OS in its press release, FibroGen only disclosed the proportion of patients who achieved surgical resection in the trial for each group: a blatant attempt at data mining to misdirect investors PLAINVIEW Pamrevlumab Doesn’t Work in Pancreatic Cancer 97
Resection Rate On Par With Historical Precedent • In the Phase 1/2 pancreatic cancer trial, resection was achieved in 33.3% (8/24) of the pamrevlumab group patients compared to 7.7% (1/13) of control group • Gillen et al 2010 conducted a meta-analysis of 111 studies (n=4,394) and found that among pancreatic cancer patients who were initially deemed non-resectable, 33.2% received resection following chemotherapy – nearly identical to the 33.3% achieved in the pamrevlumab trial • We also note that FibroGen’s ASCO poster indicates that none of the three patients in the control arm who saw RECIST responses were deemed eligible for resection. This seems unlikely unless the RECIST responses were significant enough that resection no longer made sense for those patients – in which case the resection rate in the control group is ironically obscured by efficacy PLAINVIEW Pamrevlumab Doesn’t Work in Pancreatic Cancer 98
FibroGen’s Pancreatic Cancer Theory • The theory behind combining pamrevlumab with chemotherapy is that the reason outlook is so bleak for pancreatic cancer patients is because the exceptionally dense extracellular matrix associated with pancreatic cancer blocks a chemotherapeutic or immune response in those patients • However, it is widely thought that the principal culprit behind the enhanced ECM is a constitutively active hedgehog signaling pathway, not CTGF Source: Stromal biology and therapy in pancreatic cancer PLAINVIEW Pamrevlumab Doesn’t Work in Pancreatic Cancer 99
Recommend
More recommend